Cycles of Contingency

Home , Grubbing, Peter Klopfer, Ruth Millikan, Vacuum activity

Cycles of Contingency Cover image: Of related interest Cycles of Contingency Developmental Systems and Evolution A new theoretical approach to the evolution of develop- The Evolution of Cognition Cycles of Contingency edited by Susan Oyama, Paul E. Griffiths, and mental systems and the relationship between genes edited by Cecilia Heyes and Ludwig Huber Russell D. Gray and biological form, such as advocated in this volume, Developmental Systems and Evolution also entails new tools of empirical analysis and hypoth- In the last decade, "evolutionary psychology" has come to refer exclusively to research on The nature/nurture debate is not dead. esis testing. Much of the present methodology of genet- human mentality and behavior, motivated by a nativist interpretation of how evolution Dichotomous views of development still underlie ic, developmental research is geared toward disclosing operates. This book encompasses the behavior and mentality of nonhuman as well as many fundamental debates in the biological and the very proximate roles of individual genes in single, human animals and a full range of evolutionary approaches. Rather than a collection by social sciences. Developmental systems theory mostly two-dimensional, snapshots of developmental and for the like-minded, it is a debate about how evolutionary processes have shaped (DST) offers a new conceptual framework with events. But development is a genuinely three-dimen- cognition. The debate is divided into five sections: Orientations, Categorization, Causality, which to resolve such debates. DST views ontogeny sional process of coordinating cell behavior over time, Consciousness, and Culture. as contingent cycles of interaction among a varied which requires an understanding of the dynamics of set of developmental resources, no one of which interactions among genes, cells, and tissues, which Where Biology Meets Psychology controls the process. These factors include DNA, lead to pattern formation and morphogenesis. New tools for the representation and analysis of embryonic Philosophical Essays cellular and organismic structure, and social and development must be able to take account of these edited by Valerie Gray Hardcastle ecological interactions. DST has excited interest three-dimensional (and ultimately four-dimensional) from a wide range of researchers, from molecular dynamics of development, including the topology and A great deal of interest and excitement surrounds the interface between the philosophy biologists to anthropologists, because of its ability timing of gene expression, the amounts of gene prod- of biology and the philosophy of psychology, yet the area is neither well defined nor well to integrate evolutionary theory and other disci- ucts, and cell proliferation parameters. And they must represented in mainstream philosophical publications. This book is perhaps the first to plines without falling into traditional oppositions. equally deal with the generic physical conditions pres- open a dialogue between the two disciplines. Its aim is to broaden the traditional subject ent in developing tissues, such as adhesiveness, diffu- matter of the philosophy of biology while informing the philosophy of psychology of rele- The book provides historical background to DST, sion, biomechanics, and other epigenetic parameters, vant biological constraints and insights. recent theoretical findings on the mechanisms of including distance, volume, and shape. Such new tools heredity, applications of the DST framework to of phenomic research comprise computer-based 3D behavioral development, implications of DST for the Oyama, Griffiths, and Gray, reconstruction techniques. The cover picture shows a philosophy of biology, and critical reactions to DST. 3D representation of the expression patterns of the muscle developmental regulatory gene Myf5 in a body Susan Oyama is Professor of Psychology, Emerita, segment and the limbs of a mouse embryo. These and at John Jay College, and at the CUNY Graduate other techniques of increasing sophistication (see, e.g., Center, New York City. Paul E. Griffiths is Professor http://www.univie.ac.at/GeneEMAC) not only permit pre- of History and Philosophy of Science at the cise three-dimensional analysis and database storage University of Pittsburgh. Russell D. Gray is Senior of comprehensive gene expression data, but also pro- Lecturer in Psychology at the University of vide a quantitative basis for the mathematical modeling Auckland. of context-dependent cell and tissue behaviors in development and evolution. The MIT Press Life and Mind: Philosophical Issues in Biology and Massachusetts Institute of Technology Psychology Image and text by Gerd B. Müller and Johannes Streicher from the Integrative Morphology Group at Cambridge, Massachusetts 02142 A Bradford Book the University of Vienna, Austria http://mitpress.mit.edu editors OYACH 0-262-15053-0 ,!7IA2G2-bfafdi!:t;K;k;K;k edited by Susan Oyama, Paul E. Griffiths, and Russell D. Gray CYCLES OF CONTINGENCY Life and Mind: Philosophical Issues in Biology and Psychology Kim Sterelny and Rob Wilson, editors

Cycles of Contingency: Developmental Systems and Evolution, Susan Oyama, Paul E. Gri≤ths, and Russell D. Gray, editors, 2001 Coherence in Thought and Action, Paul Thagard, 2001 Norms of Nature: Naturalism and the Nature of Functions, Paul Sheldon Davies, 2001 CYCLES OF CONTINGENCY Developmental Systems and Evolution

edited by Susan Oyama, Paul E. Gri≤ths, and Russell D. Gray

All rights reserved. No part of this book may be reproduced in any form by any electronic or mechanical means (including photocopying, recording, or information storage and retrieval) without permission in writing from the publisher.

This book was set in Times New Roman by Asco Typesetters, Hong Kong in QuarkXpress, and was printed and bound in the United States of America.

Library of Congress Cataloging-in-Publication Data Cycles of contingency: developmental systems and evolution / Susan Oyama, Paul E. Gri≤ths, and Russell D. Gray, editors. p. cm. “A Bradford book.” Includes bibliographical references and index. ISBN 0-262-15053-0 (alk. paper) 1. Developmental psychology. 2. Nature and nuture. 3. Genetic psychology. I. Oyama, Susan. II. Gri≤ths, Paul E. III. Gray, Russell D. BF713 .C93 2000 155.7—dc21 00-026951 Contents

Preface vii 10 Niche Construction, Ecological Inher- Acknowledgments ix itance, and Cycles of Contingency in Contributors xi Evolution 117 Kevin N. Laland, F. John Odling- 1 Introduction: What Is Developmental Smee, and Marcus W. Feldman Systems Theory? 1 III THE DEVELOPMENT OF I INFLUENCES 13 PHENOTYPES AND BEHAVIOR 127

2 Toward a Systems View of Develop- 11 The Ontogeny of Phenotypes 129 ment: An Appraisal of Lehrman’s H. Frederik Nijhout Critique of Lorenz 15 12 The Development of Ant Colony Timothy D. Johnston Behavior 141 3 A Critique of Konrad Lorenz’s Theory Deborah M. Gordon of Instinctive Behavior 25 13 Behavioral Development and Daniel S. Lehrman Darwinian Evolution 149 4 A Developmental Psychobiological Patrick Bateson Systems View: Early Formulation and 14 Parental Care and Development 167 Current Status 41 Peter H. Klopfer Gilbert Gottlieb 5 Gene, Organism and Environment: IV RETHINKING DEVELOPMENT A New Introduction 55 AND EVOLUTION 175 Richard C. Lewontin 15 Terms in Tension: What Do You Do 6 Gene, Organism and Environment 59 When All the Good Words Are Richard C. Lewontin Taken? 177 Susan Oyama II RETHINKING HEREDITY 67 16 Darwinism and Developmental 7 Let’s Talk about Genes: The Process Systems 195 Molecular Gene Concept and Its Paul E. Gri≤ths and Russell D. Gray Context 69 17 Generative Entrenchment and the Eva M. Neumann-Held Developmental Systems Approach 8 Deconstructing the Gene and Recon- to Evolutionary Processes 219 structing Molecular Developmental William C. Wimsatt Systems 85 18 Developmental Systems, Darwinian Lenny Moss Evolution, and the Unity of Science 239 9 The Systems of Inheritance 99 Bruce H. Weber and David J. Depew Eva Jablonka vi Contents

19 From Complementarity to Obviation: On Dissolving the Boundaries between Social and Biological Anthropology, Archaeology, and Psychology 255 Tim Ingold

V RESPONSES TO DEVELOPMEN- TAL SYSTEMS THEORY 281

20 On the Status and Explanatory Structure of Developmental Systems Theory 283 Peter Godfrey-Smith 21 Beyond the Gene but Beneath the Skin 299 Evelyn Fox Keller 22 Distributed Agency within Intersect- ing Ecological, Social, and Scientiﬁc Processes 313 Peter Taylor 23 Niche Construction, Developmental Systems, and the Extended Replicator 333 Kim Sterelny 24 Developmental Systems Theory and Ethics: Different Ways to Be Norma- tive with Regard to Science 351 Cor van der Weele

Index 363 Preface

The idea for this book emerged from sessions at the International Society for History, Philosophy and Social Studies of Biology, ﬁrst at Brandeis University in 1993, then at the University of Leuven, Belgium, in 1995 and the University of Washington in Seattle in 1997. The topic of these sessions was the developmental systems approach to development and evolution. Like the papers in those sessions, the papers collected here do not conform to any “party line.” Instead, the volume aims to explore the implications, potential and limitations of a group of ideas in which there has been a great deal of interest across a range of disciplines. It aims to locate points of contact and points of disagreement between developmental systems theory and a number of other attempts to improve the conceptual framework of the life sciences. Our main hope for the volume is that it will act as a resource for other people intrigued by these ideas, and who wish to make use of them and build upon them in their own disciplines.

Our thanks as editors must go, ﬁrst and foremost, to the twenty-two other contributors to the volume. These authors have been exemplary in their ability to meet deadlines and to engage in the productive discussion with the editors and with one another which was made possible by the timely submission of draft manuscripts. The volume was initially developed on a website, to allow authors to read and respond to one another’s work, as well as to receive comments from the editors. This has helped the chapters to talk to one another and, we hope, given a volume a sense of unity. To all these people, we say many thanks. Particular thanks are also due to Ross West of the University of Sydney, who worked as a research assistant through the development of the volume. In addition to maintaining the website and managing the physical production of the manuscript, Ross drew or redrew many of the illustrations and was always on hand to respond to the authors’ queries. Russell Gray would like to thank Nicola Gavey for her patience and support during the production of this book.

Acknowledgments

“A Critique of Konrad Lorenz’s Theory of Instinctive Behavior” by D. S. Lehrmann ﬁrst appeared in Quarterly Review of Biology 28(4): 337–363 and is reprinted by permission of the University of Chicago Press.

“Gene, Organism, and Environment” by R. C. Lewontin ﬁrst appeared in D. S. Bendall (Ed.), Evolution: From Molecules to Men (Cambridge: Cambridge University Press, 1983), pp. 273–285.

“From Complementarity to Obviation: On Dis- solving the Boundaries Between Social and Biological Anthropology, Archaeology, and Psychology” by T. Ingold ﬁrst appeared in Zeitschrift für Ethnologie 123: 21–52.

Contributors

Patrick Bateson Timothy D. Johnston Sub-Department of Animal Behaviour Department of Psychology University of Cambridge University of North Carolina at Greensboro Cambridge, England Greensboro, North Carolina David Depew Evelyn Fox Keller Communication Studies/POROI Program in Science, Technology, and Society University of Iowa Massachusetts Institute of Technology Iowa City, Iowa Cambridge, Massachusetts Marcus W. Feldman Peter H. Klopfer Department of Biological Sciences Biology Department Stanford University Duke University Stanford, California Durham, North Carolina Deborah M. Gordon Kevin N. Laland Department of Biological Sciences Sub-Department of Animal Behaviour Stanford University University of Cambridge Stanford, California Cambridge, England Peter Godfrey-Smith Daniel S. Lehrman Department of Philosophy †Deceased Stanford University Richard C. Lewontin Stanford, California Museum of Comparative Zoology Gilbert Gottlieb Harvard University Center for Developmental Science Cambridge, Massachusetts University of North Carolina Lenny Moss Chapel Hill, North Carolina Department of Philosophy Paul E. Gri≤ths University of Notre Dame Department of History & Philosophy of Science Notre Dame, Indiana University of Pittsburgh Eva M. Neumann-Held Pittsburgh, Pennsylvania European Academy for the Study of Russell Gray Consequences of Scientiﬁc and Technological Department of Psychology Advance University of Auckland Bad Neuenahr-Ahrweiler, Germany Auckland, New Zealand H. Frederik Nijhout Tim Ingold Department of Zoology Department of Sociology Duke University University of Aberdeen Durham, North Carolina Aberdeen, Scotland F. John Odling-Smee Eve Jablonka Institute of Biological Anthropology Cohn Institute for the History and Philosophy Oxford University of Science and Ideas Oxford, England Tel Aviv University Tel Aviv, Israel xii Contributors

Susan Oyama John Jay College of Criminal Justice and The Graduate Center City University of New York New York, New York Kim Sterelny Department of Philosophy Research School of the Social Sciences Australian National University Australia, and Department of Philosophy Victoria University Wellington, New Zealand Peter Taylor Critical and Creative Thinking Program University of Massachusetts Boston, Massachusetts Cor van der Weele Centre for Bioethics and Health Law Utrecht University Utrecht, The Netherlands Bruce H. Weber Department of Chemistry and Biochemistry California State University, Fullerton Fullerton, California William C. Wimsatt Department of Philosophy University of Chicago Chicago, Illinois CYCLES OF CONTINGENCY

1 Introduction: What Is Developmental Systems Theory?

Susan Oyama, Paul E. Gri≤ths, and Russell D. Gray

The nature/nurture debate is not dead. Open a the attempt to partition causal responsibility for book, read a newspaper, turn on the TV, read the formation of organisms into additive compo- Science or Nature and you will find yourself bom- nents. Such maneuvers do not resolve the nature/ barded with claims and counterclaims. Are there nurture debate; they continue it. This is typical of “genius” genes? If not those, then surely “gay” the way in which the traditional view of devel- ones? Is aggression the consequence of social and opment morphs into new forms when challenged economic conditions, or is it a product of evolu- and returns to plague current academic and so- tion? Are cognitive differences between men and cial debates. If it is no longer acceptable to ask women due to genetics or upbringing? Can we whether something is instinctive, then we ask shape our destiny, or are we robots programmed whether it has a large genetic component. If that, by our selfish genes? These are not esoteric ques- too, becomes unacceptable, then we ask if there is tions, of concern only to a few academic special- a genetic predisposition toward it. What we need ists. Their answers can have social and political is the “stake-in-the-heart move” (Oyama 1985: consequences. People are quick to draw conclu- 27)—a way of thinking about development that sions about the possibility, or even the rightness, does not rely on a distinction between privileged, of trying to subvert nature’s plans. If intelligence essential causes and merely supporting or inter- is innate, then some would say that compensa- fering causes. tory programs are a waste of effort and money. Other concepts in the heartland of contem- If sexual orientation is a biological given rather porary biology, such as inheritance and evolu- than a free choice, then, so it is argued, the lan- tion, need substantial reformulation as well. The guage of morality in that context is both inap- reliability of many aspects of development has propriate and useless. encouraged biologists, psychologists, and social scientists to postulate some central directing Underlying these vexed questions are a number agency or “master molecule.” Inheritance and of oppositions: nature or nurture, genes or envi- evolution are defined as the passing on and alter- ronment, biology or culture. Developmental sys- ation of such master molecules. Other inputs tems theory (hereafter DST) is an attempt to do to development tend to be lumped together as biology without these dichotomies. This is a more “environment” and treated as a standard back- di≤cult task and requires a greater theoretical re- ground that is not itself in need of explanation. In working of biological concepts than has so far contrast, DST views both development and evo- been realized. The standard response to nature/ lution as processes of construction and recon- nurture oppositions is the homily that nowadays struction in which heterogeneous resources are everyone is an interactionist: All phenotypes are contingently but more or less reliably reassem- the joint product of genes and environment. Ac- bled for each life cycle. It is these cycles of con- cording to one version of this conventional “in- tingency that we need to unpack, and it is these teractionist”1 position, the real debate should not cycles that are the subject of this book. be about whether a particular trait is due to nature or nurture, but rather how much each “influences” the trait. The nature/nurture debate is thus So What Exactly Is DST? allegedly resolved in a quantitative fashion. The question is no longer whether intelligence is in- What we have come to term developmental sys- nate or acquired, but instead whether intelligence tems theory is not a theory in the sense of a spe- is 50 percent or 70 percent genetic. DST rejects cific model that produces predictions to be tested 2 Introduction

against rival models. Instead, it is a general theo- is supposed to have dissolved. There are many retical perspective on development, heredity and kinds of influences on development, and there are evolution, a framework both for conducting many ways to group these interactants together. scientific research and for understanding the DST does not claim that all these sources of cau- broader significance of research findings.2 Many sal influence play the same role, nor that all are other biological theories play this dual role, per- equally important (whatever that might mean). haps most noticeably in recent years the idea of Rather, different groupings of developmental fac- the “selfish gene” (see chapter 20). tors are valuable when addressing different ques- Developmental systems theory is not attributa- tions. The distinction between genes and every ble to one person or group. It draws on insights other causal factor in development (“envi- from researchers in a wide range of areas who ronment”) is just one more grouping, possibly have been dissatisfied with crude dichotomous helpful for some purposes, much less so for many accounts of development and have attempted to others. Many developmentally constructive inter- formulate an alternative.3 Table 1.1 draws on a actions do not fit traditional categories, and for number of past attempts to specify a list of tenets this reason have largely been overlooked or mar- of DST (Gray 1992, 1997; Schaffner 1998; ginalized. Oppositions between genes (or biolo- Gri≤ths and Knight 1998; Oyama 2000b, gy) and learning, or between genes (or biology) Introduction). Programs of scientific research are and culture, are endemic to many fields but are not easily reduced to a set of precisely stated miserably inadequate for capturing the multitude axioms (Kuhn 1970), so these tenets are more of causal factors needed for any reasonable treat- like what Schaffner (1998) has called “themes” of ment of ontogeny or phylogeny. DST emphasizes developmental systems research. In the rest of crucial but often overlooked similarities among this section we expand and comment on these six resources that are usually contrasted. Pheno- themes. copying, for instance, occurs when genetic mutations, as well as changes in the outside Joint Determination by Multiple Causes world, can bring about similar alterations in the organism (Markert and Ursprung 1971; It is a truism that all traits are influenced by both Waddington 1975; see also Oyama 1981 on the genetic and nongenetic factors. According to significance of this causal parity or symmetry in DST, however, this “interactionist consensus” is phenocopying from the perspective of DST). little better than the nature/nurture dispute it There are bithorax mutants in Drosophila, but the

Table 1.1 Major themes in developmental systems theory 1. Joint determination by multiple causes—every trait is produced by the interaction of many developmental resources. The gene/environment dichotomy is only one of many ways to divide up these interactants. 2. Context sensitivity and contingency—the signiﬁcance of any one cause is contingent upon the state of the rest of the system. 3. Extended inheritance—an organism inherits a wide range of resources that interact to construct that organism’s life cycle. 4. Development as construction—neither traits nor representations of traits are transmitted to offspring. Instead, traits are made—reconstructed—in development. 5. Distributed control—no one type of interactant controls development. 6. Evolution as construction—evolution is not a matter of organisms or populations being molded by their environments, but of organism-environment systems changing over time. What Is Developmental Systems Theory? 3

bithorax phenotype can also be induced by ether. havior geneticists these days are quick to admit Genes and ether shocks turn out to be develop- that high heritability scores do not show that it mentally equivalent in this respect. Phenomena is hard to alter the trait by nongenetic means. that are usually contrasted to one another can be Whenever a number of causal factors interact to equivalent in evolution, too. Developmental produce an outcome, we should expect that the influences may follow a lineage equally closely effect of changing one factor will depend on what through evolution, even though one is genetic is happening to the others. However, current the- and the other “environmental”—genes and oretical frameworks encourage much more re- dietary cues, for example (see chapters 16 and search on “genes for” traits than on statistical 23). These often overlooked similarities form part interactions among developmental factors in nat- of the evidence for DST’s claim of causal parity ural populations (see Schlichting and Pigliucci between genes and other factors of development. 1998 for a refreshing change). The “parity thesis” (Gri≤ths and Knight 1998) The persistent tendency to minimize context does not imply that there is no difference between sensitivity and developmental contingency when the particulars of the causal roles of genes and studying genetic factors in development is con- factors such as endosymbionts or imprinting nected to the prevalence of information meta- events. It does assert that such differences do not phors in contemporary biology (Keller 1985; justify building theories of development and evo- Oyama 1985). As long as the DNA is thought of lution around a distinction between what genes as containing information about developmental do and what every other causal factor does. outcomes, it will seem sensible to inquire whether outcomes occur because they are represented in Context Sensitivity and Contingency the chromosomes. Once an outcome is seen as an expression of the genetic information that con- The demands of interactionism in its conven- trols development, it acquires a special status. It tional form can often be satisfied by merely ad- represents what the organism is “meant to be,” mitting that every organism must have some and deviations from it are misrepresentations of genes and some environment. With that out of the true nature of the organism—its inner es- the way, the real business of settling what is due sence, which was conferred on it at the moment to nature and what is due to nurture can con- of conception (at least in those organisms that tinue. Typically, this work proceeds by inferring have such moments). In such an intellectual more or less directly the extent of the correlation framework, context sensitivity is often treated between genes and phenotype in one or more as interference with the basic pattern of biolog- populations. These methods can be as direct as ical causation. For DST, contingency is basic, molecular screening or as indirect as a study of whether the results are expected or surprising. monozygotic twins raised apart. The underlying logic is very similar across this whole range of Extended Inheritance methods: The stronger the correlation, the more the genes are said to be responsible for the trait. A traditional way to privilege genes over other However, as Richard Lewontin has argued, heri- causes in development is to argue that genes are tability estimates are not measures of global cau- the only things organisms inherit from their an- sal importance, nor do they indicate how much a cestors. Hence the biological nature of organisms trait can be modified by environmental changes must be in the genes. DST insists on a defini- (Lewontin 1974). Heritability estimates measure tion of inheritance that explicitly recognizes the the proportion of variation in a specific popula- wide range of resources that are “passed on” and tion that is attributable to genetic differences. Be- are thus available to reconstruct the organism’s 4 Introduction

life cycle. Some of these resources are familiar— in size, not the generation of biological order chromosomes, nutrients, ambient temperatures, (Pinto-Correia 1997). Modern preformationism childcare. Others are less familiar, despite the re- is subtler: The organism is not preformed in the cent explosion of work on “epigenetic inheri- egg, but the information that programs its de- tance” (see chapter 9). These include chromatin velopment is preformed in the genes. By contrast, marks that regulate gene expression, cytoplasmic an epigenetic account of development is one that chemical gradients and gut- and other endosym- never sidesteps the task of explaining how a de- bionts. Another important topic in recent biology velopmental outcome is produced. The claim that is the participation of the organism in the con- development occurs because it is programmed to struction of its niche. Hence a further aspect of occur or because it has been selected by evolution inheritance is the local physical environment, is merely a promissory note redeemable against altered by past generations of the same species future developmental biology. Similar views and other species as well as the organism’s own about biological explanation have been described activities (see chapters 10 and 12). Many of elsewhere as “constructivist interactionism” these inherited resources have distinctive roles. (Callebaut and Stotz, forthcoming; Oyama 2000b; DNA is unique in acting as templates for protein see also chapter 15, this volume). Despite its synthesis. Membranes are unique in acting as clumsiness, this phrase succinctly expresses two templates for the assembly of proteins into more major themes of developmental systems theory. membrane. Chemical traces from foraging play a The life cycle of an organism is developmentally characteristic role in diet choice in many rodents. constructed, not programmed or preformed. It A written text, interpreted in an enormously com- comes into being through interactions between plex personal and cultural setting, is distinctive in the organism and its surroundings as well as in- yet other ways. DST explores these diverse roles, teractions within the organism. acknowledging differences but resisting any attempt to divide them into, say, one set contribut- The conviction that development involves multi- ing to the organism’s “nature” and another that ple, interdependent causes is entirely compatible inﬂuences its “nurture.” Oyama has argued that with the practical requirements of research. Prac- if these vexed terms are to be retained, then ticing researchers have to draw boundaries “nature” should refer simply to the outcomes of around the system to be studied, placing certain development and “nurture” to the processes that factors in the foreground while taking others for produce, maintain, and alter those outcomes granted. In this book the contributions of Gilbert (Oyama 1985). Gottlieb, Frederik Nijhout, Deborah Gordon, Pat Bateson, and Peter Klopfer show that a stra- Development as Construction tegic concern with complexity and context dependence need not interfere with the process Developmental systems theory is a thoroughly of identifying individual causal contributions by epigenetic account of development. Current use controlling other variables. The practice of of term “epigenetic” is ambiguous. It is often changing one variable at a time while holding used to mean “in addition to the genes,” as in the others constant is important, but it is incom- phrase “epigenetic inheritance” (see chapter 9). plete. Additional investigation is required, both Its basic meaning, however, derives from a con- to show how a causal factor is coupled in a sys- trast between preformationist and epigenetic tem of causes and to reveal the ways in which theories of development. Classical preformation- these links change over time. It does not require ism held that the egg contains a tiny organism, so considering everything at once, as some seem to that embryologists only had to explain increases fear, but can be done by coordinating diverse What Is Developmental Systems Theory? 5

investigations. In fact, having a richer strategic are held constant, the outcomes of development vision should allow researchers to make more in- can give us information about the genes, but if telligent and ﬂexible “tactical” use of reduction- the genes are held constant, outcomes give us istic research strategies (Wimsatt 1980; see also information about whichever other resource we chapter 17). have decided to let vary. Every resource whose state affects development could thus be consid- Distributed Control ered a “source of developmental information.”

Taking a systems perspective on developmental Maynard Smith (2000) has recently suggested processes means, among other things, attending that this does not capture what is really meant to the ways in which the developing organism when biologists talk about genetic information. functions as a resource for its own further de- According to him, biologists use information in velopment. The organism helps determine which an intentional or semantic sense—a gene has an other resources will contribute to that develop- intended meaning as well as causal consequences, ment, as well as the impact they will have. The and this intentional sense of information reveals roles played by the vast and heterogeneous as- the true asymmetry between genes and other sembly of interactants that contribute to a life- developmental resources. Sterelny and Gri≤ths course are system-dependent and change over say: “A distinctive test of intentional or semantic time. So DST creates an inhospitable context for information is that talk of error or misrepresenta- moves that preempt the investigation of actual tion make sense. A map of Sydney carries seman- processes by identifying one type of resource as tic information about the layout of Sydney. controlling or directing the process, leaving other Hence it makes sense to say of any putative map interactants to function as background condi- that it is wrong or that it has been misread” tions, raw materials, or sources of disturbance. (1999: 104). We believe that despite the widespread talk of Maynard Smith draws on Ruth Millikan’s genetic blueprints and programs in contemporary (1984) attempt to explain intentional informa- biology, there is no scientiﬁcally defensible sense tion in evolutionary terms. This “teleosemantic” in which a subset of developmental resources theory says that things carry intentional informa- contains a program or set of instructions for tion about whatever evolution has selected them development.4 to represent; a gene contains intentional information about a phenotype that it has been selected The most obvious way to defend talk of genetic to produce. But the same can be said of any programs and genetic information is to argue that developmental resource whose presence can be it is in some unproblematic sense related to infor- given an evolutionary explanation. As we argued mation theory. “Information” in this statistical in the section on extended inheritance, many sense is the systematic dependence of a signal on kinds of developmental resources are inher- a source, a dependence that is created by a set of ited and evolve. The teleosemantic account does channel conditions. In the case of development, not show that only genes carry developmental the genes are typically taken to be the source, so information. the channel conditions are all the other resources We believe that the heuristic value of the idea needed for development to occur. But in infor- of developmental information in certain contexts mation theory the role of source and channel con- is more than outweighed by its misleading condition can be reversed. A source is simply one notations. Locating information in a single type channel condition whose current state the signal of developmental resource obscures the context- is being used to investigate. If all other resources dependency of causation by localizing control. 6 Introduction

Some contributors to the present volume, how- we believe that it does not go far enough (Oyama ever, are happier to deploy information concepts 1987). Like Levins and Lewontin (1985), we than we are, and do so in their chapters. think that the fundamental problem lies in the way causation is viewed in biological systems. Evolution as Construction This book does not therefore focus on debates about the (mis)application of biological concepts The idea of construction through the interaction and techniques in arguments about intelligence, of many different factors is applicable to evolu- aggression, and gender, but rather on the con- tion as well as development, and it highlights cepts themselves: on the kinds of understand- striking similarities between the two processes ings of biology, development, and evolution that (Oyama 1992). Just as there are no preexisting make poor practice likely and render critiques of representations or instructions that shape organ- that practice less than optimally effective. The isms from within, there are no preexisting niches purpose of the present collection is thus to pro- or environmental problems that shape popula- vide a forum for exploring DST’s alternative tions from without (see chapter 6). Evolutionary conceptions of development, heredity, and evolu- change is the result of interactions in which out- tion. The contributors did not necessarily have to comes are codetermined, or co-constructed, by identify themselves as proponents of DST, nor populations and environments with their own, were they required to adhere to any party line. often intricately interrelated, histories and char- In fact, some were selected because they could acteristics; outcomes are not imposed by or pre- offer informed criticism. We selected contributors figured in only certain interactants. Extended whose writing promised to extend, illustrate, inheritance both increases the range of develop- challenge, parallel, or contrast with issues raised mental outcomes that can be given evolutionary in DST. Many were invited to present their own explanations and alters our view of evolutionary empirical work, but they were asked to confront dynamics (see chapter 16). If evolution is change developmental systems ideas in a substantive way in developmental systems, then, as just noted, it is as well. In addition, authors were sometimes en- no longer possible to think of evolution as the couraged to address key concepts such as genetic shaping of the organism to fit an environmental information, program, and transmission. niche. Rather, the various elements of the devel- Many of the papers collected here aim at noth- opmental systems coevolve. Organisms construct ing less than a root-and-branch transformation their niches both straightforwardly by physically of contemporary biological thought. We do not transforming their surroundings and, equally im- expect such large hopes to be fulfilled in the importantly, by changing which elements of the mediate future, but we will be gratified if we can external environment are part of the develop- convince some readers that there are some real mental system and thus able to influence the evo- objections to ideas that have previously seemed lutionary process in that lineage. unproblematic—ideas such as the gene/environment dichotomy and the genetic program for development. We also hope that by drawing Aim of This Book attention to alternative ways of thinking about biological systems, and to some programs of A common response to the confusion and polit- research that embody them, we will persuade ical mischief caused by the nature/nurture debate others to try out some of these alternatives in is to scold the participants for their empirical their own work. sloppiness and technical errors (Kitcher 1985, Part I deals with key influences on develop- 2001). Though this is often a valuable exercise, mental systems thinking, starting with Daniel S. What Is Developmental Systems Theory? 7

Lehrman’s classic 1953 critique of the concept of and describing the relationship between the genes innateness. As Timothy Johnston shows in his of classical transmission genetics and those of introduction to this essay, aspects of what we molecular biology has proved extremely di≤cult would now call a systems view of development (Falk 1984, 1986; Kitcher 1982, 1984; Sarkar can be discerned in Lehrman’s work and in that 1998). Neumann-Held and Moss show how a de- of other animal behavior researchers such as Kuo velopmental perspective can move this debate (1967) and Schneirla (1966). The work of these forward, in part by revealing how the neglect of early critics of traditional understandings of in- development has distorted theoretical concep- stinct and maturation remains relevant because tions of genes and gene action and marginalized those understandings are still with us. Eschewing other developmental factors. Despite these simi- pseudo-explanatory notions such as instinct and larities in approach, Moss and Neumann-Held innateness allows us to ask a host of questions propose quite different reconceptualizations of about the actual causal influences at each stage of the gene. Both are grounded in the practice of development. Gilbert Gottlieb has devoted his contemporary genetics, and both provide ways career to elucidating these influences. In chapter to integrate developmental systems thinking into 4 he looks back on this long and fruitful history genetics. The future dialectic between these two of research and outlines the conceptual frame- views should prove illuminating. work he has helped build. The ongoing research The remaining chapters in part II reflect two tradition stemming from Lehrman, Kuo, and exciting innovations in recent evolutionary biol- Schneirla and ably represented by Gottlieb (see ogy. Eva Jablonka’s book with Martha Lamb, Michel and Moore 1995 for a recent text in this Epigenetic Inheritance and Evolution (Jablonka tradition) provides a standing reply to the chal- and Lamb 1995), makes it di≤cult to go on min- lenge that developmental dichotomies are the imizing the theoretical and empirical significance only way to render complex developmental sys- of extragenetic mechanisms of cellular heredity. tems amenable to empirical study. Until now, phenomena such as DNA imprint- In the remainder of part I, Richard Lewontin ing have been assimilated by conventional neo- examines the implications of his constructionist Darwinism by invoking developmental programs view of development and evolution for research in the cell or by arguing that their impact on evo- in genetics in a new introduction to his classic lution is small when compared to genetic inheri- paper “Gene, Organism and Environment” tance. Needless to say, Jablonka herself does not (1983) which is reproduced here as chapter 6. In accept such maneuvers, but rather traces out the this and other papers published around the same full implications of these phenomena for theories time, Lewontin questioned the traditional model of development and evolution. of adaptation in which the environment acts on a Kevin Laland, John Odling-Smee, and Marcus passive organism and fits it to a preexisting eco- Feldman introduce their work on the signifi- logical niche. Lewontin’s discussion of the role of cance of niche construction, giving many exam- organism in constructing its own niche is elabo- ples of the reciprocal influence of organisms and rated elsewhere by Kevin Laland, John Odling- their surroundings that Lewontin described in Smee, and Marcus Feldman (chapter 10) and by the early 1980s. Laland and his coauthors show Paul Gri≤ths and Russell Gray (chapter 16). how models of gene-culture coevolution can be Part II looks at attempts to reformulate the adapted to explore such phenomena. The theme idea of heredity so as to do justice to the facts of of niche construction runs through much of this development. The essays by Eva Neumann-Held volume, and it is treated in a variety of ways and Lenny Moss focus on the concept of the gene (compare chapters 10, 16, and 19, for instance). itself. Tracking the changing definition of “gene” There is still some uncertainty about the best 8 Introduction

way to integrate this important concept into our plications. She suggests that although such his- picture of evolutionary change. tories complicate the theoretician’s (and reader’s) Even readers who are sympathetic to the con- task, they can also be put to good use in elabo- cept of a developmental system and aware of the rating an account of life processes that is ade- problems it helps to resolve sometimes ask what quate to the phenomena of developmental and difference it makes in practice. The chapters in evolutionary biology and the social sciences. part III, like Gottlieb’s chapter in part I, show In “Darwinism and Developmental Systems,” how the ideas of DST are deployed in actual Paul Gri≤ths and Russell Gray systematically research. The papers in this section make it clear redefine in developmental systems terms the key that similar ideas have been generating important concepts of evolutionary theory: inheritance, nat- work in a number of fields for many years. One ural selection, adaptation, and lineage. They aim aim of the present volume is to bring these re- to show that the DST formulation of evolution search traditions into closer contact with one can do all the explanatory work of the conven- another and to show how they complement one tional ones and can actually extend the range of another. H. Frederik Nijhout provides a develop- phenomena that can be given adaptive/historical mental perspective on the genotype/phenotype explanations. Continuing the focus on evolution, relationship. He shows what happens when a William Wimsatt’s chapter considers the implica- simple but realistic developmental model of how tions of the fact that evolution must operate by genes influence phenotypic traits is added to a producing systems capable of reliably reconstruct- conventional population genetic treatment of ing themselves, but also capable of evolutionary evolution. The results of this modeling exercise change. He argues that this has immediate im- challenge the conventional idea that the selection plications for the sorts of systems that can evolve. of phenotypes results in the selection of “genes Like Wimsatt, Bruce Weber and David Depew for” those phenotypes. Deborah Gordon gives a are longstanding advocates of the need for a DST-style treatment of an entity above the level resynthesis of evolution and development. In of the individual organism (an ant colony and its their chapter they explore the prospects for DST nest). She shows how behavioral patterns that are as a source of that new synthesis and also as a both complex and flexible can be regulated with- way to forge links between biology and the phys- out a central locus of control. Transindividual ical sciences. units also figure in Peter Klopfer’s contribution. Part IV concludes with Tim Ingold’s examina- Klopfer examines relations between infants and tion of biology/culture oppositions in anthropol- parents, and evaluates several popular metaphors ogy. He shows how much of our understanding of development. Patrick Bateson reviews some of human life relies on ignoring our role in the pervasive confusions over heritability and devel- construction of our environment, and hence in opment, and gives examples of the kind of adap- the construction of ourselves. Ingold also ex- tive developmental flexibility that has been a plores links between the gene/environment and focus of his long-term interest in fusing the devel- biology/culture oppositions and, perhaps the opmental and functional perspectives. most famous of all these dichotomies, that of Part IV deals with the overall impact that de- mind and body. velopmental systems ideas can have on biological Part V contains essays more concerned to theory. Susan Oyama’s “Terms in Tension” con- explore and evaluate DST than to add to it or siders the di≤culties of employing terms such as provide arguments in its support. Peter Godfrey- interaction, system, and construction, which have Smith confronts the question raised earlier: Is complex histories and conflicting theoretical im- DST a contribution to biological science or What Is Developmental Systems Theory? 9

something more like a “philosophy of nature”? introduction, problems that have motivated In doing so, he provides valuable insights into many of the authors collected here to pay closer the variety of functions performed by scientific attention to the ways in which we conceptualize approaches in general. Kim Sterelny evaluates our natures. DST’s attempt to extend the concept of inheritance, taking up the challenge implicit in DST’s claim to provide a principled definition of Notes inheritance. He argues that conventional neo- 1. The term interactionism is widely used but has such Darwinism has principled reasons for exclud- a broad spectrum of meanings as to be almost useless. ing some of what DST wants to include. Central In chapter 15, Oyama examines some of these multiple DST themes of historical contingency, distrib- senses and the problems caused by their coexistence. uted control, and heterogeneity of developmental 2. “Approach” or “perspective” might thus be prefer- interactants are taken up in Peter Taylor’s chap- able, but the DST label seems to have stuck. Fausto- ter. Like several other authors, Taylor discusses Sterling’s (2000) excellent treatment of research on sex- the problem of allocating causal responsibility in uality shows this perspective at work in a field especial- complex systems, and, in addition, he addresses ly laden with just the sorts of social freight alluded to the crucial questions of agency and the limi- earlier. tations of trying to bring about change solely 3. In the areas of animal behavior and psychobiology through the medium of ideas. these include Bateson (1984, 1991), Gottlieb (1997), Hinde (1968), Johnston (1982, 1987), Klopfer (1973), The work of Evelyn Fox Keller also resonates Kuo (1967), Lehrman (1953, 1970), and Schneirla strongly with many DST themes. This is surely (1966). In genetics, Lewontin (1974) and Waddington the case in her analysis of the metaphors that (1975); in developmental biology, Nijhout (1990); and have sustained dichotomous thinking in modern in molecular biology, Stent (1981). biology (Keller 1985). In her contribution to this 4. Argument and evidence for this view, which will book, however, Keller argues that contrary to strike many people as surprising, can be found in DST’s emphasis on the multiplicity of possible Godfrey-Smith (1999); Gray (2001); Gri≤ths and Gray boundaries, some boundaries really are special. (1994); Johnston (1987); Moss (1992); Nijhout (1990); In particular, she argues that DST is in danger of Oyama (1985); Sarkar (1996); Sterelny and Gri≤ths neglecting the importance of the cell membrane (1999); Strohman (1997). and of the skin that comes to define the body. In the final chapter of part V, Cor van der References Weele takes up the issue of the relationship between scientific approaches and ethical sys- Bateson, P. P. G. (1984). Genes, evolution, and learn- tems, making explicit some of the concerns about ing. In P. Marler and H. S. Terrace (Eds.), The Biology science/society relations that are implicit in other of Learning, pp. 75–88. Berlin: Springer-Verlag. chapters. She also discusses the “critical science” Bateson, P. P. G. (1991). Are there principles of be- tradition to which some of the authors collected havioural development? In P. Bateson (Ed.), The De- here have contributed. Her concept of an ethics velopment and Integration of Behaviour, pp. 19–39. of attention is relevant to the use of a systems Cambridge: Cambridge University Press. perspective to counteract the hierarchies of im- Callebaut, W., and K. Stotz. (forthcoming). Bio- portance built into prevailing ways of thinking. epistemology and the Challenge of Development and In raising these issues, van der Weele brings us Sociality. Cambridge, MA: MIT Press. back to the social, economic, and political prob- Falk, R. (1984). The gene in search of an identity. lems that were alluded to in the beginning of this Human Genetics 68: 195–204. 10 Introduction

Falk, R. (1986). What is a gene? Studies in History and Keller, E. F. (1985). Refiguring Life: Metaphors of Philosophy of Science 17: 133–173. Twentieth Century Biology. New York: Columbia Fausto-Sterling, A. (2000). Sexing the Body: Gender University Press. Politics and the Construction of Sexuality. New York: Kitcher, P. (1982). Genes. British Journal of Philosophy Basic Books. of Science 33: 337–359. Godfrey-Smith, P. (1999). Genes and codes: Lessons Kitcher, P. (1984). 1953 and all that: A tale of two from the philosophy of mind? In V. Hardcastle (Ed.), sciences. Philosophical Review 93: 335–373. Biology Meets Psychology: Constraints, Conjectures, Kitcher, P. (1985). Vaulting Ambition: Sociobiology and Connections, pp. 305–331. Cambridge, MA: MIT Press. the Quest for Human Nature. Cambridge, MA: MIT Gottlieb, G. (1997). Synthesizing Nature-Nurture: Pre- Press. natal Roots of Instinctive Behavior. Mahwah, NJ: Kitcher, P. (2001). Battling the undead: How (and how Lawrence Erlbaum. not) to resist genetic determinism. In R. Singh, K. Gray, R. D. (1992). Death of the gene: Developmental Krimbas, J. Beatty, and D. Paul (Eds.), Thinking about systems strike back. In P. E. Gri≤ths (Ed.), Trees of Evolution: Historical, Philosophical and Political Per- Life: Essays in Philosophy of Biology, pp. 165–209. spectives, pp. 396–414. Cambridge: Cambridge Univer- Dordrecht: Kluwer Academic. sity Press. Gray, R. D. (1997). “In the Belly of the Monster”: Klopfer, P. H. (1973). On Behavior: Instinct Is a Feminism, developmental systems, and evolutionary Cheshire Cat. Philadelphia: Lippincott. explanations. In P. A. Gowaty (Ed.), Evolutionary Biol- Kuhn, T. (1970). The Structure of Scientific Revolutions. ogy and Feminism, pp. 385–413. New York: Chapman 2d ed. Chicago: University of Chicago Press. and Hall. Kuo, Z.-Y. (1967). Dynamics of Behavior Development. Gray, R. D. (2001). Selfish genes or developmental (2d ed. 1976.) New York: Plenum. systems? Evolution without replicators and vehicles. In Lehrman, D. S. (1953). A critique of Konrad Lorenz’s R. Singh, C. Krimbas, J. Beatty, and D. Paul (Eds.), theory of instinctive behavior. Quarterly Review of Thinking about Evolution: Historical, Philosophical Biology 28: 337–363. and Political Perspectives, pp. 184–207. Cambridge, England: Cambridge University Press. Lehrman, D. S. (1970). Semantic and conceptual issues in the nature-nurture problem. In L. R. Aronson, E. Gri≤ths, P. E., and R. D. Gray (1994). Developmental Tobach, D. S. Lehrman, and J. S. Rosenblatt (Eds.), systems and evolutionary explanation. Journal of Phi- Development and Evolution of Behavior: Essays in losophy 91: 277–304. Memory of T. C. Schneirla, pp. 17–52. San Francisco: Gri≤ths, P. E., and R. D. Knight (1998). What is the W. H. Freeman. Philosophy of Science developmentalist challenge? 65: Levins, R., and R. Lewontin. (1985). The Dialectical 253–258. Biologist. Cambridge, MA: Harvard University Press. Hinde, R. A. (1968). Dichotomies in the study of devel- Lewontin, R. C. (1974). The analysis of variance and opment. In J. M. Thoday and A. S. Parkes (Eds.), the analysis of cause. American Journal of Human Genetic and Environmental Influences on Behavior, pp. Genetics 26: 400–411. 3–14. New York: Plenum. Lewontin, R. C. (1983). Gene, Organism and Environ- Epigenetic Inher- Jablonka, E., and M. J. Lamb. (1995). ment. In D. S. Bendall (Ed.), Evolution: From Molecules itance and Evolution: The Lamarkian Dimension. Ox- to Men, pp. 273–285. Cambridge: Cambridge Univer- ford: Oxford University Press. sity Press. Johnston, T. D. (1982). Learning and the evolution Lewontin, R. C., S. Rose, and L. J. Kamin. (1984). Not of developmental systems. In H. C. Plotkin (Ed.), in Our Genes. New York: Pantheon. Learning, Development, and Culture, pp. 411–442. New Develop- York: Wiley. Markert, C. L., and H. Ursprung. (1971). mental Genetics. Englewood Cliffs, NJ: Prentice-Hall. Johnston, T. D. (1987). The persistence of dichotomies in the study of behavioral development. Developmental Maynard Smith, J. (2000). The concept of information Philosophy of Science . Review 7: 149–182. in biology. 67: 177–194 What Is Developmental Systems Theory? 11

Michel, G. F., and C. L. Moore. (1995). Developmental Schneirla, T. C. (1966). Behavioral development and Psychobiology: An Interdisciplinary Science. Cam- comparative psychology. Quarterly Review of Biology bridge, MA: MIT Press. 41: 283–302. Millikan, R. G. (1984). Language, Thought, and Other Stent, G. (1981). Strength and weakness of the genetic Biological Categories. Cambridge, MA: MIT approach to the development of the nervous system. In Press. W. M. Cowan (Ed.), Studies in Developmental Neuro- Moss, L. (1992). A kernel of truth? On the reality of biology, pp. 288–320. Oxford: Oxford University Press. the genetic program. In D. Hull, M. Forbes, and K. Strohman, R. C. (1997). The coming Kuhnian revolu- Okruhlik (Eds.), Proceedings of the Philosophy of tion. Nature Biotechnology 15: 194–200. Science Association, 1992, vol. 1, pp. 335–348. East Sterelny, K., and P. E. Gri≤ths. (1999). Sex and Death: Lansing, MI: Philosophy of Science Association. An Introduction to the Philosophy of Biology. Chicago: Nijhout, H. F. (1990). Metaphors and the roles of genes University of Chicago Press. in development. Bioessays 12: 441–446. Waddington, C. H. (1975). The Evolution of an Evolu- Oyama, S. (1981). What does the phenocopy copy? tionist. Ithaca, NY: Cornell University Press. Psychological Reports 48: 571–581. Wimsatt, W. C. (1980). Reductionistic research strate- Oyama, S. (1985). The Ontogeny of Information: Devel- gies and their biases in the units of selection contro- opmental Systems and Evolution. (2d ed., 2000.) Cam- versy. In T. Nickles (Ed.), Scientiﬁc Discovery: Case bridge: Cambridge University Press. Durham, NC: Studies, pp. 213–259. Dordrecht: Reidel. Duke University Press. Oyama, S. (1987). Review of P. Kitcher, “Vaulting Ambition.” Canadian Philosophical Review 7: 203– 205. Oyama, S. (1992). Ontogeny and phylogeny: A case of metarecapitulation? In P. E. Gri≤ths (Ed.), Trees of Life: Essays in Philosophy of Biology, pp. 211–240. Dordrecht: Kluwer. Oyama, S. (2000a). Causal democracy and causal contributions in DST. Philosophy of Science, 67 (proceedings): 332–347. Oyama, S. (2000b). Evolution’s Eye: A Systems View of the Biology-Culture Divide. Durham, NC: Duke Uni- versity Press. Pinto-Correia, C. (1997). The Ovary of Eve. Chicago: University of Chicago Press. Sarkar, S. (1996). Biological information: A sceptical look at some central dogmas of molecular biology. In S. Sarkar (Ed.), The Philosophy and History of Molecu- lar Biology: New Perspectives, pp. 187–232. Dordrecht: Kluwer. Sarkar, S. (1998). Genetics and Reductionism. Cam- bridge: Cambridge University Press. Schaffner, K. F. (1998). Genes, behavior, and developmental emergentism: One process, indivisible? Philoso- phy of Science 65: 209–252. Schlichting, C. D., and M. Pigliucci. (1998). Phenotypic Evolution: A Reaction Norm Perspective. Sunderland, MA: Sinauer.

I INFLUENCES

Toward a Systems View of Development: An Appraisal of Lehrman’s Critique 2 of Lorenz

Timothy D. Johnston

The work of Daniel S. Lehrman provides the behavior was relatively uncontroversial in psy- conceptual foundation for a great deal of the em- chology. Inherited behavior (instinct) was under- pirical and theoretical work on behavioral devel- stood to be an inherent part of the individual’s opment that has been undertaken in the last half makeup that resulted from the evolutionary his- century. Most of his publications are empirical tory of the species to which it belonged. Instincts contributions to the literature on behavioral were especially important in nonhuman animals, endocrinology, and many of those are among although they were also thought to account for at the most widely cited papers in that field. His least some human behavior. Although the mech- broader impact on our thinking about behavioral anisms of inheritance were not well understood development and evolution comes primarily from (see Maienschein 1987; Sapp 1983), it was gener- two theoretical publications: his 1953 critique of ally, if implicitly, assumed that one could speak Konrad Lorenz’s theory of instinct, excerpted of the inheritance of behavior as straightfor- in the chapter that follows, and a later chapter wardly as one could speak about the inheritance (Lehrman 1970) that reiterates and extends the of a morphological or physiological trait, about ideas presented in the earlier paper. Between which no controversy existed. them, those two publications contain what is still Evolutionary writers like Charles Darwin one of the clearest statements of the systems view (1871), George Romanes (1888), and Conwy of development, modern articulations of which Lloyd Morgan (1895) used the concept of in- are represented by the chapters in the present vol- stinct to explain the behavior and mental abilities ume. Lehrman himself did not use the termi- of human and nonhuman animals, but it was nology of systems thinking, but his analysis of William James and William McDougall whose development illustrates many of the most sign- writings were most responsible for stimulating a ificant features of modern developmental systems debate over the explanatory utility of the concept theory (DST). in psychology. James (1890) drew on Darwinian As implied by the title of his 1953 paper, evolutionary theory in establishing functionalism Lehrman articulated his developmental views as an important force in American psychology in reaction to Lorenz’s presentation of classical around the turn of the last century. James defined ethological instinct theory, which emerged in the instinct as “the faculty of acting in such a way as writings of Lorenz and his collaborators, espe- to produce certain ends, without foresight of the cially Nikolaas Tinbergen, during the 1930s and ends, and without previous education in the per- 1940s (see Lehrman’s bibliography for citations formance” (James 1890, vol. 2: 383), emphasizing to this literature). Lorenz’s was not, of course, the both the teleological and the nativistic aspects of first theory of instinct to engage the attention of instinct. In James’s wide-ranging and rather dif- psychologists, nor was Lehrman’s the first criti- fuse psychology, instinct was just one among cism of the concept of instinct or of the categories many determinants of behavior. In McDougall’s of inherited and acquired behavior as useful ways system, conversely, instinct became the founda- to approach the analysis of behavior. tion on which all behavior was based. McDougall (1908) argued that all human behavior has an instinctive core that provides both motivation and The Inherited and the Acquired in Behavior, direction for behavior. Learning may play an 1890–1953 important role in determining what particular objects and behaviors become associated with Before about 1920, use of the categories “inher- various instincts, but the instincts themselves are ited” and “acquired” to explain the origins of inherited, unalterable, and essential to the adap- 16 Timothy D. Johnston

tive organization of the behavioral repertoire. For that instinctive behavior exists in human or non- about the first two decades of the twentieth cen- human animals, and that all supposed instincts tury, the concept of instinct was repeatedly in- can in principle be explained as the outcome of voked as an explanation for behavior, most often environmental influences of one kind or another. not on the basis of careful experimentation and Although some of Kuo’s arguments are over- theoretical analysis, but by simply postulating an blown, he made an important fundamental point: instinct whenever a type of behavior seemed in Unless we know in detailed, mechanistic terms need of explanation. James had legitimized this what it means to say that instinctive behavior approach (he listed more than twenty human patterns are “inherited,” the use of instinct as an instincts, including instincts of shyness, fear, explanatory category produces a “finished psy- acquisitiveness, play, and modesty). Although chology,” or at least a “finished” developmental McDougall offered a more fully developed analy- psychology. That is, it tends to block further insis, he too provided lists of instincts that were vestigation into the ontogeny of the behavior by intended to explain the whole range of human purporting to explain when all it really does is to behavior. Indeed, such lists were common in label. It may be correct to identify certain elemen- writings about instinct of this period: One writer tary movement patterns, what Kuo called “un- counted nearly 850 major types of instinct pro- learned reaction units,” as inherited, but more posed in the psychological literature between complex coordinations of behavior (including 1900 and 1920 (Bernard 1924). those usually identified as instincts) offer too This cavalier use of the concept, coupled many opportunities for the influence of experi- with the vitalistic and teleological character ence to be attributed to heredity. of McDougall’s instinct theory (which made it In his later papers, Kuo (e.g., 1924) went even anathema to the mechanistic S-R psychology of farther, arguing that “in a strictly behavioristic Watsonian behaviorism; see Boakes 1984, chap. psychology, with its emphasis on laboratory proce- 8) soon provoked a backlash. Beginning with a dure and with its insistence on physiological expla- paper by Dunlap (1919), the “anti-instinct move- nation of behavior, there is practically no room for ment” in American psychology criticized the con- the concept of heredity” at all (p. 428, original ital- cept of instinct on a number of grounds—that it ics). In this paper, he withdrew his earlier conces- invoked vitalistic forces, that instinct theorists sion that some simple reactions may be inherited frequently offered no evidence for the existence of because “so long as there are inherited reactions, the instincts they proposed, and that attributing simple as they may be, there is justification for the instinctive behavior to “inborn dispositions” use of the term instinct” (p. 439). “The traditional leaves unanswered the question of how the be- sharp distinction between inherited and acquired havior comes into being. The last of these is responses,” he wrote, “should be abolished. All most directly relevant to understanding the im- responses must be looked upon as the direct re- pact of the anti-instinct movement on develop- sult of stimulation, as interactions between the mental thinking. animal and its environment” (p. 439). Kuo in- Perhaps the most persistent and unforgiving of sisted that the idea of inherited behavior was the anti-instinct critics was Zing-Yang Kuo, who simply too vague and nonspecific to do more wrote five theoretical articles criticizing the con- than obscure questions about the origins of the cept of instinct between 1921 and 1930, in addi- behavior in question. tion to a series of experimental reports analyzing Kuo’s arguments were based more on theo- the prenatal development of behavior in chicks. retical principle than on empirical evidence, of In his first two papers, Kuo (1921, 1922) argued which very little pertaining to the early develop- that there is no evidence supporting the claim ment of behavior was then available. However, Toward a Systems View of Development 17

Carmichael (1925) bolstered Kuo’s position by the physiological basis of behavior” (p. 447). pointing out that even anatomical structure does Other authors used behavioral and physiological not develop independently of environmental in- data to support their claims that visual percep- fluences. He summarized embryological research tual organization (Hebb 1937) and sexual behav- showing that many presumably “inherited” struc- ior in rats (Beach 1951; Stone 1951) are innate, tures in fish and amphibians require particular although Hebb and Beach later modified their environmental conditions to develop normally views (Beach 1955; Hebb 1953). and concluded: “Heredity and environment are In 1951, Tinbergen published his classic and in- not antithetical, nor can they expediently be sep- fluential account of ethological theory, The Study arated; for in all maturation there is learning: of Instinct. Although the ethological approach in all learning there is hereditary maturation” to behavior had been a prominent feature of the (p. 260). European scientific scene since the 1920s, it only Debate over the utility of the instinct concept, became widely known to American psychologists and over the related concept of maturation to starting around 1950 (Dewsbury 1984: 131ff). explain the development of instinctive behavior, Much of the earlier literature was written in continued throughout the next twenty-five years. German or Dutch, and what was published From the outset, contributors to the debate had in English had appeared in biological journals recognized that the term might be used in more or conference proceedings not usually read by than one way. Indeed, Dunlap (1919) opened the American psychologists (e.g., Tinbergen 1942; debate by arguing that it was important “to dis- Lorenz 1950). Nonetheless, the European lit- tinguish between the instinct as a group of activ- erature, and especially the work of Lorenz, ities teleologically defined, and the instinct as a described a theory of instinct that avoided physiological group” (p. 307; original emphasis). McDougall’s teleology, while providing an ac- Most of the anti-instinct writers aimed their crit- count that was more fully developed theoreti- icisms at the former use of the term, most closely cally than any available in the work of American associated with McDougall’s theory, although, as comparative and physiological psychologists. we have seen, others (like Kuo) wished to reject Furthermore, the theory was based on extensive any kind of inherited behavior at all. But some observations, and some experiments, on a large authors, especially those trained in comparative number of species under natural conditions. and physiological psychology, were much more These features of Lorenz’s theory made it possi- sympathetic toward Dunlap’s second use, “in- ble for Lehrman to articulate a critique of the stinct as a physiological group,” by which he concept of instinct that revealed its deficiencies meant “a certain definite group of muscular and for the analysis and understanding of behavior glandular performances … resulting from a defi- more clearly than did earlier criticisms. nite stimulus or complex of stimuli” (p. 308). For example, Lashley (1938) rejected the use of instinct in the teleological sense (which he referred Systems Thinking in Lehrman’s 1953 Critique of to as “a dynamics of imaginary forces” [p. 447]) Lorenz’s Instinct Theory but accepted its value when referring to specific, identifiable behavior and its underlying Lehrman’s paper was only one of several criti- physiology—Dunlap’s “physiological” instincts. cisms of the concept of instinct that appeared Lashley made no apologies for using the concept around 1953, although it was by far the most of instinct in this sense, asserting that “the dis- comprehensive. Lehrman’s fluency in German tinction between genetic and environmental in- allowed him to work with the original papers fluences … is of real significance for problems of on which Lorenz’s ideas were based (Silver and 18 Timothy D. Johnston

Rosenblatt 1987), rather than relying on sum- rable element of the organism, dependent on its maries and secondhand accounts written in own specific physiological substrate. For exam- English, and this enhanced the credibility and imple, he suggested that the gradual postnatal impact of his criticism. Drawing on the earlier work provement of pecking in chicks “is very probably of Kuo and Schneirla (e.g., 1949), Lehrman pro- due in part to an increase in strength of the leg vided a framework for thinking about devel- muscles and to an increase in balance and stab- opment that he and others could build into a ility of the standing chick, which results partly coherent alternative to instinct theory. Although from this strengthening of the legs and partly Lehrman did not identify his account as a sys- from the development of equilibrium responses” tems approach to development, it embodies some (Lehrman 1953: 344), rather than to the ma- of the most significant elements of current sys- turation of specific underlying neural circuits. tems thinking. This is quite similar to the dynamical systems view The most readily evident “systems” features of of human locomotor development presented Lehrman’s critique were his repudiation of the by Thelen (e.g., Thelen, Kelso, and Fogel 1987; distinction between innate and acquired behav- Thelen 1995), which explains development not iors, or elements of behavior, and his recognition just in terms of increasing behavioral coordina- that behavior cannot be isolated from the rest of tion, but also in terms of the growing strength the organism’s physiological and anatomical and mass of the infant’s limbs. makeup. He provided both theoretical arguments Although the term system, in the sense in and empirical evidence to show that behavior which it is employed by modern developmental cannot be neatly divided into the categories of systems theorists, does not appear in Lehrman’s learned and innate. Instead, he argued, we must paper, there is no doubt that he would have been analyze the development of every pattern of entirely sympathetic to the aims of systems the- behavior in terms of a continuing interaction ory as it is represented in this volume. The ideas between the organism and its environment (not that we should seek an explanation of develop- between the genotype and the environment, as is ment in the interactions that occur within the de- sometimes proposed). Although the mechanisms veloping organism and between organism and its of learning sometimes are involved in this in- environment, and that the organism is more than teraction, other, less obvious contributions of just a nervous system that processes information, experience play a part as well. In particular, he are central to the account of development that rejected the isolation (or deprivation) experiment Lehrman advanced against Lorenz and the clas- as an adequate tool for analyzing development, sical ethological theory of which he was the chief pointing out that self-stimulation is still possible architect and foremost advocate. for an isolated animal. Later work, especially that of Gottlieb (e.g., 1971, 1991) has revealed the importance of self-stimulation for normal behav- The Response to Lehrman’s Critique ioral development. Lehrman did not go so far as to consider organism and environment as part of Lehrman’s paper was one of three published at a single system, but he clearly understood that we about the same time that criticized the concepts cannot partition behavior into elements and give of instinct and innate behavior as useful bases for a separate account of the development of each of developmental thinking. Both Hebb (1953) and them, an insight that provides the core of the sys- Beach (1955) marshaled similar arguments to tems view of development. those presented by Lehrman, in the process re- Lehrman also criticized the ethologists for treating from earlier positions in which they had treating each innate behavior as a neatly sepa- defended the utility of the concept. Hebb (1953: Toward a Systems View of Development 19

44) explicitly noted that he had earlier accepted least some ethologists in the 1950s and 1960s. the existence of innate perceptual organization in Lehrman did argue that the development of all part because he had overlooked the significance behavior involves the influence of experience, of some of his own data and he took issue with but, like his mentor Schneirla, he urged a much Tinbergen’s (1951) proposal that we should study broader interpretation of “experience” than the the organization of innate behavior before study- very narrow set of circumstances generally sub- ing learning, an approach that Hebb concluded sumed under the term “learning.” In several pa- would be logically impossible. All three papers pers, ethologists attempted to counter Lehrman’s were influential in shaping subsequent discussion assertion that some particular behavior is not in- about the concept of instinct. nate by showing that it could not be explained as As might be expected, many ethologists re- the outcome of standard conditioning procedures jected the arguments of Lehrman and his psycho- (“trial-and-error learning”) and thus concluding logical colleagues, although others were more that it must, in fact, be innate. It seemed di≤cult receptive. For example, speaking at a conference for these ethologists to recognize that Lehrman that brought ethologists and their critics together was proposing to abandon the learned-innate dis- soon after publication of Lehrman’s critique, tinction in toto, not reclassifying purportedly in- Tinbergen (1955) conceded: “Let me say right at nate behaviors as learned (Lehrman 1957). the beginning, that I admit that we must drop this The most important and influential response to use of the word ‘innate’ for the reasons already Lehrman’s criticisms came from Lorenz himself given, namely that the word can be applied only in the form of a long paper in German (Lorenz to differences, not to characters, and also because 1961), later translated into English and expanded tests such as ours [i.e., the deprivation experi- into a book (Lorenz 1965). In this book there is ment] exclude only part of all possible environ- an important shift in Lorenz’s position regard- mental influences [p. 102] … instead of drawing ing the defining feature of instincts. Originally, a positive conclusion or a pseudo-conclusion by Lorenz (e.g., 1937) had drawn the learned/innate saying this response is innate, I want to spe- distinction on the basis of whether the devel- cify the little bit we have found out about the opment of a behavior is determined by genes or ontogeny of this response, saying that, at the environment. (This was a very strict distinc- moment when we studied it, it was not yet condition; indeed, where a behavior appeared to be a tioned” [p. 106]. Similarly, in his influential text- blend of learning and instinct, Lorenz argued book on animal behavior, Hinde (1966) adopted that it would always be possible to identify purely Lehrman’s developmental approach and largely learned and purely innate elements, interlocked dispensed with any attempt to draw sharp dis- or intercalated to form the behavior pattern.) But tinctions between learned and innate behavior. he now proposed that the real difference lies in Other ethologists, however, refused to accept the the source of the information that determines its developmentalists’ critique. adaptiveness. If adaptiveness is determined by One of their responses was to read Lehrman’s information acquired during an individual’s position as claiming that all behavior is the result development, then the behavior is learned; if of trial-and-error learning and that heredity has it is determined by information acquired during no effect on behavior (e.g., Eibl-Eibesfeldt 1961; the evolutionary history of the species, then the Eibl-Eibesfeldt and Kramer 1958; Hess 1962; behavior is innate. This information meta- Klinghammer and Hess 1964; Lorenz 1956). This phor formed the core of Lorenz’s long reply is manifestly not what Lehrman and others were to Lehrman (1965) and echoed points made by saying, but it seems to have been a widely ac- other ethologists who wished to preserve the cepted paraphrase of their position among at learned-innate distinction (e.g., Hess 1962; 20 Timothy D. Johnston

Thorpe 1961, 1963; see Beer 1973). It is a seri- Lorenz’s and Tinbergen’s work (especially in the ously ﬂawed metaphor (Oyama 1985; Johnston three European journals), a few direct rebut- 1987) but its attractiveness for Lorenz helps to tals of Lehrman’s criticisms of Lorenz (mainly clarify a large part of the disagreement between in the Zeitschrift), and frequent references to him and Lehrman: Whereas Lehrman’s primary Lehrman’s empirical studies of behavioral en- interest was in understanding the development of docrinology by those working in that ﬁeld. Many behavior, Lorenz’s interest was in understanding papers that seem to fall squarely into the domain its adaptiveness. The deprivation experiment, crit- of Lehrman’s analysis ignore his 1953 paper en- icized so effectively by Lehrman, is for Lorenz tirely, and others mention it only in passing. only incidentally concerned with explicating de- After 1953, most of Lehrman’s own publi- velopment; its primary goal is to reveal the source cations dealt with his empirical research on the of the information that makes behavior adaptive relations among hormones, behavior, and envir- (Hess 1962: 223; Lorenz 1965: 83ff). onment in the ring dove, exemplifying the kind of behavioral research that he had advocated, but not elaborating his theoretical position very The Legacy of Lehrman’s Critique much (see Lehrman 1962). In 1970, he published another major theoretical statement in a vol- In the decade or so immediately after the pub- ume prepared in memory of his mentor, T. C. lication of Lehrman’s critique, discussion of his Schneirla, who had died in 1968. This chapter ideas appeared in various theoretical papers de- (Lehrman 1970) both expanded on the position voted to the analysis and reanalysis of the con- articulated in 1953 and also responded directly to cept of instinct and the relation between learned Lorenz’s (1965) defense of instinct theory. I do and innate behavior. Many of these (cited earlier) not think that there was any fundamental change defended the concept of instinct, but others built in Lehrman’s theoretical position between 1953 on Lehrman’s arguments and developed his posi- and 1970, but he provided additional examples tion further (Hinde 1966; Jensen 1961; Lehrman to support his view and emphasized some points 1956; Ross and Denenberg 1960; Schneirla 1956, that were only implicit in his earlier paper. For 1966). The debate seemed, however, to engage a example, he discussed the problems that arise rather limited group of writers, and the impact of when questions about developmental mecha- his ideas was slow to be felt in the mainstream nisms are confused with questions about the phe- of research on comparative psychology and be- nomena of adaptation to the environment. Both havioral development. For example, none of the are entirely legitimate kinds of questions about eight chapters on comparative psychology that behavior, but they require different methods for appeared in the Annual Review of Psychology their investigation; most importantly, a particu- from 1953 to 1963 dealt seriously with the theo- lar answer to a question in one domain does not retical issues he raised. There are also few empir- imply anything about answers to questions in the ical papers from this period that use Lehrman’s other domain. “As I hope the discussion so far ideas as a framework for designing experiments has made clear, I have not been trying to avoid and interpreting data. An examination of the the concepts of survival value and phylogenetic four journals that published most of the research adaptation, but only to prevent them from being on animal behavior between 1954 and 1964 merged with the concepts of the causal analysis of (Animal Behaviour, Behaviour, the Journal of development” (Lehrman 1970: 37; original em- Comparative and Physiological Psychology, and phasis). The same distinction had also been made Zeitschrift für Tierpsychologie) found almost no by Tinbergen (1963), who made it clear that these papers that draw on Lehrman’s analysis in any two kinds of questions are entirely complemen- substantive way. There are frequent citations of tary in the study of behavior. Toward a Systems View of Development 21

Lorenz (1965) had claimed that the adaptive (Eds.), Comparative Psychology: A Modern Survey, pp. information specifying an animal’s innate behav- 21–77. New York: McGraw-Hill. ior is in the form of a blueprint coded in the genes Bernard, L. L. (1924). Instinct: A Study in Social Psy- that controls the unfolding (“strictly determined chology. New York: Henry Holt. maturation”) of those patterns that are innate. Boakes, R. (1984). From Darwin to Behaviorism: Psy- Lehrman wrote about the genetic blueprint in chology and the Minds of Animals. Cambridge: Cam- terms that are strikingly reminiscent of Kuo’s bridge University Press. (1921) when he objected to the use of instinct on Carmichael, L. (1925). Heredity and environment: Are the grounds that it tends to produce a finished they antithetical? Journal of Abnormal and Social Psy- psychology: “I believe that the comfort and satis- chology 20: 245–260. faction gained from disposing of the problems of Darwin, C. R. (1871). The Descent of Man, and Selec- ontogenetic development by the use of such con- tion in Relation to Sex. London: John Murray. cepts are misleading, and are based upon the Dewsbury, D. A. (1984). Comparative Psychology in the evasion or dismissal of the most di≤cult and in- Twentieth Century. Stroudsburg, PA: Hutchinson Ross teresting problems of development” (Lehrman Co. 1970: 34). Almost thirty years later the “genetic Dunlap, K. (1919). Are there any instincts? Journal of blueprint” is an idea whose use has become in- Abnormal Psychology 14: 307–311. creasingly widespread and that continues to have Eibl-Eibesfeldt, I. (1961). The interactions of unlearned a strong grip on the imaginations of scientists and behaviour patterns and learning in mammals. In J. M. Delafresnaye (Ed.), Brain Mechanisms and Learning, lay persons alike (see Nelkin and Lindee 1995). pp. 53–73. Oxford: Blackwell. The rapid growth of molecular and human be- Eibl-Eibesfeldt, I., and S. Kramer. (1958). Ethology, havioral genetics and the advent of the Human the comparative study of animal behavior. Quarterly Genome Project have made it more persuasive Review of Biology 33: 181–211. than ever to speak of behavior as being innate or Gottlieb, G. (1971). Development of Species Identifica- genetically encoded, although Lehrman’s argu- tion in Birds. Chicago: University of Chicago Press. ments against such claims are as cogent now as Gottlieb, G. (1991). Experiential canalization of behav- they were in 1953. However, as shown by the con- ioral development: Theory. Developmental Psychology tributions to this volume, the progress of DST 27: 4–13. during the same period has also been impressive, Hebb, D. O. (1937). The innate organization of visual and it is to be hoped that more scientists and activity: I. Perception of figures by rats reared in total educated laypersons will be encouraged to use darkness. Journal of Genetic Psychology 51: 101–126. this perspective for thinking about the develop- Hebb, D. O. (1953). Heredity and environment in ment of behavior. mammalian behaviour. British Journal of Animal Be- haviour 1: 43–47. Hess, E. H. (1962). Ethology: An approach toward the References complete analysis of behavior. New Directions in Psy- Beach, F. A. (1951). Instinctive behavior: Reproductive chology 1: 157–266. activities. In S. S. Stevens (Ed.), Handbook of Experi- Hinde, R. A. (1966). Animal Behavior: A Synthesis of mental Psychology, pp. 387–434. New York: John Ethology and Comparative Psychology (1st ed.). New Wiley & Sons. York: McGraw Hill. Beach, F. A. (1955). The descent of instinct. Psychologi- James, W. (1890). Principles of Psychology. New York: cal Review 62: 401–410. Henry Holt. Beer, C. G. (1973). Species-typical behavior and ethol- Jensen, D. D. (1961). Operationism and the question ogy. In D. A. Dewsbury and D. A. Rethlingshaver “Is this behavior learned or innate?” Behaviour 17: 1–8. 22 Timothy D. Johnston

Johnston, T. D. (1987). The persistence of dichotomies Lorenz, K. Z. (1961). Phylogenetische Anpassung und in the study of behavioral development. Developmental adaptive Modifikation des Verhaltens. Zeitschrift für Review 7: 149–182. Tierpsychologie 18: 139–187. Klinghammer, E., and E. H. Hess. (1964). Parental Lorenz, K. Z. (1965). Evolution and Modification of feeding in ring doves (Streptopelia roseogrisea): innate Behavior. Chicago: University of Chicago Press. or learned? Zeitschrift für Tierpsychologie 21: 338–347. Maienschein, J. (1987). Heredity/development in the Kuo, Z. Y. (1921). Giving up instincts in psychology. United States, circa 1900. History and Philosophy of the Journal of Philosophy 18: 645–664. Life Sciences 9: 79–93. Kuo, Z. Y. (1922). How are our instincts acquired? McDougall, W. (1908). Introduction to Social Psychol- Psychological Review 29: 344–365. ogy. London: Methuen & Co. Kuo, Z. Y. (1924). A psychology without heredity. Psy- Morgan, C. L. (1895). An Introduction to Comparative chological Review 31: 427–448. Psychology. London: Walter Scott. Lashley, K. A. (1938). Experimental analysis of instinc- Nelkin, D., and M. S. Lindee. (1995). The DNA tive behavior. Psychological Review 45: 445–471. Mystique: The Gene as a Cultural Icon. New York: W. Lehrman, D. S. (1953). A critique of Konrad Lorenz’s H. Freeman. theory of instinctive behavior. Quarterly Review of Biol- Oyama, S. (1985). The Ontogeny of Information: ogy 28: 337–363. Developmental Systems and Evolution. Cambridge: Lehrman, D. S. (1956). On the organization of mater- Cambridge University Press. (2d rev. ed., Durham, NC: nal behavior and the problem of instinct. In P. P. Duke University Press, 2000.) Grassé (Ed.), L’Instinct dans le comportement des ani- Romanes, G. J. (1888). Mental Evolution in Man. maux et de l’homme, pp. 475–520. Paris: Masson. London: Kegan Paul & Trench. Lehrman, D. S. (1957). Nurture, nature, and ethology Ross, S., and V. H. Denenberg. (1960). Innate behav- (Review of W. H. Thorpe, “Learning and instinct in ior: The organism in its environment. In R. H. Waters, animals,” Cambridge University Press, 1956). Contem- D. A. Rethlingshaver, and W. E. Caldwell (Eds.), porary Psychology 2: 103–104. Principles of Comparative Psychology, pp. 43–73. New Lehrman, D. S. (1962). Interaction of hormonal and ex- York: McGraw-Hill. periential influences on development of behavior. In E. Sapp, J. (1983). The struggle for authority in the field L. Bliss (Ed.), Roots of Behavior: Genetics, Instinct, and of heredity, 1900–1932: New perspectives on the rise of Socialization in Animal Behavior, pp. 142–156. New genetics. Journal of the History of Biology 16: 311–342. York: Hoeber. Schneirla, T. C. (1949). Levels in the psychological Lehrman, D. S. (1970). Semantic and conceptual issues capacities of animals. In R. W. Sellars, V. J. McGill, in the nature-nurture problem. In L. R. Aronson, E. and M. Farber (Eds.), Philosophy for the Future, pp. Tobach, D. S. Lehrman, and J. S. Rosenblatt (Eds.), 243–286. New York: Macmillan. Development and Evolution of Behavior, pp. 17–50. San Schneirla, T. C. (1956). Interrelationships of the “in- Francisco: W. H. Freeman. nate” and the “acquired” in instinctive behavior. In Lorenz, K. Z. (1937). Über die Bildung des P. P. Grassé (Ed.), L’Instinct dans le comportement des Instinktbegriffes. Naturwissenschaften 25: 289–300, animaux et de l’homme, pp. 387–452. Paris: Masson. 307–318, 324–331. [Translated as Lorenz (1957). The Schneirla, T. C. (1966). Behavioral development and nature of instinct. In C. H. Schiller (Ed.), Instinctive comparative psychology. Quarterly Review of Biology Behavior: Development of a Modern Concept, pp. 41: 283–302. 129–175. New York: International Universities Press.] Silver, R., and J. S. Rosenblatt. (1987). The develop- Lorenz, K. Z. (1950). The comparative method in ment of a developmentalist: Daniel S. Lehrman. De- studying innate behaviour patterns. Symposia of the velopmental Psychobiology 20: 563–570. Society for Experimental Biology 4: 221–268. Stone, C. P. (1951). Maturation and “instinctive” func- Lorenz, K. Z. (1956). The objectivistic theory of in- tions. In C. P. Stone (Ed.), Comparative Psychology (3d stinct. In P. P. Grassé (Ed.), L’Instinct dans le com- ed.), pp. 30–61. New York: Prentice-Hall. portement des animaux et de l’homme, pp. 51–76. Paris: Masson. Toward a Systems View of Development 23

Thelen, E. (1995). Motor development: A new synthesis. American Psychologist 50: 79–95. Thelen, E., J. A. S. Kelso, and A. Fogel. (1987). Self- organizing systems and infant motor development. Developmental Review 7: 39–65. Thorpe, W. H. (1961). Comparative psychology. An- nual Review of Psychology 12: 27–50. Thorpe, W. H. (1963). Ethology and the coding problem in germ cell and brain. Zeitschrift für Tierpsy- chologie 20: 529–551. Tinbergen, N. (1942). An objectivistic study on the innate behavior of animals. Bibliotheca Biotheoretica 1: 39–98. Tinbergen, N. (1951). The Study of Instinct. Oxford: Oxford University Press. Tinbergen, N. (1955). Psychology and ethology as supplementary parts of a science of behavior. In B. Schaffner (Ed.), Group Processes, pp. 75–167. New York: Josiah Macy Jr. Foundation. Tinbergen, N. (1963). On aims and methods of ethology. Zeitschrift für Tierpsychologie 20: 404–433.

3 A Critique of Konrad Lorenz’s Theory of Instinctive Behavior

Daniel S. Lehrman

Beginning about 1931, Konrad Lorenz, with and ecologists, and partly through the recep- his students and collaborators (notably N. tive audience provided for Lorenz and his col- Tinbergen), has published numerous behavioral league, Tinbergen, by American ornithologists. and theoretical papers on problems of instinct The ornithologists were interested from the start, and innate behavior which have had a wide- especially because a great part of the material on spread influence on many groups of scientific which Lorenz based his system came from studies workers (Lorenz 1931, 1932, 1935, 1937; Lorenz of bird behavior, but the range of interest in & Tinbergen 1938; Lorenz 1939; Tinbergen 1939; America has widened considerably. Lorenz and Lorenz 1940, 1941; Tinbergen 1942, 1948, 1950; his theories were recently the subject of some dis- Lorenz 1950; Tinbergen 1951). Lorenz’s influence cussion at a conference in New York at which is indicated in the founding of the Zeitschrift für zoologists and comparative psychologists were Tierpsychologie in 1937 and in its subsequent de- both represented (Riess 1949), and are promi- velopment, and also in the journal Behaviour, nently represented in the recent symposium on established in 1948 under the editorship of an animal behavior of the Society of Experimental international board headed by Tinbergen. Biologists (Armstrong 1950; Baerends 1950; Lorenz’s theory of instinctive and innate Hartley 1950; Koehler 1950; Lorenz 1950; behavior has attracted the interest of many in- Tinbergen 1950), and extensively used in several vestigators, partly because of its diagrammatic chapters of a recent American handbook of ex- simplicity, partly because of its extensive use of perimental psychology which will be a standard neurophysiological concepts, and partly because sourcebook for some years to come (Beach 1951; Lorenz deals with behavior patterns drawn from Miller 1951; Nissen 1951). the life cycle of the animals discussed, rather than Because Lorenz’s ideas have gained wide atten- with the laboratory situations most often found tion, and in particular because a critical dis- in American comparative psychology. These fac- cussion [337–338] of these matters should bring tors go far toward accounting for the great atten- usefully into review Lorenz’s manner of dealing tion paid to the theory in Europe, where most with basic problems in the comparative study of students of animal behavior are zoologists, phys- behavior, a consideration of Lorenz’s system and iologists, zoo curators or naturalists, unlike the school seems very desirable at this time. psychologists who constitute the majority of American students of animal behavior (Schneirla [Editor’s Note: In the section omitted here (pp. 1946a). 338–340) Lehrman uses egg-rolling behavior in In recent years Lorenz’s theories have attracted the gray goose to illustrate some of the features of more and more attention in the United States a classically defined “instinctive” behavior. When as well, partly because of a developing interest a goose sees an egg that has rolled out of its nest, in animal behavior among American zoologists it stands up, reaches out with its neck so that its bill is hooked over the far side of the egg, and pulls the egg back into the nest, using side-to-side This paper originally appeared in the Quarterly Review movements of its bill to prevent the egg from of Biology 28: 337–363 (1953). Page breaks and omis- sions are indicated in the text as (e.g.) [337–338]. Only rolling away. This example illustrates appetitive those references cited in the excerpts reprinted here are behavior—stretching the neck out toward the egg; included in the references. The references have been the instinctive act or consummatory behavior—the reformatted in the style of other contributions to this highly stereotyped movement of the bill that pulls volume, and a few errors have been corrected. 26 Daniel S. Lehrman

the egg back toward the nest; the innate releasing Tinbergen (1942), closely following Lorenz, pattern (or simply releaser)—the appearance of speaks of instinctive acts as “highly stereo- the egg outside the nest and the hard feeling of typed, coordinated movements, the neuromotor the egg against the bill, which together elicit the apparatus of which belongs, in its complete instinctive movement by triggering an innate form, to the hereditary constitution of the ani- releasing mechanism; and the taxis, or orienting mal.” Lorenz (1939) speaks of characteristics of movement—the side-to-side adjustment of the behavior which are “hereditary, individually position of the bill that prevents the egg from fixed, and thus open to evolutionary analysis.” rolling away.] Lorenz (1935) also refers to perceptual patterns (“releasers”) which are presumed to be innate . . . [338–340] because they elicit “instinctive” behavior the first time they are presented to the animal. He also refers to those motor patterns as innate which Problems Raised by Instinct Theories occur for the first time when the proper stimuli are presented. Lorenz’s student Grohmann Even this brief summary brings to light several (1938), as well as Tinbergen and Kuenen (1939), questions which ought to be critically examined speak of behavior as being innately determined with reference to the theory. These are questions, because it matures instead of developing through furthermore, which apply to instinct theories in learning. general. Among them are: (1) the problem of It is thus apparent that Lorenz and Tinbergen, “innateness” and the maturation of behavior; by “innate” behavior, mean behavior which is (2) the problem of levels of organization in an hereditarily determined, which is part of the orig- organism; (3) the nature of evolutionary levels of inal constitution of the animal, which arises quite behavioral organization, and the use of the com- independently of the animal’s experience and en- parative method in studying them; and (4) the vironment, and which is distinct from acquired manner in which physiological concepts may be or learned behavior. properly used in behavior analysis. There follows It is also apparent, explicitly or implicitly, that [340–341] an evaluation of Lorenz’s theory in Lorenz and Tinbergen regard as the major crite- terms of these general problems. ria of innateness that: (1) the behavior be stereotyped and constant in form; (2) it be characteris- “Innateness” of Behavior tic of the species; (3) it appear in animals which have been raised in isolation from others; and (4) The Problem it develop fully formed in animals which have Lorenz and Tinbergen consistently speak of been prevented from practicing it. behavior as being “innate” or “inherited” as Undoubtedly, there are behavior patterns though these words surely referred to a definable, which meet these criteria. Even so, this does not definite, and delimited category of behavior. It necessarily imply that Lorenz’s interpretation of would be impossible to overestimate the heuristic these behavior patterns as “innate” offers genuine value which they imply for the concepts “innate” aid to a scientific understanding of their origin and “not-innate.” Perhaps the most effective way and of the mechanisms underlying them. to throw light on the “instinct” problem is to In order to examine the soundness of the con- consider carefully just what it means to say that cept of “innateness” in the analysis of behavior, it a mode of behavior is innate, and how much will be instructive to start with a consideration of insight this kind of statement gives into the origin one or two behavior patterns which have already and nature of the behavior. been analyzed to some extent. Lorenz’s Theory of Instinctive Behavior 27

Pecking in the Chick acter from that suggested by the concept of a unitary, innate item of behavior. Kuo’s observa- Domestic chicks characteristically begin to peck tions strongly suggest several interpretations of at objects, including food grains, soon after the development of pecking (which, of course, are hatching (Shepard and Breed 1913; Bird 1925; subject to further clarification). For example, the Cruze 1935; and others). The pecking behavior head-lunge arises from the passive head-bending consists of at least three highly stereotyped com- which occurs contiguously with tactual stimula- ponents: head lunging, bill opening and closing, tion of the head while the nervous control of the and swallowing. They are ordinarily coordinated muscles is being established. By the time of hatch- into a single resultant act of lunging at the grain ing, head-lunging in response to tactual stimula- while opening the bill, followed by swallowing tion is very well established (in fact, it plays a when the grain is picked up. This coordination major role in the hatching process). is present to some extent soon after hatching, The genesis of head-lunging to visual stim- and improves later (even, to a slight extent, if the ulation in the chick has not been analyzed. In chick is prevented from practicing). Amblystoma, however, Coghill (1929) has shown This pecking is stereotyped, characteristic of that a closely analogous shift from tactual to the species, appears in isolated chicks, is present visual control is a consequence of the establish- at the time of hatching, and shows some improvement of certain anatomical relationships between ment in the absence of specific practice. Obvi- the optic nerve and the brain region which earlier ously, it qualifies as an “innate” behavior, in the mediated the lunging response to tactual stim- sense used by Lorenz and Tinbergen. ulation, so that visual stimuli come to elicit re- Kuo (1932a–d) has studied the embryonic sponses established during a period of purely development of the chick in a way which throws tactual sensitivity. If a similar situation obtains considerable light on the origin of this “innate” in the chick, we would be dealing with a case of behavior. As early as three days of embryonic intersensory equivalence, in which visual stim- age, the neck is passively bent when the heart- uli, because of the anatomical relationships be- beat causes the head (which rests on the thorax) to tween the visual and tactual regions of the brain, rise and fall. The head is stimulated tactually became equivalent to tactual stimuli, which in by the yolk sac, which is moved mechanically turn became effective through an already ana- by amnion contractions synchronized with the lyzed process of development, which involved heartbeats which cause head movement. Begin- conditioning at a very early age (Maier and ning about one day later, the head first bends Schneirla 1935). actively in response to tactual stimulation. At The originally diffuse connection between about this time, too, the bill begins to open and head-lunge and bill-opening appears to be close when the bird nods—according to Kuo, strengthened by the repeated elicitation of lung- apparently through nervous excitation furnished ing and billing by tactual stimulation by the yolk by the head movements through irradiation in sac. The repeated elicitation of swallowing by the the still-incomplete nervous system. Bill-opening pressure of amniotic fluid following bill-opening and closing become independent of head-activity probably is important in the establishment of only somewhat later. After about 8 or 9 days, the post-hatching integration of bill-opening and fluid forced into the throat by the bill and head swallowing. movements causes swallowing. On the twelfth day, bill-opening always follows head-movement. Maternal Behavior in the Rat [341–342] In the light of Kuo’s studies the “innateness” Another example of behavior appearing to fulfill of the chick’s pecking takes on a different char- the criteria of “innateness” may be found in the maternal behavior of the rat. 28 Daniel S. Lehrman

Pregnant female rats build nests by piling up any part of the female instead of concentrated strips of paper or other material. Mother rats will posteriorly as with normal mother rats, and that “retrieve” their pups to the nest by picking them retrieving does not occur. up in the mouth and carrying them back to the These considerations raise some questions con- nest. Nest-building and retrieving both occur cerning nativistic interpretations of nest-building in all normal rats; they occur in rats which have and retrieving in the rat, and concerning the been raised in isolation; and they occur with no meaning of the criteria of “innateness.” To begin evidence of previous practice, since both are per- with, it is apparent that practice in carrying food formed well by primiparous rats (retrieving may pellets is partly equivalent, for the development take place for the first time only a few minutes of nest-building and retrieving, to practice in car- after the birth of the first litter of a rat raised rying nesting-material, and in carrying the young. in isolation). Both behavior patterns therefore Kinder (1927) has shown that nest-building activ- appear to satisfy the criteria of “innateness” ity [342–343] is inversely correlated with environ- (Wiesner and Sheard 1933). mental temperature, and that it can be stopped Riess (pers. com.), however, raised rats in iso- by raising the temperature su≤ciently. This lation, at the same time preventing them from finding, together with Riess’s experiment, sug- ever manipulating or carrying any objects. The gests that the nest-building activity arises from floor of the living cage was of netting so that feces ordinary food (and other object) manipulation dropped down out of reach. All food was pow- and collection under conditions where the accu- dered, so that the rats never carried food pellets. mulation of certain types of manipulated mate- When mature, these rats were placed in regular rial leads to immediate satisfaction of one of the breeding cages. They bred, but did not build nor- animal’s needs (warmth). The fact that the rat mal nests or retrieve their young normally. They is generally more active at lower temperatures scattered nesting material all over the floor of the (Browman 1943; Morgan 1947) also contributes cage, and similarly moved the young from place to the probability that nest-building activity will to place without collecting them at a nest-place. develop. In addition, the rat normally tends to Female rats do a great deal of licking of their stay close to the walls of its cage, and thus to own genitalia, particularly during pregnancy spend much time in corners. This facilitates the (Wiesner and Sheard 1933). This increased lick- collection of nesting material into one corner of ing during pregnancy has several probable bases, the cage, and the later retrieving of the young to the relative importance of which is not yet that corner. Patrick and Laughlin (1934) have known. The increased need of the pregnant rat shown that rats raised in an environment without for potassium salts (Heppel and Schmidt 1938) opaque walls do not develop this “universal” ten- probably accounts in part for the increased lick- dency of rats to walk close to the wall. Birch’s ing of the salty body fluids as does the increased experiment suggests that the rat’s experience in irritability of the genital organs themselves. Birch licking its own genitalia helps to establish retriev- (pers. com.) has suggested that this genital licking ing as a response to the young, as does its experi- may play an important role in the development of ence in carrying food and nesting material. licking and retrieving of the young. He is raising female rats fitted from an early age with collars Maturation-vs.-Learning, or Development? made of rubber discs, so worn that the rat is effec- The Isolation Experiment tively prevented from licking its genitalia. Present These studies suggest some second thoughts on indications, based on limited data, are that rats the nature of the “isolation experiment.” It is so raised eat a high percentage of their young, obvious that by the criteria used by Lorenz and that the young in the nest may be found under other instinct theorists, pecking in the chick and Lorenz’s Theory of Instinctive Behavior 29

nest-building and retrieving in the rat are not example, when he says (1942) of certain behavior “learned” behavior. They fulfil all criteria of patterns of the three-spined stickleback: “The “innateness,” i.e., of behavior which develops releasing mechanisms of these reactions are all without opportunity for practice or imitation. innate. A male that was reared in isolation . . . Yet, in each case, analysis of the developmental was tested with models before it had ever seen process involved shows that the behavior pat- another stickleback. The . . . [stimuli] . . . had the terns concerned are not unitary, autonomously same releaser functions as in the experiments developing things, but rather that they emerge with normal males.” Such isolation is by no ontogenetically in complex ways from the previ- means a final or complete control on possible ously developed organization of the organism in effects from experience. For example, is the “iso- a given setting. lated” fish uninfluenced by its own reflection What, then is wrong with the implication of the from a water film or glass wall? Is the animal’s “isolation experiment,” that behavior developed experience with human handlers, food objects, in isolation may be considered “innate” if the ani- etc., really irrelevant? mal did not practice it specifically? Similarly, Howells and Vine (1940) have re- Lorenz repeatedly refers to behavior as being ported that chicks raised in mixed flocks of two innate because it is displayed by animals raised varieties, when tested in a Y-maze, learn to go to in isolation. The raising of rats in isolation, and chicks of their own variety more readily than their subsequent testing for nesting behavior, to those of the other variety. They concluded is typical of isolation experiments. The devel- that the “learning is accelerated or retarded . . . opment of the chick inside the egg might be re- because of the directive influence of innate fac- garded as the ideal isolation experiment. tors.” In this case, Schneirla (1946b) suggests that It must be realized that an animal raised in the effect of the chick’s experience with its own isolation from fellow-members of his species is chirping during feeding has not been adequately not necessarily isolated from the effect of pro- considered as a source of differential learning pre- cesses and events which contribute to the devel- vious to the experiment. This criticism may also opment of any particular behavior pattern. The im- be made of a similar study by Schoolland (1942) portant question is not “Is the animal isolated?” using chicks and ducklings. but “From what is the animal isolated?” The iso- Even more fundamental is the question of lation experiment, if the conditions are well ana- what [343–344] is meant by “maturation.” We lyzed, provides at best a negative indication that may ask whether experiments based on the certain specified environmental factors probably assumption of an absolute dichotomy between are not directly involved in the genesis of a par- maturation and learning ever really tell us what ticular behavior. However, the isolation experi- is maturing, or how it is maturing? When the ment by its very nature does not give a positive question is examined in terms of developmental indication that behavior is “innate” or indeed any processes and relationships, rather than in terms information at all about what the process of of preconceived categories, the maturation- development of the behavior really consisted of. versus-learning formulation of the problem is The example of the nest-building and retrieving more or less dissipated. For example, in the by rats which are isolated from other rats but not rat nest-building probably does not mature from their food pellets or from their own geni- autonomously—and it is not learned. It is not talia illustrates the danger of assuming “innate- “nest-building” which is learned. Nest-building ness” merely because a particular hypothesis develops in certain situations through a develop- about learning seems to be disproved. This is mental process in which at each stage there is an what is consistently done by Tinbergen, as, for identifiable interaction between the environment 30 Daniel S. Lehrman

and organic processes, and within the organism; differ, not only from component to component of this interaction is based on the preceding stage the pattern, but also from developmental stage to of development and gives rise to the succeeding developmental stage. What is required is a con- stage. These interactions are present from the tinuation of the careful analysis of the character- earliest (zygote) stage. Learning may emerge as istics of each developmental stage, and of the a factor in the animal’s behavior even at early transition from each stage to the next. embryonic stages, as pointed out by Carmichael Our scepticism regarding the heuristic value of (1936). the concept of “maturation” should not be inter- Pecking in the chick is also an emergent—an preted as ignorance or denial of the fact that the integration of head, bill, and throat components, physical growth of varied structures plays an im- each of which has its own developmental history. portant role in the development of most of the This integration is already partially established kinds of behavior patterns under discussion in by the time of hatching, providing a clear exam- the present paper. Our objection is to the inter- ple of “innate” behavior in which the statement pretation of the role of this growth that is implied “It is innate” adds nothing to an understand- in the notion that the behavior (or a specific phys- ing of the developmental process involved. The iological substrate for it) is “maturing.” For statement that “pecking” is innate, or that it example, the post-hatching improvement in peck- “matures,” leads us away from any attempt to ing ability of chicks is very probably due in part analyze its specific origins. The assumption that to an increase in strength of the leg muscles and pecking grows as a pecking pattern discourages to an increase in balance and stability of the examination of the embryological processes lead- standing chick, which results partly from this ing to pecking. The elements out of whose in- strengthening of the legs and partly from the teraction pecking emerges are not originally a development of equilibrium responses (Cruze unitary pattern; they become related as a con- 1935). Now, isolation or prevention-of-practice sequence of their positions in the organization experiments would lead to the conclusion that of the embryonic chick. The understanding pro- this part of the improvement was due to “matu- vided by Kuo’s observations owes nothing to the ration.” Of course it is partly due to growth “maturation-versus-learning” formulation. processes, but what is growing is not pecking abil- Observations such as these suggest many new ity, just as, when the skin temperature receptors problems the relevance of which is not apparent of the rat develop, what is growing is not nest- when the patterns are nativistically interpreted. building activity, or anything isomorphic with For example, what is the nature of the rat’s it. The use of the categories “maturation-vs.- temperature-sensitivity which enables its nest- learning” as explanatory aids usually gives a building to vary with temperature? How does the false impression of unity and directedness in animal develop its ability to handle food in spe- the growth of the behavior pattern, when actu- cific ways? What are the physiological condi- ally the behavior pattern is not primarily unitary, tions which promote licking of the genitalia, etc.? nor does development proceed in a straight line We want to know much more about the course toward the completion of the pattern. of establishment of the connections between the It is apparent that the use of the concept of chick’s head-lunge and bill-opening, and between “maturation” by Lorenz and Tinbergen as well bill-opening and swallowing. This does not mean as by many other workers is not, as it at first that we expect to establish which of the compo- appears, a reference to a process of development nents is learned and which matured, or “how but rather to ignoring the process of develop- much” each is learned and how much matured. ment. To say of a behavior that it develops by The effects of learning and of structural factors maturation is tantamount to saying that the obvi- Lorenz’s Theory of Instinctive Behavior 31

ous forms of learning do not influence it, and that must underlie the notion that some behavior pat- we therefore do not [344–345] consider it neces- terns are “inherited” as such. sary to investigate its ontogeny further. The “instinct” is obviously not present in the zygote. Just as obviously, it is present in the Heredity-vs.-Environment, or Development? behavior of the animal after the appropriate age. Much the same kind of problem arises when we The problem for the investigator who wishes to consider the question of what is “inherited.” It is make a causal analysis of behavior is: How did characteristic of Lorenz, as of instinct theorists this behavior come about? The use of “explana- in general, that “instinctive acts” are regarded tory” categories such as “innate” and “genically by him as “inherited.” Furthermore, inherited fixed” obscures the necessity of investigating de- behavior is regarded as sharply distinct from velopmental processes in order to gain insight behavior acquired through “experience.” Lorenz into the actual mechanisms of behavior and their (1937) refers to behavior which develops “en- inter-relations. The problem of development is tirely independent of all experience.” the problem of the development of new structures It has become customary, in recent discussions and activity patterns from the resolution of the of the “heredity-environment” problem, to state interaction of existing structures and patterns, that the “hereditary” and “environmental” con- within the organism and its internal environment, tributions are both essential to the development and between the organism and its outer environ- of the organism; that the organism could not ment. At any stage of development, the new fea- develop in the absence of either; and that the tures emerge from the interactions within the dichotomy is more or less artificial. (This formu- current stage and between the current stage and lation, however, frequently serves as an intro- the environment. The interaction out of which duction to elaborate attempts to evaluate what the organism develops is not one, as is so often part, or even what percentage, of behavior is said, between heredity and environment. It is genetically determined and what part acquired between organism and environment! And the [Howells 1945; Beach 1947; Carmichael 1947; organism is different at each different stage of its Stone 1947].) Lorenz does not make even this development. much of a concession to the necessity of develop- Modern physiological and biochemical genet- mental analysis. He simply states that some be- ics is fast destroying the conception of a straight- havior patterns are “inherited,” others “acquired line relationship between gene and somatic by individual experience.” I do not know of any characteristic. For example, certain strains of statement of either Lorenz or Tinbergen which mice contain a mutant gene called “dwarf.” Mice would allow the reader to conclude that they homozygous for “dwarf” are smaller than nor- have any doubts about the correctness of refer- mal mice. It has been shown (Smith and ring to behavior as simply “inherited” or “geni- MacDowell 1930; Keeler 1931) that the cause of cally controlled.” this dwarfism is a deficiency of pituitary growth Now, what exactly is meant by the statement hormone secretion. Now what are we to regard that a behavior pattern is “inherited” or “geni- as “inherited”? Shall we change the name of the cally controlled”? Lorenz undoubtedly does not mutation from “dwarf” to “pituitary dysfunc- think that the zygote contains the instinctive act tion” and say that dwarfism is not inherited as in miniature, or that the gene is the equivalent of such—that what is inherited is a hypoactive pitu- an entelechy which purposefully and continu- itary gland? This would merely push the problem ously tries to push the organism’s development in back to an earlier stage of development. We now a particular direction. Yet one or both of these have a better understanding of the origin of the preformistic assumptions, or their equivalents, dwarfism than we did when we could only say it 32 Daniel S. Lehrman

is “genically determined.” However, the pituitary fluenced by him (Delacour and Mayr 1945; function developed, in turn, in the context of the Adriaanse 1947; Baerends and Baerends-van mouse as it was when the gland was developing. Roon 1950). This type of analysis derives from The problem is: What was that context and how earlier work on the taxonomic relations of behav- did the gland develop out of it? ior patterns by Whitman (1898, 1919), Heinroth What, then, is inherited? From a somewhat (1910, 1930), Petrunkevitsch (1926), and others. similar argument, Jennings (1930) and Chein Lorenz’s brilliant approach to the taxonomic (1936) concluded that only the zygote is inher- analysis of behavior characteristics has had wide ited, or that heredity is only a stage of devel- influence since it provides a very stimulating opment. There is no point here in involving framework in which to study species differences ourselves in tautological arguments over the de- and the specific characteristics of behavior. finition of heredity. It is clear, however, that to However, it does not necessarily follow from the say a behavior pattern is “inherited” throws no fact that behavior patterns are species-specific light on its development except for the purely neg- that they are “innate” as patterns. We may em- ative implication that certain types of learning are phasize again that the systematic stability of a not directly involved. Dwarfism in the mouse, characteristic does not indicate anything about nest-building in the rat, pecking in the chick, and its mode of development. The fact that a charac- the “zig-zag [345–346] dance” of the stickleback’s teristic is a good taxonomic character does not courtship (Tinbergen 1942) are all “inherited” in mean that it developed autonomously. The shape the sense and by the criteria used by Lorenz. But of the skull bones in rodents, which is a good they are not by any means phenomena of a com- taxonomic character (Romer 1945), depends in mon type, nor do they arise through the same part upon the presence of attached muscles kinds of developmental processes. To lump them (Washburn 1947). We cannot conclude that be- together under the rubric of “inherited” or cause a behavior pattern is taxonomically stable “innate” characteristics serves to block the inves- it must develop in a unitary, independent way. tigation of their origin just at the point where In addition it would be well to keep in mind it should leap forward in meaningfulness. (Ana- that the species-characteristic nature of many stasi and Foley 1948, considering data from the behavior patterns may result partly from the fact field of human differential psychology, have been that all members of the species grow in the same led to somewhat the same formulation of the environment. Smith and Guthrie (1921) call such “heredity-environment” problem as is presented behavior elements “coenotropes.” Further, it is here.) not at all necessary that these common features of the environment be those which seem a priori Taxonomy and Ontogeny to be relevant to the behavior pattern under study. Lorenz’s frequent assumption (e.g., 1935) Lorenz (1939) has very ably pointed out the that the effectiveness of a given stimulus on first potential importance of behavior elements as tax- presentation demonstrates an innate sensory me- onomic characteristics. He has stressed the fact chanism specific for that stimulus is not based that evolutionary relationships are expressed just on analysis of the origin of the stimulus- as clearly (in many cases more clearly) by simi- effectiveness, but merely on the fact that Lorenz larities and differences in behavior as by the more has eliminated the major alternative he sees to the commonly used physical characteristics. Lorenz nativistic explanation. himself has made a taxonomic analysis of a fam- Thorpe and Jones (1937) have shown that the ily of birds in these terms (Lorenz 1941), and apparently innate choice of the larvae of the flour others have been made by investigators in- moth by the ichneumon fly Nemerites as an Lorenz’s Theory of Instinctive Behavior 33

object in which to deposit its eggs is actually a higher (phylogenetic) development of the one makes consequence of the fact that the fly larva was fed the other superfluous and stops its development. The on the larvae of the flour moth while it was devel- reaching of a higher psychic performance goes hand-in- oping. By raising Nemerites larvae upon the lar- hand with a reduction of the automatisms that take vae of other kinds of moth Thorpe and Jones part in the action, leaving a behavior pattern with the same function as the one originally existing. (Lorenz caused them, when adult, to choose preponder- 1937) antly these other moths on which to lay their eggs. The choice of flour-moth larvae for oviposi- Again: tion is quite characteristic of Nemerites in nature. In view of Thorpe and Jones’ work, it would It is a peculiarity of many behavior patterns of higher innate instinctive elements and individually- obviously be improper to conclude from this fact animals, that acquired elements immediately follow each other, within that the choice is based on innately-determined a functionally unitary chain of acts . . . I have charac- stimuli. Yet, before their paper was published, terized this phenomenon as instinct-training interlace- the species-specific character of the behavior ment. Similar interlacements occur between instinctive would have been just as impressive evidence for acts and intelligent or insightful behavior.... The “innateness” as species-specificity ever is. essence of such an interlacement is that, within a chain Taxonomic analysis, while very important, is of innate instinctive acts there is a definite point, which not a substitute for concrete analysis of the onto- point is innately determined, where a learned act is geny of the given behavior, as a source of infor- inserted. This learned act must be acquired by each mation about its origin and organization. individual in the course of its ontogenetic development. In such a case, the chain of innate acts has a gap, in which, instead of an instinctive act, there is a ‘capacity Levels of Organization to acquire’. (Lorenz 1937) [All emphases are Lorenz’s.]

Levels of “Innateness” It is apparent that Lorenz regards differences in the extent to which learning occurs as repre- Animals at different evolutionary levels show senting differences in the size of the gaps in the characteristic differences in the extent and man- chain of innate behavior. He considers any given ner of learning. In addition, within the same ani- “component” of behavior as “innate” or not mal’s behavior different activities may be more or “innate.” This is entirely consistent with his less susceptible to the influence of learning, and virtual identification of “innate” with “autono- may be affected in different ways by learning mously developing.” (Schneirla 1948, 1949). However, we have already tried to make it Lorenz explains these facts in terms of the clear that behavior patterns classified as “innate” richness of the animal’s instinctive equipment. As by any criterion do not all fall into the same cat- described above, his conception is that instinctive egory with respect to embryonic origin, develop- [346–347] behavior is sharply different from all mental history, or level of organization. Lorenz behavior leading up to the performance of the notes that more or fewer of the components of instinct. This “appetitive” behavior is conceived behavior may be “innate.” But nowhere does he of as the sole evolutionary source of all learned recognize that one component may be more or and intelligent behavior. Thus he says: less “innate,” or “innate” in one or another man- appetitive behavior, as the sole root of all ‘variable’ ner. We may call attention to an important dif- behavior, not only is physiologically something funda- ference between the pecking of the chick and the mentally different from the automatism of instinctive nest-building of the rat, both behavior patterns behavior, but . . . the two different processes appear as which develop without specific practice of the ‘substitutes’ (vikariierend) for each other, in that the patterns: A major part of the learning which 34 Daniel S. Lehrman

appears to be antecedent to the emergence of concept of “innate” behavior represents a lump- pecking in the chick occurs before hatching, ing-together of many different kinds and levels while much of the learning which is antecedent to of behavior on the basis of an essentially pheno- the emergence of nest-building in the rat occurs typic classification, and the imposition of pre- after birth. conceived categories upon that classification. Shall we call those behavior patterns “innate” which develop before birth and not those which Evolutionary Levels develop after? This would be fruitless in view of the demonstrated existence of prenatal condition- Since Lorenz does not discuss the existence of ing (Ray 1932; Gos 1933; Spelt 1948; Hunt 1949), qualitative differences with respect to modes and unsatisfactory in view of the problem of of [347–348] development within his category of the so-called postnatal “maturation” of various “innate” behavior it is not surprising that his “innate” behavior patterns (Grohmann 1938). conception of the evolution of behavior lacks any But we must recognize that different behavior notion of qualitative change. Lorenz maintains at patterns may involve learning at different ontoge- all levels a sharp distinction between “instinctive netic stages to different extents, and in different acts” and “appetitive behavior” (which includes ways. For example, much less of the behavior of all oriented, goal-directed, and variable types of the rat is directly a consequence of the specific behavior at all levels). He says: characteristics of its structure than in the case of the earthworm (Maier and Schneirla 1935). The If we consider the unbroken series of forms of corresponding modes of behavior, which extends in a involvement of learning in the development of smooth progression from protozoa to man, we must the rat’s behavior is different from and occurs at determine that we cannot distinguish between taxis, on different developmental stages from that of the the one hand, and, on the other, behavior guided by the chick. Further, some responses of the rat (such as simplest intelligence (Einsicht). We cannot here distin- licking of a painful spot) are very much less sub- guish between taxis and, in the case of our frog, an ject to change by learning than others, such as intelligence which might (anthropomorphically speak- care of young (Sperry 1945; Uyldert 1946). These ing) be limited to the knowledge: ‘There sits the fly’. are not differences in the number of behavioral (Lorenz 1937) elements which are “innate,” but rather in the This is restated in a later paper (Lorenz 1939): way in which the structures are involved in the “No sharp line can be drawn between the sim- development of behavior at different evolution- plest orienting-reaction and the highest ‘insight- ary levels and for different behavior patterns. ful’ behavior.” Lorenz does not fully utilize the idea of levels It might be pointed out that whether we can of organization of behavior, apparently because distinguish various levels of behavioral organiza- his concept of “innateness” is not the result of tion depends in part on our assiduity in attempt- analysis of the development of behavior; it is in ing to distinguish them. Preconceptions about the part the result of a preconception that “innate” number and kind of categories into which behav- and “not-innate” are the two categories into ior ought to fall naturally has an important effect which behavior logically falls. Consequently on the kind of examination we make of behavior Lorenz and his school have classified behavior as patterns and the kinds of distinctions we find our- “innate” and “not-innate” on the basis of crite- selves able to make among them. ria which when carefully examined appear to be In the quotation above we have translated arbitrary. Their category of “innate” therefore as “smooth” (progression) Lorenz’s word includes very different kinds of behavior, which “stufenlöse,” which might be more literally involve learning in many different ways. Lorenz’s Lorenz’s Theory of Instinctive Behavior 35

translated “without steps” or “without levels.” gen as evidence of the accumulation of reaction- This is a gratuitous and very misleading oversim- specific energy in the instinctive center until it plification on Lorenz’s part. The transition from “forces” its way through the inhibiting innate protozoa to man is not “stepless.” There are char- releasing mechanism and “goes off “ without any acteristic structural differences between phyletic detectable external stimulus. levels, and these differences are responsible for Lees (1949) has cited the example of the cycli- characteristic differences in the organization of cal colony activities of the ant Eciton hamatum behavior. A protozoan is not like a simpler man. (Schneirla 1938) as an example of “something It is a different kind of organism, with behavior akin to ‘vacuum activity.’” Colonies of this army which depends in different ways on its structure. ant pass regularly through statary and nomadic The analysis of behavior mechanisms at differ- phases, each lasting about 20 days. As Lees ent levels (Schneirla 1946b) shows that it is fre- points out [based on Schneirla’s (1944) descrip- quently misleading to speak of behavior patterns tion]: or elements as homologous when they seem to serve similar (or the “same”) functions and have During the statary phase the bivouac, to which the single queen is confined, remains in situ and raiding superficially similar characteristics. Analysis of activities are minimal. During the nomadic phase the structural organizations out of which the specific position of the bivouac is changed each nightfall and behavior patterns emerge shows that similar strong raiding parties emerge from the colony. This behaviors at different phyletic levels often are activity is in no way related to the abundance or end-products of evolutionary selection leading to scarcity of food in the neighborhood. the similar behavior, but deriving from different structures so that the underlying processes and This cyclic behavior thus appears to Lees to mechanisms are not the same. have the character of a “vacuum activity,” in that Lorenz’s application of the concept of evolu- it occurs periodically without any noticeable tionary change does not consist of analyzing the change in the external stimulus-conditions. This different ways in which behavior patterns at dif- is very misleading, for Schneirla’s (1938, 1944) ferent evolutionary levels depend on the struc- analysis of this behavior has shown that the ture and life of the organism. It consists rather change from statary to nomadic behavior is a of abstracting aspects of behavior, reifying them consequence of the growth of a great new brood as specific autonomous mechanisms, and then of ants. When the callow workers emerge from citing them as demonstrations of “evolution” in their cocoons, their movements stimulate the a purely descriptive taxonomic sense. Taxonom- adult workers to great activity. As the callows ically, this procedure is often extremely valuable, mature and cease to be dependent on the adults, but by its implicit assumption that “elements” their energizing effect is lessened. At this point, of behavior maintain their nature regardless of the emergence of wriggling larvae from the eggs change in the organization in which they are supplements the diminishing activating effect of embedded (more properly, we should say from the callows on the adults. When the larvae which they emerge), it hinders rather than helps pupate, and become inactive, the adults are no analysis of the behavior patterns themselves. longer subject to trophallactic (Wheeler 1928) stimulation, and the colony changes to its statary . . . [348–358] period. The point that is relevant to our discussion is “Vacuum Activities” that Schneirla’s analysis leads to a conception that is the opposite of that implied by the notion The so-called “vacuum activities” or Leerlauf- of “vacuum activity.” The periodic recurrences reaktionen are regarded by Lorenz and Tinber- 36 Daniel S. Lehrman

are not the result of the building up of energy in Conclusion any animal’s nervous system. They are the result of the periodic recurrences of inter-individual We have summarized the main points of Lorenz’s stimulating effects. The behavior is not repre- instinct theory, and have subjected it to a critical sented “in advance” in any of the animals in the examination. We find the following serious flaws: colony; it emerges in the course of the ants’ relationships with one another and with the envi- 1. It is rigidly canalized by the merging of widely ronment. There is no “reaction-specific energy” different kinds of organization under inappropri- being built up. The periodicity is a result of the ate and gratuitous categories. periodicity of the queen’s egg-laying, which is not 2. It involves preconceived and rigid ideas of a “center” having any characteristics correspond- innateness and the nature of maturation. ing to the behavior. And even this is not a direct relationship. If the number of larvae in a colony 3. It habitually depends on the transference of is experimentally reduced by 50 per cent, thus concepts from one level to another, solely on the reducing their total stimulating effect, a normal basis of analogical reasoning. nomadic phase cannot occur. Recent findings 4. It is limited by preconceptions of isomorphic (Schneirla and Brown 1950) have in fact con- [358–359] resemblances between neural and be- firmed the hypothesis that each of the regular havioral phenomena. large-scale egg-delivering episodes in the queen’s 5. It depends on finalistic, preformationist con- function basic to the cycle is a specific outcome of ceptions of the development of behavior itself. her over-feeding, due to a maximal stimulation of 6. As indicated by its applications to human psy- the colony by the brood. This event, occurring chology and sociology, it leads to, or depends on inevitably at the end of each nomadic phase, is a (or both), a rigid, preformationist, categorical “feed-back” type of function, not at all related to conception of development and organization. the implications of “vacuum activity.” The restrictive nature of such categorical theo- Any instinct theory which regards “instinct” as ries as that of Lorenz is very well illustrated by immanent, preformed, inherited, or based on Lees’ remarks on Eciton. The actual development specific neural structures is bound to divert the process leading to the periodic performances of investigation of behavior development from fun- this ant are well understood, and are known to damental analysis and the study of developmen- have no essential relationship to any “reaction- tal problems. Any such theory of “instinct” specific energy” in any nervous system; further inevitably tends to short-circuit the scientist’s they are known not to be “innate” as such investigation of intraorganic and organism- (Schneirla 1938). The processes leading to this environment developmental relationships which behavior surely have nothing to do with the underlie the development of “instinctive” processes leading to “vacuum activities” in a fish. behavior. Yet the superficial similarity is su≤cient to cause Lees to cite the ant’s behavior as an example of a type of behavior described for vertebrates. This is Acknowledgments a good example of the tendency encouraged by such theories to look for cases fitting the theoret- I am greatly indebted to Dr. T. C. Schneirla (who ical categories in many types of behavior, rather originally suggested the writing of this paper) and than analysis of the processes involved in the to Dr. J. Rosenblatt for many stimulating and development of any one behavior pattern. helpful discussions of the problems discussed here. Dr. Schneirla in particular has devoted Lorenz’s Theory of Instinctive Behavior 37

much attention to criticism of the paper at vari- Carmichael, L. (1936). A re-evaluation of the concepts ous stages. of maturation and learning as applied to the early The following people also have read the paper, development of behavior. Psychological Review 43: in part and at various stages, and have made 450–470. many helpful suggestions and comments: Drs. H. Carmichael, L. (1947). The growth of the sensory con- G. Birch, K. S. Lashley, D. Hebb, H. Klüver, L. trol of behavior before birth. Psychological Review 54: 316–324. Aronson, J. E. Barmack, L. H. Hyman, L. H. Lanier, and G. Murphy. Since these scientists dif- Chein, I. (1936). The problems of heredity and environment. Journal of Psychology 2: 229–244. fer widely in the extent of their agreement or disagreement with various points of my discussion, Coghill, G. E. (1929). Anatomy and the Problem of Behavior. London: Cambridge University Press. I must emphasize that none of them is in any way responsible for any errors of omission or com- Cuze, W. W. (1935). Maturation and learning in chicks. Journal of Comparative Psychology 19: 371–409. mission that may appear. Delacour, J., and E. Mayr. (1945). The family Ana- tidae. Wilson Bulletin 57: 3–55. References Gos, E. (1933). Les reflexes conditionnels chez l’em- bryon d’oiseau. Bulletin du Société Royal des Sciences Adriaanse, M. S. C. (1947). Ammophila campestris de Liège 4–5: 194–199; 6–7: 246–250. Latr. und Ammophila adriaansei Wilcke. Ein Beitrag Grohman, J. (1938). Modifikation oder Funktionsrei- zur vergleichenden Verhaltensforschung. Behaviour 1: fung? Ein Beitrag zur Klärung der wechselseitigen 1–34. Beziehungen zwischen Instinkhandlung und Erfahrung. Anastasi, A., and J. P. Foley, Jr. (1948). A proposed Zeitschrift für Tiersychologie 2: 132–144. reorientation in the heredity-environment controversy. Hartley, P. H. T. (1950). An experimental analysis of Psychological Review 55: 239–249. interspecific recognition. Symposia of the Society of Armstrong, E. A. (1950). The nature and function of Experimental Biology 4: 313–336. displacement activities. Symposia of the Society of Ex- Heinroth, O. (1910). Beiträge zur Biologie, namentlich perimental Biology 4: 361–394. Ethologie und Psychologie der Anatiden. International Baerends, G. P. (1950). Specializations in organs and Ornithological Congress (Berlin) 5: 589–702. movements with a releasing function. Symposia of the Heinroth, O. (1930). Ueber bestimmte Bewegungs- Society of Experimental Biology 4: 337–360. weisen bei Wirbeltieren. Gesellschaft Naturforschender Baerends, G. P., and J. M. Baerends-van Roon. (1950). Freunde zu Berlin, 1929, pp. 333–342. An introduction to the study of the ethology of cichlid Heppel, L. A., and C. L. A. Schmidt. (1938). Studies on fishes. Behaviour (Supplement) 1: 1–242. the potassium metabolism of the rat during pregnancy, Beach, F. A. (1947). Evolutionary changes in the phys- lactation and growth. University of California Publica- iological control of mating behavior in mammals. tions in Physiology 8: 189–205. Psychological Review 54: 297–315. Howells, T. R. (1945). The obsolete dogmas of hered- Beach, F. A. (1951). Instinctive behavior, reproductive ity. Psychological Review 52: 23–34. activities. In S. S. Stevens (Ed.), Handbook of Experi- Howells, T. R., and D. O. Vine. (1940). The innate dif- mental Psychology, pp. 387–434. New York: John ferential in social learning. Journal of Abnormal and Wiley & Sons. Social Psychology 35: 537–548. Bird, C. (1925). The relative importance of maturation Hunt, E. L. (1949). Establishment of conditioned and habit in the development of an instinct. Pedagog- responses in chick embryos. Journal of Comparative and ical Seminar 32: 68–91. Physiological Psychology 42: 107–117. Browman, L. G. (1943). The effect of controlled tem- Jennings, H. S. (1930). The Biological Basis of Human peratures upon the spontaneous activity rhythms of the Nature. New York: Norton & Co. albino rat. Journal of Experimental Zoology 94: 477– 489. Keeler, C. (1931). The Laboratory Mouse. Cambridge, MA: Harvard University Press. 38 Daniel S. Lehrman

Kinder, E. F. (1927). A study of the nest-building activ- Lorenz, K., and N. Tinbergen. (1938). Taxis und ity of the albino rat. Journal of Experimental Zoology Instinkthandlung in der Eirollbewegung der Graugans. 47: 117–161. I. Zeitschrift für Tierpsychologie 2: 1–29. Koehler, O. (1950). Die Analyse der Taxisanteile Maier, N. R. F., and T. C. Schnierla. (1935). Principles instinktartigen Verhaltens. Symposia of the Society of of Animal Psychology. New York: McGraw-Hill Co. Experimental Biology 4: 269–303. Miller, N. E. (1951). Learnable drives and rewards. In Kuo, Z. Y. (1932a). Ontogeny of embryonic behavior S. S. Stevens (Ed.), Handbook of Experimental Psy- in Aves. I. The chronology and general nature of the chology, pp. 435–472. New York: John Wiley & Sons. behavior of the chick embryo. Journal of Experimental Morgan, C. T. (1947). The hoarding instinct. Psy- Zoology 61: 395–430. chological Review 54: 335–341. Kuo, Z. Y. (1932b). Ontogeny of embryonic behavior Nissen, H. W. (1951). Phylogenetic comparison. In S. S. in Aves. II. The mechanical factors in the various stages Stevens (Ed.), Handbook of Experimental Psychology, leading to hatching. Journal of Experimental Zoology pp. 347–386. New York: John Wiley & Sons. 62: 453–489. Patrick, J. R., and R. M. Laughlin. (1934). Is the wall- Kuo, Z. Y. (1932c). Ontogeny of embryonic behavior seeking tendency in the white rat an instinct? Journal of in Aves. III. The structure and environmental factors Genetic Psychology 44: 378–389. Journal of Comparative Psy- in embryonic behavior. Petrunkevitsch, A. (1926). The value of instinct as a chology 13: 245–272. taxonomic character in spiders. Biological Bulletin 50: Kuo, Z. Y. (1932d). Ontogeny of embryonic behavior 427–432. in Aves, IV. The inﬂuence of embryonic movements Ray, W. S. (1932). A preliminary study of fetal condi- Journal of Compa- upon the behavior after hatching. tioning. Child Development 3: 173–177. rative Psychology 14: 109–122. Riess, B. F. (1949). The isolation of factors of learning Lees, A. D. (1949). Modern concepts of instinctive and native behavior in ﬁeld and laboratory studies. Science Progress in the Twentieth Century behaviour. Annals of the New York Academy of Sciences 51: 1093– 37: 318–321. 1102. Lorenz, K. (1931). Beiträge zur Ethologie sozialer Romer, A. S. (1945). Vertebrate Paleontology. Chicago: Journal für Ornithologie Corviden. 79: 67–127. University of Chicago Press. Lorenz, K. (1932). Betrachtungen über das Erkennen Schneirla, T. C. (1938). A theory of army-ant behavior Journal für der arteigenen Triebhandlungen der Vögel. based upon the analysis of activities in a representative Ornithologie 80: 50–98. species. Journal of Comparative Psychology 25: 51–90. Lorenz, K. (1935). Der Kumpan in der Umwelt des Schneirla, T. C. (1944). The reproductive functions of Journal für Ornithologie Vogels. 83: 137–213, 289– the army-ant queen as pacemakers of the group behav- 313. ior pattern. Journal of the New York Entomological Lorenz, K. (1937). Ueber den Begriff der Instink- Society 52: 153–192. Folia Biotheoretica thandlung. 2: 17–50. Schneirla, T. C. (1946a). Contemporary American ani- Lorenz, K. (1939). Vergleichende Verhaltensforschung. mal psychology in perspective. In P. L. Harriman (Ed.), Zoologische Anzeitung 12 (Suppl. band): 69–102. Twentieth Century Psychology, pp. 306–316. New Lorenz, K. (1940). Durch Domestikation verursachte York: Philosophical Library. Störungen arteigenen Verhaltens. Zeitschrift für Schneirla, T. C. (1946b). Problems in the biopsychol- Angewandte Psychologie und Charaktarkunde 59: 2– ogy of social organization. Journal of Abnormal and 81. Social Psychology 41: 385–402. Lorenz, K. (1941). Vergleichende Bewegungsstudien an Schneirla, T. C. (1948). Psychology, Comparative. Anatinen. Journal für Ornithologie 89 (Sonderheft): Encyclopaedia Britannica 18: 690–708. 194–294. Schneirla, T. C. (1949). Levels in the psychological Lorenz, K. (1950). The comparative method in study- capacities of animals. In R. Sellars, V. J. McGill, and ing innate behavior patterns. Symposia of the Society of M. Farber (Eds.), Philosophy for the Future, pp. Experimental Biology 4: 221–268. 243–286. New York: Macmillan & Co. Lorenz’s Theory of Instinctive Behavior 39

Schneirla, T. C., and R. Z. Brown. (1950). Army-ant m. merula L., und T. e. ericetarum Turton). Zeitschrift life and behavior under dry-season conditions. 4. für Tierpsychologie 3: 37–60. Further investigation of cyclic processes in behavioral Uyldert, I. E. (1946). A conditioned reflex as a factor and reproductive functions. Bulletin of the American influencing the lactation of rats. Acta Brevia Neerlan- Museum of Natural History 95: 263–354. dica de Physiologia, Pharmacologia, Microbiologia 14: Schooland, J. B. (1942). Are there any innate behavior 86–89. tendencies? Genetic Psychology Monographs 25: 219– Washburn, S. L. (1947). The relation of the temporal 287. muscle to the form of the skull. Anatomical Record 99: Shepard, J. F., and F. S. Breed. (1913). Maturation and 239–248. use in the development of an instinct. Journal of Animal Wheeler, W. M. (1928). The Social Insects. New York: Behaviour 3: 274–285. Harcourt, Brace & Co. Smith, P. E., and E. C. MacDowell. (1930). An heredi- Whitman, C. O. (1899). Animal behavior. Biological tary anterior-pituitary deficiency in the mouse. Anatom- Lectures, 1898, pp. 285–338. ical Record 46: 249–257. Whitman, C. O. (1919). The behavior of pigeons. Smith, S., and E. R. Guthrie. (1921). General Psy- Publications of the Carnegie Institute 257: 1–161. chology in Terms of Behavior . New York: Appleton. Wiesner, B. P., and N. M. Sheard. (1933). Maternal Spelt, D. K. (1948). The conditioning of the human Behaviour in the Rat. London: Oliver & Boyd. fetus in utero. Journal of Experimental Psychology 38: 338–346. Sperry, R. W. (1945). The problem of central nervous reorganization after nerve regeneration and muscle transposition. Quarterly Review of Biology 20: 311–369. Stone, C. P. (1947). Methodological resources for the experimental study of innate behavior as related to environmental factors. Psychological Review 54: 342– 347. Thorpe, W. H., and F. G. W. Jones. (1937). Olfactory conditioning in a parasitic insect and its relation to the problem of host selection. Proceedings of the Royal Society of London, B 124: 56–81. Tinbergen, N. (1939). On the analysis of social organization among vertebrates, with special reference to birds. American Midland Naturalist 21: 210–234. Tinbergen, N. (1942). An objectivistic study of the innate behaviour of animals. Bibliotheca Biotheoretica 1: 39–98. Tinbergen, N. (1948). Physiologische Instinktforsch- ung. Experientia 4: 121–133. Tinbergen, N. (1950). The hierarchical organization of nervous mechanisms underlying instinctive behaviour. Symposia of the Society of Experimental Biology 4: 305–312. Tinbergen, N. (1951). The Study of Instinct. Oxford: Oxford University Press. Tinbergen, N., and D. J. Kuenen. (1939). Ueber die auslösenden und die richtunggebenden Reizsitua- tionen der Sperrbewegung von jungen Drosseln (Turdus

A Developmental Psychobiological Systems View: Early Formulation and 4 Current Status

Gilbert Gottlieb

[August Weismann’s] proposition, that effective varia- ing). I reasoned that if the prenatal period of tion is due to the influence of nutrition upon the determi- development was important, then the two-day nants contained in the reserve germ-plasm, seems to span of hatching had to have consequences for throw too much stress on the nutrition and environment, the ducklings’ neurosensory, neuromotor, per- too little on the inherent activities of living matter. But if ceptual, and behavioral competence at the time it be regarded as an expression of the fact that all effec- they were trained (exposed) to the imprinting tive variation is a joint product of the inherent activities of germinal cells and the conditioning effects of their en- object. Indeed, when I plotted the imprintability vironment, it is a self-evident proposition which may be of the ducklings as a function of posthatch age cheerfully accepted. at training, I did not find a sensitive period, but —Morgan 1893–94 when I plotted the same behavioral results according to “developmental age” (timed from the It remains inscrutable why we have had to labor onset of incubation to training) I did find a criti- so diligently in the recent past to attain a cheerful cal period before which they did not imprint and acceptance of what was self-evident over a hun- after which they did not imprint. Imprinting was dred years ago. But I suppose that is what this defined as learning the visual characteristics of book is about. I have been given the pleasant task the imprinting object as judged by a later be- of describing the history and currently status of havioral preference test with the familiar object the concept of probabilistic epigenesis (or a devel- versus an unfamiliar object. opmental psychobiological systems view). Soon after I published these results (Gottlieb 1961), which suggested developmental age as Behavioral Embryology a more precise age baseline than posthatch age for measuring the sensitive period, I realized that Two ideas that I held from my graduate student others were interpreting developmental age as days were (1) the overriding importance of pre- “maturational age,” a too-narrow connotation natal development and (2) the bidirectional in- that I expressly wished to avoid in explicitly fluences of structure and function (construed choosing the term developmental age over matu- very broadly to include the psychological as well rational age. So, Peter Klopfer and I immedi- as the behavioral, neuroanatomical, and neuro- ately embarked on an experiment to show that physiological levels). When I did my doctoral dis- the best developmental age for imprinting could sertation on the sensitive period for imprinting be moved around, depending on the sensory ex- in ducklings, I noticed that, although all of the perience to which the ducklings were exposed eggs were set in the incubator at the same time, prior to the training trial (Gottlieb and Klopfer the ducklings hatched over a two-day period. 1962). This experiment resulted from my implicit Therefore, in addition to timing the sensitive belief in the bidirectionality of structure and period from the time of hatching, I timed the function. If the ducklings were reared with visual ducklings’ age from the onset of incubation. My and social experience from hatching till training, reasoning was that those that hatched early were their sensitive periods for auditory and visual im- less mature at the standard times of training than printing were different from ducklings that were those that hatched later, the latter being more reared in darkness and in social isolation from mature as the birds were all trained at one of five hatching till training. The sensitive period for posthatch age ranges (either 4–7 hrs, 8–12 hrs, imprinting was not exclusively a function of mat- 13–17 hrs, 18–22 hrs, or 23–27 hrs after hatch- uration but depended also on the nature and extent of the birds’ prenatal and postnatal ex- 42 Gilbert Gottlieb

periences prior to entering into the imprinting Shortly after Kuo and I completed our descrip- situation. tive study of the behavior of the duck embryo in At this point in my career, having had the the fall of 1963 and he had returned to his home benefit of assistance from Zing-Yang Kuo, who in Hong Kong to begin work on his final book, kindly consented to work in my laboratory in The Dynamics of Behavior Development, I re- 1963 to teach me how to do behavioral observa- ceived an invitation from Ethel Tobach and her tions of embryos, I embarked on the experimen- colleagues in the Animal Behavior Department at tal analysis of the role of normally occurring the American Museum of Natural History, to embryonic auditory experience in structuring the write a chapter for a book in honor of the career species-specific auditory perceptual capabilities and imminent retirement of T. C. Schneirla. I observed after hatching. As a way of further- decided to write a theoretical review of the his- ing my education in the techniques of behavi- tory and present state of thinking about behav- oral embryology, Kuo and I visited Viktor ioral embryology. Since Schneirla had been an Hamburger’s laboratory in 1963. Hamburger, admirer of Kuo’s embryological observations a distinguished neuroembryologist, was now and he, and his student Daniel Lehrman (1953), turning his efforts to the study of motor behavior had used those observations to good advantage in the chick embryo, with the assistance of two in their critiques of instinct theory, it seemed psychologists, postdoctoral associate Martin especially fitting to do a review of behavioral Balaban and Ronald Oppenheim, who was a embryology for Schneirla’s Festschrift. graduate student at that time. Hamburger was claiming the spontaneous autonomy of motor behavior in the early chick embryo, whereas Predetermined and Probabilistic Epigenesis of Kuo’s observations through the “windows” in Behavior the egg shell suggested to him that, while the early chick embryo was indeed motorically active, the In my literature review, I found two rather dif- extent, kind, and frequency of movement was ferent conceptualizations of behavioral embryol- intimately related to the environmental contin- ogy, one I called predetermined epigenesis and gencies of the egg and the chick embryo’s own the other probabilistic epigenesis. The various body parts at each phase of development. To adherents to the predetermined view had in com- make the difference in views on this matter even mon “the assumption that the behavioral se- starker, Balaban (not Hamburger) told me it was quence is predetermined by factors of neural possible to remove the embryo from the shell and growth and differentiation (maturation), which observe its normal behavior unimpeded without have an essentially invariant schedule of appear- the necessity of Kuo’s windowing procedure. In ance. In this view, nonsensory environmental my naïve way, I was delighted and began freeing factors merely support development or allow be- embryos from their prison with abandon, while havioral development to occur and sensory stim- Kuo stared in silent disbelief. Pretty quickly ulation does not influence or determine the it became clear that the embryos began dying course of behavioral development in any sig- rather than exhibiting their species-typical motor nificant way” (Gottlieb 1970: 112). In this view, movements, so with Balaban’s assent (he had behavior was an epiphenomenon of neural matu- only done a few embryos before I got there), we ration and did not in itself contribute to develop- agreed this was not a usable experimental proce- ment. For example, Coghill (1929: 16) wrote that dure and Kuo and I would go back to the old embryonic behavior patterns become “organized tried and true way of observing the embryos in through a regular order of sequence of definite situ through windows made in the egg shell. phases in the growth of the nervous system.” A Developmental Psychobiological Systems View 43

Later, Anokhin (1964: 85) wrote, “It is true that of the structure-function relationship is one of the main the systemogenetic type of the maturation and . . . assumptions of predetermined epigenesis. The bidirec- growth is . . . evidently inborn . . . preformed, and tional version of the structure-function relationship is in fact, in the process of the ontogenesis, they cor- a logical consequence of the view that the course and outcome of behavioral epigenesis is probabilistic: it en- respond demonstrably to the ecological factors of tails the assumption of reciprocal effects in the relation- that species of animal.” This latter point on the ship between structure and function whereby function importance of prenatal development to prepare, (exposure to stimulation and/or movement or muscu- as it were, the neonate for survival in its infantile loskeletal activity) can significantly modify the develop- environment cannot be gainsaid, but that can ment of the peripheral and central structures that are also be regularly achieved with the help of nor- involved in these events. (Gottlieb 1970: 123) mally occurring species-typical behavioral and environmental influences. Although the edited book in which my chapter I used the term probabilistic epigenesis to de- appeared was not published until 1970, the words signate the alternative view to predetermined were penned in 1965. Even at this relatively early epigenesis. Probabilistic epigenesis holds that date I was able to find some evidence for the behavioral development of individuals within a influence of stimulative factors (as broadly de- given species does not follow an invariant or fined above) on prenatal development, and also inevitable course, and, more specifically, that the for the bidirectionality of the structure-function sequence and outcome of individual behavioral relationship. Although the concepts of predeter- development is probable (with respect to norms) mined and probabilistic epigenesis would seem rather than certain. to fall in a general way into the ancient nature/ Further, I wrote: nurture dichotomy, what I was most pleased with in the formulation of predetermined and prob- A degree of uncertainty in behavioral development abilistic epigenesis was explicitly avoiding the is demanded by the overwhelming importance ascribed genes/environment dichotomy. It was clear to me to stimulative factors at almost all stages of prenatal that genetic activity was involved in both for- development. In this view, stimulative factors are re- mulations, so I just left out that aspect entirely. garded, in a weak sense, as merely facilitating the The questions dividing the two camps were the development of behavior along certain lines or, in a all-su≤ciency of neural maturation versus the much stronger sense, as channeling or forcing the development of behavior along certain lines. Stimulative contribution of stimulative and behavioral fac- events are broadly construed and include six main factors in facilitating and channeling prenatal neural tors: (1) presensory or nonsensory mechanical agita- maturation, and these were empirically testable tion; (2) interoceptive stimulation; (3) proprioceptive questions. stimulation; (4) exteroceptive stimulation; (5) neuro- Due to Schneirla’s sudden death in 1968, I did chemical (e.g., hormonal) stimulation; and (6) the mus- not have the opportunity to discuss probabilistic culoskeletal effects of use or exercise. epigenesis and the bidirectional S-F hypothesis with him (the Festschrift was supposed to be a A second way in which probabilistic epigenesis differs from predetermined epigenesis is that certain theorists surprise), but I feel confident that he would have [I had in mind E. B. Holt, Z.-Y. Kuo, and T. C. welcomed these notions as entirely compatible Schneirla] implicitly assume that probabilistic epigene- with his own thinking on these matters, as exem- sis necessitates a bidirectional structure-function plified, for example, in his 1956 review, in which hypothesis. The conventional version of the structure- he extensively discusses maturation and experi- function hypothesis is unidirectional in the sense that ence interrelationships in prenatal development structure is supposed to determine function in an essen- and in his 1966 paper, in which he mentions tially nonreciprocal relationship. The unidirectionality “maturation-experience integrations of the em- 44 Gilbert Gottlieb

bryo” (p. 290) in referring to Kuo’s research. neural plasticity. Even here, however, function Elsewhere, he wrote of the “fusion” of matura- (experience) is thought to become important only tion and experience (Schneirla 1965). after, say, all the neural connections are made. I did send a prepublication copy of my chap- Synapses are lost if they are not functionally val- ter to Kuo in 1965, at which time he was writing idated. Recently, this idea has become somewhat his book The Dynamics of Behavior Development, more refined to say that there is an early “exu- originally published in 1967. In that book he was berant” overproduction of synapses which are generous enough to credit me with the explicit then pruned down by experience (and the lack notion of S-F bidirectionality, so I felt that I had thereof). This is different from saying that, in the indeed made some contribution by explicating first instance, the synapses are built in part by what I believed to be inherent in Kuo’s (and function. It is only very recently that the appro- Holt’s and Schneirla’s) views on early behavioral priate experiments have been done to show that development. normally occurring spontaneous neural activity is Probabilistic epigenesis and S-F bidirection- necessary to get synapses into the correct area of ality rely heavily on prenatal sensory (as well as the brain in the first place (Catalano and Shatz motor) function as an important contributor to 1998). species-typical development, so in the next reviews I undertook to examine the full range of evidence for sensory functioning in bird embryos The Developmental Manifold Concept (Gottlieb 1968) and in a variety of mammalian fetuses including humans (Gottlieb 1971a). What In 1971, my own research program with duck I found was that each sensory system begins to embryos and hatchlings paid off in what I function while still undergoing maturation (i.e., believed to be a conceptually valuable way. (I while cell division, migration, growth, and differ- hoped others would agree, but this agreement entiation are still going on), so each system could proved to be a rockier and longer road than I contribute to its own normal prenatal (as well as anticipated [Gottlieb 1997].) In 1965, I had postnatal) development. Thus, normally occur- shown that ducklings and chicks hatched in incu- ring spontaneous and evoked function (experi- bators, and thus deprived of maternal contact, ence broadly defined) could play a role in the could nonetheless identify the maternal assembly rate at which a sensory system becomes com- call of their own species after hatching. The only pletely mature as well as contributing to the over- vocal-auditory experience they had was exposure all competence of the system by, for example, to their own and sib vocalizations prior to enter- influencing the number of neurons, as well as the ing the test situation. In 1966, I was able to show size of neurons and their axonal and dendritic that enhancing exposure to sib vocalizations fields. Thirty years later, this is now well estab- lowered the latency and increased the duration lished, of course, but it was not at the time of their behavioral response to their own species (review in Black and Greenough 1998). The maternal call. However, it was necessary to de- positive influence of functional activity on the vise an embryonic devocalization procedure to nervous system is not restricted to the prenatal truly rule in the critical importance of the embry- period but continues into adulthood (Kemper- onic vocalizations in perfecting the perceptual mann, Kuhn, and Gage 1997). selectivity of the response that was evident after I viewed the functional contribution to matur- hatching. With the help of John Vandenbergh, I ing (but not yet completely mature) systems as an was able to devise an embryonic muting ope- especially important feature of S-F bidirectional- ration that did not otherwise interfere with the ity. It is now a common finding in studies of health of the embryo and hatchling (Gottlieb and A Developmental Psychobiological Systems View 45

Vandenbergh 1968). Now, the selectivity of the conventionally regarded as instinctive or innate. If this postnatal response to the species’ maternal call prediction turns out to be correct, the nature-nurture could be examined in ducklings that had not controversy may all but evaporate, and a consensus experienced their own or sib vocalizations. Lo will have been reached on the idea that structure only fully realizes itself through function. (Gottlieb 1971b: and behold, the devocalized mallard duckling’s 156–157) usual auditory selectivity was not in place—they could not distinguish the mallard maternal call A full understanding of the development (and from the chicken maternal call. The control birds evolution) of behavior will necessarily include that had been allowed to hear their own embry- genetic factors, so I went on to say, “One of the onic vocalizations for eighteen to twenty-three main questions in the ontogeny of behavior con- hours before being devocalized did show the cerns the contribution of genetic (molecular) and usual preference for the mallard maternal call other biochemical factors. While there is agree- over the chicken maternal call (Gottlieb 1971b: ment that such factors provide an indispensable 141–142). impetus to neural maturation, the exact nature or The outcome of these experiments led to the mechanics of the molecular and biochemical con- formulation of the developmental manifold control of early neural maturation is not yet known” cept, a forerunner to West and King’s (1987) (Gottlieb 1971b: 157). notion of “settling nature and nurture into an From the beginning of my work in the early ontogenetic niche”: 1960s I was eager to include the genetic contribution and seized upon an early opportunity to The present results indicate that the epigenesis of collaborate with two neurologists, one of whom species-specific auditory perception is a probabilistic phenomenon, the threshold, timing, and ultimate per- (George Paulson) was adept at brain dissection fection of such perception being regulated jointly by in birds and the other of whom (Stanley Appel) organismic and sensory stimulative factors. In the nor- was running a laboratory in which protein syn- mal course of development, the manifest changes and thesis in the nervous system was being studied. improvements in species-specific perception do not rep- According to what was then already known pro- resent merely the unfolding of a fixed or predetermined tein synthesis was the result of messenger RNA organic substrate independent of normally occurring activity and mRNA activity was a consequence sensory stimulation. With respect to the evolution of of DNA activity (DNA}mRNA}protein). So, species-specific perception, natural selection would protein synthesis could be used as an indirect seem to have involved a selection for the entire devel- measure of genetic activity. Accordingly, I pre- opmental manifold, including both the organic and normally occurring stimulative features of ontogeny. pared two groups of duck embryos for Paulson With respect to the heavy emphasis placed here on the and Appel by exposing them for several days role of normally occurring stimulation in perfecting before hatching either to tape-recorded sib vocal- species-specific perception, it is pertinent to point out izations or to extra visual stimulation by incu- that the same condition may also hold for the develop- bating them in a lighted incubator chamber. The ment of species-typical action patterns. In the past it control group was incubated in the dark and in has been usual in sensory isolation or motor depriva- acoustic isolation from other embryos. The point tion studies to ask merely whether the deprived animal of the experiment was to look for enhanced pro- can subsequently perform the species-typical behavior, tein synthesis in the auditory and visual parts of not when or how well the animal performs it. As we the brain in the treated groups. move into an era of increasingly sophisticated analyses of the development of behavior, it will not be alto- Subsequently, Appel told us that both experi- gether surprising to find that normally occurring sen- mental groups showed an enhancement of pro- sory stimulation or motor movement is essential to the tein synthesis in the synaptic regions of the brain normal threshold, timing, and perfection of behavior stem, in which the auditory nuclei are located, 46 Gilbert Gottlieb

and in the optic lobes of the brain, which mediate Probabilistic Epigenesis auditory as well as visual stimulation. This was Bidirectional Structure-Function Development a clear indication that the extra auditory and Genetic activity (DNAMRNAMProtein)M visual stimulation had enhanced gene expression structural maturationMfunction, activity, or in the embryo. This, of course, implied a bidirec- experience tional S-F relationship all the way to the genetic level during the embryonic period, and it meant As it applies to the nervous system, structural that genetic activity could be influenced by nor- maturation refers to neurophysiological and neu- mally occurring exteroceptive sensory stimula- roanatomical development, principally the struction and thus result in an enhancement of neural ture and function of nerve cells and their synaptic maturation. The experiment was completed in interconnections. The unidirectional S-F view 1965, and Paulson and I pleaded with Appel to assumes that genetic activity gives rise to struc- complete the data analysis so we could share the tural maturation that then leads to function in a results with interested colleagues by publishing nonreciprocal fashion, whereas the bidirectional them. Alas, overly hectic research, clinical, and view holds that there are reciprocal influences other duties took precedence in Appel’s life, so among genetic activity, structural maturation, Paulson and I were frustrated in our desire to see and function. In the unidirectional view, the ac- this work to completion in the form of publica- tivity of genes and the maturation process are tion in a refereed journal. At the time there was pictured as relatively encapsulated or insulated, only one other study in the literature implicat- so that they are uninfluenced by feedback from ing exteroceptive influences on genetic activity, the maturation process or function, whereas the a study involving the influence of vestibular- bidirectional view assumes that genetic activity mediated learning on changes in nuclear RNA and maturation are affected by function, activity, base ratios in vestibular nerve cells in rats (Hydén or experience. The bidirectional or probabilistic and Egyházi 1962). It was not until 1967 that view applied to the usual unidirectional formula Rose published a study showing enhanced RNA calls for arrows going back to genetic activity to and protein synthesis in the visual cortex of rats indicate feedback serving as signals for the turn- as a consequence of visual stimulation. ing on and off of genetic activity. The usual view, as in the central dogma of molecular biology described below, calls for genetic activity to be Normally Occurring Environmental and regulated by the genetic system itself in a strictly Behavioral Influences on Gene Activity feedforward manner. It was in 1998 that I was able, at last, to write It was in 1976 that I first extended the unidirec- a comprehensive empirical review of bidirection- tionality and bidirectionality of S-F function to ality at the genetic level, in the context of a cri- the genetic level, and elaborated on this topic in tique of the central dogma of molecular biology, reviews published in 1983 (p. 13) and 1991 (p. 13) a highly influential and outstanding example of as follows: predetermined epigenesis. Predetermined Epigenesis Unidirectional Structure-Function Development The Central Dogma Genetic activity (DNA}RNA}Protein)} structural maturation}function, activity, or The central dogma asserts that “information” experience flows in only one direction from the genes to the structure of the proteins that the genes bring A Developmental Psychobiological Systems View 47

plan for the construction and determination of the organism. In this view, the genome is not seen as part of the holistic, bidirectional develop- Figure 4.1 mental-physiological system of the organism, re- Central dogma of molecular biology. The right-going sponsive to signals from internal cellular sources arrows represent the central dogma. The discovery of retroviruses (represented by the left-going arrow from such as the cytoplasm of the cell, cellular ad- RNA to DNA) was not part of the dogma but, after the hesion molecules (CAMs), or to extracellular in- discovery, was said by Crick (1970) not to be prohib- fluences such as hormones, and certainly not to ited in the original formulation of the dogma (Crick extraorganismic influences such as stimuli or sig- 1958). (From Gottlieb 1998. Copyright © by the nals from the external environment. American Psychological Association. Reprinted with The main point of my review was to extend the permission.) normally occurring influences on genetic activity to the external environment, thereby further demonstrating that the genome is not encapsu- about through the formula DNA}RNA} lated and is in fact a part of the organism’s gen- Protein. (Messenger RNA [mRNA] is the inter- eral developmental-physiological adaptation to mediary in the process of protein synthesis. In the environmental stresses and signals: Genes express lingo of molecular biology, the DNA}RNA is themselves appropriately only in responding to called transcription and the RNA}Protein is internally and externally generated stimulation. called translation.) After retroviruses were dis- Further, in this holistic view, while genes partic- covered in the 1960s—in which RNA reversely ipate in the making of protein, protein is also transcribes DNA through the enzyme reverse subject to other influences, and protein must be transcriptase—Crick wrote a postscript to his further stimulated and elaborated to become part 1958 paper in which he congratulated himself of the nervous system (or other systems) of the for not claiming that reverse transcription was organism, so that genes operate at the lowest impossible: “In looking back I am struck not only level of organismic organization and they do not, by the brashness which allowed us to venture in and of themselves, produce finished traits or powerful statements of a very general nature, but features of the organism. The organism is a prod- also by the rather delicate discrimination used in uct of epigenetic development, which includes selecting what statements to make” (Crick 1970: the genes as well as many other supragenetic 562). Any ambiguity about the controlling fac- influences. Since this latter point has been the tors in gene expression in the central dogma was subject of numerous contributions (reviewed in removed in a later article by Crick, in which he Gottlieb 1992) I shall not deal with it further specifically says that the genes of higher organ- here, but, rather, restrict myself to documenting isms are turned on and off by other genes (Crick that the activity of genes is regulated in just the 1982: 515). Figure 4.1 shows the central dogma of same way as the rest of the organism, being called molecular biology in the form of a diagram. forth by signals from the normally occurring external environment, as well as the internal en- The Genome According to Central Dogma vironment. While this fact is not well known in the social and behavioral sciences, it is surprising The picture of the genome that emerges from the to find that it is also not widely appreciated in central dogma is (1) one of encapsulation, setting biology proper (Strohman 1997). In biology, the the genome off from supragenetic influences, and external environment is seen as the agent of nat- (2) a largely feedforward informational process ural selection in promoting evolution, not as a in which the genes contain a blueprint or master crucial feature of individual development (van 48 Gilbert Gottlieb

der Weele 1995). Many biologists subscribe to the environment. This holistic developmental system notion that “the genes are safely sequestered of bidirectional, coactional influences is captured inside the nucleus of the cell and out of reach of schematically in figure 4.2. In contrast to the uni- ordinary environmental effects” (Wills 1989: 19). directional and encapsulated genetic predeter- minism of the central dogma, a probabilistic view of epigenesis holds that the sequence and out- Normally Occurring Environmental Influences on comes of development are probabilistically de- Gene Activity termined by the critical operation of various endogenous and exogenous stimulative events. As can be seen in table 4.1, a number of differ- The probabilistic-epigenetic framework pre- ent naturally occurring environmental signals can sented in figure 4.2 is based not only on what we stimulate gene expression in a large variety of now know about mechanisms of individual devel- organisms from nematodes to humans. To opment at all levels of analysis, but also derives understand the findings summarized in table from our understanding of evolution and natural 4.1, the nongeneticist will need to recall that the selection. As everyone knows, natural selection sequence of amino acids in proteins is determined serves as a filter and preserves reproductively by the sequence of nucleotides in the gene that successful phenotypes. These successful pheno- “codes” for it, operating through the inter- types are a product of individual development, mediary of mRNA. So there are three levels of and thus are a consequence of the adaptabil- evidence of genetic activity in table 4.1: protein ity of the organism to its developmental condi- expression or synthesis, mRNA activity, and ge- tions. Therefore, natural selection has preserved netic activity itself. As noted in table 4.1, there (favored) organisms that are adaptably respon- are important neural and behavioral correlations sive to their developmental conditions, both be- to genetic activity, even though the activity of the haviorally and physiologically. Organisms with genes is quite remote from these effects. The post- the same genes can develop very different pheno- translational expression of genes beyond the ini- types under different ontogenetic conditions. tial synthesis of protein involves the intervention Because the probabilistic-epigenetic view pre- of many factors before the end product of gene sented in figure 4.2 does not portray enough de- activity is realized. tail at the level of genetic activity, it is useful The fact that normally occurring environmen- to flesh that out in comparison to the previ- tal events stimulate gene activity during the usual ously described central dogma of molecular biol- course of development in a variety of organisms ogy. As shown in figure 4.3, the original central means that genes and genetic activity are part of dogma explicitly posited one-way tra≤c from the developmental-physiological system and do DNA}RNA}Protein, and was silent about any not stand outside of that system. other flows of “information” (Crick 1958). In the bottom of figure 4.3, probabilistic epigenesis, From the Central Dogma of Molecular Biology to being inherently bidirectional in the horizontal Probabilistic Epigenesis and vertical levels (figure 4.2) has information flowing not only from RNA}DNA but between M } Once again, the main purpose here is to place Protein Protein and DNA DNA. The only } genes and genetic activity firmly within a holis- relationship that is not yet supported is Protein tic developmental-physiological framework, one RNA, in the sense of reverse translation (Protein in which genes not only affect each other and altering the structure of RNA), but there are mRNA but are affected by activities at other lev- other influences of Protein on RNA activity (not els of the system, up to and including the external its structure) which would support such a direc- A Developmental Psychobiological Systems View 49

Table 4.1 Normally Occurring Environmental and Behavioral Influences on Gene Activity Environmental Species Signal or Stimulus Result (alteration in) Author(s) Nematodes absence or presence of neuronal daf-7 gene mRNA Ren et al. 1996 food expression, inhibiting or provoking larval development Fruit flies heat stress during larval transient elevated heat shock Singh & Lakhotia 1988 development proteins and thermotolerance Fruit flies light-dark cycle PER and TIM protein Lee et al. 1996; Myers expression and circadian et al. 1996 rhythms Various reptiles Incubation temperature sex determination Reviewed in Bull 1983; van der Weele 1995 Songbirds conspecific song forebrain mRNA Mello, Vicario, & (canaries, Clayton 1992 zebra finches) Hamsters light-dark cycle pituitary hormone mRNA Hegarty et al. 1990 and reproductive behavior Mice acoustic stimulation c-fos expression, neuronal Ehret & Fisher 1991 activity, tonotopy in auditory system Mice light-dark cycle c-fos mRNA expression in Smeyne et al. 1992 suprachiasmatic nucleus of hypothalamus, circadian locomotor activity Rats tactile stimulation c-fos expression and number Mack & Mack 1992 of somatosensory cortical neurons Rats learning task involving nuclear RNA base ratios in Hydén & Egyházi 1962 vestibular system vestibular nerve cells Rats visual stimulation RNA and protein synthesis in Rose 1967 visual cortex Rats Environmental complexity brain RNA diversity Uphouse & Bonner 1975; review in Rosenzweig & Bennett 1978 Rats prenatal nutrition cerebral DNA (cerebral cell Zamenhof & van number) Marthens 1978 Rats infantile handling; separa- Hypothalamic mRNAs for Meaney et al. 1996 tion from mother corticotropin-releasing hormone throughout life Cats visual stimulation visual cortex RNA complexity Grouse et al. 1979 (diversity) Humans academic examinations interleukin 2 receptor mRNA Glaser et al. 1990 taken by medical students (immune system response) (psychological stress)

Note: mRNA, messenger RNA; PER and TIM are proteins arising from per (period) and tim (timeless) gene activity. (From Gottlieb 1998. Copyright © 1998 by the American Psychological Association. Reprinted by permission.) 50 Gilbert Gottlieb

Figure 4.2 Probabilistic-epigenetic framework: Depiction of the completely bidirectional and coactional nature of genetic, neural, behavioral, and environmental influences over the course of individual development. (From Individual Development and Evolution: The Genesis of Novel Behavior by Gilbert Gottlieb. Copyright © 1992 by Oxford University Press, Inc. Used by permission of Oxford University Press, Inc.) tional flow. For example, a process known as Finally, one hopes that the emphasis here on phosphorylation can modify proteins such that normally occurring environmental influences on they activate (or inactivate) other proteins gene activity does not raise the spectre of a new, (Protein}Protein) which, when activated, trigger subtle form of “environmentalism.” If I were to rapid association of mRNA (Protein}RNA say organisms are often adaptably responsive to activity). When mRNAs are transcribed by DNA their environments, I do not think that would they do not necessarily become immediately label me an environmentalist. So, by calling at- active but require a further signal to do so. The tention to genes being adaptably responsive to consequences of phosphorylation could provide their internal and external environments, I am that signal (Protein}Protein}mRNA activity} not being an environmentalist but merely in- Protein). A process like this appears to be including genetic activity within the probabilistic- volved in the expression of “fragile X mental epigenetic framework that characterizes the retardation protein” under normal conditions organism and all of its constituent parts. and proves disastrous to neural and psychologi- In view of the findings reviewed here, in the cal development when it does not occur.1 future it would be most important to eschew both In summary, the central dogma lies behind genetic determinism and environmental deter- the persistent trend in biology and psychology to minism, as we now should understand that it is view genes and environment as making identi- truly correct (not merely a verbalism) to say that fiably separate contributions to the phenotypic environments and genes necessarily cooperate in outcomes of development. Quantitative behavior bringing about any outcome of individual devel- genetics is based on this erroneous assumption. opment. (For exceptions see note 1.) The remain- Although genes no doubt play a constraining role der of the book will no doubt enlarge on this in development, the actual limits of these con- theme many times over and in various contexts, straints are quite wide and, most importantly, both evolutionary and developmental. I appreci- can not be specified in advance of experimental ate being given the opportunity to recount my manipulation or accidents of nature. There is no contribution to this still-maturing synthesis. In doubt that development is constrained at all lev- addition to the writings of the psychologists els of the system (figure 4.2), not only by genes Zing-Yang Kuo, T. C. Schneirla, and Daniel and environments. Lehrman, the probabilistic-epigenetic view builds A Developmental Psychobiological Systems View 51

on the open-systems view of development cham- pioned by the biologists Ludwig von Bertalanffy (1962), Paul Weiss (1939), and Sewall Wright (1968), as I have described more fully elsewhere (e.g., Gottlieb 1992, 1997; Gottlieb, Wahlsten, and Lickliter 1998). In this brief overview I have not covered certain features of probabilistic epigenesis: nonobvious experiential contingencies in instinctive behavior, developmental causality, and the signiﬁcance of coaction for individual development. These topics are covered in Gott- lieb (in press).

Figure 4.3 Acknowledgment Different views of influences on genetic activity in the central dogma and probabilistic epigenesis. The filled The author’s work is funded in part by NIMH arrows indicate documented sources of influence, while the open arrow from Protein back to RNA remains a grant P50-MH-52429. theoretical possibility in probabilistic epigenesis and is prohibited in the central dogma (as are ProteinM Note Protein influences). Protein}Protein influences occur when (1) prions transfer their abnormal conformation 1. The label of “fragile X mental retardation protein” to other proteins and (2) when, during normal develop- makes it sound as if there is a gene (or genes) that pro- ment, proteins activate or inactivate other proteins as in duces a protein that predisposes to mental retardation phosphorylation example described in text. The filled whereas, in actual fact, it is this protein that is missing arrows from Protein to RNA represent the activation (absent) in the brain of fragile X mental retardates, and of mRNA by protein as a consequence of phosphoryla- thus represents a failure of gene (or mRNA) expression M tion, for example. DNA DNA influences are termed rather than a positive genetic contribution to mental “epistatic,” referring to the modification of gene expres- retardation. The same is likely true for other “genetic” sion depending on the genetic background in which disorders, whether mental or physical: These most often they are located. In the central dogma genetic activity is represent biochemical deficiencies of one sort or an- } dictated solely by genes (DNA DNA), whereas in other due to the lack of expression of the requisite genes probabilistic epigenesis internal and external environ- and mRNAs to produce the appropriate proteins nec- mental events activate genetic expression through pro- essary for normal development. Thus, the search for } teins (Protein DNA), hormonal, and other influences. “candidate genes” in psychiatric or other disorders is To keep the diagram manageable, the fact that behav- most often a search for genes that are not being ex- ior and the external environment exert their effects on pressed, not for genes that are being expressed and DNA through internal mediators (proteins, hormones, causing the disorder. So-called cystic fibrosis genes and etc.) is not shown; nor is it shown that the protein prod- manic-depression genes, among others, are in this cate- ucts of some genes regulate the expression of other gory. The instances that I know of in which the pres- genes. (Further discussion in text.) (From Gottlieb ence of genes causes a problem are Edward’s syndrome 1998. Copyright © 1998 by the American Psychological and trisomy 21 (Down syndrome), wherein the presence Association. Reprinted with permission.) of an extra, otherwise normal, chromosome 18 and 21, respectively, causes problems because the genetic system is adapted for two, not three, chromosomes at each location. In some cases, it is of course possible that the expression of mutated genes can be involved in a dis- 52 Gilbert Gottlieb

order, but, in my opinion, it is most often the lack of Gottlieb, G. (1970). Conceptions of prenatal behavior. expression of normal genes that is the culprit. In L. R. Aronson, E. Tobach, D. S. Lehrman, and J. S. Rosenblatt (Eds.), Development and Evolution of Behavior: Essays in Memory of T. C. Schneirla, pp. References 111–137. San Francisco: W. H. Freeman. Gottlieb, G. (1971a). Ontogenesis of sensory function Anokhin, P. K. (1964). Systemogenesis as a general in birds and mammals. In E. Tobach, L. R. Aronson, regulator of brain development. Progress in Brain Re- and E. Shaw (Eds.), The Biopsychology of Development, search 9: 54–86. pp. 111–137. New York: Academic Press. Black, J. E., and W. T. Greenough. (1998). Devel- Gottlieb, G. (1971b). Development of Species Identifi- opmental approaches to the memory process. In J. L. cation in Birds: An Inquiry into the Prenatal Determi- Martinez and R. P. Kesner (Eds.), Learning and Mem- nants of Perception. Chicago: University of Chicago ory: A Biological View (3rd ed.), pp. 55–88. New York: Press. Academic Press. Gottlieb, G. (1976). Conceptions of prenatal develop- Bull, J. J. (1983). Evolution of Sex Determining Mecha- ment: Behavioral embryology. Psychological Review 83: nisms. Menlo Park, CA: Benjamin/Cummings. 215–234. Catalano, S. M., and C. J. Shatz. (1998). Activity- Gottlieb, G. (1983). The psychobiological approach to dependent cortical target selection by thalamic axons. developmental issues. In M. M. Haith and J. J. Campos Science 281: 559–562. (Eds.), Handbook of Child Psychology: Infancy and De- Coghill, G. E. (1929). Anatomy and the Problem of velopmental Psychobiology (vol. 2, 4th ed.), pp. 1–26. Behaviour. Cambridge: Cambridge University Press. New York: Wiley. Crick, F. H. C. (1958). On protein synthesis. Symposia Gottlieb, G. (1991). Experiential canalization of behav- of the Society for Experimental Biology: No. 12. The ioral development: Theory. Developmental Psychology Biological Replication of Macromolecules, pp. 138–163. 27: 4–13. Cambridge: Cambridge University Press. Gottlieb, G. (1992). Individual Development and Evo- Crick, F. (1970). Central dogma of molecular biology. lution: The Genesis of Novel Behavior. New York: Ox- Nature 227: 561–563. ford University Press. Crick, F. (1982). DNA today. Perspectives in Biology Gottlieb, G. (1997). Synthesizing Nature-Nurture: and Medicine 25: 512–517. Prenatal Roots of Instinctive Behavior. Mahwah, NJ: Ehret, G., and R. Fisher. (1991). Neuronal activity and Erlbaum. tonotopy in the auditory system visualized by c-fos gene Gottlieb, G. (1998). Normally occurring environmental expression. Brain Research 567: 350–354. and behavioral influences on gene activity: From cen- Glaser, R., S. Kennedy, W. P. Lafuse, R. H. Bonneau, tral dogma to probabilistic epigenesis. Psychological C. Speicher, J. Hillhouse, and J. K. Kiecolt-Glaser. Review 105: 792–802. (1990). Psychological-stress-induced modulation of in- Gottlieb, G. (in press). Probabilistic epigenesis of de- terleukin 2 receptor gene expression and interleukin 2 velopment. In J. Valsiner and K. Connolly (Eds.), production in peripheral blood leukocytes. Archives of Handbook of Developmental Psychology. London: Sage. General Psychiatry 47: 707–712. Gottlieb, G., and P. H. Klopfer. (1962). The relation of Gottlieb, G. (1961). Developmental age as a baseline developmental age to auditory and visual imprinting. for the determination of the critical period in imprint- Journal of Comparative and Physiological Psychology ing. Journal of Comparative and Physiological Psycho- 55: 821–826. logy 54: 422–427. Gottlieb, G., and J. G. Vandenbergh. (1968). Ontogeny Gottlieb, G. (1965). Imprinting in relation to parental of vocalization in duck and chick embryos. Journal of and species identification by avian neonates. Journal of Experimental Zoology 168: 307–325. Comparative and Physiological Psychology 59: 345–356. Gottlieb, G., D. Wahlsten, and R. Lickliter. (1998). The Gottlieb, G. (1966). Species identification by avian significance of biology for human development: A neonates: Contributory effect of perinatal auditory developmental psychobiological systems perspective. In stimulation. Animal Behaviour 14: 282–290. A Developmental Psychobiological Systems View 53

R. M. Lerner (Ed.), Handbook of Child Psychology: Ren, P., C-S. Lin, R. Johnson, P. S. Albert, D. Pilgrim, Theoretical Models of Human Development vol. 1, pp. and D. L. Riddle. (1996). Control of C. elegans lar- 233–273. (5th ed.) New York: Wiley. val development by neuronal expression of a TGFß Grouse, L. D., B. K. Schrier, C. H. Letendre, and P. G. homolog. Science 274: 1389–1391. Nelson. (1980). RNA sequence complexity in central Rose, S. P. R. (1967). Changes in visual cortex on first nervous system development and plasticity. Current exposure of rats to light: Effect on incorporation of tri- Topics in Developmental Biology 16: 381–397. tiated lysine into protein. Nature 215: 253–255. Hegarty, C. M., J. A. Jonassent, and E. L. Bittman. Rosenweig, M. R., and E. L. Bennett. (1978). Expe- (1990). Pituitary hormone gene expression in male riential influences on brain anatomy and brain chem- golden hamsters: Interactions between photoperiod istry in rodents. In G. Gottlieb (Ed.), Early Influences, and testosterone. Journal of Neuroendocrinology 2: pp. 289–327. New York: Academic Press. 567–573. Schneirla, T. C. (1956). Interrelationships of the Hydén, H., and E. Egyházi. (1962). Nuclear RNA “innate” and the “acquired” in instinctive behavior. In changes of nerve cells during a learning experiment in P-P. Grassé (Ed.), L’Instinct dans le comportement des rats. Proceedings of the National Academy of Sciences, animaux et de l’homme, pp. 387–452. Paris: Masson. USA 48: 1366–1373. Schneirla, T. C. (1965). Aspects of stimulation and Kempermann, G., H. G. Kuhn, and F. H. Gage. organization in approach/withdrawal. Processes under- (1997). More hippocampal neurons in adult mice living lying vertebrate behavioral development. Advances in in an enriched environment. Nature 386: 493–495. the Study of Behavior 1: 1–74. Kuo, Z-Y. (1967). The Dynamics of Behavior Devel- Schneirla, T. C. (1966). Behavioral development and opment. New York: Random House. comparative psychology. Quarterly Review of Psychol- Lee, C., V. Parikh, T. Itsukaichi, K. Bae, and I. Edery. ogy 41: 283–302. (1996). Resetting the Drosophila clock by photic regula- Singh, A. K., and S. C. Lakhotia. (1988). Effect of low- tion of PER and a PER-TIM complex. Science 271: temperature rearing on heat shock protein synthesis 1740–1744. and heat sensitivity in Drosophilia melanogaster. Devel- Lehrman, D. S. (1953). A critique of Konrad Lorenz’s opmental Genetics 9: 193–201. theory of instinctive behavior. Quarterly Review of Biol- Smeyne, R. J., K. Schilling, L. Robertson, D. Luk, J. ogy 28: 337–363. Oberdick, T. Curran, and J. I. Morgan. (1992). Fos- Mack, K. J., and P. A. Mack. (1992). Induction of lacZ transgenic mice: Mapping sites of gene induction transcription factors in somatosensory cortex after in the central nervous system. Neuron 8: 13–23. tactile stimulation. Molecular Brain Research 12: 141– Strohman, R. C. (1997). The coming Kuhnian revolu- 147. tion in biology. Nature Biotechnology 15: 194–200. Meaney, M. J., J. Diorio, D. Francis, J. Widdowson, P. Uphouse, L. L., and J. Bonner. (1975). Preliminary evi- LaPlante, C. Caldji, S. Sharma, J. P. Seckl, and P. M. dence for the effects of environmental complexity on Plotsky. (1996). Early environmental regulation of fore- hybridization of rat brain RNA to rat unique DNA. brain glucocorticoid receptor gene expression: Impli- Developmental Psychobiology 8: 171–178. cations for adrenocortical responses to stress. Devel- van der Weele, C. (1995). Images of Development: Envi- opmental Neuroscience 18: 49–72. ronmental Causes in Ontogeny. Doctoral dissertation, Mello, C. V., D. S. Vicario, and D. F. Clayton. (1992). Vrije University. Song presentation induces gene expression in the song- von Bertalanffy, L. (1962). Modern Theories of Devel- bird forebrain. Proceedings of the National Academy of opment: An Introduction to Theoretical Biology. New Sciences, USA 89: 6818–6822. York: Harper. Morgan, C. L. (1893–94). Dr. Weismann on heredity Weiss, P. (1939). Principles of Development: A Text in and progress. The Monist 4: 30. Experimental Embryology. New York: Holt, Rinehart Myers, M. P., K. Wager-Smith, A. Rothenfluh- and Winston. Hilfiker, and M. W. Young. (1996). Light-induced West, M., and A. King. (1987). Settling nature and nur- degradation of TIMELESS and entrainment of the ture into an ontogenetic niche. Developmental Psycho- Drosophila circadian clock. Science 271: 1736–1740. biology 20: 549–562. 54 Gilbert Gottlieb

Wills, C. (1989). The Wisdom of the Genes: New Pathways in Evolution. New York: Basic Books. Wright, S. (1968). Evolution and the Genetics of Pop- ulation: Vol. 1. Genetic and Biometric Foundations. Chicago: University of Chicago Press. Zamenhof, S., and E. van Marthens. (1978). Nutri- tional inﬂuences on prenatal brain development. In G. Gottlieb (Ed.), Early Inﬂuences, pp. 149–186. New York: Academic Press. 5 Gene, Organism and Environment: A New Introduction

Richard C. Lewontin

The essay “Gene, Organism and Environment” metaphor would be a complete description of the takes up two fundamental metaphors that inform object of interest, in which case it is no longer theoretical, experimental, and natural historical metaphorical. We are constantly being urged to practice in biology. These are the metaphors of see the world of living phenomena in a different development, which carries the implication of and better way, using new organizing principles: an unfolding or unrolling of an internal program the organism as a computer program, the organ- that determines the organism’s life history from ism as a Boolean network, the organism as a self- its origin as a fertilized zygote to its death, and organizing machine. But the question that must the metaphor of adaptation, which asserts that always be asked of such proposals is how the evolution consists in the shaping of species to ﬁt practice of biologists would be changed in a way the requirements of an autonomous external en- that would allow us to answer previously unan- vironment. That is, both in developmental and swerable questions. What new experiment, what in evolutionary biology, the inside and the out- new testable prediction, what explanation of a side of organisms are regarded as separate previously mysterious or contradictory observa- spheres of causation with no mutual dependence. tion will ﬂow from a change in point of view? The burden of the essay is that these metaphors What will the dialectical point of view about mislead the biologist because they fail to take organism and environment contributed to the account of the interactive processes that link the doing of biology? The biological philosopher inside and the outside. The changes that occur in only interprets the world; the point, however, is an organism during its life from conception to to change it. death depend uniquely on both the cell constit- When we consider how the understanding of uents that are present in the fertilized egg and genes and environment in “development” (we on the sequence of environments through which have no other word for it) has been effective in the organism passes in its lifetime. But the nature biology, we see a curiously split history. In ap- of these environments is not independent of the plied biology, especially in plant and animal organism because at every instant the life activ- breeding, the understanding that there is a unique ities of the organism determine what constitutes interaction between genotype and environment the relevant combinations of external physical in development has been of fundamental practi- states and they simultaneously cause changes in cal importance for nearly one hundred years. The those states. The question that remains unad- standard method of breeding crops for, say, in- dressed by the essay is, “So what?” crease in yield, is to grow the various varieties It is easy enough to criticize the metaphors that under test in several years and in several locations are used in science. None is perfect and all cap- in the region of production. The variety chosen ture only some parts of reality while falsifying for release is not necessarily the one with high- others. The “billiard ball” model of molecules in est average yield over environments because uni- a gas calls attention to the elastic collisions and formity of result over years and locations is given the preservation of momentum in the kinetic strong weight. A seed company that sells a vari- theory of gases, but real billiard balls do not ety with slightly higher average yield, but that undergo perfectly elastic collisions, and they have causes palpable loss to a quarter of the farmers, different colors and make a clicking sound when will soon go out of business. The result of the they strike each other. Yet these truths about bil- breeding practice that takes account of genotype- liard balls are familiar to every physicist, and no environment interactions has been the evolution one is led astray by them. Indeed, the only perfect in crop plants of less and less environmental sen- 56 Richard C. Lewontin

sitivity, but a significant amount always remains quantitative geneticists and seems not susceptible and is taken account of in agricultural genetics. to knockout gene experiments. If developmental In contrast, nearly every developmental geneticist genetics is to be really relevant to understanding working on morphogenesis in laboratory model the production of form, it cannot constrain itself organisms ignores completely the effect of the in- to the universalities of form, but must also ex- teraction of genes and environment. One source plain the variation, and in particular why an of the difference in practice is the difference in increase in temperature causes an increase in size problematic. For the developmental geneticist for one genotype, while it causes a decrease in size using Drosophila or Caenorhabditis, the question for another genotype and is nonmonotonic in its being asked is how the transcription from one effect on a third. Presumably the way in which gene by the cell influences the transcription of this is to be done is, at first, to continue the study other genes. But this leads to a second difference, of the same signaling pathways, but to create which is in the nature of the genetic variation that experiments that will cause small perturbations, is used. To study the signaling pathways in the both genetic and environmental, to those path- transcription of genes, mutants of major qual- ways and then to analyze at a molecular level itative effect are employed, and there is a strong how those perturbations result in small changes bias toward gene loci and their mutations whose in morphogenesis. But such experiments have a effects can be seen reliably. The environment greater importance than the explanation of the enters in only as a set of laboratory conditions observed genotype/environment interactions in necessary for the survival and maximal expres- development. The ultimate purpose of develop- sion of the abnormal mutations. Every theoret- mental genetics must be to provide an articu- ical and experimental agricultural geneticist, on lated story of the coming into being of particular the other hand, is concerned with continuously macromolecular structures at particular places at and subtly varying physiological and morpholog- particular moments in the life of an organism. ical traits in which the interaction with the envi- The existence of nonlinearities and nonmonot- ronment is critical. The Drosophila geneticist is onicities in genotype/environment interactions at leisure to choose the best genotypes and phe- tells us that the correct story cannot be given notypes for the purposes of a “knockout” ex- purely as a diagram of gene-gene signaling. A periment, in which the transcription of a gene is crucial material mechanism is missing for the prevented and the signaling pathway of which it establishment of morphogenetic gradients and is a part is blocked. The soybean geneticist is the localization of qualitatively different macro- stuck with the trying to find multilocus geno- molecules when only internal autonomous types that will increase seed yield by a couple of processes are considered. The realization of the percent. program of explanation of developmental biol- The standard program of developmental ge- ogy cannot be accomplished without including netics is reinforced by the discovery that the gene the influence of the external. signaling pathways that differentiate the front The effect of taking the dialectic between or- end from the back end of an animal, or that are ganism and environment seriously is, if anything, active in the laying down of body segmentation, even more profound for evolutionary biology are distributed throughout the entire range of than for studies of development. If it is true that animal forms and may even have homologs in the life activities of an organism enter into the plants. Thus the developmental geneticist has the specification of its environment, then different satisfaction of studying a system that has been phenotypes will make different environments. preserved since the pre-Cambrian. But a great Because evolution occurs by a change in the fre- deal of evolution is precisely of sort studied by quencies of heritable types in a population, it fol- Gene, Organism and Environment: Introduction 57

lows that during the evolution of a change in the model-making with no constraints beyond the frequencies of heritable types in a population the elements of genetics. environments experienced by the population are The program of experimental and observa- changing. But this, in turn, means that selective tional evolutionary genetics does not escape the forces, which are a consequence of relations be- implications of constructionism. The form of tween organisms and their environments, are also the classical experiment on the determination of changing in a way that is sensitive to the change the fitnesses of different genotypes is to measure in population composition. The result is that, real- some components of the age specific mortality istically, selection coe≤cients are frequency- and fecundity schedules of each genotype in iso- dependent, and theoretical modeling of the lation in a physical environment or environments effect of natural selection must use frequency- determined by the investigator. The resultant dependent formulations. In the history of the fitness estimates are then used to explain what development of population genetic theory some has happened or to predict what will happen small attention has been paid to frequency- during the evolution of the population. But if the dependent models, but for the most part they are fitness of a genotype is dependent upon the mix- regarded as special cases, to be considered in ture of other genotypes in the population, then textbooks and monographs under the rubric of nothing has been learned by measuring the “Some Complexities.” Nearly every model of “fitnesses” of the genotypes in isolation. On the natural selection assumes constant fitnesses or, other hand, it is hopeless to measure the net at the most, fitnesses that vary with some au- fitnesses of many genotypes in an immense array tonomous extrinsic force or stochastically. The of different frequency combinations. A realis- conclusions from frequency-independent models tic evolutionary genetics that takes context de- of selection are, in general, invalid for frequency pendence of fitness seriously cannot continue to dependent cases. For example, natural selection operate at the purely superficial level of directly does not maximize average fitness and may even measuring fitnesses, of characterizing outcomes minimize it. The first consequence of taking a without understanding their mediation. What constructionist view seriously would be the is required is an experimental program of un- abandonment of frequency independent selection packing “fitness.” This involves determining models and their replacement by more realistic experimentally how different genotypes juxtapose ones. There is, however, an immense cost to be different aspects of the external world, how they paid. Once we admit frequency dependence into alter that world and how those different envi- our modeling, there are no purely theoretical ronments that they construct affect their own bio- constraints on what may be predicted. Every fre- logical processes and the biological processes of quency dependent model makes its own peculiar others. The understanding of living systems can- predictions. Every biological situation will need not be achieved merely through a description of to be specifically modeled, and small errors in the their details, but neither can that understanding claims about biology may lead to radical errors in be obtained by ignoring them. The truth may not evolutionary prediction. Moreover, the theoreti- be in the details, but the details matter and we cian becomes inextricably tied to the experimen- must have them in hand for our program of talist and natural historian, because without the explanation. Ironically, the demand for a dialec- biological details there is no applicable model. It tical view of the evolution of organisms and their is not clear that theoretical population geneticists environment is, at first pass, a demand that evo- and evolutionists are prepared to give up the dis- lutionary biologists take Descartes’s original ciplinary autonomy that they currently possess, machine metaphor seriously. an autonomy that licenses them to engage in

6 Gene, Organism and Environment

Richard C. Lewontin

The modern theory of evolution is, as is so coming from internal factors, and causes of phy- often said, a fusion of the two great insights of logenetic variation, as being imposed from the nineteenth-century biology: Darwin’s realization external environment by way of natural selec- that the variation among species arises from the tion. (I gloss over Darwin’s later flirtation with conversion of variation between individuals with- Lamarckism, since the proposition that the envi- in species, and Mendel’s discovery of the segre- ronment specifically engenders heritable adaptive gation of discrete factors as the basis for the variation is at total variance with the Darwinian inheritance of differences between individuals. We mechanism of evolution.) It is from this view of are constantly reminding ourselves and others environment as the cause of organisms that the that the immense progress made in biology in entire corpus of modern evolutionary biology the present century rests firmly on these two arises. major discoveries of a previous time. What is not We cannot appreciate fully the nature of the always appreciated, however, is that the legacies change in biology wrought by Mendel and of Darwin and Mendel are also responsible for Darwin unless we understand the historical im- certain di≤culties in biology, di≤culties that pre- portance of the objectification of the organism. vent us from some kinds of further progress and Descartes’ metaphor of the organism as machine which keep us locked into a rigid framework of had virtually no impact on biology for two hun- thought about the development and evolution dred years. So, for example, even Harvey’s of organisms. These di≤culties arise, ironically, mechanical description of the circulation of from the very source of Mendel and Darwin’s blood was not really accepted until the beginning success as biologists, their separation of internal of the nineteenth century, and Caspar Friedrich from external forces acting on organisms. For Wolff’s epigenetic theory and his remarkably Mendel the internal “factors” were the causes of modern distinction between genotype and phe- the form of the organism and were, in the end, notype had no effect in embryology until the the proper objects of study. The “factors,” what Entwicklungsmechanik of the latter part of the we now call “genes,” were the subjects and the last century. Biology lacked clear notions of sep- organisms the objects of developmental forces. arable causes and effects and, more important, a From this view of gene as the cause of organism systematic commitment to the analysis of biolog- has flowed the entire corpus of modern mechanical systems along mechanistic lines. Lamarck’s ical and molecular genetics. For Darwin, the ex- view that information from the external world ternal world, the environment, acting on the could become permanently incorporated in or- organism was the cause of the form of organisms. ganisms and their progeny through the mediation The environment, the external world with its of a living being’s needs and will was quintes- autonomous properties, was the subject and the sentially representative of pre-Darwinian biol- organism was, again, the object acted upon. In ogy. As Charles Gillespie (1959) so cogently puts Darwinism, the organism is the interaction of it: “Lamarck’s theory of evolution belongs to the two causal sequences, autonomous in their dy- contracting and self-defeating history of subjec- namics. Internal forces produced the variation tive science, and Darwin’s to the expanding and among organisms, and autonomous external conquering history of objective science.” forces molded the species on the basis of these By making organisms the objects of forces autonomous internally caused variations. The es- whose subjects were the internal heritable factors sence of Darwin’s account of evolution was the and the external environment, by seeing organ- separation of causes of ontogenetic variation, as isms as the effects whose causes were internal 60 Richard C. Lewontin

and external autonomous agents, Mendel and lations and the new observations of radioactiv- Darwin brought biology at last into conformity ity, of light, and of astronomy. Without relativity with the epistemological meta-structure that and quantum theory, physics and theoretical already characterized physics since Newton and chemistry would have ground to a halt. On the chemistry since Lavoisier. This change in world other hand, biology, far more diverse in its sub- view was absolutely essential if biology was to ject matter, far more loosely tied together into a progress by making contact with physical science coherent science, has undergone a radically un- and by becoming quantitative and predictive. even development. Some branches such as molec- The mechanistic reductionism and the clear sepa- ular biology have made extraordinary progress ration of internal and external were as necessary by concentrating on just those questions for in the nineteenth century for the creation of a which the simple mechanical reductionism of the scientific biology as Newton’s ideal bodies and nineteenth century is the perfect epistemology. perfect determinism were for the physics of the Developmental biology, the study of cognition seventeenth. But we must not confuse the histor- and memory, and evolutionary biology, on the ically determined necessity of a particular episte- other hand, have profited only marginally from mological stance at one stage in the development these rapid advances. Rather, they are stalled by of a science with a perfect model that will guar- their attempt to use outdated concepts to con- antee all future progress. On the contrary, the front a rich phenomenology to which these con- very progress made possible by certain revolu- cepts clearly do not apply. Evolutionary biology tionary formulations may lead eventually to suffers particularly because it is the nexus of all results that are in contradiction with those earlier other biological sciences, so that a lack of pro- formulations and which can be resolved only by gress in developmental biology, in ecology, in their reexamination. Yet those reexaminations behavioral science, all are fatal to a proper under- are themselves rooted in the past formulations. standing of evolution. Newtonian mechanics and classical optics were Specifically, evolutionary biology must con- in serious contradiction with the newer observa- front two issues about the forms of organisms. tions of physics at the end of the nineteenth cen- One is the ontogenetic process by which the se- tury, but anyone familiar with the development quence of forms that comprise an individual’s of the Special Theory of Relativity immediately life history come into being. The second is the sees how that theory, in one aspect a negation phylogenetic process by which species as collec- of Newtonian principles, is built entirely on a tive entities form and change based on the varia- Newtonian framework and could never have tions among the individuals that make them up. been developed in the absence of classical Classical, post-Darwinian, post-Mendelian biol- physics. ogy has settled on two metaphors through which As in physics, so too in biology. As time has the processes are seen. The first, ontogenetic, passed, Mendel’s view of organisms as the mani- process is seen as an unfolding of a form, already festation of autonomous internal “factors” with latent in the genes, requiring only an original their own laws, and Darwin’s view of organisms triggering at fertilization and an environment as passive objects molded by the external force of adequate to allow “normal” development to con- natural selection, have become increasingly in tinue. The second, phylogenetic, process is seen contradiction with the known facts of develop- as problem and solution. The environment “poses mental and population biology. But the situation the problem”; the organisms posit “solutions,” of in biology has developed rather differently than which the best is finally “chosen.” The organism in physics. All of physical science seemed blocked proposes; the environment disposes. These two by the repeated conflict between classical formu- metaphors are simply the forms of the original Gene, Organism and Environment 61

Mendelian view that internal factors make the genetics (and of evolution) consistently describe individual organism and the Darwinian view that organisms as “determined” by their genes. It external forces determine the collectivity. In the is a short step to seeing organisms as “lumber- balance of this chapter, I want to make clear why ing robots” created by their genes “body and these two metaphors are wrong. Individual devel- mind.” It is not hard to see why such unbiological opment is not an unfolding, and evolution is not rubbish is taken seriously. Yet, the vast majority a series of solutions to present problems. Rather, of morphological, behavioral, and physio- genes, organisms, and environments are in recip- logical differences among individuals do not rocal interaction with each other in such a way “Mendelize.” It is simply not possible to read off that each is both cause and effect in a quite the genotypic differences between tall and short complex, although perfectly analyzable, way. individuals of Rumex acetosella from their The known facts of development and of natural individual phenotypes or from any number of history make it patently clear that genes do controlled test crosses involving them or their rel- not determine individuals nor do environments atives, not to speak of the genotypic difference determine species. between faithful spouses and philanderers. The reaction of many evolutionists (encouraged, again, by textbooks) to this obvious fact is to Gene, Environment, and Organism in consign such characters to “polygenic control,” Development invoking the multiple factor hypothesis, and thus saving the basic model that genes determine I will begin with the obvious. It is well known organisms. But quantitative variation of a char- that Mendel solved the basic problem of the laws acter is not prima facie evidence that it is of inheritance by investigating the heredity of influenced by many genes. Single gene mutations very special sorts of differences, those in which affecting eye shape, wing variation and bristle there was a determinate correspondence between number in Drosophila, or enzyme activity in genotype and phenotype. Given the genotype, humans, all show quantitative variation in phe- the phenotype corresponding to it was unam- notype and, unless care is taken to control the biguously defined, at least under the condition of condition of development, considerable overlap Mendel’s experimental garden. Indeed, it is the between genotypes in their phenotypic distribu- essence of the Mendelian methodology that one tions. The fundamental general fact of pheno- can read off the genotype given either the indi- genetics is that the phenotype of organisms is a vidual’s phenotype or, in the case of complete consequence of non-trivial interaction between dominance, the phenotypes of the progeny of genotype and environment during development. a single test cross. It is not always explicitly All that genes ever do is to specify a norm of reac- pointed out that all of modern biochemical, tion over environments. Moreover, fitness too is molecular, and developmental genetics also has a phenotype and varies from environment to depended for progress on finding genetic dif- environment, both because other aspects of the ferences that make clear-cut, non-overlapping phenotype develop differently in different envi- phenotypic classes under easily controlled condi- ronments, and because a given shape or behav- tions. It is amusing to contemplate where bacte- iour or physiology will confer different fitnesses riophage genetics (and, a fortiori, all of molecular in different environments. During the fifteen genetics) would be now if Benzer’s bacteriophage years between 1950 and 1965, population geneti- mutants had only differed from each other by 1% cists, largely under the influence of plant and ani- in the number of progeny phage they produced mal breeders and of Schmalhausen’s seminal on restrictive hosts. One consequence of this work, Factors of Evolution (1949), devoted con- methodological history is that textbooks of 62 Richard C. Lewontin

siderable attention to the environmental contin- edge is usually to attempt to assign the relative gency of phenotype and to the effects of selection weights to genotype, environment, and devel- in different environmental regions (see, for ex- opmental noise, as for example in a heritability ample, Lerner 1954, Dobzhansky and Spassky study or some other form of the analysis of vari- 1944, Falconer 1960, and Robertson 1960), but ance. The second fact that evolutionary biologists this work passed out of fashion with the advent must cope with is that the influence of each factor of molecular population genetics. Indeed, except depends upon the influence of the other factors so for the famous work of Clausen, Keck and that no assignment of fixed weights to genetic, Hiesey (1958), no study of the norms of reaction environmental, and noise components of varia- of naturally occurring heterozygous genotypes tion is possible (Lewontin 1974). There are no appeared until the experiments of Gupta on more important experimental results for evolu- Drosophila (Gupta and Lewontin 1982). Thus, tionary biology than those of Rendel (1967) on the basic data for judging the effects of selection canalization, yet the early impact of these find- in particular genotypes are simply lacking. The ings on evolutionary theory seems now to be for- little that is known shows clearly that the de- gotten. What Rendel and others have shown is velopmental responses of different genotypes to the reciprocity of effects of genetic state on envi- varying environments are non-linear and do not ronmental sensitivity and of environmental state allow the simple ordering of genotypes along on genetic sensitivity of the developing organism. a one-dimensional scale of phenotype. Norms Moreover, the genotype is itself hierarchically of reaction cross each other so that no geno- organized so that the environmental sensitivity of type gives a phenotype unconditionally larger, a given gene substitution will depend upon genes smaller, faster, slower, more or less different than at other loci. The consequence is that selection another. These well-known facts seem, however, can change the average expression of a trait, inde- to have made no impact on evolutionary theorists pendently from the variation of that expression in who continue to speak about selection for a char- response to variable environment and to develop- acter and about genes that are selected because mental noise. Thus, by selection, an organism can they produce that character. be made developmentally insensitive or highly A second, less well-known, feature of develop- sensitive to perturbations of its genotype, its envi- ment is that phenotype is not given even when the ronment, developmental accidents, or any com- genotype and the environment are completely bination of these. Internal and external facts not specified. There is a significant effect of “develop- only play a role in development, but each deter- mental noise” (Waddington 1957) in producing mines the role played by the other. phenotype. The two sides of a Drosophila have Thus far, my description of development is the same genotype, and no reasonable definition still in terms of the organism as the object, the of environment will allow that the left and right effect of causes both internal and external to it, sides of a pupa developing halfway up the side of although the causes themselves interact with each a glass milk bottle in the laboratory are in dif- other. The final step in the integration of devel- ferent environments. Yet, the number of sterno- opmental biology into evolution is to incorporate pleural bristles and the number of eye facets differ the organism as itself a cause of its own develop- between the two sides of an individual fly. Small ment, as a mediating mechanism by which exter- events at the level of thermal noise acting during nal and internal factors influence its future. To cell division and differentiation have large effects describe phenotype as the consequence of gene, on the final developmental outcome. environment, and accident leaves out of account It is only the beginning of understanding to entirely the element of temporal order which is agree that internal and external factors con- of the essence in a developmental process. The tribute to phenotype. The reaction to that knowl- organism’s phenotype is in a continual state of Gene, Organism and Environment 63

change from fertilization to death. The pheno- mental problems. What then is the motive power type at any instant is not simply the consequence of further evolution? The solution proposed by of its genotype and current environment, but also Van Valen (1973) is that the environment is of its phenotype at the previous instant. That is, constantly moving and that species are sim- development is a first order Markov process in ply running to keep up. In that case, it is the which the next step depends upon the present autonomous forces of environmental change that state. The temporal order of environments is thus govern the rate of evolution, and we would be a critical, but not a su≤cient, prediction of future well advised to study the laws of environmen- development. If a Drosophila adult hatches from tal rather than organismic change if we want to the pupal case as an unusually small fly, it does understand what has been happening. More par- not matter what the contributing causes were to adoxical is the necessity of defining environments making it small, it now has an unusually high sur- without organisms. To make the metaphor of ad- face to volume ratio, which will play an impor- aptation work, environments or ecological niches tant role over the rest of its life history. Small must exist before the organisms that fill them. changes in ambient temperature or in its own There must be a preferred, denumerable set of activity will be felt by it in a way different from its combinations of factors that make “environ- larger sibs and will have different effects on its ments” and a non-denumerable infinity of combi- reproductive rate. The way in which its genotype nations of factors that are not. The history of life and its future environmental sequence influence is then the history of the coming into being of the fly are themselves effects of the organism as new forms that fit more and more closely into cause. The organism is not simply the object of these preexistent niches. But what laws of the developmental forces, but is the subject of these physical universe can be used to pick out the pos- forces as well. Organisms as entities are one of the sible environments waiting to be filled? In fact, we causes of their own development. only recognize an “environment” when we see the organism whose environment it is. Yet so long as we persist in thinking of evolution as adap- Organism and Environment in Evolution tation, we are trapped into an insistence on the autonomous existence of environments indepen- Our usual description of evolution by natural dent of living creatures. In fact, we should be selection is framed in terms of the process of able to list the environments on Mars, being care- adaptation. A species’ environment exists and ful all the while not to be influenced by our changes as a consequence of some autonomous knowledge of earthly life! If, on the other hand, forces outside the species itself. This outside we abandon the metaphor of adaptation, how world poses problems for the species, problems of can we explain what seems the patent “fit” of acquiring space, consumables, light, and individ- organisms and their external worlds? Fish have uals of the opposite sex. Those most successful in fins, and not only fish but whales, seals, penguins solving the problems, because, by chance, their and even sea snakes have some sort of flattened morphologies, physiologies, and behaviors make body part that the animal uses for swimming. them mechanically the best fit to do so, leave the Moles have long claws as do anteaters; birds and most offspring and thus the species adapts. This bats have wings, and so on. The marvellous fit of view of evolution, however, has certain paradox- organisms to their environments seems obvious. ical features (Lewontin 1978). One is that all What is left, then, but some concept of a progres- extant species are said to be already adapted to sive fitting of organisms to predetermined adap- their environments. A good deal of evolutionary tive peaks? biology is taken up with demonstrating that their What is left out of this adaptive description of features represent optimal solutions to environ- organism and environment is the fact, clear to all 64 Richard C. Lewontin

natural historians, that the environments of or- Ecological succession is precisely the history of ganisms are made by the organisms themselves as self-destruction of species by alterations of their a consequence of their own life activities. How do own environment. I know that stones are part of the environment of (3) Organisms transduce the physical signals thrushes? Because thrushes break snails on them. of the external world. Changes in external tem- Those same stones are not part of the environ- perature are not perceived by my liver as thermal ment of juncos who will pass by them in their changes but as alterations in the concentrations search for dry grass with which to make their of certain hormones and ions. The photon en- nests. Organisms do not adapt to their environ- ergy impinging on my retina and the vibrational ments; they construct them out of the bits and energy at my ear drums when I see and hear a pieces of the external world. This construction rattlesnake are immediately changed through process has a number of features: the mediation of my central nervous system into (1) Organisms determine what is relevant. changes in adrenalin concentration. But this While stones are part of a thrush’s environment, change is in part a consequence of my biology, tree bark is part of a woodpecker’s, and the since another rattlesnake would presumably react undersides of leaves part of a warbler’s. It is the rather differently. life activities of these birds that determine which (4) Organisms create a statistical pattern of parts of the world, physically accessible to all of environment different from the pattern in the them, are actually parts of their environments. external world. Organisms, by their life activ- Moreover, as organisms evolve, their environ- ities, can damp oscillations, for example in food ments, perforce, change. All animals are covered supply by storage, or in temperature by chang- with a thin boundary layer of warm moist air ing their orientation or moving. They can, on as a consequence of their metabolism. Small ecto- the contrary, magnify differences by using small parasites may be completely immersed in that changes in abundance of food types as a cue for boundary layer which thus determines their envi- switching search images. They can also integrate ronmental temperature and humidity. But if nat- and differentiate. Plants may ﬂower only when ural selection should increase the body size of a su≤cient number of days above a certain tem- these parasites, they may emerge through the perature have been accumulated. Cladocera can layer into the stratosphere of a colder drier change from asexual to sexual reproduction in world. It is the genes of sea-lions that make the response to a large alteration in oxygen supply, sea part of their environment and the genes of temperature or food supply, irrespective of the lions that make the savannah part of theirs, yet absolute level. Even the period of external oscil- those genes are descended from a common carni- lation can be modulated, as when cicades count a vore ancestor. prime number of seasonal ﬂuctuations. It might (2) Organisms alter the external world as it be objected that the notion of organisms con- becomes part of their environments. All organ- structing their environments leads to absurd isms consume resources by taking up minerals, results. After all, hares do not sit around con- by eating. But they may also create the resources structing lynxes! But in the most important sense for their own consumption, as when ants make they do. First, the biological properties of lynxes fungus farms, or trees spread out leaves to catch are presumably in part a consequence of selection sunlight. They may create an environment more for catching prey of a certain size and speed, i.e. hospitable to their own species, as for example hares. Second, lynxes are not part of the environ- when beavers raise the water level of a pond, but, ment of moose while they are of hares, because of in contrast, white pine in New England creates biological differences between moose and hares. a dense shade that prevents its own reseeding. Then what about laws of physical nature? Or- Gene, Organism and Environment 65

ganisms did not pass the law of gravity. Yet, If environments do not preexist, how are we to whether gravitation is an aspect of the environ- explain the remarkable similarities that evolved ment of an organism depends upon biological independently in different groups? Not only do properties of the organism, for example size. animals as unrelated as fish, mammals, and birds Bacteria are “outside” gravity, but their very all have fin-like appendages when they are aquat- small size makes them subject to a very different ic, but a whole marsupial fauna developed with physical phenomenon, Brownian motion of mol- “wolves,” “moles,” “mice,” and “cats” (although ecules, which larger organisms do not notice. no ungulates, marine forms, elephants, bats, or The metaphor of construction rather than ad- horses). The error is to suppose that because aptation leads to a different formulation of natu- organisms construct their environments they can ral selection and evolution. On the adaptive view construct them arbitrarily in the manner of a evolution can be represented formally as a pair of science fiction writer constructing an imaginary differential equations. The first, describing the world. The coupled equations of coevolution change in organisms, O, as a function of organ- of organism and environment are not uncon- ism and environment, E, dO!dt;f(O, E), and strained. Some pathways through the organism- a second law of the autonomous change of environment space are more probable than environment, dE!dt;g(E). A constructionist others, precisely because there are real physical view makes this into a pair of coupled differen- relations in the external world that constrain tial equations in which organism and environ- change. The construction of an environment that ment coevolve, each as a function of the other, includes living in a fluid medium of a certain den- dO!dt;f(O, E), and dE/dt;g(O, E). sity and viscosity places certain quite loose con- The parallel in population genetics is between straints on morphology. Either the organism will frequency-independent and frequency-dependent be sessile, as many are, and wait for food, or it selection. In the first case, one postulates a set of will be vagile and chase it. If it is vagile it may adaptive peaks which may or may not change in propel itself by jet action like the squid or by time as a consequence of external forces, and a umbrella sculling like the jelly-fish or by flattened set of gene frequencies that change in response to appendages. But even in the last case, there are the potential field represented by the adaptive those who undulate up and down (whales), side peaks. In the second case, the location and exis- to side (sharks), fly in the water (rays), beat tiny tence of the peaks are themselves functions of the wings rapidly (sea horses), and a variety of other genetic composition of the evolving population. fin movements. Where there is strong conver- The model of the first process is climbing a moun- gence as in certain marsupial-placental pairs, this tain peak, of the second walking on a trampoline. should be taken as evidence about the nature of While population geneticists usually model selec- constraints on development and physical relation as a frequency-independent process, adding tions, rather than as evidence for pre-existing frequency dependence as an added complication niches. of marginal interest, the actual situation is the The nagging problem of adaptation has always reverse. Most selective processes are frequency- been what one imagines the progression of events dependent, as for example are any processes in to be during the process itself, since we always which fitness depends on relative position in an observe only the finished product. If a seal’s ordered series, or in which there is competition flippers are an adaptation to water, at what stage for resources in short supply with different rela- in the evolutionary history of seals did swim- tive success of different types. ming in water become the “problem” which seals A problem seems to be posed for the construc- “solved” by losing their legs? No one imagines tionist view by the phenomenon of convergence. that a whole group of terrestrial carnivores sim- 66 Richard C. Lewontin

ply plunged into the water one day, experiencing to lean so heavily on an impoverished view of the a new major adaptive problem, and then pro- relation between gene, environment, and organ- ceeded to adapt to it by the usual route of natu- ism. If the hundredth anniversary of the death of ral selection for small increases in flipper-like Darwin is not to mark the death of Darwinism, morphology. Nor, alternatively, can we say that we need to struggle for its transfiguration. swimming has always been a major problem for carnivores. There is, in fact, no reason why we should not, a priori, reverse the entire scenario. References Perhaps an early pinniped ancestor acquired Clausen, J., D. D. Keck, and W. W. Hiesey. (1958). slightly flipper-like appendages for an entirely Experimental studies on the nature of species, Vol. 3: different reason—genetic drift or some pre- Environment responses of climatic races of Achillea. adaptation. Partly aquatic life then became an Carnegie Institution of Washington Publication 581: opportunity rather than a problem. This is not a 1–129. very satisfactory theory, at least as a typical one Dobzhansky, T., and B. Spassky. (1944). Manifestation for evolution, but for precisely the same reason of genetic variants in Drosophila pseudoobscura in dif- that the standard adaptive story is unaccept- ferent environments. Genetics 29: 270–290. able. Both uncouple organism and environment Falconer, D. S. (1960). Selection of mice for growth on in such a way that we must regard one as under- high and low planes of nutrition. Genetical Research, going significant autonomous change and the Cambridge 1: 91–113. other as responding. Gillespie, C. C. (1959). Lamarck and Darwin in the Concentrating as we always have on the prob- history of science. In B. Glass, O. Temkin, and W. L. lem of how the organisms change under natural Straus (Eds.), Forerunners of Darwin, pp. 265–291. Baltimore: Johns Hopkins University Press. selection, we have neglected to ask seriously how the environment to which the organism is sup- Gupta, A. P., and R. C. Lewontin. (1982). A study of reaction norms in natural populations of D. pseudo- posedly responding has come to be a problem in obscura. Evolution 36: 934–948. the first place. Presumably the non-aquatic carni- Lerner, T. M. (1954). Genetic Homeostasis. Edinburgh: vore ancestors of the Pinnipedia slowly incorpo- Oliver and Boyd. rated the water as a more and more energetically Lewontin, R. C. (1974). The analysis of variance and significant aspect of their environment while their the analysis of causes. American Journal of Human morphologies and physiologies changed to make Genetics 26: 400–411. that appropriation more energetically reward- Lewontin, R. C. (1978). Adaptation. Scientific Amer- ing. Were a complete reconstruction possible, we ican 239(9): 156–169. would be unable to find the moment at which Rendel, J. M. (1967). Canalization and Gene Control. swimming was for the first time a “problem” to London: Academic Press. be “solved” by the animal. Robertson, F. W. (1960). The ecological genetics of Organisms, then, both make and are made by growth in Drosophila. II. Selection for large body size their environment in the course of phylogenetic on different diets. Genetical Research, Cambridge 1: change, just as organisms are both the causes and 305–318. consequences of their own ontogenetic develop- Schmalhausen, I. I. (1949). Factors of Evolution: The ment. The alienation of internal and external Theory of Stabilizing Selection. Philadelphia: causes from each other and of both from the Blakeston. organism, seen simply as passive result, does not Van Valen, L. (1973). A new evolutionary law. stand up under even the most casual survey of Evolutionary Theory 1: 1–30. our knowledge of development and natural his- Waddington, C. H. (1957). The Strategy of the Genes. tory. It is a tribute to the power of long-held London: Allen and Unwin. ideology that the study of evolution continues II RETHINKING HEREDITY

7 Let’s Talk about Genes: The Process Molecular Gene Concept and Its Context

Eva M. Neumann-Held

The Problem with the “Gene” “arguable that the old term gene . . . is no longer useful, except as a handy and versatile expres- It is almost common knowledge among biologists sion, the meaning of which is determined by the and philosophers of biology (e.g: Burian 1986; context.” Sterelny and Gri≤ths (1999: 144) iden- Carlson 1991; Falk 1984; Keller 2000; Kitcher tify at least two different usages of the term gene 1982, 1992; Portin 1993) that the classical molec- in a molecular genetics textbook. They comment: ular gene concept is not su≤cient anymore in the “This suggests that the gene does not really name face of the complex interactive processes being a unit of molecular biology, but is shorthand for reported by molecular biology. The classical any of several different units. . . . Gene is used in molecular gene concept defined a gene as that molecular biology as a shifting tag rather than as segment of DNA which codes for a polypeptide. a name for a specific molecular kind” (p. 145). Thereby, this gene concept implied a structural They add: “Molecular biologists often seem to and functional unity which allows for a one-to- use genes to mean ‘sequences of the sort(s) that one relationship between DNA and correspond- are of interest in the process that I am working ing polypeptide. That seemed to justify the on’ ” (p. 133). wording that the “information” for the polypep- What do we do with this situation? Three posi- tide resides in this particular segment of DNA. tions can be taken. One is to relinquish the term Today, however, we learn from molecular biol- gene altogether. Alternatively, one might allow ogy on the one hand about numerous genomic for several different usages, thereby accepting the organizational features such as genomic imprint- status quo. But there is also a third possibility, ing and “overlapping genes,” and on the other which is to strive to develop precise definitions hand about complicated gene expression fea- for gene. tures, such as alternative mRNA splicing and A look at the literature reveals that all three mRNA editing. Both types of features have led to positions have their supporters. Kitcher (1992: the realization that the assumption of a kind of 130), for example, writes: “Indeed, it is hard to “static” one-to-one relationship between DNA see what would be lost by dropping talk of genes and polypeptide was an oversimplification. from molecular biology and simply discussing the In the face of this situation, it might seem a properties of various interesting regions of nucle- contradiction that the talk of biologists and the ic acid.” In contrast to this position, I think that general public contains as many references to more than a term would be lost by dropping gene genes as in the good old days when there was no from the biological talk. In addition, it would no reason to question the classical molecular gene longer be possible to name a particular relation- concept. However, a closer look on the usage of ship between DNA segments and the course of the word gene in communities of biologists shows events that leads to the production of polypep- that gene is used to mean very different things. tides. This point will be discussed in more detail Falk (1984: 203), for example, observed: “A ca- below. Furthermore, it will be argued that for sual glance at the current genetic literature would any particular context the important question is be enough to reveal that although the term ‘gene’ precisely, what makes particular regions of nucle- is very much in use, it means different things for ic acid interesting? Neither a subsumption of different people. . . . Today the gene is . . . a unit, a “interesting regions” under the name gene nor a segment that corresponds to a unit-function, focus on these regions in themselves (promoter, as defined by the individual experimentalist’s coding regions, etc.) can solve this question. I needs.” Portin states (1993: 208) that it is quite therefore prefer a clarification of the relationship 70 Eva M. Neumann-Held

between DNA and polypeptide production to a 1995, Rheinberger 2000) gene concepts. This policy of (supposedly) benign neglect. position is represented by those, who, for differ- The second approach to the multiple usages of ent reasons, want to try to find a precise meaning gene might be called the liberal standpoint. It just for the term gene, which nevertheless incorpo- admits that the concept of the gene is “fuzzy.” rates the new findings of molecular biology. Within this approach one can distinguish a num- These efforts are in the realm of biological sci- ber of lines of argumentation. The earlier quo- ences as well as in philosophy of biology. The tations from Falk, Portin, and Sterelny and journal Nature recently published a suggestion on Gri≤ths, for example, share with each other the the gene concept by Epp (1997), and almost any claim that the different meanings of the term gene textbook of molecular biology and genetics will do not pose a problem, because the actual object include some efforts of trying to define the gene. of reference could be deduced by the context. In the realm of philosophy of biology the require- According to Sterelny and Gri≤ths (1999: 133) ment for the precise introduction of scientific the “usage” of the term gene is “perfectly satis- terms motivated Fogle (1990) to suggest a gene factory” because it is grounded in a “rich back- concept, whereas Schaffner’s (1993) and Waters’s ground of shared assumptions.” Rheinberger (1994) efforts in this regard are designed to help (2000) defends the vagueness of the term gene them to defend the possibility that a Mendelian even more emphatically by stressing that the term gene concept could be reduced to a molecular gene proved historically to be of heuristic value, gene concept.1 so that vagueness itself would not be grounds for In their efforts to define “genes,” all of these its rejection. Similarly, Keller (2000) argues that approaches focus on the nucleic acid sequences it might be just its “disarray” which today may that are involved in the synthesis of a polypeptide provide “conceptual value” to the term gene. (note here the similarity to Kitcher’s approach). I agree that in certain research contexts the dif- Within these limits, then, the approaches suggest ferent usages of the term gene seem not to pose a different conceptualizations by including differ- problem. However, at least when these contexts ent, functionally relevant parts of DNA in their are left and assumptions about the nature of definition or by allowing that different stages of genes cross disciplinary boundaries or reach the mRNA processing can be called a gene, as long public sphere, different usages of the term gene as they are involved in the course of events that can become a problem. One example can be lead to the synthesis of a polypeptide. One might found in the writings of Dawkins (1982, 1989), call these approaches, therefore, “nucleic-acid when so-called evolutionary genes are introduced centered gene concepts.” However, although this as arbitrary DNA sequences, and then, in a next means a certain lack of precision in terminology, step, become confused with polypeptide produc- I would prefer to call these suggestions “DNA- ing molecular genes (for a critical analysis, see centered” gene concepts. If it should occur that a Gri≤ths and Neumann-Held 1999; Neumann- differentiation between “DNA” and “RNA cen- Held 1998). Although a certain amount of teredness” is required, for the goal of this article, vagueness might have heuristic value, when I would point that out. straightforward ambiguities arise in a single dis- Although the suggestions in the framework of cussion there is a clear need for more precision. this third approach are quite sophisticated, one However, greater precision is not only needed in should keep in mind that all definitions of genes, popular and interdisciplinary contexts, but also which only focus on nucleic acid sequences actu- in the context of molecular research. ally focus on the question: “Which DNA se- To explain this point, let us take a brief look at quences causally participate in the polypeptide the third position in regard to “fuzzy” (Fischer synthesis?” Conceptualizing genes in such a way Let’s Talk about Genes 71

is particularly interesting when the research focus best given by looking in more detail at the short- is on the question, whether or not variations in comings of recent efforts for defining the gene. polypeptide sequences are caused by mutations in The common feature of all these definitions is particular DNA segments. It should be stressed that a gene is defined as one or more pieces of here that such a question certainly reflects a legit- DNA which code for a polypeptide (for exam- imate research interest, and I make this point not ples, see Levin 1985: 685; Knippers 1997: 287, only to express some kind of tolerance. Rather, I 283ff). Efforts to recognise the complex molecu- intend to point out that any effort to define gene lar mechanisms involved are found in Singer and (or other scientific terms) should start with a Berg (1991: 440). After explaining their uneasi- clarification of the purposes, the contexts of re- ness with any gene definition, they nevertheless search interests, for which a term or an empiri- adopt the following definition of the gene “for the cal approach is designed. Thus, instead of asking purposes of this book”: for the causal influence of DNA sequences on polypeptide production, for example, one might A combination of DNA segments that together constitute an expressable unit, expression leading to the for- also ask whether there are not other, context mation of one or more specific functional gene products specific causal influences in addition to DNA. that may be either RNA molecules or polypeptides. The This leads to the further question of how certain segments of a gene include (1) the transcribed region DNA segments become involved in polypeptide (the transcription unit), which encompasses the coding production at all. The observation that polypep- sequences, intervening sequences, any 5’ or 3’ trailer tide synthesis is time and tissue regulated makes sequences that surround the ends of the coding particularly obvious the importance of under- sequences, and any regulatory segments included in the standing the regulation mechanisms of transcrip- transcription unit, and (2) the regulatory sequences that tion and translation. Those who defend the need flank the transcription unit and are required for specific for precise gene definitions try to include the pos- expression. sibility of regulated expression by including in the This definition, like most others, concentrates definition those segments of DNA (and mRNA) only on DNA sequences, and only on those DNA which can be shown to be necessary for these reg- sequences that are involved in the synthesis of a ulation mechanisms. However, the (regulated) polypeptide chain. Although such a definition, as production of a polypeptide needs more than stated above, can be read as reflecting specific only DNA sequences. One might wonder, there- research interests in the role of DNA sequences, fore, whether a definition of gene should not also another reading cannot be excluded. For exam- include more than only DNA sequences, if the ple, there is no mentioning of all other entities, major interest is to understand the (regulated) which are not located on the DNA but still might production of polypeptides. It is exactly at this be necessary to allow expression of DNA seg- point and to address these research interests, that ments. Therefore, it is easy to read this definition I want to suggest a new gene concept. The rein such a way that the causal arrow leading to search contexts in which this gene concept will the synthesis of a polypeptide has only one impor- be of value are mostly, although not exclusively, tant starting point: “a combination of DNA in developmental genetics. sequences.” This attribution of causal primacy would be appropriate in a system where all other Genes for Development environmental, non-DNA located parameters are kept constant, as could be obtained in, for Grounds for suggesting a new gene concept for example, in vitro experiments. However, it seems the context of developmental genetics might be that this is not the (only) context that Singer and Berg have in mind, and it is clearly not the only 72 Eva M. Neumann-Held

context that I am interested in. When Singer and coding sequence for a polypeptide. Among the Berg specify these DNA sequences that “are other mechanisms that are important for using required for specific expression,” they obviously mRNAs as templates for polypeptide synthe- want to include in their definition the observation sis are so-called capping and polyadenylation of that in experimental systems, derived from living the primary mRNA. Although examples for the beings, not all DNA sequences are expressed all “under-determination” of polypeptides by DNA the time. Rather, it is the time and tissue regu- sequences are discussed at length in Neumann- lated expression that allows for ordered devel- Held (1999a, which see for detailed references), opmental processes at all. Furthermore, it is a two examples from this former paper should be well-known problem in in vitro translation sys- briefly mentioned here because, in my opinion, tems that numerous polypeptide fragments are they are particularly fascinating examples of synthesized from a given mRNA, and sophisti- molecular regulation mechanisms. The first ex- cated techniques are required to identify the ample is alternative mRNA splicing. During the polypeptide of interest. Even if these expression processes of alternative mRNA splicing differ- systems work very well in vitro, it is still often a ent parts of the primary mRNA of a particular matter of luck for gene-technologists whether such DNA are spliced together. Therefore, different gene constructions would work as anticipated templates are created as “mature mRNA” for the when brought back into an in vivo system. I will process of polypeptide synthesis. The obvious re- come back to the “in vivo challenges” later. sult is that different polypeptides are synthesized What can be learned from this so far is that upon use of the same DNA sequence. Alternative the regulated synthesis of a polypeptide clearly splicing is quite common in eukaryotes. needs more than a coding region alone. It needs RNA editing, the second example, edits quite regulatory regions, as Singer and Berg point out literally the primary mRNA by adding, deleting rightly, but it also needs something that does the or exchanging particular nucleotides so that dif- regulation, using those DNA sequences as regu- ferent mRNAs are created than the correspond- latory regions. We can refer to all these influences ing DNA would have allowed one to predict. and entities that perform regulatory functions Once again, the same piece of DNA is used here by means of DNA (or mRNA) sequences as in different processes which lead to different poly- the non-DNA located, or “environmental” enti- peptide products. Furthermore, both alternative ties, factors, parameters, or influences in gene mRNA splicing and mRNA editing depend on expression. the developmental state of a particular organism, The significance of these non-DNA located and both mechanisms are not at all rare. entities for the synthesis of polypeptide chains Alternative splicing is quite often found in becomes most obvious in those cases in which eukaryotes, and mRNA splicing happens very the same DNA sequence is part of very differ- frequently in organelles. Furthermore, the extent ent expression processes, leading to different of editing is developmental and tissue dependent. polypeptides. One example of this is the human gene for For example, activation and termination apolipoprotein B (Strachan and Read 1996: 198). points, respectively, on the DNA (for transcrip- In the liver apoB codes for a mRNA of 14.1 kb tion) and on the mRNA (for translation) are and a protein of 4536 aminoacids, whereas in the not just there, but need to be recognized as intestines, the mRNA has a length of only 7kb, such. That is shown by the observable fact that coding for 2152 amino acids. The shorter length those sequences which at some point are recog- of the resulting amino acid chain is caused by a nized as, for example, a promoter (transcription stop codon that is introduced into the mRNA by starting sequence), serve at another point as a mRNA editing, which changes a cytosine into an Let’s Talk about Genes 73

uracile at position 6666. Although some evidence She writes: “Considered as a functional unit, the indicates that editing of cytoplasmic mRNA gene is no longer a static entity, set above and depends on the environment of the nucleic acid apart from the processes that specify cellular and sequence (Knippers 1997: 400, 482), this does not inter-cellular organization, but itself a part and explain the tissue specifity of mRNA editing. parcel of these processes, defined and brought In plant mitochondria mRNA editing is appar- into existence by the action of a complex self- ently not even dependent on the sequence com- regulating dynamical system in which and for position. So far, no conclusive models of the which the inherited DNA provides the crucial mechanisms have been developed here, and the raw material.” And further: “genes function in physiological significance of the sometimes ex- these debates to remind modern geneticists of tremely complicated mRNA editing processes, what it is that makes a region of nucleic acid for example in protozoan mitochondria, remain a ‘interesting’ or of what constitutes ‘meaningful point of speculation (Knippers 1997: 482ff). structure’ in the genome” (Keller 2000). To summarize: In these cases of alternative Therefore, Keller suggests: “One way of posing mRNA splicing and mRNA editing, no one-to- the particular di≤culty that is here made explicit one relationship between DNA and polypeptide might be to paraphrase Howard Pattee’s ques- can be established, and the same piece of DNA tion . . . and ask: How does a sequence become a can be used for different functional and structur- gene?” (Keller 2000). al units. The sequence of nucleotides on the DNA It is by posing exactly this question that Keller does not even allow prediction of the sequence of suggests a difference which I want to empha- amino acids in the polypeptide products. This size. If one wants to conceptualize the complex empirical evidence shows that it is not only the molecular mechanisms of polypeptide synthesis presence of a DNA sequence that determines the processes, then, so I argue, it is necessary to discourse of events that lead to the synthesis of a tinguish between DNA sequences on the one polypeptide but, in addition, specific non-DNA hand, and the embedding of DNA sequences in located factors must act on DNA and derived the course of events of polypeptide synthesis on mRNA to determine the particular processing the other hand. mechanisms. The fate of the DNA thus depends DNA sequences are undoubtedly necessary in on the developmental state and tissue. One could the process of making a polypeptide. But in ad- say in some metaphoric sense that in the in- dition, any process of polypeptide synthesis re- teractions with specific environmental factors quires factors other than this DNA. Non-DNA “meaning” is attributed to particular nucleic located entities are required to determine how acid segments, which, depending on the context, DNA segments are used to make one or more are then used, for example, on one occasion as transcriptional units. From there, further in- a promoter and on another as one of several teractive processes determine the course of the putative coding regions.2 Therefore, what counts translation processes. This becomes particularly as a significant structure to perform a particular obvious in cases of tissue and/or time regulated function is determined by the context in which polypeptide synthesis. that segment is being used. Strictly speaking, the Therefore, in summary, the synthesis of a lin- consensus sequence (TATAAT) of a promoter is ear polypeptide chain is caused by numerous fac- only such a consensus sequence when it is used as tors which interact with each other in a particular a promoter, and not, for example, as part of a order. Among these factors are particular DNA coding region. sequences. Which DNA segments participate in Recently, Evelyn Fox Keller reached very sim- the synthesis of a polypeptide at a particular ilar results in her analysis of the molecular gene. point cannot be determined by looking at the 74 Eva M. Neumann-Held

nucleotide sequence alone, and a particular DNA to admit that I hesitate to include these RNAs in sequence can become part of different courses of the suggested gene concept. And why should we polypeptide syntheses. want to summarize both kinds of processes under A very important point is, however, that once the same heading anyway? One argument would polypeptide synthesis has occurred, it is possi- be that one should name everything a gene, which ble to clearly identify particular DNA segments includes a DNA segment that is involved in a which featured in the course of events that re- transcription process. Maybe it was an argument sulted in the synthesis of this particular poly- of such a kind which allowed classical molecular peptide. Therefore, in retrospect, a particular gene definitions to include “RNA producing” DNA-polypeptide relationship can be specified. DNA sequences as “genes.” I would hold against This sounds very similar to Keller’s statement this argument, and therefore against this gene that the definition of a gene as that which is definition, the fact that it emphasizes another “expressible” allows “for an after-the-fact desig- point than I have in mind. An important feature nation of the particular combination of DNA of my suggested gene definition is that it binds segments actually employed” (Keller 2000). together processes of transcription and transla- However, in this quotation Keller seems to tion. Under this perspective the main emphasis limit the use of the word gene to DNA segments is not on the question, in which kind of produc- again, whereas I want to emphasize that an tion processes particular DNA sequences play an “after-the-fact designation” is, in principle, possi- important role, but the emphasis is on the ques- ble for any entity that causally participates in the tion, how are DNA sequences used in the process. From this then, we arrive at a distinction processes of polypeptide production. The whole between DNA and gene, and find a new concep- purpose of trying to suggest PMG as a gene con- tualization of the gene for developmental pro- cept for developmental processes is to gain some cesses. This concept can be stated as follows: clarification of the concept gene for at least one clearly defined application area. If we include “Gene” is the process (i.e., the course of events) that processes under the heading gene, which are lim- binds together DNA and all other relevant non-DNA ited to only transcription, it is my feeling that we entities in the production of a particular polypeptide. The term gene in this sense stands for processes which lose some of the just-gained clarity again, and are specified by (1) specific interactions between specific we might wonder later on, whether other DNA DNA segments and specific non-DNA located en- regions, which are involved in regulatory func- tities, (2) specific processing mechanisms of resulting tions, should be called genes as well. Therefore, I mRNA’s in interactions with additional non-DNA tend to exclude what is commonly called “RNA located entities. These processes, in their specific tem- genes” from the suggested developmental gene poral order, result (3) in the synthesis of a specific concept, although I have not arrived at a final polypeptide. This gene concept is relational, and it stand-point in this matter myself. always includes interactions between DNA and its Coming back to point (3) in the suggested gene (developmental) environment. definition, it is this last feature in particular, the At this point one might wonder whether I emphasis on the interaction between DNA and should not include so-called “RNA genes” in this its environment, which allows this gene concept definition as well, thereby changing point (3). The to be viewed as an extension of the developmen- term “RNA genes” is usually used to designate tal systems approach (DSA), as it was developed those parts of DNA sequences which do not lead by Susan Oyama (1985). This claim might be best to an mRNA, which then is translated into a supported by briefly explaining how DSA actu- polypeptide, but which lead to other kinds of ally motivated me to search for a developmental RNAs, for example rRNAs and tRNAs. I have gene concept. Furthermore, such an excursion Let’s Talk about Genes 75

will allow for further clarification of the context polypeptide chains. The need for further clar- of scientific inquiry for which this gene concept ification in this regard was seen by Oyama was developed. (1988: 261) herself, when she wrote, for example: A central point in Oyama’s DSA is the claim “But even leaving aside the complexities of over- that, if one wants to understand the development lapping and discontinuous genes and the prob- of any phenotypic trait, a distinction between lems they pose for the concept of coding, the “genetic” and “nongenetic” causes is not suitable. information-as-sequence must be reconstituted as Although I found the arguments provided by well.” Therefore, I made it my goal to work out a DSA very inspiring, I nevertheless wondered why gene concept that extends DSA for application to DSA seemed to stop short when approaching the developmental processes on those molecular levels molecular level itself. Oyama, for example, con- of interactions, which have to do with DNA and ceded that “it makes sense in general to say that end with the synthesis of linear polypeptide the primary structure of a polypeptide is encoded chains. It has been pointed out before, but should on the chromosomes” (Oyama 1985: 70). But be stressed again, that all the italics in the last what does it mean to make “sense in general”? It sentence indicate points of emphasis and at the could be interpreted as assuming a (more or less) same time points of exclusion, which define the simple correspondence between a DNA segment purposes of the search for a gene concept for and a linear polypeptide chain. Therefore, DSA development. The suggested gene concept is not obviously could be understood as being willing to for an evolutionary context and not about other concede that at least on the level of polypeptide kinds of molecular interactions than those men- synthesis DNA would have “more causal pow- tioned. The limits of this gene concept will be dis- ers” than non-DNA located influences. One may cussed further in the next paragraph. But first grant at this point that such an interpretation I need to say a word about naming this gene cannot be deduced analytically from Oyama’s concept. statement, because one has to differentiate When I first presented my gene concept I spoke between the kind and the size of causal role. But of conceptualizing the gene in “constructionist because the concept of causality remains un- ways.”3 The term constructionism was used by specified here, Oyama’s statement can lead to Gray (1992: 175 and 203, note 23), who applied it misunderstandings, particularly in combination to the new viewpoint on development and evolu- with statements of other DSA representatives. tion as it was exemplified in the writings of Gray granted, for example, that “the nucleotide Lewontin and Oyama. Today, I do not favor the sequence does specify the primary structure of term constructionist because, for several reasons, a protein” (Gray 1992: 190). Again, one might it is likely to lead to misunderstandings. One argue that in a certain sense it does, and again, reason is its phonetic closeness to the term misunderstandings are possible according to constructivism, which is being used in the nam- which this statement seems to represent a rather ing of such different approaches as “radical- standard view point on a more prominent causal constructivism” (for example, Maturana 1985; role of DNA. What I saw here was, on the one Maturana and Varela 1990; Schmidt 1987), hand, a gap in DSA that pointed to a need for “social-constructivism” (for example, Berger clarification, a blind spot in DSA when applying and Luckmann 1980; Luhmann 1988) and “me- the approach to the molecular level of develop- thodical constructivism” (Janich et al. 1972–73; mental processes. On the other hand, I thought Lorenzen 1987; Mittelstrass 1973). Another that DSA would provide a suitable framework downside of the use of the term constructionism for the reconceptualization of those molecu- is that it suggests a programmatic unity, where lar contexts that allow the synthesis of linear there really are a diversity of very different ap- 76 Eva M. Neumann-Held

proaches, which have (only?) an “antireduction- I have not set these limits to PMG because I ist consensus” in common (Sterelny and Gri≤ths regard the following processes as unimportant 1999: 137ff). Developmental systems theory in the production of phenotypic traits (or trait (DST) or approach (DSA) (for a discussion of the differences). On the contrary, I find them so im- distinction between theory and approach in this portant that they should be considered indepen- context, see chapter 20 of this volume) is one for- denly, particularly independently from the level mulation in the antireductionist community, and of DNA sequences. To make this point clear, I I understand the gene concept I suggest as an need to present my argument in more detail. addition or extension to DSA for conceptualizing I have shown that the process of polypeptide the molecular course of events in the synthesis of production needs numerous other factors besides polypeptides. However, because I am no longer DNA. A linear polypeptide chain, however, is using the name constructionist, I decided that a then used for numerous further processes, in more appropriate name for this gene concept which this polypeptide is folded and interacts would be the process molecular gene concept with cofactors, other polypeptides and other cell (PMG). components. Here, this polypeptide, and possibly its functional loss, becomes a factor in the development of a phenotypic trait or a so-called genet- The Limits of PMG ic disease. These levels of process, producing a linear polypeptide chain on one hand and the fur- PMG has relevance only for particular contexts ther interactions of this chain with other compo- of research interests. Some distinctions have been nents on the other, are often not differentiated. discussed in the preceding paragraph. For exam- Instead, it is assumed (!) that a mutation on a ple, PMG was not designed as an evolutionary DNA sequence that influences the functional gene concept, and it remains arguable whether effect of the polypeptide is su≤cient to explain it is possible to find a unified gene concept for variations in phenotypic traits. Thus, the human both developmental and evolutionary research geneticist and anthropologist Ulrich Wolf states: contexts.4 PMG is designed for contexts that focus on the . . . In particular cases, ontogenetic modification relationship between DNA and a polypeptide (or appears to be of minor significance, so that the pheno- an RNA), because this relationship indicates a type of a mutation can be predicted with consider- process of production that is central for develop- ably accuracy. This is no surprise if, depending on the mental processes on the molecular level. This nature of the mutation and the physiological function of the gene affected, the genotype-phenotype relation- might be the reason that Singer and Berg (1991) ship is direct. . . . It is assumed that the total of the non- want to preserve this relationship in their sugges- genetic influences (epigenetic, environmental) are tion of a molecular gene concept. Furthermore usually so similar or are compensated by the organism (this is rather a minor point), it was my pur- to such an extent that the respective mutations act as pose to extend DSA by the conceptualization of the major variable during ontogenetic development. the relationship between DNA and polypeptide (Wolf 1995: 127) synthesis, and therefore it was my purpose to focus on these rather than on other molecular However, it is Wolf’s point that very often the processes. genotype-phenotype relationship is more com- But if these processes between DNA and syn- plex and does not necessarily allow for the sim- thesis of a linear polypeptide are viewed as plifying assumption to which he refers. Wolf important for understanding development on the stresses that “the phenotype is . . . the product of molecular level, why then, one might ask, stop at ontogenetic development rather than the mere the level of linear polypeptide synthesis? consequence of the genetic constitution of the Let’s Talk about Genes 77

Figure 7.1 Schematic representation of some transcription, editing, and translation processes, highlighting the context- dependency of the expression of gene products. (This ﬁgure ﬁrst appeared in Gri≤ths and Neumann-Held 1999; copyright © 1999 American Institute of Biological Sciences.) 78 Eva M. Neumann-Held

zygote” (p. 127), and in his 1997 paper, Wolf dis- again: “The hope that the mutant genotype at the cusses numerous examples where identical mu- major locus would predict the variant phenotype tations do not exclude phenotypic variation. originated in the new ability to study mutant Similar viewpoints are also held by some other allelic expression in reductive in vitro expression human geneticists. Scriver and Waters (1999: systems and in transgenic animals.” 267), for example, summarize: “It was a hope In other words, these experiments were per- that delineation of genotypes [as in the human formed under highly controlled conditions, in genome project] with new methods for the detec- which all parameters but the particular gene (i.e. tion of mutations would enable the prediction of DNA sequences) could be held more or less con- variant phenotypes; in the case of human genetic stant. Scriver and Waters (p. 271ff) continue: disease, this would have added value for progno- “But genomes function in vivo, where much more sis and treatment.” than the major gene (here the PAH gene) is ex- But these hopes were an oversimplification, as pressed and where the whole organismal pheno- Scriver and Waters show even for the case of type is more than the sum of its parts.” the “classic ‘monogenic’ autosomal recessive These results of the “in vivo challenges” sup- disease” Phenylketonurie (PKU). Scriver and port my conviction that it is necessary to limit Waters (1999: 267)5 remark that a mutation “at PMG to those processes that lead from DNA to the human PAH locus was deemed su≤cient to linear polypeptide chains, and to distinguish explain the impaired function of the enzyme these clearly from processes in which the resulting phenylalanine hydroxylase (enzymic phenotype), polypeptides participate in the construction of a the attendant hyperphenylalaninemia (metabolic phenotypic trait. phenotype) and the resultant mental retardation (cognitive phenotype).” However, they continue, “expectations for a consistently close correlation The Openness of PMG between the mutant genotype and variant phenotype has been somewhat disappointed.” After having discussed the limits of PMG, the What has caused these disappointments? Scri- next thing to do is to focus on the openness of this ver and Waters (p. 268) write that “two impor- gene concept. The incorporation of DNA as well tant challenges to the hypothesis that a PAH as of non-DNA located factors into PMG might genotype would consistently predict a ‘mono- raise questions whether PMG would allow scien- genic’ phenotype had emerged: (1) patients, tific research. These kind of doubts present them- even sibs, sharing identical mutant PAH geno- selves often under the headline holism. Holism is types could have greatly different cognitive and not an unproblematic term, and at least from the metabolic phenotypes; (2) there are many in- standpoint of philosophy of science, one would stances of discordance between the mutant PAH have to ask for further clarification. However, genotype, its predicted effect on enzyme function, that might not be necessary here. I will therefore and the associated metabolic phenotype.” substitute the term openness for holism and Althought the linear polypeptide chain that explain openness by paraphrasing the criticism as gives rise to the enzyme phenylalanine hydroxy- follows: In emphasizing all developmental causes lase is effected due to a mutation of the corre- for the construction of a polypeptide, PMG sponding DNA sequence, even the activity of the seems to require consideration of everything (in enzyme itself, not to mention the further effects, principle) or at least of a nonaccountable number vary due to the environment in which this enzyme of factors. However, since one cannot do empiri- is located. To quote from Scriver and Waters cal research on “everything,” PMG cannot offer new insights or research strategies to biologists. Let’s Talk about Genes 79

Therefore, DNA-centered gene concepts are make such an empirical investigation easier than superior, because here we find a concentration on PMG? one causal factor, which then can be tested in One possible answer to this question is an regard to its influence on developmental pro- assumption, according to which a result that cesses. DNA-centered gene concepts, but not “clarifies the role of DNA in the making of a PMG, offer biologists something they can trans- polypeptide,” could also be expressed by substi- form into empirical questions and research in tuting “DNA” by “gene.” But such a substitution their laboratories. could only be defended when arguing for a fuzzy In my view, these doubts are based on a mis- gene concept, in which everything that refers to understanding, not only in regard to PMG, but DNA segments may be called gene. The empi- also in regard to DNA-centered gene concepts. In rical methods of molecular biology, however, particular, there seems to be a misunderstanding allow for finer distinctions, which then require with regard to the consequences that PMG, but finer distinctions in definitions. Depending on the not its alternatives (!), would have for empirical experimental setup, it is possible to distinguish research. Therefore, I want to begin my defense between DNA sequences that (in a particular of PMG with a brief look on some methodologi- context) function as promoters, for example, and cal requirements of empirical science. those that function as enhancers. Both DNA seg- It seems trivial that an assumption, according ments will (mostly) influence the production of to which some factor causally influences the pro- a polypeptide independently of one another. duction of an object of scientific inquiry, has to Because they can be distinguished from each be tested (1) by varying the parameters under other empirically, it makes sense to distinguish observation, holding all other parameters con- between them in terminology as well. However, stant, and (2) by observing putative changes in both entities might very well causally influence the pattern of production of the object of interest. the synthesis of the same polypeptide, which is In addition, it might be required that some mech- the reason that in DNA-centered gene concepts anism be given by which the factor under inves- they are summarized as being parts of one gene. tigation can influence the observed outcome, in So whereas empirical research will result in the order to justify attributing causality rather than identification of different DNA segments, these mere correlation. might very well belong to the same functional Applying this to research on the effects of unit in regard to the production of a particular DNA sequences on polypeptide production, it polypeptide. Here, the need to distinguish be- is obvious—and reflects empirical methods and tween DNA and gene becomes obvious even in practice—that an experimental setup has to be the DNA-centered gene concept. Another point, developed, which allows for an answer to the however, is more important in the context of this question: What is the causal relationship between section. In analogy to the doubts toward PMG variations in DNA sequences on one hand and one might also question DNA-centered gene con- variations in polypeptide synthesis on the other cepts, and ask whether in the end it might not be hand? In such an experiment, then, at least in the the entire genome which has a causal influence on ideal case, all other parameters except the DNA, the synthesis of every single polypeptide. It seems are kept constant. The result of such an approach to me to be a very appropriate answer to say: will clarify the role of DNA in the making of a Yes, but who cares? The point is that of course polypeptide. the choice of which DNA segment is tested for its Why should PMG not allow for such an relevance for polypeptide synthesis, will be deter- approach? Or to pose the question a different mined by pragmatic considerations, research in- way: Why should DNA-centered gene concepts terests, methods, and skills. To summarize those 80 Eva M. Neumann-Held

DNA segments under a common heading of DNA sequences are somehow “more causal,” “gene” does not seem to pose a problem in the “more real,” or “more significant” for the synthe- context of DNA-centered gene concepts. sis of a polypeptide than other factors. Kelly PMG differs in this respect only in that it Smith (2000) has called this interpretation of allows for inclusion of not only DNA, but also genetic results “gene-centrism”: “the view that non-DNA located entities, thereby integrating the genes are the most important explanatory fac- into the gene concept those relevant entities that tors in biology because of their unique causal are necessary for the functional specification of powers—whether in controlling individual onto- the DNA sequences involved. Here again, of geny (development), accounting for abnormal course, it will be pragmatic considerations, re- functioning in adults (disease), or explaining search interests, methods, and skills that decide changes in populations over time (evolution).” which factor is tested for its relevance for I have already provided evidence as to why, on polypeptide synthesis. Therefore, PMG’s open- the one hand, such an interpretation of molecular ness does not pose any more problems for the research is not defensible, and why, on the other application of empirical methods, and the inter- hand, the same evidence allows a conceptualiza- pretation of the results, than any DNA-centered tion of the gene in terms of PMG. In this section gene concept. I briefly want to highlight some further benefits that PMG might have when gaining more attention among scientists and the general public. PMG in Context It should be remembered that, after all, scientists apply, work, and discuss their (ontological) The analysis of the preceding section reveals an assumptions and interpretations not only in pri- important insight. I have argued that the goals of vate, but publicly, in front of society as a whole. the researchers (and their funding sources), the That alone is reason enough to ask how these technical, and the methodical tools play a deci- assumptions can be justified, and particularly if sive role, when considering whether DNA or any reasons are presented to doubt them. Here, scien- other parameters should be investigated in the tists are responsible for their research. Certainly search for an explanation of the synthesis of a since physicists became aware of the con- polypeptide chain. These goals are not them- sequences of their “value-free” research, when selves part of the scientific method, but goals and their “formulas” exploded over Hiroshima and tools set the stage for the actual scientific inquiry, Nagasaki, (some) scientists started to accept that and for the design of the experiment. Goals and they cannot deny all responsibility for the conse- tools allow for a preselection of possible scientific quences of their research.7 But also philosophy of inquiries from possible alternative questions. science, which is constituted by reflection on the Therefore, those preselections have to be justified different levels of the scientific discourse, has 6 on grounds other than scientific ones. Here, we the potential for criticism of science and should see now an important difference between PMG use it (Janich et al. 1972–73; Mittelstrass 1973; and DNA-centered gene concepts, in that DNA- Weingarten 1998). I would claim that such criti- centered gene concepts already reflect a preselec- cism need not be detrimental to the scientific tion by focusing on the effect of DNA in the enterprise itself. synthesis of polypeptides. Although I would not With this in mind, we might now ask in which defend it, I might still be willing to concede that way alternative gene concepts and their underly- this might be unproblematic as long as it is not ing assumptions effect areas outside of the actual assumed that the concentration on DNA factors scientific process. alone can be justified by ontological claims. By As has been discussed at length, the develop- “ontological claims” I mean here the idea that ment of phenotypic traits is very complex, and a Let’s Talk about Genes 81

prediction of its outcome can only be a more or Neumann-Held 1997, 1999b) and even human less appropriate estimation, if it is (only) based on physical performance (Montgomery et al. 1998), an analysis of DNA sequences. It is empirical evi- by reference to specific DNA sequences seems dence that determines in the end the quality of more convincing than, for example, a research predictions which can be made in any single case. project on the causal influences of the womb This should have consequences for genetic coun- environment on health later in life (Barker 1998). seling, as has been demanded by philosophers One might view these preferences for “genetic and ethicists like Rehmann-Sutter (1998) and research programs” as a temporary choice of Cor van der Weele (see chapter 24 of this vol- society, which cannot financially support all ume), and also by human geneticists. Scriver and more or less promising research programs. I Waters (1999: 272) state clearly that in the case of think, however, that one should offer more con- PKU: “The biological view has implications for vincing arguments for such a choice. the investigation, counselling and treatment of The defense of such choices becomes more crit- patients with genetic disease, because each per- ical when they are grounded in gene-centrism, on son, perhaps excepting monozygotic twin pairs, the assumption that these research strategies can with a nominal monogenic disorder is likely to supply a fundamentally deeper or truer knowl- have a particular phenotypic form of it.” They edge of developmental pathways (and their ab- continue: “The ideas expressed here are not errations) than other research approaches can really new. . . . Lionel Penrose, in his inaugural do. The argument as to why to support genetic address as Galton Professor in 1946 . . . observed, research is then not a scientific argument, nor an among other things, that IQ was not predictable argument that points to some (assumed) prag- from genotype within PKU families, and opined matic reasons (for example: easier access to that a multidisciplinary approach would benefit manipulations), but is an ideological argument. our understanding of this disease. His original Insofar as it is believed by scientists and by the paper has been reprinted to remind us that there general public that this argument is not ideologi- were giants on whose shoulders we stand to see a cal but logically derived from a scientific argu- little further.” ment, the consequences can be that DNA will The framework of this article does not provide become the most important factor because that room to discuss whether or not science really will produce the most dramatic effects, not with produces progress. Rather, I want to point out regard to developmental pathways, but with that more than fifty years after Penrose’s talk his regard to the construction of human self- followers seem to have forgotten the earlier understanding. Such changes can already be ob- recognition of the complexity of traits. served, as has been demonstrated, for example, Judging from the number of publications, their by Jackie Leach Scully.8 Following this line of titles, and the discussion sections (for example, in reflection, humans might not become “homo Science, Nature, and certainly more specialized geneticus” (Rehmann-Sutter 1999) by genetic journals like the German Medizinische Genetik), manipulations, but because humans would it is obviously still far easier to obtain research believe that they, their own lives, their children, money for investigations that claim to be able to the human race itself, could be styled genetically. locate the genetic causes of phenotypic traits and Reproductive choices in, for example, preimplan- their variations than for alternative approaches. tation diagnosis, might not be taken based on For scientific journals and communities, dis- scientifically backed knowledge, but rather on tributors of research money, popular science scientist-supported ideologies, which are rooted magazines, and the public in general, the “ex- in particular background assumptions. If these planation” of traits such as aggressivity (Brunner assumptions, however, prove on examination to et al. 1993a, 1993b; for a critical discussion, see be somewhat outdated in the light of new em- 82 Eva M. Neumann-Held

pirical evidence, then it would seem sensible to regulated synthesis of a polypeptide chain. Examples replace them. are complicated methylation processes that selectively Here, PMG offers a promising alternative. inhibit expression from particular DNA segments. In Following PMG, one cannot claim “more causal- genetic imprinting, for example, the methylation of ity” for DNA in processes of polypeptide synthe- DNA segments depends on their origin from the maternal or the paternal side. The consequence is that sis, because it can be shown that the synthesis of although the maternally and the paternally inherited a polypeptide requires in addition specific non- DNA segments do not differ in the order of their nu- DNA factors. DNA sequences are “the cause” cleic acid sequences, they are nevertheless treated dif- of the synthesis of a polypeptide only if the ex- ferently, and therefore have different effects. Heutink et perimental setup does not allow the investiga- al. (1992) report, for example, that paragangliomas, a tion of anything other than the effects of DNA specific kind of mostly benign tumors, will effect indi- sequences. viduals only when the “disease gene” is inherited from In the framework of PMG, it is obvious that the father, whereas “expression of the phenotype is not such a result may not be interpretated as indicat- observed in the offspring of an affected female.” For ing the causally specific role of DNA but rather further examples of genetic imprinting see Strachan and Read (1996: 201ff). as pointing to the limitations of the scientific approach as such. This then might become rea- 3. The talk was given at the conference Developmental Systems, Competition, and Cooperation in Socio- son to ask further questions. One might wonder, biology and Economics, held at Marienrode Monas- for example, whether claims of validity can be tery, Hildesheim-Marienrode, Germany, on April redeemed for those assumptions, according to 24–28, 1996. which science is granted so much significant 4. Recently, Peter Beurton (1998) suggested a new gene influence on human self-understanding. But this concept which should unify evolutionary and develop- will have to be discussed in another place. mental gene concepts. For a discussion of his approach, see Neumann-Held and Rehmann-Sutter (1999). 5. My special thanks go to Prof. Dr. Ulrich Wolf, Acknowledgments Freiburg, who draw my attention to this article. 6. These justifications, which are asked for here, should I like to express my thanks to Mathias Gutmann not be confounded with a historical or sociological and to Christoph Rehmann-Sutter for many analysis of what is accepted as an explanation, as has helpful discussions, and to Barbara Alexius for been presented, for example, by Keller (1999). her help in translating this article. This work was 7. For a case study on the German physicist and partly supported by a grant of the foundation philosopher Carl Friedrich von Weizsaecker, see MGU, University of Basel. Drieschner (1992), and von Weizsaecker (1977: 101ff: “The political role of science in our culture”). 8. I am quoting here from “Nothing like a Gene,” a Notes paper presented by Jackie Leach Scully at the Genes 1. For a detailed discussion of the positions of Fogle and Development Symposium at the University of (1990) and Waters (1994), see Neumann-Held (1999a); Basel, March 19–20, 1999. of the position of Epp (1997), see Gri≤ths and Neumann-Held (1999). For extensive discussion of References reductionist positions, see Sterelny and Gri≤ths (1999) and Sarkar (1998). For a criticism of reductionism in Barker, D. J. P. (1998). Mothers, Babies and Health in genetics from the standpoint of “methodical cultural- Later Life. Edinburgh: Churchill Livingstone. ism,” see Gutmann (1998). Berger, P., and T. Luckmann. (1980). Die gesell- 2. In addition to those mentioned here, there are many schaftliche Konstruktion der Wirklichkeit. Eine Theorie more molecular mechanisms known that influence the der Wissenssoziologie. Frankfurt: Fischer. Let’s Talk about Genes 83

Beurton, P. J. (1998). Was sind Gene heute? Theory Leben? pp. 141–156. Loccum: Evangelische Akademie Bioscie. 117: 90–99. Loccum. Beurton, P. J. (2000). A unified view of the gene, or how Heutink, P., A. G. L. van der Mey, L. A. Sandkuijl, et to overcome reductionism. In P. J. Beurton, H-J. al. (1992). A gene subject to genomic imprinting and Rheinberger, and R. Falk (Eds.), The Concept of the responsible for hereditary paragangliomas maps to Gene in Development and Evolution, Cambridge: Cam- chromosome 11q23-qter. Human Molecular Genetics 1: bridge University Press. 7–10. Brunner, H. G., M. R. Nelen, P. Iandvoort, et al. Janich, P., F. Kambartel, and J. Mittelstrass. (1993a). X-linked borderline mental retardation with (1972–73). Wissenschaftstheorie als Wissenschaftskri- prominent behavioral disturbance: Phenotype, genetic tik. (Series of articles in nine contributions in Aspekte localization, and evidence for disturbed monoamine V, issues 9–12; Aspekte IV, issues 1–5). metabolism. American Journal of Human Genetics 52: Keller, E. F. (1999). Making sense of life: Explanation 1032–1039. in developmental biology. In J. Maienschein and R. Brunner, H. G., M. R. Nelen, X. O. Breakefield, et al. Creath (Eds.), Biology and Epistemology, pp. 244–260. (1993b). Abnormal behavior associated with a point Cambridge: Cambridge University Press. mutation in the structural gene for monoamine oxidase Keller, E. F. (2000). Is there an organism in this text? A. Science 262: 578–580. In P. Sloan (Ed.), Controlling Our Destinies, pp. Burian, R. M. (1986). On conceptual change in biology: 273–290. South Bend, IN: Notre Dame University The case of the gene. In D. J. Depew and B. H. Weber Press. (Eds.), Evolution at a Crossroads, pp. 21–42. Cam- Kitcher, P. (1982). Genes. British Journal of Philosophy bridge, MA: MIT Press. of Science 33: 337–359. Carlson, E. A. (1991). Defining the gene: An evolv- Kitcher, P. (1992). Gene: Current usages. In E. Fox ing concept. American Journal of Human Genetics 49: Keller and E. A. Lloyd (Eds.), Keywords in Evolu- 475–487. tionary Biology, pp. 128–131. Cambridge, MA: Har- Dawkins, R. (1982). The Extended Phenotype: The vard University Press. Gene as the Unit of Selection. Oxford: Freeman and Co. Knippers, R. (1997). Molekulare Genetik. Stuttgart: Dawkins, R. (1989). The Selfish Gene (2d ed.). Oxford: Thieme. Oxford University Press. Levin, B. (1985). Genes. New York: Wiley. Drieschner, M. (1992). Carl Friedrich von Weizsäcker Lorenzen, P. (1987). Lehrbuch der konstruktiven zur Einführung. Hamburg: Junius. Wissenschaftstheorie. Mannheim: Bibliographisches Epp, C. D. (1997). Definition of a gene. Nature 389: 537. Institut. Falk, R. (1984). The gene in search of an identity. Luhmann, N. (1987). Soziale Systeme. Grundriss einer Human Genetics 68: 195–204. allgemeinen Theorie. Frankfurt: Suhrkamp. Fischer, E. P. (1995). What’s in a gene? On the advan- Maturana, H. R. (1985). Erkennen: Die Organisation tages of imprecision and the success of an open-ended und Verkoerperung von Wirklichkeit. (2d ed.) Braun- idea. In P. Bernhard and C. Cookson (Eds.), Genethics: schweig: Vieweg. Debating Issues and Ethics in Genetic Engineering, pp. Maturana, H. R., and F. J. Varela. (1990). Der Baum 6–9. Basel: H. P. Bernhard. der Erkenntnis: Die biologischen Wurzeln des men- Fogle, T. (1990). Are genes units of inheritance? schlichen Erkennens: Goldmann. Biology and Philosophy 5: 349–371. Mittelstrass, J. (1973). Das praktische Fundament der Gray, R. (1992). Death of the gene: Developmental sys- Wissenschaft und die Aufgabe der Philosophie. In F. tems strike back. In P. Gri≤ths (Ed.), Trees of Life, pp. Kambartel and J. Mittelstrass (Eds.), Zum normati- 165–209. Dordrecht: Kluwer Academic. ven Fundament der Wissenschaft, pp. 1–69. Frankfurt: Gri≤ths, P., and E. M. Neumann-Held. (1999). The Suhrkamp. many faces of the gene. BioScience 49: 656–662. Montgomery, H. E., R. Marshall, H. Hemingway, et al. Gutmann, M. (1998). Information, Gene und Meta- (1998). Human gene for physical performance. Nature phern. In A. Daly (Ed.), Was wissen Biologen schon vom 393: 221–222. 84 Eva M. Neumann-Held

Neumann-Held, E. M. (1997). Gene können nicht alles Schaffner, K. (1993). Discovery and Explanation in erklären. Zur Interpretation genetischer Daten. Uni- Biology and Medicine. Chicago: University of Chicago versitas 52: 469–479. Press. Neumann-Held, E. M. (1998). Jenseits des “genetischen Schmidt, S. J. (Ed.), (1987). Der Diskurs des Radikalen Weltbildes.” In E-M. Engels, T. Junker, and M. Konstruktivismus. Frankfurt: Suhrkamp. Weingarten (Eds.), Ethik der Biowissenschaften, pp. Scriver, C. R., and P. J. Waters. (1999). Monogenetic 261–280. Berlin: Verlag für Wissenschaft und Bildung. traits are not simple: Lessons from phenylketonuria. Neumann-Held, E. M. (1999a). The gene is dead—long Trends in Genetics 15: 267–272. live the gene: Conceptualizing genes the constructionist Singer, M., and P. Berg. (1991). Genes and Genomes: A way. In P. Koslowski (Ed.), Sociobiology and Bio- Changing Perspective. Mill Valley, CA: University economics: The Theory of Evolution in Biological and Science Books. Economic Thinking, pp. 105–137. Berlin: Springer. Smith, K. (2000). What is a genetic trait? In D. Magnus Neumann-Held, E. M. (1999b). Von Genen, (Ed.), Contemporary Genetic Technology: Scientiﬁc, Merkmalen und Kontexten. Eine philosophisch- Ethical, and Social Challenges. Melbourne: Krieger. wissenschaftstheoretische Analyse zum Begriff der Sterelny, K., and P. E. Gri≤ths. (1999). Sex and Death: “genetischen Verursachung.” In T. Braun and M. An Introduction to Philosophy of Biology. Chicago: Gene und Gesellschaft, Elstner (Eds.), pp. 85–94. University of Chicago Press. Heidelberg: Deutsches Krebsforschungszentrum. Strachan, T., and A. P. Read. (1996). Molekulare Neumann-Held, E. M., and C. Rehmann-Sutter. Humangenetik. Heidelberg: Spektrum Akademischer (1999). Individuation and reality of genes. A comment Verlag. to Peter J. Beurton’s article “Was sind Gene heute?” Der Garten des Theory Bioscie. 118: 85–95. von Weizsäcker, C. F. (1992). Menschlichen. München: Hanser Verlag. Oyama, S. (1985). The Ontogeny of Information: Phil- Developmental Systems and Evolution. Cambridge: Waters, C. K. (1994). Genes made molecular. osophy of Science Cambridge University Press. (2d rev. ed., Durham, NC: 61: 163–185. Duke University Press, 2000.) Weingarten, M. (1998). Wissenschaftstheorie als Wissenschaftskritik. Oyama, S. (1988). Stasis, development and heredity. In Bonn: Pahl-Rugenstein. M-W. Ho and S. Fox (Eds.), Process and Metaphors in Wolf, Ulrich (1995). The genetic contribution to the the New Evolutionary Paradigm, pp. 255–274. London: phenotype. Human Genetics 95: 127–148. Wiley. Wolf, Ulrich (1997). Identical mutations and pheno- Portin, P. (1993). The concept of the gene: Short his- tyopic variation. Human Genetics 100: 305–321. tory and present status. The Quarterly Review of Biol- ogy 68: 173–223. Rehmann-Sutter, C. (1998). DNA-Horoskope. In M. Düwell and D. Mieth (Eds.), Ethik in der Human- genetik. Die neueren Entwicklungen in der genetischen Frühdiagnostik aus ethischer Perspektive, pp. 415–443. Tübingen: Francke Verlag. Rehmann-Sutter, C. (1999). Das Jahrhundert der Genetik und der homo geneticus. Novalis, Nov., pp. 11–14. Rheinberger, H-J. (2000). Gene concepts. Fragments from the perspective of molecular biology. In P. Beurton, R. Falk, and H-J. Rheinberger (Eds.), The Concept of the Gene in Development and Evolution, pp. 219–239. Cambridge: Cambridge University Press. Sarkar, S. (1998). Genetics and Reductionism. Cam- bridge: Cambridge University Press. 8 Deconstructing the Gene and Reconstructing Molecular Developmental Systems

Lenny Moss

The articulation of a “developmental systems avoid confusing apples and oranges. This will perspective” has entailed both deconstructive and constitute the deconstructive aspect of the chap- reconstructive undertakings. The deconstructive ter. “Genes,” I will argue, are, and can be, pro- side has consisted largely in critiques of nativism ductively conceived of in two different ways, in ethology, and psychology (e.g., Lehrman, albeit with nothing good resulting from the Gottlieb, Johnstone, Gray, and Oyama), cri- conflation of the two. And while these different tiques of genetic determinism in cell, molecular, senses of the gene bear an important historical developmental and population biology (e.g., relationship to the classical versus molecular dis- Keller, Lewontin, Moss, Nijhout), and critiques tinction, they are not identical with it, and indeed of the incoherence of the gene-environment may now both involve molecular-level referents. dichotomy itself (e.g., Lewontin, Oyama). Re- Next, the second of the two senses of the gene, constructively, the conceptualization of organ- having become unencumbered by the first, will be isms qua developmental systems has been given a taken up and reembedded in the context of func- certain structure by way of its characterization as tionally conserved, contingently linked, multi- systems of developmental resources (both inside molecular modules. and outside the skin of the organism) the interac- In a recent text entitled Cell, Embryos and tions of which are stably reproduced in succeed- Evolution, John Gerhart and Marc Kirshner ing generations (Gri≤ths and Gray). Inside “the (1997), two of our leading contemporary investi- skin,” the stock of relevant developmental re- gators in cell and developmental biology, have sources has been shown to include epigenetic moved against the prevailing genomic tide to systems, such as chromosomal imprinting, cyto- accomplish exactly what good scientists are plasmic structure and steady-state dynamics, meant to do: they have thought about the most which, along with nucleic acid sequence, are her- di≤cult and important problems of their field itable and susceptible of variation (Jablonka and and have offered new concepts and explanatory Lamb). Also in the reconstructive vein, Weber models based upon a comprehensive and insight- and Depew have begun to bring the DST per- ful analysis of the available evidence. I will argue spective into a productive encounter with work that Gerhart’s and Kirschner’s notions of con- on the self-organizing properties of autocatalytic, served functional modules, which have become thermodynamically “dissipative,” non-equilibrium multiply contingent in their linkages and thereby systems. The present chapter addresses both the central to exploratory and adaptive ontogenetic deconstructive and the reconstructive dimension pathways, are exactly what the DST doctor of DST. If successful, this chapter will highlight requires for overcoming inner/outer dichotomies an intermediate level of biological organization in favor of self-organizing, causally reciprocal that, by its very nature, will serve to recontextu- systems of interactants. alize the understanding of genes, thus obviating even naive temptations toward gene/environment dichotomies, and even more important still, will Gene-P or Gene-D: That is the Question open up a very rich area of empirical investigation to examination and conceptualization in Wilhelm Johannsen coined the term gene in 1907 developmental-systems terms. and introduced the phenotype/genotype distinc- This chapter will proceed in two parts. First, it tion by 1910 (Carlson 1966). His motivation for will be necessary to introduce a basic distinction the latter was to step away from the preforma- with respect to the concept of a gene so as to tionist assumptions of early Mendelism, which 86 Lenny Moss

failed to distinguish between that which is phy- significance from having a definite relationship to sically transmitted between generations and the some unit of phenotype. There are, of course, characteristics that come to appear in the mature “genes” that satisfy this definition—that is, genes form of progeny. By 1923 Johannsen’s skepticism whose identity is specified by some “unit” of and disenchantment with genetic dogmatism had phenotype that is heritably expressed along reached a fever pitch and his inspired reflections Mendelian lines. The research program to which provide a nice point of departure for making Johanssen was addressing himself assumed that a key distinction. To begin with, Johannsen all of the phenotype would come to be geneti- raised a question about the scope of Mendelian cally specified in this way. Johanssen hereby decomposability. expresses his disbelief in this being the case. He thought that only a limited set of characteris- Certainly by far the most comprehensive and most deci- tics, none of which get to the core of a species sive part of the whole genotype does not seem to be able identity, would separate out that way . . . and he to segregate in units; and as yet we are mostly operating with “characters,” which are rather superficial in com- was right. But why would this be the case? parison with the fundamental specific or generic nature Johanssen’s reflections on the meaning of allelic of the organism. This holds good even in those frequent variation provide further insight on this matter. cases where the characters in question may have the greatest importance for the welfare or economic value When we regard Mendelian “pairs,” Aa, Bb and so on, normal of the individuals. it is in most cases a reaction (character) that is abnormal We are very far from the ideal of enthusiastic the “allele” to an . Yellow in ripe peas is nor- Mendelians, namely the possibility of dissolving genomal, the green is an expression for imperfect ripeness as types into relatively small units, be they called genes, can easily be proven experimentally, e.g., by ether- allelomorphs, factors, or something else. Personally I ization. . . . The rich material from the American Drosophilia believe in a great central “something” as yet not divisi- —researches of Morgan’s school has sup- ble into separate factors. The pomace-flies in Morgan’s plied many cases of multiple allelisms—most of all of splendid experiments continue to be pomace flies even them being different “abnormalities” compared with if they lose all “good” genes necessary for a normal fly- the characters of the normal wild fly. . . . To my mind life, or if they be possessed with all the “bad” genes, the main question in regard to these units is this: Are detrimental to the welfare of this little friend of the experimentally demonstrated units anything more than geneticist. (Johannsen 1923: 137) expressions for local deviations from the original (“normal”) constitutional state in the chromosomes? Depending on how one hears this passage, one Is the whole of Mendelism perhaps nothing but an may be tempted either to dismiss it as simply establishment of very many chromosomal irregulari- being out of date, or to respond sympathetically ties, disturbances or diseases of enormously practical and theoretical importance but without deeper value with the sense that, yes, there is something right for an understanding of the “normal” constitution of about Johannsen’s intuition. Upon inspection, natural biotypes? (Johannsen 1923: 138–140). one can see that what distinguishes these two “hearings” ultimately has much to do with the Eight decades years after Johanssen’s paper, it nature of the genotype/phenotype relationship. is still surprisingly common for educated people In one sense, we want to say that Johannsen is to believe that alternative alleles at a single locus just wrong because we now know that all genom- carry substantive instructions for qualitatively ic DNA segregates into separate chromosomes. different traits. The canonical example for this But to speak that way is to set aside any consid- misunderstanding would probably be eye color eration of the direct relationship of a unit of where one allele is “for brown eyes” and one genotype to a unit of phenotype. Johanssen went allele is “for blue eyes.” But, just as Johanssen to pains to distinguish genotype from phenotype, suggests, what this locus actually entails is a nec- but the gene that he has in mind still gets its essary (but not su≤cient) resource for making Deconstructing the Gene 87

brown eye pigment. The “allele for blue eyes” is not and expressivity meant to denote the extent to just any deviation from the sequence necessary which a gene expresses itself, i.e., in a phenotype). for the brown pigment resulting in its absence in Context-dependent variations in phenotype were the organism. There is no “information” for blue thereby treated as the result of nothing but the pigments, let alone blue eyes, in “the gene for” intrinsic properties of alleles. blue eyes. Blue eyes are produced by the develop- The “whole of Mendelism” is circumscribed by ing organism in the absence of this particular those conditions that will allow something to brown-eye genetic resource. The allele (or gene) “show up” as a Mendelian gene. If allelic varia- for blue eyes is indeed a deviation from a norm, tions do not result in predictably different pheno- just in the sense that Johanssen suggests. Of types, either because all the variants at a locus course, this is not to say that a deviation from the su≤ce as “normal” resources or because the phe- norm cannot become important in a positive notypic consequences of alternative alleles is too sense, or that what “counts” as normal cannot context-sensitive, or if the absence of the nor- shift; it can. The norm is determined by the con- mal resource is developmentally lethal, then such text. In an agricultural context where commodity alleles will not “show up” as genes within the value and not ecological versatility becomes the sphere of Mendelism. Johanssen’s notion that motor force of reproduction the loss of an abil- Mendelian genes (no matter how economically ity to make a seed, for example, can become and medically important) are limited to relatively the norm. In the case of eye color, the inability superficial traits can readily be seen and appreci- to make brown presumably became a selec- ated through our contemporary eyes. tive advantage within certain human breeding The basic notion of the gene that Johanssen populations. described and delimited still has currency. It is Johanssen suggested that the “whole of still in use and still useful, although it is certainly Mendelism” is nothing but a set of allelic devia- no longer the only basis for our understanding of tions from a norm. It is that set of alleles associ- what it means to be a gene. This gene concept is ated with the Mendelian inheritance of specific what I will refer to as Gene-P. Gene-P is defined phenotypes. It should not be surprising that the by its relationship to a phenotype, albeit with vast majority of these are cases in which the pheno requirements as regards specific molecular notypic difference is the result of the functional sequence nor with respect to the biology involved absence of some “normal” resource. When the in producing the phenotype. Gene-P is the ex- absence of some specific sequence results in a pre- pression of a kind of instrumental preformation- dictable phenotype then one can speak of a “gene ism (thus the “P”).1 When one speaks of a gene in for this phenotype.” But whether it is blue eyes, the sense of Gene-P, one simply speaks as if it albinism, or cystic fibrosis, the resulting pheno- causes the phenotype. A gene for blue eyes is a type is the result of what organisms do in the Gene-P. What makes it count as a gene for blue absence of the normal sequence. Viewed this way eyes is not any definite molecular sequence (after it is hardly surprising that what organisms do will all, it is the absence of a sequence-based resource be context-dependent—varying with the presence that matters here), nor any knowledge of the and absence of all the other developmental developmental pathway that leads to blue eyes resources, for example, genomic, gestational, (to which the “gene for blue eyes” makes a negli- nutritional, sunlight, and so on. This context gible contribution at most), but only the ability to dependence was assimilated in the gene-centrism track the transmission of this gene as a predictor of the Mendelian model by use of the terms pen- of blue eyes. Thus far Gene-P sounds purely clas- etrance and expressivity (with penetrance meant sical, that is, Mendelian as opposed to molecular. to denote the gene’s propensity to express itself or But a molecular entity can be treated as a Gene- 88 Lenny Moss

P as well. BRCA1, the gene for breast cancer, is a the larger context, and not germane to the status Gene-P, as is the gene for cystic fibrosis, even of N-CAM as a Gene-D. The expression of an though in both cases phenotypic probabilities embryonic form (highly sialylated) in the mature based upon pedigrees have become supplanted by organism is associated with neural plasticity in probabilities based upon molecular probes. What the adult brain (Walsh and Doherty 1997) but these molecular probes do is to verify that some could well have pathological consequences if normal DNA sequence is absent, by confirming expressed in other tissues—yet it would not affect the presence of one, out of many possible, devia- the identity of the N-CAM sequence as a Gene- tions from that normal sequence that has been D. So where a Gene-P is defined strictly on the shown to be correlated (to a greater or lesser basis of its instrumental utility in predicting a extent) with some phenotypic abnormality. To phenotypic outcome and is most often based satisfy the conditions of being a gene for breast upon the absence of some normal sequence, a cancer, or a gene for cystic fibrosis, does not Gene-D is a specific developmental resource, entail knowledge about the biology of healthy defined by its specific molecular sequence, and breasts nor of healthy pulmonary function, nor is thereby functional template capacity, and yet it it contingent upon an ability to track the causal is indeterminate with respect to ultimate pheno- pathway from the absence of the normal se- typic outcomes. quence resource to the complex phenomenol- A Gene-P allows one to speak predictively ogy of these diseases. The explanatory “game” about phenotypes, but only (as Johanssen real- played by Gene-P is thus not confined to purely ized) in a limited number of cases and within classical methods, which unfortunately has made some contextually circumscribed range of prob- it all the easier to conflate this meaning of the abilities. In the absence of, for example, a full “gene” with the one I will refer to as Gene-D. molecular-developmental understanding of the Quite unlike Gene-P, Gene-D is defined by its processes resulting in the pathophysiology of cys- molecular sequence. A Gene-D is a developmental tic fibrosis, it can be prognostically useful to resource (hence the “D”) which in itself is in- speak of “the gene for cystic fibrosis.” The “nor- determinate with respect to phenotype. To be a mal resource,” that is, the Gene-D located at the Gene-D is to be a transcriptional unit (extend- cystic fibrosis locus for the great majority of in- ing from start to stop codons) within which are dividuals who do not have a family history of contained molecular template resources. These cystic fibrosis a≥iction, is not thereby a gene for templates typically serve as resources in the pro- normal pulmonary function (any more than the duction of various “gene-products”—directly in roughly thirty thousand other genes involved in the synthesis of RNA, and indirectly in the synth- normal pulmonary function); rather, it is a mem- esis of a host of related polypeptides. To be a gene ber of a “family” of transmembrane ion-channel for N-CAM, the so-called “neural cell adhesion templates. As a developmental resource, it is one molecule,” for example, is to contain the spe- among very many that play a direct role in pul- cific nucleic acid sequences from which any of a monary development and function (as well as hundred potentially different isoforms of the N- many other things). To speak of, and direct one’s CAM protein may ultimately be derived (Zorn attention to, this gene for a transmembrane-ion and Krieg 1992). Studies have shown that N- conductance regulator protein is to become CAM molecules are (despite the name) expressed involved in an entirely different kind of explana- in many tissues, at different developmental tory game, that is, that of a Gene-D. There is no stages, and in many different forms. The pheno- preformationist story to be had at this level. To types of which N-CAM molecules are co-consti- study the biological role and function of this gene tutive are thus highly variable, contingent upon for a chloride channel involves locating it within Deconstructing the Gene 89

all of the contexts in which it is biologically active the identification of the “normal” sequence at the and attempting to elucidate the causal pathways BRCA1 locus. This sequence provides the tem- in which it is an interactant (see Kerem and plate from which a family of proteins are ulti- Kerem 1995; Jilling and Kirk 1997). And as with mately synthesized, and these proteins are (like it any developmental resource, its status with or not) referred to as BRCA1 proteins. BRCA1 respect to cause and effect in any given interac- proteins are expressed ubiquitously across tis- tion will be contextual and perspectival (i.e., its sue types and developmental stages (Welsch, actions will be viewed as either the cause of some- Schubert, and King 1998). The BRCA1 proteins thing or as the result of something else, depend- are expressed in different tissues at different times ing upon how a particular inquiry is framed). As and appear to be associated with DNA repair, a molecular-level developmental resource, Gene- transcriptional transactivation of other Genes-D, D is ontologically on the same plane as any num- and cell-cycle control (albeit all in the context ber of other biomolecules, that is, proteins, RNA, of a complex and variable set of other fac- oligosaccharides, and so forth, which is only to tors) (Irminger-Finger, Siegel, and Leung 1999). say that it warrants no causal privileging before Where the probes for BRCA1 as a Gene-P pro- the fact. Gene-P and Gene-D are distinctly differ- vide an instrument for predicting breast cancer ent concepts, with distinctly different conditions risk, molecular studies on the BRCA1 sequence of satisfaction for what it means to be a gene. as a Gene-D have indicated that proteins derived They play distinctly different explanatory roles. from the “normal” sequence, in the cases of There is nothing that is simultaneously both a woman with “sporadic” breast cancer, fail to Gene-D and a Gene-P. That the search for one migrate to the cell nucleus and are rather found can lead to the discovery of another does not in the cytoplasm (Chen et al. 1995). Studies on change this fact. Finding the Gene-P for cystic BRCA1 as a developmental resource, that is, as a fibrosis led to the identification of a Gene-D for Gene-D, may well lead to important insights into a chloride-ion conductance channel template se- the cellular dynamics of breast cancer on the quence. But the latter is not a gene for an organ- basis of the elucidation of complex developmen- ismic phenotype. Its explanatory value is not tal processes that involve proteins derived from realized (and cannot be realized) in the form of the “normal” BRCA1 sequence as well as many an “as if” preformationist tool for predicting others. The potential medical significance of this phenotypes. Rather, the explanatory value of a is not easily overestimated inasmuch as 90 per- Gene-D is realized in an analysis of developmen- cent to 95 percent of breast cancers are not asso- tal and physiological interactions in which the ciated with any germline mutation, but rather are direction and priority of causal determinations sporadic. Thus for the great majority of breast are experimentally first revealed. Cystic fibrosis cancers there is no Gene-P to treat as a marker provides one example of the relationship between before the fact, as if it independently determined these two gene concepts, the BRCA1 gene pro- the phenotype. Studies on the dynamics of vides another. In this case a Gene-P for use in BRCA1 as a developmental resource, that is, identifying heritable risk for breast cancer was as a Gene-D, with no phenotypic outcome pre- identified through studies on families in which inscribed, have, not surprisingly, become wide- such risk had been epidemiologically established. spread. This resulted in the targeting of a locus and the The common use of two qualitatively different construction of molecular probes for identifying explanatory strategies, that is, that of Gene-P and some of the aberrant sequences the presence of that of Gene-D, which yet share the name which correlate with heightened incidence of “gene” predictably lends itself to much easy con- breast cancer (Miki et al. 1994). It also resulted in fusion. The notion that there exists such a thing 90 Lenny Moss

as a gene that is simultaneously a specific nucleic of a particular polypeptide. But if conceived as a acid template and a preformationistic determi- Gene-D (or any other developmental resource) nant of an organismic phenotype (i.e., the genet- this would be fine. It is only a residual association ic blueprint we hear so much about) is based upon with the Gene-P concept that leads one to expect exactly this conflation. The Gene-P/Gene-D dis- more. Neumann-Held’s solution is to pack every- tinction is new to developmental systems theory thing into a revised gene concept that is causally literature. One of the benefits of this analysis is necessary for the production of a particular that it enables DST thinkers to acknowledge the polypeptide (which can only be known in retro- fact of a valid role for Gene-P as tool, albeit spect) and call this the Process Molecular Gene coarse, yet useful under certain circumstances (PMG).2 The consequences of doing this are quite (such as clinical genetics counseling). As a con- problematic. A typical eukaryotic Gene-D (as cept, Gene-P, however, falls outside of the efforts discussed below for the case of N-CAM) may of DST to formulate a perspective which does not provide the template resource for, let’s say, ten presume the causal (or ontological) priority of different polypeptide forms,3 based upon differ- any particular kind of entity and thereby main- ential mRNA splicing. This would already tains an explanatory openness on all empirical increase the number of PMG names necessary by fronts. It is not Gene-P as such which should be an order of magnitude. The PMG would have condemned, but rather the conflation of the two to include, for starters, both the complexes asso- gene concepts. Once distinguished from Gene-P ciated with transcription and the complexes asso- and cut free of its baggage, Gene-D, the develop- ciated with splicing, thus the PMG would jump mental resource gene, emerges as exactly the right up to a different level in the biological hierar- gene concept for the purposes of DST. chy, including as it would those intermediate lev- In the absence of the Gene-P/Gene-D distinc- els of organization I refer to as multimolecular tion, predictable consequences can be seen. modules. But the composition of transcriptional Certain attempts to clarify “the” gene concept complexes already are highly variable (with tissue proceed by focusing on what amounts to one of type, developmental stage, physiological state of the above while ignoring the existence of the a cell, etc). And the regulatory impact of a tran- other. Waters (1994) demonstrated how “genes scriptional complex is not based solely on its became molecular” by focusing on how, in effect, composition but also upon the kinetics of its Gene-P became molecularized while ignoring assembly (Miklos and Rubin 1996). Further, Gene-D. Godfrey-Smith (1999), conversely, has transcriptional complexes are also buffered so recently offered a treatment of the gene concept that variations in composition do not necessarily in terms of coding which amounts to identifying result in different transcriptional results (Gerhart “the gene” with Gene-D through ignoring the and Kirshner 1997). So just at the level of at- history of Gene-P. Both of these studies have tempting to include the most proximate factors valuable things to offer on their respective sides involved in producing a polypeptide, there is an of the divide—but attempts to seek clarity or explosion of complexity and contingency that coherence that fail to recognize the Gene- obviates the possibility of a manageable, let alone P/Gene-D distinction run the risk of bringing perspicuous, taxonomy. contaminating baggage across the border. Such is A Gene-D denotes something, that is, a tran- the case, I would suggest, for Neumann-Held’s scriptional unit of DNA, albeit a something that attempt (see chapter 7 of this volume) to provide is only a resource from which different results DST with a unified gene concept. Her critical may arise in different contexts. In trying to pack point of departure is the idea that the “classical some causally complete set of reactants into a molecular gene” cannot account for the making revised gene concept, Neumann-Held falls victim Deconstructing the Gene 91

to a conflationary temptation. PMGs would not any particular explanatory story to be consistent denote anything more durable than singularities with the tenets of the DST perspective. in the life history of an organism and thus would But just how tractable are explanatory stories be of negligible biological value. Even the at- based solely upon the contingent causal relations tempt to limit an expanded gene concept to “just of Genes-D and other contextually relevant that” which pertains to the linear polypeptide developmental resources? What the sad en- backbone is unacceptably arbitrary, inasmuch as durance of that tired old dichotomy consisting of nascent polypeptides are enzymatically modified (conflated) genes and (ill-defined) environment in the lumen of the endoplasmic reticulum simul- has helped to obscure, are the many levels of bio- taneous with their translational elaboration at logical ordering that mediate between individual the surface of the rough ER. In other words, biomolecules and whole developmental systems. To logically there is no temporal boundary between give up the preformationist umbilical cord is not polypeptide backbone synthesis and polypeptide to drop into an abyss of limitless complexity but folding and modification. The PMG concept rather to remain empirically open to discovering sacrifices both what the Gene-P and the Gene-D what level of biological ordering is most relevant concepts provide, and offers nothing of value in for one’s explanatory purposes. Those who have their place. The solution to the Gene-P/Gene-D limited their explanatory sights to “genes and incompatibility problem is not to try to split the environment” have fallen behind the research difference. These gene concepts do entirely differ- tide. Functionally conserved, multimolecular ent kinds of work in entirely different ways. modules have emerged as new units of develop- The Gene-P/Gene-D distinction manages to ment, morphology, innovation, and variation, at help address certain criticisms of DST. Critics an intermediate level of biological ordering. It is of the developmental systems perspective, most time to turn our attention to the nature of these emphatically Phillip Kitcher (in press), but also modules. Sterelny, Smith, and Dickison (1996), have questioned whether DST proffers an extreme and unwieldy holism that, while possibly true to the Modules in the Middle multidirectional causal complexity of reality, obviates the possibility of actual scientific prog- Most conserved processes in metazoa operate under contingencies, that is, their activity depends on condi- ress. But here we can show that Kitcher and tions, both external and internal to the organism. like-minded scholars are guilty of mixing up (Gerhart and Kirshner 1997) apples and oranges. If DSTers insisted that every causal story, for whatever explanatory purpose, The biological use of the term module is in itself which happens to involve any Gene-D had to take rather ambiguous. Modules have been spoken of into account every other developmental resource in many different places and with different which might play a causal role at any time, or any explanatory intents, some of which would have place, in the life history of an organism, he/they little in common with present purposes. The would be justified in their criticisms. But the cri- sense that runs through any ascription of modu- teria of causal relevance for explanatory stories larity is that of a unit that is a component part of for DST are contingent upon the contexts of a larger system and yet possessed of its own struc- interest—indeed, that is much of the point. To tural and/or functional identity. But modules separate out the instrumental as-if approach of have been posited from both “top-down” and Gene-P concepts, from the mechanistic basis of “bottom-up” research directions, sometimes as the DST approach is yet to say nothing about the explanans and sometimes as explanandum. When scope of causal factors that must be invoked for one begins with the phenomenal properties of a 92 Lenny Moss

system as a whole, and then attempts to explain was famously found that “genes come in pieces” its working on the basis of a postulated modu- (Gilbert 1978), which is to say that the typical lar organization, as cognitive scientists such as transcriptional unit extending from start to stop Minsky have done with the brain, then modules codons, turns out to be a sea of nonprotein tem- are arrived at from a top-down direction. Such plate sequences (“introns”) within which islands modules have not been empirically observed but of template-sequence “exons” are embedded. speculatively brought forth as explanans (not as What shores up the sense in which exons are explanandum). Modules of this sort may well indeed modules is the ample evidence that the express a preformationist viewpoint, although protein fragments which exon templates “code- that does not necessarily need to be the case. One for” have structural and functional integrity could, for example, begin “at the top” and pos- (Gilbert and Glynias 1993). This is well illus- tulate an ad hoc formation of modularized struc- trated by the example of N-CAM. The Gene-D ture contingent upon circumstance. But in for N-CAM, that is, that particular transcrip- evolutionary psychology, where a new form of tional unit, contains nineteen different exons module-talk is in vogue, it is a clear spirit of which serve as modular resources from out of preformationism that carries the new day. For which many different N-CAM protein forms example, evolutionary pyschologists have be- can ultimately be fashioned. Any N-CAM mole- come prone to begin with evidence for some cog- cule, at the linear peptide level, is indebted to nitive phenotype, (e.g., social cheat detection; some specific subset of N-CAM Gene-D exon depressive/socially submissive behavior, etc.) and modules (but never all of them). NCAM proteins then proceed to argue for its evolutionary origins are classed into four main groups based on size as a proto-human Pleistocene adaptation. The (110KDa, 120KDa, 140KDa, 180KDa) and cognitive capacity/phenotype (whether still of plasma membrane attachment (see figure 8.1). adaptive value or not) is then construed to be the Where the 140KDa and 180KDa classes traverse expression of a developmentally invariant, pre- through the plasma membrane the 120KDa formationistically transmitted module that has class is linked to the membrane only superficially been passed along from generation to generation through an auxiliary connector. No N-CAM ever since. The cautionary point is just that the does both. And the basis for this difference is the concept of module itself does not specifiy its place either/or inclusion of one of two different exon along a preformationism-epigenesis axis. Having modules into the final RNA “transcript.” The evidenced the slipperiness of the module idea and 140KDa and 180KDa must be derived from a the fact that it is not just any old biological mod- transcript which includes exon d16 in order to be ule that is being embraced, it is time to clarify able traverse the plasma membrane, but must that sense of module which is of interest. lack exon d15, as where the 120KDa NCAM The “genetic revolution” of the twentieth cen- class is derived from a transcript that must tury did not result in a search for any form of include exon d15 in order to be able to associate subcellular modules, nor any expectation of by auxiliary connector, but must lack exons finding such. Rather, the recognition of modular- d16–19 (figure 8.1). The process by which the ity came as a surprise. What was expected, by complete set of exons, at the level of the messen- virtue of conflating Gene-P with Gene-D, was ger RNA transcript, is reduced to a functional set something like direct relationships, able to be through the excision of selected exons, is referred mechanistically elucidated, between transcrip- to as “splicing.” tional units and organismic phenotypes. What Modularity, at the level of individual genes has come to be found are modular patterns of (Gene-D), which is the rule not the exception for organization both “above” and “below” the level the eukaryotic cell and all metacellular organ- of the full transcriptional unit. Of the latter, it isms, provides for developmentally contingent Deconstructing the Gene 93

Figure 8.1 A schematic diagram of the four main classes of N-CAM protein. See text for details.

flexibility in the expression and realization of matrices, where at least three such Gene-D “gene-products” from out of the resource base homology groups have been identified: the inte- which any Gene-D represents. N-CAM is just grins, the cadherins, and the immunoglobulin such a modularized Gene-D resource, but it is superfamily. A necessary molecular concomitant also just one member of a “superfamily” of mod- of organismal complexity appears to be that of ularized genetic resources whose kinship is great developmental versatility in the resources defined by the possession of homologous mod- available for constructing cell-to-cell and cell-to- ules. The “immunoglobulin superfamily” consists matrix linkages. The significance of this for de- of many different genetic open reading frames velopmental systems thinking will be addressed (Genes-D), including those that provide the tem- further below. It is also more than suggestive plate resources for the various immunoglobulins, that during the course of 3.5 billion years, meta- immune cell receptors, certain growth factor re- cellular life has not evolved from prokaryotes ceptors, and a variety of cell-adhesion molecules whose modular structure at the DNA level in addition to N-CAM (Walsh and Doherty appears to have been lost (Gilbert and Glynias 1997). Modularity and homology have come 1993). Can it be that a necessary prerequisite for together as complementary themes arising out differentiated multicellular organization is a of research into subcellular biochemistry and capacity for contingent cell-to-cell linkages that molecular biology. Much of the evolutionary requires a proliferation of linker-molecule vari- novelty associated with increasing organismic ants generated by modular reshu≥ing? complexity, it turns out, has been achieved In their quasi-manifesto of 1996, developmen- through the reshu≥ing and mixing and matching tal biologists Gilbert, Opitz, and Raff hailed the of modular exon units to form families of homol- return of the three key concepts that had been ogous genetic (Gene-D) resources. This has been shunted aside by geneticists. These concepts, particularly pronounced with respect to those which in a strong sense are interlinked, are that of molecules associated with developmentally and macroevolution, homology, and morphogenetic functionally contingent associations between fields. Homology, which had formerly provided a cells and other cells, and cells and extracellular pivotal linkage between ontogeny and phylo- 94 Lenny Moss

geny, was cast aside by population genetics. If are fundamentally different. What can properly evolution was to be thought of strictly in terms be gleaned from these findings is that there are of changes in gene frequencies, then there would biological processes that are more fundamental be no place for relations of similarity in explain- than even these differences. Not only genetic ing evolutionary change. All could be accounted sequences, but also whole processes, including for on the basis of allelic difference, selection, and activation circuits are modular in nature and population dynamics. However, where nine- highly conserved. The vertebrate eye and the teenth century biology discovered relations of invertebrate eye, the vertebrate segmentation homology on the macroscopic morphological scheme and the invertebrate segmentation level, later twentieth century biology has discov- scheme, have homologous modules in common. ered relations of homology on the subcellular and And these process modules are likely to go back molecular levels. Unlike nineteenth century to the one-celled eukaryotic stage. Novel mor- homology, the new understanding is based upon phologies appear to be the result of novel link- modular units that appear to predate the diver- ages and novel timing, that is, novel develop- gence of any of the major body plans. mental patterns, between modular components The strongest impetus for acknowledging the established within and across cellular bound- role of such ground-level homologies has come aries. The developmental emergence of dynamic from further investigations into homeosis. patterns of modular organization across cells re- Homeosis pertains to the progressive segmenta- sults in morphogenetic fields. The recognition of tion of body parts that had been genetically ana- highly conserved, dynamic, multimolecular mod- lyzed in Drosophilia. Homeotic genes were found ules at many intermediate levels of organization to all contain a homologous sequence module provides developmental systems theory, I would called a “homeobox,” to be arrayed along a sin- suggest, the materials for articulating its view. gle chromosome, and to be activated sequentially Where Gilbert, Opitz, and Raff have been the in sequential segments indicating a direct regula- harbingers of a new rapprochement with con- tory involvement in the processes of developmen- cepts of embryology past, Gerhart and Kirshner tal segmentation. The big surprise came with the have contributed much toward an understanding discovery that all of the above holds for verte- of the dynamics of modularity and contingency. brates as well as invertebrates. The segmentation The core idea is just this. Evolution of complex of invertebrates such as Drosophilia and that of multicellular organisms has proceeded through vertebrates was thought only to be analog- the pulling-apart of highly conserved processes ous, not homologous. The finding that genetic whose linkages have become contingent. When resources, their relative position on chromo- linkages are contingent they can be “regulated” somes and the activation cascade processes asso- and made sensitive and responsive to surround- ciated with developmental segmentation in both ing conditions. Self-organizing sequential con- invertebrates and vertebrates are homologous, structions of modular unions in regulatory was startling. Similarly it was found that the Pax- linkage are truly cycles of contingency. Adaptive 6 gene homologs, which are functionally (experi- developmental flexibility is achieved through the mentally) interchangeable across vertebrate and capacity to generate variable alternative linkages invertebrate taxa, appear to play a comparable that are subject to selective stabilization in con- role in eye morphogenesis, although vertebrate text sensitive ways. A marvelous example of this and invertebrate eyes are certainly not homolo- can be seen in Kirshner’s work on the micro- gous in classical morphological terms. tubule system. Vertebrate and invertebrate bodily segmen- Microtubule structures are central to cellular tation, and vertebrate and invertebrate eyes morphogenesis: to the general formation of cell Deconstructing the Gene 95

shape and internal compartmentalization, to the text sensitivity—in the reciprocity of centripe- formation of axons and dendrites in neural tissue tal and centrifugal vectors of causal influence— and of apical and basolateral domains in epithe- emanating from and entering into individual lia, to the directedness of cellular secretions, and cells. so on. The microtubule system is also the source The microtubule system causes cell-morpho- of the mitotic spindle fibers that pull homologous genesis to be contingent upon the availability of chromosomes apart during mitosis. Microtubules cues in the larger morphogenetic field (which is grow radially out from a microtubule organizing itself the product of processes of interactive self- center (MTOC) and are found to be in a state of assembly). There are, however, even more gen- “dynamic instability.” Microtubules emanating eral mechanisms for pulling core functional from the MTOC are simultaneously growing out processes apart and rendering their connections toward the cell periphery and depolymeriz- contingent upon internal and external factors. ing back to the center, albeit at different rates. The most general mechanism for doing this is Polymerization to the cell periphery from the cell that of phosphorylation. By adding a highly center takes about ten minutes, whereas complete charged (negative) phosphate group to a protein, depolymerization takes only approximately two the enzymatic or other activity of that protein minutes (Gerhart and Kirschner 1997). The ac- can generally be disrupted, and made contingent celerated rate of depolymerization is dependent upon removal of the phosphate group. The en- upon GTP hydrolysis. Why cells would expend zymes that attach phosphate groups onto pro- significant amounts of stored energy (in the form teins are called “kinases.” The enzymes that of GTP) on destabilizing microtubule structures remove them are “phosphatases.” There are an had been a mystery. Cellular morphogenesis is estimated two thousand genes (Gene-D) that are a central part of organismic development, and template resources for kinase synthesis and microtubule filament assembly is a key feature of another one thousand for phosphatases (Hunter cellular morphogenesis. Kirshner’s solution was 1995). Thus 3 percent to 6 percent of the hu- to see processes of variation and selection as cen- man genome is associated with this particular tral to how an organism solves problems of self- contingency-making process alone. Modularity organizing ontogeny. The expendure of GTP is built of the ability to uncouple core functions energy buys directional variation. It prevents and make the linkages to one another contingent microtubules from lingering too long in the upon context specifics. The proliferation of link- absence of contextual reinforcement (Mitchison age possibilities is seen, among other places, in and Kirschner 1984; Kirschner and Mitchison the elaboration, by way of superfamilies, of 1986). The microtubule system becomes a kind Gene-D resources for cell adhesion molecules as of context sensor. The ability to generate stable mentioned above. Many multiple cell-linkage filamentous structures comes from within, but the devices, built of modular variation, make for sen- orientational cues are contextual. Microtubule sitive context discrimination and the transmis- filaments quickly depolymerize unless they are sion of developmentally contingent vectors of selectively stabilized by contacts or signals at the influence in all directions. Any particular trans- cellular periphery (Gerhart and Kirshner 1997). membrane cell-adhesion molecule will serve both Microtubule filament stabilization leads to cellu- selectively to stabilize bonds to other cells, or lar differentiation. Once a cell becomes spatially extracellular matrices, and to transmit signals to differentiated, it becomes itself an orientational within the cell, which feed into common phos- force within a morphogenetic field. The micro- phorylation pathways that influence the state of tubule system is in effect a subcellular module the cell with respect to cell division, selective tran- that participates in context formation and con- scriptional activiations or deactivations, secre- 96 Lenny Moss

tory activity, and more. Multicellular assemblages Dawkinsonian “evolutionary gene” are discussed and of self-organized, mutually selectively stabilized aptly criticized. However, in both cases the remedy that cells constitute morphogenetic fields. It is little they propose is to save the “gene” from the inadequacy surprise that many of the Genes-P that have been to which it would be condemned if understood as only implicated in tumor pathogenesis, as in the case some piece of DNA. While I am in sympathy with the diagnosis, I take issue with the proposed cure. of colorectal cancer, have led to the identification of Genes-D associated with cell-adhesion protein 3. This is a conservative estimate. There are more than a hundred possible splicing isoforms (Zorn and Krieg production. The loss of the stabilizing influence 1992), and twenty-seven alternatively spliced forms are, of a morphogenetic field could well result in a for example, reported to be expressed during rat heart cellular trajectory that, while adaptive for the development alone (Reyes, Small, and Akeson 1991). cell, becomes perilous for the organism (Moss in press). Modularity has arisen as a theme in subcellular References biology unexpectedly, first of all, as an explanan- Carlson, E. A. (1966). The Gene: A Critical History. dum not as an explanans. As investigators such as Philadelphia: Saunders. Gerhart and Kirshner and Gilbert, Opitz, and Chen, Y., C. F. Chen, D. J. Riley, et al. (1995). Raff (and many others) focus upon the meaning Aberrant subcellular Localization of BRCA1 in breast and significance of this, new and promising styles cancer. Science 270: 789–791. of explanation have begun to emerge. These Falk, R. (1991). The dominance of traits in genetic module-centered explanatory approaches ideally analysis. Journal of the History of Biology 24: 457–484. complement the intuitions and explanatory Falk, R. (1995). The struggle of genetics for indepen- strategies of DST and the Gene-P/Gene-D dis- dence. Journal of the History of Biology 28: 219–246. tinction that I have put forth. With Gene-P Gerhart, J., and M. Kirschner. (1997). Cells, Embryos, properly defined, demarcated and set aside to and Evolution: Toward a Cellular and Developmental play its own role as an instrumentally simplified Understanding of Phenotypic Variation and Evolution- predictor device, Gene-D need no longer be bur- ary Adaptability. Malden, MA: Blackwell Science. dened with responsibility for providing shortcuts Gilbert, W. (1978). Why genes in pieces? Nature 271: to the phenotype but may be left to simply play 501. the role of a resource designator. DST prescribes Gilbert, W., and M. Glynias. (1993). On the ancient an explanatory outlook that is open to all fields nature of introns. Gene 135: 137–144. and all directions of influence—choosing its level Godfrey-Smith, P. (1999). Genes and codes: Lessons of gaze as befits circumstance. For many devel- from the philosophy of mind? In V. Hardcastle (Ed.), opmental accounts, it now appears, a story about Biology Meets Psychology: Constraints, Conjectures, the life history of context-sensitive modular asso- Connnections. Cambridge, MA: MIT Press. ciations may prove to be most perspicuous. Gri≤ths, P., and E. Neumann-Held. (1999). The many faces of the gene. Biosciences 49: 656–662. Hunter, T. (1995). Protein kinases and phosphatases: Notes The yin and yang of protein phosphorylation and sig- nalling. Cell 80: 225–236. 1. The idea that the Mendelian gene constitutes an expression of “instrumental preformationism” was Irminger-Finger, I., B. Siegel, and W. Leung. (1999). introduced by Rafael Falk (1991, 1995). The functions of breast cancer susceptibility gene 1 (BRCA1) product and its associated proteins. Biol- 2. Neumann-Held’s position expressed in the current ogical Chemistry 380: 117–128. volume was adumbrated in Gri≤ths and Neumann- Held (1999). In this slightly earlier paper the limits Jilling, T., and K. Kirk. (1997). The biogenesis, tra≤c, of both the “classical molecular gene” and the and function of the cystic fibrosis transmembrane con- Deconstructing the Gene 97

ductance regulator. International Review of Cytology 172: 193–241. Johannsen, W. (1923). Some remarks about units in heredity. Hereditas 4: 133–141. Kerem, E., and B. Kerem. (1995). The relationship between genotype and phenotype in cystic ﬁbrosis. Current Opinion in Pulmonary Medicine 1: 450–456. Kirschner, M., and T. Mitchison. (1986). Beyond self- assembly: From microtubules to morphogenesis. Cell 45: 329–342. Kitcher, P. (in press). Battling the undead: How (and how not) to resist genetic determinism. In R. Singh, C. Krimbas, J. Beatty, and D. Paul (Eds.), Thinking about Evolution: Historical, Philosophical, and Political Per- spectives. Cambridge: Cambridge University Press. Miki, Y., J. Swenson, D. Shattuck-Eidens, et al. (1994). A strong candidate for the breast and ovarian cancer susceptibility gene BRCA1. Science 266: 66–71. Miklos, G., and G. Rubin. (1996). The role of the genome project in determining gene function: Insights from model organisms. Cell 86: 521–529. Mitchison, T., and M. Kirschner. (1984). Dynamic instability of microtubule growth. Nature 312: 237–242. Moss, L. (in press). What Genes Can’t Do: Prolegomena to a Philosophy beyond the Modern Synthesis. Cam- bridge, MA: MIT Press. Reyes, A., S. Small, and R. Akeson. (1991). At least 27 alternatively spliced forms of the neural cell adhesion molecule mRNA are expressed during rat heart development. Molecular Cell Biology 11: 1654–1661. Sterelny, K., K. C. Smith, and M. Dickison. (1996). The extended replicator. Biology and Philosophy 11: 377–403. Walsh, F., and P. Doherty. (1997). Neural cell adhesion molecules of the immunoglobulin superfamily: Role in axon growth and guidance. Annual Review of Cell and Developmental Biology 13: 425–456. Waters, C. K. (1994). Genes made molecular. Philosophy of Science 61: 163–185. Welsch, P., E. Schubert, and M. King. (1998). Inherited breast cancer: An emerging picture. Clinical Genetics 54: 447–458. Zorn, A., and P. Krieg. (1992). Developmental regulation of alternative splicing in the mRNA encoding Xenopus laevis neural cell adhesion molecule (NCAM). Developmental Biology 149: 197–205.

9 The Systems of Inheritance

Eva Jablonka

During the last two decades, most thinking about Development is something that happens to vehi- inheritance and evolution has been deeply in- cles (and is controlled by replicators) to ensure fluenced by what has been learned about the the further propagation of replicators. While molecular nature of the gene. The structural replicators are units of heritable variation, vehi- organization of the gene, the conditions for its cles are targets of selection. The generation of transmission, the way in which it is transmitted, new variant replicators is assumed to be inde- and the way it varies, have shaped the modern pendent of the selective environment (which acts view of heredity and have been very influential in on vehicles), and of the developmental process molding ideas about evolution. This influence that vehicles undergo. The replicator is clearly went beyond the strictly biological realm and very similar to the gene, the unit of Johannsen’s affected ideas about the evolution of culture. genotype, and it carries much of the latter’s bag- However, for a gene-like concept to be used in gage. explanations of nongenetic evolution, a more The view of inheritance embodied in the repli- general concept was necessary. Such a concept, cator concept affects the way in which evolution the “replicator,” was suggested by Dawkins is understood, and leads to a view of evolution (1976). The replicator was defined as “anything that reflects the modern neo-Darwinian version in the universe of which copies are made” of Darwin’s original selection theory. According (Dawkins 1982: 83). This definition seems, at first to Darwin’s theory, in a world in which there are sight, broad enough to accommodate different interacting entities with the properties of multi- types of heredity and reproduction, since “copy- plication, heredity, and heritable variation that ing” can be understood to include many types of affects the chances of multiplication, natural processes. However, as Dawkins, Hull, and many selection will necessarily occur, and in the long others made clear, the replicator entails a very term, adaptive evolution will follow (Maynard special kind of copying, which presupposes that Smith 1986). In this general form, Darwinian only instructions or representations (which is selection theory does not specify what the entities what replicators embody) rather than the imple- should be, how they multiply, how variations are mentations of representations, can be mean- inherited, or how they are generated. It also does ingfully “copied” or inherited. Following the not make a priori assumptions about the rela- distinction between genotype and phenotype, tionship between heredity and development. It is which was suggested by Johannsen at the turn the generality of Darwin’s selection theory that of the twentieth century and molded the theory gives it its great explanatory power and its poten- of the emerging discipline of genetics (Johannsen tial applicability to different domains of historical 1911), Dawkins suggested a distinction between change. replicators and vehicles. He defined the vehicle For Darwinian selection theory to be fruitfully as “any unit, discrete enough to seem worth nam- applied to a particular domain, its major con- ing, which houses a collection of replicators and cepts have to be specified for that domain. The which works as a unit for the preservation and replicator seems to fit particularly well the molec- propagation of those replicators” (Dawkins ular, neo-Darwinian version of Darwinism (or 1982: 114). The vehicle was called “interactor” by genic neo-Darwinism). According to genic neo- Hull, to emphasize its active functional role as a Darwinism, nucleic acids are the sole units of her- propagator of replicators (Hull 1980). Vehicles or itable variation, the transmission of these units is interactors are, of course, not only carriers of independent of their expression, and the genera- replicators, but they are also their products. tion of genetic variations is not adaptively guided 100 Eva Jablonka

by the selective environment or the developmen- I shall then discuss the transmission of parental tal history of the organism. and group legacies through niche construction, This replicator-centered, gene-derived view of and argue that it is the whole developmental sys- heredity is, however, not only severely limited, tem, with all its different and interacting inheri- but also severely misleading. There are multiple tance systems, that has to be considered when we inheritance systems, with several modes of trans- think about the transmission of variations from mission for each system, that have different prop- one generation to the next (Oyama 1985; Gri≤ths erties and that interact with each other. They and Gray 1994). This means that the replica- include the genetic inheritance system (GIS), cel- tor/vehicle dichotomy has to be discarded, and lular or epigenetic inheritance systems (EISs), the we must go back to a single (though complex) systems underlying the transmission of behavior minimal unit—a unit that is simultaneously a patterns in animal societies through social learn- unit of development, multiplication, and herita- ing (BISs), and the communication system em- ble variation—the reproducer (Griesemer 2000). ploying symbolical languages (SIS) (Jablonka, I start with the most fundamental and best Lamb, and Avital 1998). These systems all carry understood inheritance system in living organ- information, which I shall deﬁne here as the isms, the genetic inheritance system, which is transmissible organization of an actual or potential based on DNA replication. state of a system. In addition to the intrinsic properties of the different inheritance systems, the feedback loops The Genetic Inheritance System (GIS) formed between the organism’s activities and its ecological and social environment often create The information in the genetic inheritance system conditions for the reconstruction of ancestral is organized in the sequence of nucleotides in phenotypes in descendant generations. Devel- nucleic acids, which in most extant organ- opmental and ecological legacies may be said to isms is DNA. The gene is a template made up of be passed on between generations. Inheritance nucleotides whose sequential organization can systems with replicator-like properties are very be transformed through a complex process of unusual, and certainly do not represent or sum decoding into functional RNA and proteins. up the many ways in which heritable variations Genetic information is thus encoded. Encoding are transmitted across generations. I use “trans- means that one system of transmissible elements mission” in a general way, to denote all the (signs) represents not just itself, but also another processes leading to the regeneration of the same system of elements that combine to form the type of organization-states across generations. actual, functional, messages. In the GIS, nu- This includes the direct transfer of resources, as cleotide triplets in a structural gene are elements well as the activities that lead to the reconstruc- of the DNA system, and they represent amino tion of ancestral phenotypes. In what follows I acids in a protein, which is the functional “mes- shall discuss different inheritance systems and sage.” In natural language, utterances represent compare them with respect to those properties actual objects and events in the world, as well as that seem to me most pertinent to the under- other words and meaning-relations. standing of inheritance: the type of variation Information is also carried in DNA regions transmitted; whether or not information is that can control the decoding of other DNA encoded; the type of mechanism leading to the re- sequences. The noncoding but regulatory regions generation of variations in the next generation; in DNA cannot be said to encode information in the relationship between development and the the same sense as the coding regions. However, generation of new heritable variations (table 9.1). particular sequences (of varying length) are spread throughout the genome and perform The Systems of Inheritance 101

Niche construction—variant interaction of organism and environment can be transmitted Unlimited at the cell level Unlimited at the organism and lifestyle level some vertical Modular(methy-lation) Can be encoded Modular (methy- Mostly vertical Limited? lation) patterned (gene)metabolic cycle patterned patterned behaviorand content (patterned) (patterned) & holistic & holistic horizontal, horizontal Variation transmitted Information Mode of transmission Physiological Blind & Holistic Nonencoded Holistic Mostly vertical Limited (genetic) DNA sequence Blind & some Modular Encoded Modular Mostly vertical Unlimited (cellular) (behavioral) (symbolical) Symbolic form Learned Modular Encoded Modular Mostly Unlimited StructuralChromatin 3D markingcomplex Pattern of chro- Blind & Blind & mosome marks patterned Holistic Holistic Nonencoded Nonencoded Holistic Holistic Mostly vertical Mostly vertical Limited Limited Steady-state Activity state of Blind & Holistic Nonencoded Holistic Mostly verticalInducing Limited substanceNon-imitativesocial learning Pattern of Pattern of behaviorImitation behavior Often induced Often learned Pattern of (patterned) Holistic (patterned) Holistic Nonencoded Often learned Nonencoded Holistic Holistic Holistic Mostly vertical Vertical Nonencoded& Limited Holistic Limited Vertical & Limited horizontal Table 9.1 Types of information and modes of transmission systems for of different inheritance Inheritance GIS EIS Epigenetic BIS SIS System Unit Origin Alteration Encoding(organismal)Inducing substance Type state Direction patterned Type of Heredity 102 Eva Jablonka

sequence-typical regulatory functions, so general of the meiotic reshu≥ing of genes (in sexually types of functions can be inferred from sequence reproducing organisms), and of several classes of organization. Such regulatory sequences thus errors in DNA maintenance. Errors can be due to form a kind of higher order “code.” physico-chemical damage to the DNA, they can The organization of information in DNA is occur during DNA replication and repair, and modular (or digital), that is, it is decomposable they can result from the activity of genomic par- into separate discrete units drawn from a stan- asites: genetic elements that multiply excessively dard set, (the units in DNA are the nucleotides A, and move from site to site in the genome. Errors C, T, G), and the information is alterable digit by that are not removed or repaired accurately by digit. Following Szathmáry (1995), a replication the DNA maintenance machinery were assumed process that proceeds digit by digit will be called to be the ultimate raw material for evolution by modular replication. The genetic system is the natural selection. Although there is no doubt that prime example for a system that is modularly a lot of variation in DNA is indeed random in replicated. The enzymatic machinery that repli- this sense, the view that all variation is random cates the DNA, or that edits and repairs it, is has been challenged. largely indifferent to its sequence organization. The challenge has come from several direc- This means that a sequence that has beneficial tions. It has been shown that different nucleotide effects when decoded will be replicated and sequences differ in the likelihood that they will be repaired with the same fidelity as one with dele- damaged, invaded by genomic parasites or repli- terious effects, or a sequence that is completely cated inaccurately. The rate and type of new vari- nonfunctional. Furthermore, the transmissibility ation may thus depend on how the nucleotides in of the template remains unaltered following its the sequence are organized, and this organization replication. Usually transmission is vertical, from may be adaptive. For example, Moxon and his parents to offspring, but occasionally it can be colleagues have shown that in the pathogen horizontal, so genetic information can be trans- Haemophila influenza the genes that influence its mitted between nonrelated individuals, including antigenicity are highly mutable because the short individuals belonging to different species. tandem repeats in them make them prone to The modular nature of the replication and mutation by recombination and strand slippage. alteration of information allows for the inheri- The high mutation rate in these genes is advanta- tance of many combinations of modules—a geous, because it enables this pathogenic organ- DNA molecule with ten linearly linked nucle- ism to evade the immune system of the host otides has more than a million possible variant (Moxon et al. 1994). The sites in which mutations sequences. This means that the evolutionary po- preferentially occur are the result of adaptive tential of a modularly alterable and transmitted evolution. Moreover, mutation rate can increase unit, such as a gene that consists of hundreds of selectively not only at sites but also in conditions nucleotides, is very large. The number of possible in which a higher mutation rate is selectively sequences greatly exceeds the number of individ- beneficial. Wright (1997) has shown that amino uals in any realistic system. Such a system can be acid starvation in E. coli increases the transcrip- said to have unlimited heredity (Szathmáry and tion of genes that help the cells survive longer, Maynard Smith 1993; Maynard Smith and and concurrently increases the mutation rate in Szathmáry 1995). these genes. This condition-dependent increase in Until recently, the generation of variations in mutation rate is adaptive since such targeted DNA has been assumed to be random with mutation in the relevant genes may “rescue” the respect to the selecting environment. Variations cell without greatly increasing the load of muta- were assumed to be exclusively the consequence tion. It seems that through natural selection the The Systems of Inheritance 103

mechanisms that allow selective control of gene ment, and evolution. The generation of patterned expression have been coupled with mechanisms variation is part of the developmental process no that determine the fidelity of copying, so that the less than changes in transcriptional activation of inducible system that turns genes on and off also genes, although the effect of changes in DNA turns the production of mutations on and off. may often last longer than changes in trans- Such “targeted” mutations cannot be said criptional activity. The process of generating to be random in the classical sense, since adap- patterned variation is part of both develop- tively advantageous mutations are preferentially ment and evolution. Although there is a certain (though not exclusively) induced under the ap- (short-term) degree of autonomy of heredity propriate conditions and in the relevant domains. and development if mutations are random, if Randomness has not been eliminated, but it has they are patterned, heredity loses this partial been restricted and channeled. However, the mu- independence. tations are not goal-directed in any teleological sense, and their targeted production is the consequence of natural selection that had acted on ran- The Epigenetic Inheritance Systems (EISs) dom variations. Variation has been targeted by selection to be preferentially generated in a subset Epigenetic inheritance systems are the systems of sites, under particular conditions. It is di≤cult underlying cellular heredity. It is well known that to know how to define such variations. The term once cells become determined during develop- patterned variation, which has been suggested by ment, they often maintain their functional and the economist Ekkehart Schlicht with respect to structural characteristics through many cell divi- cultural evolution, is the one I choose to use in sions, even though the stimuli that first induced this paper (Schlicht 1997). It is better than previ- their determined state early in development were ously suggested terms such as directed, adaptive, transient, and are no longer present. Kidney cells induced, and guided variation because it does not and fibroblasts within the same organism have carry the teleological connotation of premedi- identical DNA base-sequences, yet each cell type tated design, yet does carry the connotation of “breeds true”: Kidney cells transmit their func- some degree of preexisting structuring (by past tional state to daughter kidney cells, while skin natural selection). Once a system for generating fibroblast cells transmit their very different cel- patterned variation has evolved, it channels and lular phenotype to their descendants. The me- guides evolution. chanisms that are responsible for this cellular From an evolutionary point of view the exis- inheritance have been termed epigenetic in- tence of a cellular system for the production of heritance systems. The transmission of heritable patterned variations makes good sense. It would epigenetic variations is possible not only within be remarkable if a cellular system for targeting individuals, but also between generations of in- the generation of variations had not evolved dur- dividuals, so EISs can have direct evolutionary ing the four billion years since life appeared on importance. earth. It is quite easy to see how the enzyma- Three types of epigenetic inheritance systems tic genetic engineering kit that all cells use to (EISs) have been described (Jablonka and Lamb rearrange, amplify, and delete pieces of their 1995). The first type of EIS is the steady-state DNA could have been modified by selection to system, which is based on the activity of self- allow the genome to respond to different reoccur- sustaining feedback loops. It was first described ring types of environmental stress (Shapiro 1997). theoretically by Wright (1945), and has been The ability to generate patterned variations found in many biological systems. In its simplest forges direct links between heredity, develop- form, a gene produces a product as a result of 104 Eva Jablonka

Figure 9.1 A steady-state system showing the perpetuation of an induced active state through cell division. (a) After induction, gene A is turned on and its product, a, positively regulates its own activity. The regulator a need not be a direct protein-product of gene A, but can be the metabolic product of the direct (protein) product, a small metabolite with regulatory function. The box shows the self-regulation of the genetic circuit. (b) The regulatory product a diffuses into the environment, enters into inactive cells, turns on gene A, and hence leads to the self-sustaining activity of the circuit in these cells. induction by an external developmental or envi- cally identical cells in the very same environment ronmental stimulus, and this product then stimu- may be heritably different because of the prior, lates further activity of the gene (through positive different, developmental history of their ancestor self-regulation) even when the original external cells. As long as the concentration of the prod- inducing stimulus has disappeared (ﬁgure 9.1). ucts of the self-sustaining cycle does not fall Once switched on, the cell lineage continues to below a critical threshold, the active, “on,” state produce the gene-product unless its concentra- is maintained following cell division; once the tion falls below some critical threshold value. concentration falls below the threshold, the cycle Two genetically identical cells can therefore be in is in the “off” state, which is also maintained. The two alternative states (“on” and “off”), and both states of activity and inactivity are reproduced in states can be self-perpetuating, even when the daughter cells as an automatic consequence of inducing environment changes. Thus two geneti- cell division, and transmission is an integral part The Systems of Inheritance 105

of growth and multiplication. The generation of Thus, at the level of the cell, the inheritance of the activity state is part of development, yet the functional states may be practically unlimited, developmental states can be faithfully transmitted and cumulative evolutionary change may occur. within the cell lineage for many generations. In this case, of course, many of the variations are The information reproduced in this type of clearly induced by the environment (although system is nonmodular or holistic (here I follow random environmental fluctuations may also the distinctions, but not the terminology, of generate some variants). The environment both Szathmáry 1995). Although the cycle can be induces a set of different adaptive variant states divided into discrete modules (modular gene A, and fine-tunes the adaptation by selecting the modular product a, modular regulatory domain), most appropriate ones. In this inheritance system the functional state cannot be transmitted module both the reproduction of the activity states in by module. It can only be transmitted when the daughter cells and the generation of variations processes of interactions among components are are part of the cell’s development, and it is the regenerated in the daughter cells. Changes in any phenotype (a dynamic activity state, a process) one component usually prevent the transmission that is reproduced. of the whole cycle. It is only the state of activity The second EIS is that of structural inheri- of the whole cycle that can vary. However, cellu- tance, where existing cell structures are used to lar states may also be transmitted horizontally, guide, or template, the formation of new simi- between lineages. If the positively regulating lar structures. Variant complexes or architectures product is not only transferred to daughters cells made up of the same components can be stably as an inevitable part of cell division, but also dif- inherited. Inheritance is through some kind of fuses to the cell’s environment, it may “infect” three-dimensional templating, with existing neighboring cells from another lineage and in- structural patterns facilitating the construction of duce its own activity state in them. Rather than similar “daughter” patterns. For example, in cili- inheriting the phenotype through descent, the ates, genetically identical cells can have different nondescendant cells are interacting with the envi- patterns of cilia on their cell surfaces, and these ronment that the “infecting” cells have modi- different patterns are inherited. Prions seem to fied and become phenotypically identical to them be another example of such structural inheri- through this interaction (figure 9.1b). tance (Grimes and Aufderheide 1991, Tuite and Often each individual self-sustaining cycle can Lindquist 1996). In this structural inheritance have only two states (“on” or “off”), and the sys- system there are clear modules (the modular tem can move only between two states, so noth- components of the complex), but transmission is ing evolutionarily very interesting can occur at holistic: The complex is not transmitted module this level. The number of variant, functional, and by module, nor are the modules alterable unit by heritable states that every single cycle can show unit. The structural information may be trans- is very small, much smaller than the number mitted by the fragmentation of the original com- of individuals the population can include. The plex, followed by growth, as in a crystal, or by system therefore can be said to show limited other means where the interacting units within heredity. However, within a cell there are often the complex form the conditions for self- several independent cycles. More than a million organization of free floating units. There is no variant cell states are possible if a cell has general, autonomous system of transmission in- twenty different cycles! New developmental con- dependent of the structural properties of the par- ditions can induce changes in the activity states of ticular complex. The reliability of transmission several cycles in cells, producing many variant will be specific to each structural complex and states, which can then be subject to selection. depend on its unique properties. Variations in the 106 Eva Jablonka

organization of the units into self-perpetuating on the coding properties of the triplet in which complex-variants can be affected by environmen- it participates, but can affect transcriptional acti- tal conditions, so variations are often patterned vation in the chromosomal region in which it (figure 9.2). As with the steady-state EIS, struc- occurs. With this EIS there is a dedicated, tures are likely to be passed on vertically, by function-independent, copying machinery (the descent. However, horizontal transmission is also enzyme methyl-transferase) that can copy pat- possible, as testify some prion diseases where the terns of methylation irrespective of their past or pathogenic prions are transmitted to nonrelatives present function. Information is organized in a and even to individuals of other species. The modular way (a nucleotide can be in two states— number of heritable states of each complex may methylated or nonmethylated), so that methyla- be very limited, but in a cell with tens of com- tion sites are alterable unit by unit, and transmis- plexes, there are practically unlimited heritable sion proceeds module by module (figure 9.3). architectural states. “Copying” of complexes is However, the reproduction of methylation pat- part of development and multiplication; there is terns is not always modular and does not always no specialized machinery that can copy different depend on the special enzymatic machinery. architectures. Variation, when patterned, is both Methylation patterns can be transmitted sexually developmental and evolutionary. between organisms through sperms and eggs. As In the third EIS, the chromatin-marking EIS, the germline becomes determined, there are wide- states of chromatin that affect gene expression spread and sometimes radical changes in chro- are clonally inherited. Genetically identical cells mosome marks, including patterns of cytosine can have variant and heritable chromatin marks. methylation on chromosomes. However, paren- Marks are protein or RNA complexes associated tal patterns of methylation can still be regener- with DNA, or small chemical groups, such as ated in the offspring because some traces of the methyl groups, that bind to certain nucleotides. past are retained, as partial (protein or methyla- The type, the density, and the pattern of marks tion) marks, and these partial traces or “foot- on a chromosome region affect its potential tran- prints” are reconstituted into full marks during scriptional state, and changes in marks can be the embryogenesis of the offspring. There seems induced by the change in the environment. When to be a cycle of changes in chromosomal marks the marks are protein complexes, their reproduc- during germline formation and during early tion in daughter cells is probably similar to the embryonic development that leads to the recon- reproduction of other three-dimensional com- stitution of parental methylation marks (figure plexes, although the DNA sequence to which 9.4). Because changes in methylation marks, like protein marks bind may constrain variation and changes in other types of heritable chromosome enhance the fidelity of reproduction. However, marks, can be induced by the environment and the best-understood chromatin marking EIS, the the variation can be inherited, some of the varia- methylation-marking EIS, is somewhat unusual tion is patterned. Heredity in this system is un- in its modular organization and mode of trans- limited when we consider the whole genome or mission. Nucleotides in many organisms can be several large chromosomal domains, but limited in a methylated or nonmethylated state, and the when a short DNA sequence is considered. alternative states can be clonally inherited. The Unlike the GIS, with all EISs the generation of most commonly methylated nucleotide is cyto- new variation is typically patterned (although it sine, and in most eukaryotes it is the cytosine in can also be completely accidental), and cannot be CpG doublets or CpNpG triplets that can be in divorced from the physiological development of an either methylated or nonmethylated state. The the cell as it interacts with the environment. In methylated state of the nucleotide has no effect most cases, the transmission of information is The Systems of Inheritance 107

Figure 9.2 The perpetuation of two alternative organizations of ciliary structures in Paramecium. (a) The organization and perpetuation of normal ciliary rows through cell division (horizontal line). (b) The perpetuation of an experimentally inverted ciliary row. 108 Eva Jablonka

Figure 9.3 The inheritance of alternative patterns, (a) and (b), of DNA methylation. The black dots represent methyl groups. The dotted boxes show hemimethylated sites following the replication of DNA. These sites are preferential targets for a methylating enzyme, which methylates the opposite nonmethylated site in the DNA duplex. Different methylation patterns can therefore be perpetuated through cell division. holistic. The processes that allow the faithful transmitted unit itself. Of course, when we are transmission of variant functional or struc- looking at the functioning of the cell, the different tural states in the cell lineage do not utilize a dedi- inheritance systems interact and cannot be treat- cated, specialized, function-independent copying ed as autonomous: For example, products of a machinery (with the exception of the methylation steady state EIS can affect heritable chromatin EIS in somatic cells). Instead, these processes are marks and 3D structures, and vice versa. by-products of general growth and multiplication If we move from the level of the cell to the level processes. The ﬁdelity of reproduction depends of the multicellular organism, there is ample evi- on the speciﬁcs of the cycle, or the three- dence showing that the cells that begin new dimensional structure of the complex. At the cel- organisms, the egg and the sperm, can carry epi- lular level heredity is unlimited, although it may genetic information, and that variations in epige- be very limited at the level of the functional, netic information are often inherited (Jablonka The Systems of Inheritance 109

distinctive reproductive pattern. Without any genetic differences, two maternal lineages may differ, consistently, over many generations, in the sex ratio of the offspring they produce. Another example of phenotypic transmission at the organismal level is the transmission of microorganisms between generations through feces. Young of many species of mammals consistently eat their own and other individuals’ feces, a habit known as coprophagy. Most of the Figure 9.4 A normal cycle of changes in chromatin marks (e.g., mammals that practice coprophagy are herbi- methylation marks) during gametogenesis and early vores, who consume cellulose-rich plant material embryogenesis. As germ cells proceed through gameto- and have a dense symbiotic bacterial and proto- genesis, the chromatin marks on DNA sequence A zoan gut flora that helps them to break down and change (from m3 to m1). In the zygote, the mark on A digest cellulose. The young of many herbivorous is changed to m2, and during early embryogenesis to species eat their mother’s feces, and in this way m5. When segregation of soma and germ line occurs, they directly inoculate their own guts with the some cells with m5 marks become germline cells and maternal flora of useful microorganisms. Differ- again acquire mark m3. How an induced change in ences between the gut floras of different mothers, marks may alter the cycle is not shown here (for a figure will be transferred to their offspring, and may be and discussion of self-perpetuating cycles of changed marks, see Jablonka and Lamb 1995: 154–156). perpetuated for many generations. In many cases, these parental legacies affect behavior. and Lamb 1995, 1998). There is also another type The Regeneration of Behavior: The Behavioral of phenotypic information transfer between gen- Inheritance Systems (BISs) erations, which is more di≤cult to categorize because it does not occur at the cellular level, but Behavior that can be transmitted has been cate- at the level of the whole organism. The mater- gorized in many different ways. With social nal environment in which the mammalian fetus learning alone, more than thirty terms and dis- develops sometimes has effects that can be carried tinctions have been suggested. For the purpose over to later generations. For example, if female of this essay, which concentrates on the type and Mongolian gerbil embryos develop in a uterine transmission of information, I will distinguish environment in which most of the embryos are three general types of transfer of behavioral male, and they are therefore exposed to high level information. of testosterone, they mature late, are more terri- The first is very similar to the whole-organism torial than other females and, in turn, produce transgenerational reproduction of phenotypes litters with a greater proportion of males than discussed in the last section, but focuses on the the normal 1 : 1 sex ratio. The result is that reproduction of behavior. In this system the pro- their daughters, who usually also develop in a cesses that lead to similarity between the behav- testosterone-rich uterine environment, also ma- iours depend on the transfer of behavior-affecting ture late, and produce mainly male offspring substances between interacting individuals. I (Clark, Karpiuk, and Galef 1993; Clark and therefore call this type of transfer the inducing- Galef 1995). The developmental legacy of the substance transfer. Unlike the other two BISs, mother is transferred to her daughters, so there is transmission is not dependent on learning. a nongenetic transmission and repetition of this 110 Eva Jablonka

The transmission of food preferences via the transfer of substances through the placenta and the milk in mammals is a good example of this type of BIS. Mammal fetuses are able to smell semivolatile liquids transferred to them across the mother’s placenta, and later show preference or aversion for food items carrying these smells (Smotheran 1982; Hepper 1988). Transmission of substances through milk has similar effects. The results of cross-fostering and other simple experiments with mice have shown that the food the mother prefers, and therefore frequently eats, biases the food preferences of the young so that those who feed on her milk tend to have the same preferences. Such results are typical for many mammals, including other rodents and ruminants (Galef and Sherry 1973; Provenza and Balf 1987). There are other channels through which Figure 9.5 inducing-substances that bias behavioral prefer- A blue tit opening a milk bottle by tearing the foil cap ences can be transferred (Avital and Jablonka (from Hinde 1982). 2000). Information transferred in inducing substances is not encoded, and its transmission is snakes. However, what they learn is not the holistic. Usually (but certainly not always), trans- motor flight behavior patterns of the experienced mission is vertical. The variation generated is com- adults, but rather that snakes have to be avoided. monly patterned (induced), and heredity is limited, The second example is the cultural spread of the although the number of variant preferences and blue tit’s and the great tit’s habit of opening milk the behaviors they influence may be quite large. bottle tops, a famous case of “cultural” transmis- The second type of transfer of behavioral in- sion of behavior (Fisher and Hinde 1949). Tits formation occurs through nonimitative social learnt by observation the habit of removing the learning. This has received a great deal of atten- cap and getting at the cream at the top of the bot- tion from experimental psychologists who differ- tles (figure 9.5). This is probably another case of entiate between several different types of social non-imitative social learning. The spread of the learning that do not involve imitation and/or behavior from experienced to naive tits can be direct instruction (Zentall and Galef 1988; Heyes explained as the result of naive tits having their 1994). I call this type of social learning nonimita- attention drawn to the milk bottle as a source of tive social learning. In the cases covered by this food, commonly through the behavior of an category of social learning, the naive, observing experienced individual (Sherry and Galef 1984). individual (or “observer”) learns about the envi- The method by which the top was removed was ronmental circumstances (including the objects, not imitated—each individual tit focused its stimuli, and events) that elicit a particular behav- attention on the milk bottle as a potential source ior in the experienced individual. Two examples of food and, after it own trial-and-error learning, will help to illustrate the nature of such social finally learned how to remove the top in its own mediation. When young monkeys become fearful style. Such social mediation leads, in most cases, of snakes after observing the panic-stricken re- to similarity between the behaviors of the action of adults to snakes, they too will avoid “observer” and the “model.” The model guides The Systems of Inheritance 111

or enhances the attention of the observer to the types of social learning, patterned and nonen- environmental stimulus (such as a milk bottle, or coded, and is transmitted both vertically and ho- a dangerous predator), which elicits a similar rizontally. Heredity is often limited although, in emotional and behavioural response to its own. theory, if a behavioral act is made of many indi- In this type of behavioral inheritance, informa- vidually alterable and transmissible modules (for tion is not encoded. Variation is generated by the example, if the song of a songbird is made of inventor of the new behavior through asocial many types of phrases), heredity may be unlim- learning. It is therefore patterned rather than ited. However, a huge number of combinations accidental. It is holistically transmitted through leads to a huge amount of nonsense-messages— social learning, and can be transferred both verti- to functionally useless or even positively harmful cally and horizontally. Heredity is rather limited, information. Only if there is some patterning or since the number of variants the behavior pattern ordering of the combinatorial process can the can assume may be restricted. However, at the search for functional sequences in the infinite level of the overall lifestyle, heredity may be prac- space of possibilities yield functionally meaning- tically unlimited, since different variant patterns ful results (Schlicht 1998). It is only when there is of behavior may combine to form many types of a reasonable probability that variant behavioral lifestyle. modules combine to form different yet functional The third type of BIS is learning by imitation sequences of behavior that the modular transmis- and/or instruction. I consider it to be another sion of sequences opens up truly wide evolution- type of BIS because of its modular way of trans- ary possibilities. Rule-bound organization and mission. As Heyes (1993) has argued, there is no transfer of information is clearly necessary. We compelling evidence to suggest that imitation is see this kind of rule-bound organization in sys- inherently more cognitively demanding than sev- tems of symbolic communication. eral other types of social learning. However, the modular way of transmitting and altering behavior during imitation or instruction—the parsing Symbolical Systems of Inheritance (SIS) of behavioral acts—sets it apart from other types of social learning. During imitation, the naive As T. W. Deacon stresses in his 1997 book on the individual reproduces not only similar responses evolution of language, symbols are not simple. to the environment, but also the model’s actual The American philosopher Charles Peirce dis- acts. Vocal imitation is very common among tinguished between three ways in which a sign many species of bird, whereas motor imitation (defined as information communicated between has been validated beyond reasonable doubt in sender and receiver) can refer to something. First, only a few species. Humans, chimpanzees, dol- a sign can refer to an object by resembling it. This phins, budgerigars, rats, and a few other birds type of sign is called an icon, and an example is and mammals have been shown to be able to imi- a picture of a house, which refers to an actual tate motor acts. However, because relatively few house, or the pattern on a mimetic butterfly’s experiments have been designed to distinguish wings, which resembles (and can be said to refer between imitative and nonimitative learning of to) the pattern on a model’s wings. Second, a sign motor acts, the extent of motor imitation may be can be an index and refer to an object by associ- underestimated. Intentional instruction seems to ation, through being linked to the object in space be very rare in the animal world, but again, this or time. For example, the size and brightness of a issue has not yet been systematically studied. male peacock’s tail is an index of its health and The information acquired during imitation vigor. Finally, a sign can refer to an object by and nonsymbolical instruction is, as with other convention, or according to a reference-rule that enables it to refer to other signs in the system. 112 Eva Jablonka

Such signs are symbols. Symbols must represent Table 9.1 summarizes the different proper- objects, operations, and relations among signs (as ties of the four types of inheritance systems and in natural language and, in the purest way, in allows a comparison among them. What is clear mathematical notations). The category to which is that a system based on encoded information, a sign belongs depends on the interpretive system modular transmission, and modular alteration of of which it is part, rather than on the isolated the composing modules is a very special type of sign; a portrait, for example, though iconic, is inheritance system. The two inheritance systems also a part of a symbolical system, and should that have these properties and are closest to each therefore be interpreted as a symbol. Natural other in this respect are the GIS and the SIS. human language is another example of a symbol- However, the SIS is evolutionarily derived from ical system. In the sentence I am writing just now, the BISs, and it shares important characteristics most words refer to other words rather than to with them. It is nevertheless significant that both objects in the world. the GIS and the SIS have unlimited heredity at From the point of view adopted in this chap- the level of the transmitted units themselves, and ter, symbolical systems are transmitted by social not, as with other inheritance systems, only at learning, which often involves imitation and a a higher level of organization. Because of the abi- greater or lesser degree of intentional instruction. lity to encode information, both the GIS and Symbols are transmitted both modularly and SIS transmit a lot of unexpressed information. holistically. For example, in the case of natural Nonfunctional genes are transmitted, as also are language, the narrative, the sentence, the word, nonimplemented ideas. This provides a huge re- the phoneme are all transmitted, but it is quite servoir of variation, which may become useful in clear that a spoken narrative is (unless a story is new conditions. I believe that this ever-present learned by heart) more holistically transmitted potential gives these systems a particularly im- than a single new word. Interpretation depends portant role in long-term evolution. However, no on the rules of the system (for example, gram- inheritance system acts in isolation: inheritance matical rules), so symbolical systems are or- systems interact both directly and indirectly. For ganized by those rules. Sometimes, as in natural example, the social animal, with its BISs, deter- language or mathematics, the organization is mines the selective regime in which genes are ulti- easily formalized (rules of language-specific mately selected. grammar, mathematical axioms), but it can be Another point suggested by the table is that more fuzzy (as in dancing, music, and the visual by considering a higher level of organization, lim- and motor arts). Information is (by definition) ited inheritance systems may become unlimited. encoded and is almost invariably transmitted Hence we see that EISs are limited inheritance horizontally. Vertical transmission is common in systems at the level of the unit of transmitted some systems, however. For example, early lan- information (cycle of activity, 3D complex, local guage learning usually involves vertical parent/ pattern of marks), but may be unlimited at the offspring interactions. In other cases, such as the level of the cell phenotype. A practically unlim- transmission of painting skills, it is almost always ited number of cell phenotypes can be gener- nonvertical from master to student. Symbolical ated. The same is true of BISs—at the level of a systems have unlimited heredity and huge evolu- single behavior pattern there may be few vari- tionary potential. The rules of symbolical systems ants, but the lifestyle as a whole can display many organize the systems and order them, so variation more variations. Although biological informa- is inherently constrained and patterned by these tion at the lower level is holistically organized, at internal rules. New variations arise as a result of the higher level each state is treated as a module insight, trial-and-error-learning and accident. The Systems of Inheritance 113

that can combine with others and produce prac- tions in which they have lived will be regenerated tically unlimited variation. and reexperienced by their descendants (Odling- It seems that as a system becomes more func- Smee 1988, 1995; Odling-Smee, Laland, and tionally cohesive during evolution, evolving Feldman 1996). For example, males of some repair and compensatory mechanisms, its hered- species of bowerbirds build small huts to attract ity becomes increasingly more limited. There is females, bringing fruits, seeds, and fungi to de- less selectable variation, and the result may be corate them. These decorations are often able evolutionary stasis. There are two situations in to grow, so by their behavior bowerbirds also which escape from such stasis is possible. One ensure the long term supply of the materials occurs when selection acts at higher level of which they, and their descendants, will choose biological organization (at the level of many as decorations (Diamond 1986, 1987, 1988). combining units), that is, when a higher level Caching seeds is another example of a habit that of individuality emerges (Jablonka 1994; Jab- may be reinforced through the effect it has on the lonka and Lamb 1995). The second occurs when local environment. By caching seeds, animals a system of encoding the information evolves. provide themselves with a source of food for Both situations have occurred during evolution- harsh winters, but because some of the cached ary history. seeds germinate, caching also provides new plants that will form seeds and create future caching opportunities (Källander and Smith The Transmission of Organism/Environment 1990; Smith and Reichman 1984). Variations: Niche Construction and Niche Even more obvious examples of niche con- Regeneration struction are the propagation of dialects in bird or whale groups, where the dialect of the previous The right-hand side of table 9.1 shows that generation is the condition for the acquisition of organisms often transfer variations in their epige- this dialect by the younger generation. Similarly, netic characteristics or their behavior patterns in learning to speak by human children is guar- an indirect way. By providing their descendants anteed by the child’s developing in a preexisting with the initial conditions that allow the repeti- linguistic community. Such ecological or social tion of their own developmental processes, sim- niche construction ensures that the ecological ilarity between generations is enhanced. Both and social milieu is transmitted. The conditions Waddington (1959) and Lewontin (1983) stressed eliciting the ancestral behavior are reconstructed, that living organisms are not passive entities, but and selection for the maintenance of the behavior ones that actively choose and construct their en- pattern that ﬁts the constructed niche occurs. vironment, and hence also the selective regime in The regeneration of ancestral niches and selec- which they live and in which they breed. The tive regimes can occur at different levels of bio- most obvious examples are the nests of birds and logical organization. At the cellular level, we saw the dams of beavers. Such artifacts are often also that when the regulatory product of a steady- passed on to the next generation. state cycle can diffuse into the environment it Odling-Smee developed these ideas further, changes it, thereby creating the conditions that stressing the multigenerational transfer of many induce a cycle of self-perpetuating activity in types of variations in niches. He argued that neighboring cells. This is a simple form of niche because through their activity and behavior construction. All types of niche construction de- organisms construct the ecological and social pend on the formation of self-sustaining feed- niche that they occupy, this “niche construction” back loops between the developing organism and may often ensure that the environmental condi- its niche. 114 Eva Jablonka

proach, for the reproducer is simultaneously a A Different Kind of Darwinism unit of development, of multiplication, and of heritable variation, as well as a target of selection. The diversity of inheritance systems that are able The focus on units of reproduction introduces to transmit variation at different levels of biolog- back into evolution the developing individual as ical and social organization should surely prevent an active evolutionary agent. This leads to the developmental and evolutionary biologists from consideration of the different types of develop- interpreting development and evolution in terms mental processes that lead to the regeneration of genetic variation alone. Yet, not only are other and reproduction of variant characters. It in- sources of heritable variation neglected in gene- evitably leads to concurrent attention to selection centered accounts, but also the whole dynamics at different levels of organization—the gene level, of inheritance, which is an aspect of the develop- the cell level, the organism level, and so on, and mental process, is ignored. This leads to a very to different types of heritable variation—the faulty account and understanding of develop- genetic, the epigenetic, the behavioral, and the ment and of evolution, and completely misses the symbolical. It is this richer version of Darwinian complexity, possibilities, and limitations of devel- theory that needs to be adopted. opmental and evolutionary processes. Moving from the gene to the more abstract replicator, and assuming that the replicator is the References unit of variation and evolution, is also not satisfactory. The replicator/vehicle dichotomy, which Avital, E., and E. Jablonka. (2000). Animal Traditions: is fundamental to the concept of a replicator, is Behavioural Inheritance in Evolution. Cambridge: Cam- meaningless in all cases in which the transmission bridge University Press. of information or the generation of new heritable Clark, M. M., and B. G. Galef. (1995). Parental information depends on development. Yet, as influence on reproductive life history strategies. Trends in Ecology and Evolution 10: 151–153. table 9.1 illustrates, this is the usual case. The replicator-vehicle distinction cannot therefore be Clark, M. M., P. Karpiuk, and B. G. Galef. (1993). Hormonally mediated inheritance of acquired charac- used to analyze heredity, development, or evolu- teristics in Mongolian gerbils. Nature 364: 712–716. tion. However profitable the distinction between Dawkins, R. (1976). The Selfish Gene. Oxford: Oxford replicator and vehicle may be for some evolu- University Press. tionary theorizing, this distinction simply does Dawkins, R. (1982). The Extended Phenotype. Oxford: not apply to real organisms. Freeman. At the beginning of this chapter I suggested Deacon, T. W. (1997). The Symbolic Species. New that in order to have a unifying concept of hered- York: W. W. Norton. ity that encompasses all the types of inheritance Diamond, J. (1986). Biology of birds of paradise and system, we need a theoretical framework that is bowerbirds. Annual Review of Ecology and Systematics broader than that used by genic neo-Darwinism. 17: 17–37. The developmental system approach suggested Diamond, J. (1987). Bower building and decoration by by Oyama (1985) and Gri≤ths and Gray (1994) the bowerbird Amblyornis inornatus. Ethology 74: 177– provides such a framework, as it focuses on the 204. developing and interacting individual, with the Diamond, J. (1988). Experimental study of bowerbird multiplicity of its inheritance systems and self- decoration by the bowerbird Amblyornis inornatus perpetuating feedback loops. The reproducer using colored poker chips. American Naturalist 131: concept suggested by James Griesemer (2000) 631–653. provides the unit of analysis for such an ap- Fisher, J., and R. A. Hinde. (1949). The opening of milk bottles by birds. British Birds 42: 347–359. The Systems of Inheritance 115

Galef, B. G., and D. F. Sherry. (1973). Mother’s milk: Lewontin, R. (1978). Adaptation. Scientific American A medium for transmission of cues reflecting the 239(3): 156–169. flavour of mother’s diet. Journal of Comparative Maynard Smith, J. (1986). The Problems of Biology. Physiological Psychology 83: 374–378. Oxford: Oxford University Press. Griesemer, J. (2000) Reproduction and the reduction Maynard Smith, J., and E. Szathmáry. (1995). The of genetics. In P. Beurton, R. Falk, and H-J. Major Transitions in Evolution. Oxford: Freeman. The Concept of the Gene in Rheinberger (Eds.), Moxon, E. R., P. B. Rainey, M. A. Nowak, and R. E. Development and Evolution . Cambridge: Cambridge Lenski. (1994). Adaptive evolution of highly mutable University Press. loci in pathogenic bacteria. Current Biology 4: 24–33. Gri≤ths, P., and R. D. Gray. (1994). Developmental Odling-Smee, F. J. (1988). Niche constructing pheno- Journal of Phil- systems and evolutionary explanations. types. In H. C. Plotkin (Ed.), The Role of Behavior in osophy 91: 277–304. Evolution, pp. 73–132. Cambridge, MA: MIT Press. Cel- Grimes, G. W., and K. J. Aufderheide. (1991). Odling Smee, J. (1995). Biological evolution and cul- lular Aspects of Pattern Formation: The Problem of tural change. In E. Jones and V. Reynolds (Eds.), Assembly. Basel: Krager. Survival and Religion: Biological Evolution and Cultural Hepper, P. G. (1988). Adaptive fetal learning: Prenatal Change, pp. 1–43. New York: John Wiley & Sons. Animal exposure to garlic affects postnatal preferences. Odling-Smee, F. J., K. N. Laland, and M. W. Feldman. Behaviour 36: 935–936. (1996). Niche construction. American Naturalist 147: Heyes, C. M. (1993). Imitation, culture and cognition. 641–648. Animal Behaviour 46: 999–1010. Oyama, S. (1985). The Ontogeny of Information: Heyes, C. M. (1994). Social learning in animals: Developmental Systems and Evolution. Cambridge: Categories and mechanisms. Biological Review 69: Cambridge University Press. (2nd rev. ed., Durham, 207–231. NC: Duke University Press, 2000.) Hinde, R. A. (1982). Ethology: Its Nature and Relations Provenza, F. D., and D. F. Balf. (1987). Diet learning with Other Sciences. New York: Oxford University by domestic ruminants: Theory, evidence and practical Press. implications. Applied Animal Behavioural Science 18: Hull, D. L. (1980). Individuality and selection. Annual 211–232. Review of Ecology and Systematics 11: 311–332. Schlicht, E. (1997). “Patterned variation”: The role of Jablonka, E. (1994). Inheritance systems and the evolu- psychological dispositions in social and economic evo- tion of new levels of individuality. Journal of Theore- lution. Journal of Institutional and Theoretical Econom- tical Biology 170: 301–309. ics 153(4): 722–736. Jablonka, E., and M. J. Lamb. (1995). Epigenetic Schlicht, E. (1998). On Custom in the Economy. New Inheritance and Evolution: The Lamarckian Dimension. York: Oxford University Press. Oxford: Oxford University Press. Shapiro, J. A. (1997). Genome organization, natural Jablonka, E., and M. J. Lamb. (1998). Epigenetic inher- genetic engineering and adaptive mutation. Trends in itance in evolution. Journal of Evolutionary Biology 11: Genetics 13: 98–104. 159–183. Sherry, D. F., and B. G. Galef. (1984). Cultural trans- Jablonka, E., M. J. Lamb, and E. Avital. (1998). mission without imitation: Milk bottle opening by “Lamarckian” mechanisms in Darwinian evolution. birds. Animal Behaviour 32: 937–938. Trends in Ecology and Evolution 13: 206–210. Smith, C. C., and O. J. Reichman. (1984). The evolu- Johannsen, W. (1911). The genotype conception of tion of food caching by birds and mammals. Annual heredity. American Naturalist 45: 129–159. Review of Ecology and Systematics 15: 329–335. Källander, H., and H. G. Smith. (1990). Food storing Smotheran, W. P. (1982). Odor aversion learning by the in birds: An evolutionary perspective. In D. M. Power rat fetus. Physiology of Behavior 29: 769–771. (Ed.), Current Ornithology vol. 7, pp. 147–207. New Szathmáry, E. (1995). A classification of replicator and York: Plenum Press. lambda-calculus models of biological organization. 116 Eva Jablonka

Proceedings of the Royal Society of London, Series B 260: 279–286. Szathmáry, E., and J. Maynard Smith. (1993). The origin of genetic systems. Abstracta Botanica (Budapest) 17: 197–206. Tuite, M. F., and S. L. Lindquist. (1996). Maintenance and inheritance of yeast prions. Trends in Genetics 12: 467–471. Waddington, C. H. (1959). Evolutionary systems: Animal and human. Nature 183: 1634–1638. Wright, B. E. (1997). Does selective gene activation direct evolution? FEBS Letters 402: 4–8. Wright, S. (1945). Genes as physiological agents: General considerations. American Naturalist 74: 109– 124. Zentall, T. R., and B. G. Galef. (Eds.), (1988). Social Learning: Psychological and Biological Perspectives. Mahwah, NJ: Lawrence Earlbaum. Niche Construction, Ecological Inheritance, and Cycles of Contingency in 10 Evolution

Kevin N. Laland, F. John Odling-Smee, and Marcus W. Feldman

A recurrent theme of this book is the rejection of vironments (such as nests, holes, burrows, paths, dichotomous thinking characterized by emphasis webs, pupal cases, dams, and chemical environ- on processes that are regarded as either inter- ments), and destroy other components (Lewontin nal or external to living organisms. As Lewontin 1983; Odling-Smee 1988). In addition, many (1983) has pointed out, the tendency to think organisms choose, protect, and provision “nur- dichotomously is not conﬁned to developmental sery” environments for their offspring. On the biologists. Evolutionary biologists can also slip basis of this kind of evidence, Lewontin (1982, into a dichotomous mode of reasoning. One of 1983) has argued that the “metaphor of adapta- the principal dichotomies in evolutionary theory, tion” should be replaced by a “metaphor of con- to which Lewontin draws attention, stems from struction” (see also Gray 1988). We have sought the separation of the causes of ontogenetic vari- to build on Lewontin’s writings by exploring the ation, seen as coming from internal factors, es- consequences of these constructive processes, pecially Mendelian genetics, and the causes of which we have collectively termed niche construc- phylogenetic variation, seen as something that tion (Odling-Smee 1988; Odling-Smee, Laland, is imposed by natural selection pressures aris- and Feldman 1996). We argue that through niche ing from autonomous external environments construction organisms not only shape the nature (Lewontin 1983). In this chapter we shall focus of their world, but also in part determine the primarily on the second half of this dichotomy. selection pressures to which they and their de- We want to reconsider the extent to which Dar- scendants are exposed. Other authors in this vol- winian natural selection pressures in external en- ume have pursued similar themes (Gray 1988; vironments really are autonomous, that is, they Gri≤ths and Gray 1994). really are independent of the organisms they Niche construction is not the exclusive pre- select, and we shall suggest that often they are rogative of large populations, keystone species not. We shall also propose evolutionary pro- or clever animals; it is a fact of life. All living cesses that can cause naturally selected organisms organisms take in materials for growth and main- to modify their own natural selection, as well as tenance, and excrete waste products. It follows the selection of other organisms. that, merely by existing, organisms must change Classically, adaptation has been conceived of their local environments to some degree. In spite as a process by which natural selection, stemming of its universality, niche construction is virtually from an external and independent environment, never incorporated into evolutionary accounts. gradually molds organisms to ﬁt an established Niche construction is too obvious and ubiquitous environmental “template.” The environment is for biologists to be unaware of it existence. If evo- seen as posing problems, and those organisms lutionary biologists currently neglect niche con- best equipped to deal with the problems leave the struction, it is unlikely to be because they dispute most offspring (Lewontin 1982, 1983). Although that it occurs. We suspect that most evolutionary the environmental template may be dynamic, in biologists feel that they can afford to neglect the the sense that processes independent of the or- effects of niche construction, because these effects ganism may change the world to which the pop- are regarded as either trivial, inconsequential, not ulation adapts (Van Valen 1973), the changes liable to change the nature of the evolutionary that organisms themselves bring about are rarely process, or unlikely to do so su≤ciently often to considered in evolutionary analyses. Yet to vary- warrant consideration. It is convenient, and also ing degrees, organisms choose their own habitats, simpler, to regard niche construction as merely choose and consume resources, generate detritus, the product of natural selection, and not a pro- construct important components of their own en- cess shaping selection pressures, because then the 118 Laland, Odling-Smee, and Feldman

organism/environment dichotomy is preserved boundaries of the individual organism. Much of unsullied, and each evolutionary event can be an organism’s world is constructed, organized, considered a response on the part of organisms to regulated, and destroyed by the organism itself, a fresh change in the external environment. by its ancestors, and by its fellow members of an We regard niche construction as an important, ecological community. How should we charac- neglected process in evolution. To us, the idea terize this niche construction? Let us begin with that niche construction can be dismissed because a definition of the niche. A niche refers to the it is the product of natural selection makes no “occupation” of an organism, for example, to the more sense than the counterproposal that natural ways in which an organism obtains its resources selection can be disregarded because it is a prod- or defends itself in its environment, in contrast to uct of niche construction. From the beginning of its location or “address” in its environment, or life all organisms have, in part, modified their habitat (Ehrlich and Roughgarden 1987). Our selective environments, and their ability to do view is consistent with Hutchinson’s (1957) con- so is, in part, a consequence of their naturally cept of the niche as a multidimensional hypervol- selected genes. Niche construction and natural ume, provided it is realized that Hutchinson’s selection are two processes, operating in parallel, concept is fundamentally relativistic, and that but also interacting. The pertinent question, then, a tolerance space cannot be defined except in re- is whether what is gained in understanding by a lation to an organism. In theory, organisms can focus on niche construction justifies the added be decomposed into arrays of features (traits or complexity that is required by its incorporation characters), while environments can be decom- into evolutionary accounts. In this chapter we posed into arrays of factors (Bock 1980). A fea- describe why we believe there is utility to compli- ture of an organism is only an adaptation if and cating evolutionary accounts by explicitly incor- when it is matched to a specific selection pres- porating niche construction into empirical and sure arising from an environmental factor at a theoretical analyses of evolutionary phenomena. particular location, it is the product of national We illustrate, using examples from natural his- selection, and that it increases the fitness of the tory, how the effects of niche construction are organism at that address and moment, for exam- frequently nontrivial, directional, accumulatory, ple, if it permits more e≤cient acquisition of a pervasive, and liable to change the nature of the food resource (Bock 1980). We interpret Bock evolutionary process. We will also describe the (1980) as treating adaptation as a dynamic and findings of theoretical models, which suggest that historical process: current utility, that is, synergy niche construction can make a substantial differ- between a feature and a factor, is not su≤cient to ence to the evolutionary process and can generate label the feature an adaptive trait (Odling-Smee unusual evolutionary dynamics. Our emphasis 1988). Niche construction occurs when an organ- on niche construction leads us to the position ism modifies the functional relationship between that evolution proceeds in reciprocal and simul- itself and its environment by actively changing taneous cycles of selection and niche construc- one or more of the factors in its environment, tion. Evolution is characterized by these cycles of either by physically perturbing these factors at contingency. its current address or by relocating to a different address, thereby exposing itself to different factors. The niche construction of both past and pres- Niche Construction and Ecological Inheritance ent generations may influence a population’s selective environment. Spiderwebs, insect pupal The constructive and self-organizing nature of cocoons, and caddis fly larvae houses modify the developmental systems extends far beyond the selective environments of the constructors them- Cycles of Contingency in Evolution 119

selves. In contrast, bird’s nests, female insect’s Dawkins (1982) argues that genes can express oviposition site choices, and the soil transformed themselves as “extended phenotypes” outside the by earthworms also modify the environment of bodies of the organisms that carry them. For the constructor’s descendants. The latter cases instance, the beaver’s dam is an extended pheno- are all examples of ecological inheritance. We typic effect of beaver genes, while the houses of define as ecological inheritance any case in which caddis fly larvae are equally expressions of caddis an organism experiences a modified functional fly genes. Extended phenotypes play an evolu- relationship between itself and its environment tionary role by influencing the chances that the as a consequence of the niche-constructing activ- genes responsible for the extended phenotypic ities of either its genetic or ecological ancestors trait will be passed on to the next generation. (Odling-Smee 1988). Ecological inheritance is However, this is just one aspect of the evolution- built into our theoretical models (Laland, Odling- ary feedback from niche construction. To go Smee, and Feldman 1996, 1999), and, as we will back to the beaver, its dam sets up a host of selec- describe, the addition of this second kind of tion pressures that feed back to act not only on inheritance system can make a considerable dif- the genes that underlie dam building, but also on ference to the evolutionary process. other genes that may influence the expression of At first sight it may be tempting to conclude other traits in beavers such as their teeth, tail, that the impact that most organisms have on feeding behavior, their susceptibility to predation their environments is trivial, a mere drop in the or disease, their social system, and many other ocean compared with the action of major geo- aspects of their phenotypes, including, probably, physical, chemical, or meteorological processes. the expression of another extended phenotype, A closer inspection reveals that countless organ- the building of lodges by beavers in the lakes they isms across the breadth of all known taxonomic create by their dams. groups significantly modify their local environ- Dam construction may also affect many future ments (Lewontin 1983; Hansell 1984; Odling- generations of beavers that may “inherit” the Smee 1988; Jones, Lawton, and Shachak 1994, dam, its lodge, and the altered river, as well as 1997; Odling-Smee, Laland, and Feldman 1996; many other species of organisms that now have Laland, Odling-Smee, and Feldman 1996). Fol- to live in a world with a lake in it. For example, lowing Lewontin (1983), and in agreement with beavers can create wetlands that can persist for Jones, Lawton, and Shachak’s (1994, 1997) con- centuries. They can modify the structure and dy- cept of “ecosystem engineering,” we argue that namics of riparian zones, and they can influence organisms not only adapt to environments, but the composition and diversity of both plant and in part also construct them. They may also do so animal communities. Some of these consequences across a huge range of temporal and spatial scales of niche construction may be just ecological, but stretching, for example, from a hole bored in a some are likely to have evolutionary conse- tree by an insect, to the contribution of cynobac- quences as well (Naiman, Johnston, and Kelley teria to the earth’s 21 percent oxygen atmos- 1988; Jones, Lawton, and Shachak 1994, 1997). phere, as a consequence of millions of years of Our general point here is that niche construction photosynthesis (Odum 1989). Niche construction generates forms of feedback in evolution that are starts to take on a new significance when it is not yet fully appreciated by contemporary evo- acknowledged that, by changing their world, or- lutionary theory (Lewontin 1983; Odling-Smee ganisms modify many of the selection pressures 1988; Odling-Smee, Laland, and Feldman 1996; to which they and their descendants are exposed, Laland, Odling-Smee, and Feldman 1996). and that this may change the nature of the evo- There are numerous examples of organisms lutionary process. choosing or changing their habitats, or of con- 120 Laland, Odling-Smee, and Feldman

structing artifacts, leading to an evolutionary example, as a result of the accumulated effects of response (Odling-Smee 1988; Jones, Lawton, past generations of earthworm niche construc- and Shachak 1994; Laland, Odling-Smee, and tion, present generations of earthworms inhabit Feldman 1996). For instance, orb-web spiders radically altered environments where they are ex- construct webs, which have led to the subsequent posed to modified selection pressures (Darwin evolution of camouflage, defense, and commu- 1881; Lee 1985). This is a good example of an nication behavior on the web (Preston-Mafham ecological inheritance. Perhaps one of the most and Preston-Mafham 1996). Similarly, ants, bees, frequently documented cases of ecological inher- wasps, and termites, construct nests that are itance is provided by female insects. Females of themselves the source of selection for many nest the vast majority of the millions of species of regulatory, maintenance, and defense behavior insect are oviparous, and usually the eggs are patterns (Hansell 1984; Holldobler and Wilson deposited on or near the food required by the off- 1994). The construction of artifacts is equally spring upon hatching (Gullan and Cranston common among vertebrates. Many mammals 1994). The offspring of virtually all insects inherit (including badgers, gophers, ground squirrels, from their mother a legacy of a readily available, hedgehogs, marmots, moles, mole rats, opos- nutritious larval food source. sums, prairie dogs, rabbits, and rats) construct Figure 10.1 shows how niche construction and burrow systems, some with underground pas- ecological inheritance interact with natural selec- sages, interconnected chambers, and multiple tion and genetic inheritance. Figure 10.1a repre- entrances (Nowak 1991). Here, too, there is evi- sents the standard evolutionary perspective: dence that burrow defense, maintenance, and Populations of organisms transmit genes from regulation behaviors have evolved in response one generation to the next, under the direction of to selection pressures that were initiated by the natural selection. Figure 10.1b extends this per- construction of the burrow (Hansell 1984; spective to acknowledge that phenotypes modify Nowak 1991). The same applies to reptiles and their local environments through niche construc- amphibians. tion. Genes are transmitted by ancestral organ- Of course, this will be no surprise to the bio- isms to their descendants, exactly as the standard logically minded, yet the breadth and scale of theory describes, but in addition, phenotypically niche construction will surprise many. Few peo- selected habitats, phenotypically modified habi- ple realise that there are more than 34,000 species tats, and artifacts, persist, or are actively or effec- of spider that construct silken egg sacs, burrows, tively “transmitted,” by these same organisms to or webs (Preston-Mafham and Preston-Mafham their descendants via their local environments. 1996). There are more than 9,000 species of birds, The environments encountered by descendent or- the vast majority of which construct nests ganisms are not just “templates” to which organ- (Forshaw 1998), and probably as many fish that isms adapt. Environments are partly determined do the same (Paxton and Eschmeyer 1998). There by independent environmental events (for in- are 9,500 known species of ants, and 2,000 stance, climatic, geological, or chemical events), known species of termites, all living in social but also partly by ancestral niche construction. colonies, and almost all building some kind of nest (Holldobler and Wilson 1994; Gullan and Cranston 1994). Niche construction is all- The Consequences of Niche Construction pervasive. Most cases of niche construction, however, do The evolutionary significance of niche construc- not involve the building of artifacts, but merely tion hangs primarily on the feedback that it gen- the selection or modification of habitats. For erates. Several topics in population biology, such as habitat selection, frequency- and density- Cycles of Contingency in Evolution 121

dependent selection, and coevolution, are concerned with the evolutionary consequences of feedback from changes that organisms bring about in their own and in other populations’ selective environments (Maynard Smith 1989). However, most analyses of these subjects have focused only on those loci that influence the production of the niche-constructing phenotype itself. What is missing is an exploration of the feedback effects on other loci. We have begun the development of a body of theory that sets out to explore the evolutionary consequences of niche construction in a systematic manner (Laland, Odling-Smee, and Feldman 1996, 1999). Our theoretical analyses, which employ two-locus, population-genetic models, have uncovered a number of interesting evolutionary consequences of the feedback from niche construction. The first consequence is that traits whose fitness depends on sources of selection that are alterable by niche construction (recipient traits) coevolve with traits that alter sources of selection (niche-constructing traits). This results in very different evolutionary dynamics for both traits from what would occur if each had evolved in isolation. Selection resulting from niche construction may drive populations along alternative evolutionary trajectories, may initiate new evolutionary episodes in an unchanging external environment, and may influence the amount of genetic variation in a population, by affecting the stability of polymorphic equilibria. Moreover, because of the multigenerational properties of ecological inheritance, niche construction can generate unusual evolutionary

Figure 10.1 (a) The standard evolutionary perspective: Populations of organisms transmit genes from one generation to the next, under the direction of natural selection. (b) With niche construction: Phenotypes modify their local environments (E) through niche construction. Each generation inherits both genes and a legacy of modiﬁed selection pressures (ecological inheritance) from ancestral organisms. 122 Laland, Odling-Smee, and Feldman

dynamics. This is because when ecological inher- they may now also be responsible for modifying itance is involved, the evolution of the recipient some of the sources of natural selection in their trait depends on the frequency of the niche- environments that subsequently feed back to se- constructing trait over several generations. For lect their own genes. However, relative to this sec- instance, timelags were found between the onset ond role of phenotypes in evolution, there is no of a new niche-constructing behavior, and the re- requirement for the niche-constructing activities sponse of a population to a selection pressure of phenotypes to result directly from naturally modified by this niche construction (Laland, selected genes before they can influence the se- Odling-Smee, and Feldman 1996). These time- lection of genes in populations. Animal niche lags generated an evolutionary inertia, where un- construction may depend on learning and other usually strong selection is required to move a experiential factors, and in humans it may de- population away from an equilibrium, and a mo- pend on culture. mentum, such that populations continue to The Galápagos woodpecker finch provides a evolve in a particular direction even if selection specific example (Alcock 1972). These birds cre- pressures change or reverse. Although these find- ate a woodpecker-like niche by learning to use ings are novel, they are consistent with those a cactus spine or similar implement to peck for of related theoretical analyses (Feldman and insects under bark. While true woodpeckers’ Cavalli-Sforza 1976; Kirkpatrick and Lande (Picidae) bills are adaptive traits fashioned by 1989; Robertson 1991). Theoretical population natural selection for grubbing, the finch’s capac- genetic analyses have established that processes ity to use spines to grub for insects is not an adap- that carry over from past generations can also tation. Rather, the finch, like countless other change the evolutionary dynamic to generate op- species, exploits a more general and flexible adap- posite responses to selection, and sudden cata- tation, namely, the capacity to learn, to develop strophic responses to selection (Feldman and the skills necessary to grub, in environments reli- Cavalli-Sforza 1976; Kirkpatrick and Lande ably containing cactus spines and similar imple- 1989; Robertson 1991). This growing body of ments. We emphasize that we are not suggesting theory supports our view that, in the presence that the grubbing of woodpeckers is innate while of niche construction, adaptation ceases to be a that of woodpecker finches is learned. Rather, we one-way process, exclusively a response to en- suggest that the grubbing of woodpeckers and vironmentally imposed problems, and instead not woodpecker finches has been fashioned by becomes a two-way process, with populations natural selection. The finch’s capacity to use of organisms setting as well as solving prob- spines is not guaranteed by the presence of rele- lems (Lewontin 1983; Odling-Smee, Laland, and vant genes, yet develops reliably as a consequence Feldman 1996). Our findings support Lewontin’s of its ability to interact with the environment in a (1983) original intuition that niche construction manner that allows it to benefit from its own may be a major source of adaptation in evolu- experience. Moreover, the finch’s learning cer- tion, and that its evolutionary effects deserve seri- tainly opens up resources in the bird’s environ- ous consideration. ment that would be unavailable otherwise and is A second difference that incorporating niche therefore an example of niche construction. This construction makes in evolution is that it allows behavior probably created a stable selection pres- acquired characteristics to play a role in the evo- sure favoring a bill able to manipulate tools lutionary process, in a non-Lamarkian fashion, rather than the sharp, pointed bill and long by their influence on selective environments tongue characteristic of woodpeckers. Because through niche construction. When phenotypes tool manipulation can itself in part depend on niche construct, they become more than simply learning, there is a further twist to this example. “vehicles for their genes” (Dawkins 1989) since Niche-constructing skills influenced by learning Cycles of Contingency in Evolution 123

could modify natural selection in favor of an if learned information, stored in neural tissue, enhanced learning ability. This point is obviously is expressed in niche construction in a manner speculative, however it is worth making because that subsequently modifies natural selection pres- it leads us to reconsider another well-known ex- sures and results in the selection of genes that ample of dichotomous thinking in biology, that change the learner’s brains, then a simple distinc- between nature and nurture, which, in this in- tion between population genetics and learning stance, is exemplified by genes versus learning. becomes untenable. An example, where some- Let us try to be more precise about the nature thing like this seems to have happened is in food- of the information that guides niche construc- storing birds, such as the marsh tit, where the tion. As a consequence of the differential survival niche-constructing act of storing and hiding food and reproduction of individuals with distinct items, seems to have selected for a bigger hip- genotypes, genetic evolution results in the acqui- pocampus and better memory (Sherry et al. sition, inheritance and transmission of genetically 1989). To clarify, the distinction we have made is encoded “knowledge” by individuals in popula- not between innate and learned traits, but be- tions, relative to particular environments. This tween information acquired through population information, expressed throughout development, genetic processes (expressed under the influence underpins much niche construction, and is com- of countless developmental factors in particular mon to all species. Although we describe this environments), and information acquired through information as genetically encoded, we recognize specialised information-acquiring subsystems in that it is expressed as phenotypic traits that also individual organisms (expressed under the influ- depend on a whole range of developmental fac- ence of genetic information and other develop- tors. In addition, many species have also evolved mental factors in particular environments). Each a set of complicated ontogenetic processes, that class of information guides niche construction. allow individual organisms to acquire additional Going beyond individual learning, a few spe- information that is not acquired through popu- cies, including many vertebrates, have also lation genetic processes, and is not encoded in evolved a capacity to learn from other individu- genes. These processes are themselves adaptive als and to transmit some of their own learned traits and products of genetic evolution, and their knowledge to others by social learning. The re- functioning is dependent on genetically encoded sulting “proto-cultural” information may also information, but they are nevertheless distinct underlie niche construction. An informative from genetic evolution. They comprise “faculta- example is the spread of milk bottle top opening tive” or “open” developmental processes that are in British tits (Hinde and Fisher 1951). These based on specialized information-acquiring sub- birds learned to peck open the foil cap on milk systems in individual organisms, such as brain- bottles, and to drink the cream, and this behavior based learning in animals or the immune system spread right throughout Britain and into conti- in vertebrates. We regard these subsystems as nental Europe. Hinde and Fisher found that this particularly interesting forms of phenotypic plas- behavior probably spreads by local enhance- ticity because they are capable of additional, in- ment, where the tits’ attention is drawn to the dividually based information acquisition, again milk bottles by a feeding conspecific, and after relative to particular environments. Unlike other this initial tipoff, they subsequently learn on their developmental influences on the phenotype, these own how to open the tops. However, further systems are adaptive traits selected precisely analysis by Sherry and Galef (1984) revealed because of their information-gathering quality. that, in addition to social learning by local en- This allows learned knowledge to guide niche hancement, milk bottle-top opening could be construction in many animal species. However, acquired by another means. They found that this 124 Laland, Odling-Smee, and Feldman

behavior could also spread if the birds were engineering” and it is currently receiving increas- merely exposed to opened milk bottles, even if ing recognition by ecologists (Jones, Lawton, and there were no other birds present to watch per- Shachak 1997). Thus a third difference that informing the opening behavior. In this example, corporating niche construction makes to under- the birds’ niche-constructing behavior is propa- standing biological systems is that it modulates gated by local enhancement. However, by creat- and may partly control the flow of energy and ing opened milk bottles, this niche construction matter through ecosystems. Such modifications biases the probability that other birds will learn can have profound affects on the distribution and to open bottles. Moreover, the selection pressures abundance of organisms, the influence of key- acting on genetic variation at loci affected by milk stone species, the control of energy and material bottle opening are modified in essentially the same flows, residence and return times, ecosystem manner as if genes underpinned opening. For resilience, and specific trophic relationships example, this self-induced selection might influ- (Jones, Lawton, and Shachak 1997). ence selection acting on the birds’ learning capac- Jones, Lawton, and Shachak (1997) point out ities, foraging behavior, or digestive enzymes. that a major ecological consequence of the niche This social transmission reaches its zenith as construction of organisms is that it establishes human culture, where this ability to learn from “engineering webs,” or control webs, in both others is facilitated by a further set of processes communities and ecosystems. Engineering webs (e.g. language, complex cognition). We treat cul- do not conform to the same principles of mass tural processes as a third category of information flow, stoichiometry (the proportions of chemical acquisition, distinct from, but also partly depen- elements in organisms), and the conservation of dent on, population-genetic processes, and spe- energy that govern the more familiar energy and cialized individual-based information-acquiring material flows and trophic relations among subsystems, such as learning. This third type of organisms. This makes it di≤cult to understand information can also guide niche construction. how engineering webs achieve their control, or to Consider the example of Kwa-speaking yam cul- predict which organisms are likely to have the tivators in West Africa, who increased the fre- biggest effect on an ecosystem. However, ecosys- quency of a gene for sickle-cell anemia in their tem engineering often depends on the adapta- own population as a result of the indirect effects tions of organisms, which may drive evolution of yam cultivation. These people traditionally cut when they modify natural selection pressures by clearings in the rainforest, creating more stand- niche construction, and when they generate lega- ing water and increasing the breeding grounds cies of modified natural selection pressures for for malaria carrying mosquitoes. This, in turn, subsequent generations. Thus, regardless of intensifies selection for the sickle-cell allele, whether organisms are themselves members of because of the protection offered by this allele a specified trophic web, their niche-constructing against malaria in the heterozygotic condition adaptations may qualify them as members of an (Durham 1991). Here human cultural niche con- associated engineering web, which may allow struction has resulted in a well-established and them to exert a degree of evolutionary as well as quantifiable evolutionary response. ecological control over ecosystems. The consequences of environment modification by organisms, however, are not restricted to evolution, and organisms can affect both their Organism-Environment Coevolution own and each other’s ecology by modifying sources of natural selection in their environments It is common to portray evolution as a process (Jones, Lawton, and Shachak 1997). In ecosys- characterized by descent with modification of tem ecology this process is known as “ecosystem entities in a lineage. Traditionally, organisms are Cycles of Contingency in Evolution 125

regarded as the entities that evolve, while their environmental component. Under such circum- environments are described as changing but not stances, the evolutionary process could once as evolving. Environments are only regarded as again be described using a genetic currency, and evolving to the extent that they are composed of it would also satisfactorily include those evolving other populations of organisms that clearly do aspects of the environment that are this type of evolve. It is also generally assumed that the dis- extended phenotype. tribution of characters of organisms in an evolv- In reality, there is a more di≤cult problem to ing population reflects the frequency of particular solve. The problem is that ancestral organisms naturally selected genes currently found in that not only bequeath genes to their descendants, rel- population. A consequence of this approach is ative to their selective environments in the stan- that evolutionary biologists are often content to dard way (figure 10.1a) but also bequeath a use only a genetic currency when they describe legacy of modified natural selection pressures, as evolution. an ecological inheritance, relative to those same The emphasis on evolving developmental sys- genes (figure 10.1b). The dichotomy between tems by the authors of this volume is one wel- evolving organisms and nonevolving environ- come departure from the traditional perspective. ments has clearly broken down in the evolu- We would like to propose another one by sug- tionary scheme illustrated in figure 10.1b. The gesting that when organisms niche construct, it is separation of organism and environment in evo- not just the organisms that evolve, because they lution does not sit well with the evolutionary are also likely to cause a more general coevolu- process once the implications of niche construction in organism-environment systems by their tion are spelled out. niche construction. (Gray 1988 and Gri≤ths and Gray 1994 present a similar perspective). In arguing thus, we do not advocate the mere redescrip- Acknowledgements tion of environmental change as evolution, which would constitute a purely semantic substitution. This work is supported by a Royal Society Instead we maintain that niche-constructed com- University Research Fellowship to Kevin N. ponents of the environment are both products of Laland and by National Institutes of Health the prior evolution of organisms and, in the form Grant GM28016. of modified natural selection pressures, causes of the subsequent evolution of organisms, and that References as both products and causes of evolution, these environmental components need to be incorpo- Alcock, J. (1972). The evolution of the use of tools by rated in evolutionary theory more fully than they feeding animals. Evolution 26: 464–473. are at present. It is in this sense that we see organ- Bock, W. J. (1980). The definition and recognition of isms and their environments as comprising biological adaptation. American Zoologist 20: 217–227. coevolving systems. Neither are we merely advo- Darwin, C. (1881). The Formation of Vegetable Mold cating a broadening of the concept of the organ- through the Action of Worms, with Observations on ism so that extended phenotypes are regarded as Their Habits. London: Murray. part of the evolving system (Dawkins 1982), with Dawkins, R. (1982). The Extended Phenotype. Oxford: the rest of the environment treated as nonevolv- Freeman. ing. Such a position might be justified were there Dawkins, R. (1989). The Selfish Gene. (2d ed.) Oxford always a direct correspondence between the fre- University Press. quency of particular genes in a population and Durham, W. H. (1991). Coevolution: Genes, Culture and the incidence of a particular niche-constructed Human Diversity. Palo Alto: Stanford University Press. 126 Laland, Odling-Smee, and Feldman

Ehrlich, P. R., and J. Roughgarden. (1987). The Science Lewontin, R. C. (1982). Organism and environment. of Ecology. New York: Macmillan. In H. C. Plotkin (Ed.), Learning, Development and Feldman, M. W., and L. L. Cavalli-Sforza. (1976). Culture, pp. 151–172. New York: Wiley. Cultural and biological evolutionary processes, selec- Lewontin, R. C. (1983). Gene, organism and environ- tion for a trait under complex transmission. Theoretical ment. In D. S. Bendall (Ed.), Evolution from Molecules Population Biology 9(2): 238–259. to Men, pp. 273–285. Cambridge: Cambridge Univer- Forshaw, J. (1998). Encylopedia of Birds. (2d ed.) San sity Press. Diego: Academic Press. Maynard Smith, J. (1989). Evolutionary Genetics. Gray, R. D. (1988). Metaphors and methods: Oxford: Oxford University Press. Behavioral ecology, panbiogeography, and the evolv- Naiman, R. J., C. A. Johnston, and J. C. Kelley. (1988). ing synthesis. In M-W. Ho and S. W. Fox (Eds.), Alteration of North American streams by beaver. Evolutionary Processes and Metaphors, pp. 209–242. Bioscience 38: 753–762. Chichester: Wiley. Nowak, R. M. (1991). Walker’s Mammals of the World. Gri≤ths, P. E., and R. D. Gray. (1994). Developmental (5th ed.) Baltimore: Johns Hopkins University Press. systems and evolutionary explanation. Journal of Phil- Odling-Smee, F. J. (1988). Niche constructing pheno- osophy 91: 277–304. types. In H. C. Plotkin (Ed.), The Role of Behavior in Gullan, P. J., and P. S. Cranston. (1994). The Insects: Evolution, pp. 73–132. Cambridge, MA: MIT Press. An Outline of Entomology. London: Chapman & Hall. Odling-Smee, F. J., K. N. Laland, and M. W. Feldman. Hansell, M. H. (1984). Animal Architecture and (1996). Niche construction. American Naturalist 147: Building Behavior. New York: Longman. 641–648. Hinde, R. A., and J. Fisher. (1951). Further observa- Odum, P. E. (1989). Ecology and Our Endangered Life- tions on the opening of milk bottles by birds. British Support Systems. Sunderland, MA: Sinauer. Birds 44: 393–396. Paxton, J. R., and W. N. Eschmeyer. (1998). En- Holldo55bler, B., and E. O. Wilson. (1995). Journey to the cyclopedia of Fishes. San Diego: Academic Press. Ants: A Story of Scientiﬁc Exploration. Cambridge, Preston-Mafham, R., and K. Preston-Mafham. (1996). MA: Belknap. The Natural History of Insects. Malborough, England: Hutchinson, G. E. (1957). Concluding remarks. Cold Crowood Press. Spring Harbor Symposia on Quantitative Biology 22: Robertson, D. S. (1991). Feedback theory and 415–427. Darwinian evolution. Journal of Theoretical Biology Jones, C. G., J. H. Lawton, and M. Shachak. (1994). 152: 469–484. Organisms as ecosystem engineers. Oikos 69: 373–386. Sherry, D. F., and B. G. Galef. (1984). Cultural trans- Jones, C. G., J. H. Lawton, and M. Shachak. (1997). mission without imitation—milk bottle opening by Positive and negative effects of organisms as physical birds. Animal Behaviour 32: 937–938. ecosystem engineers. Ecology 78: 1946–1957. Sherry, D. F., A. L. Vaccarino, K. Buckenham, and R. Kirkpatrick, M., and R. Lande. (1989). The evolution S. Herz. (1989). The hippocampal complex of food- of maternal characters. Evolution 43: 485–503. storing birds. Brain, Behavior and Evolution 34: 308– Laland, K. N., F. J. Odling-Smee, and M. W. Feldman. 317. (1996). On the evolutionary consequences of niche con- Van Valen, L. (1973). A new evolutionary law. Evo- struction. Journal of Evolutionary Biology 9: 293–316. lutionary Theory 1: 1–30. Laland, K. N., F. J. Odling-Smee, and M. W. Feldman. (1999). The evolutionary consequences of niche construction and their implications for ecology. Proceed- ings of the National Academy of Sciences, USA 96(18): 10242–10247. Lee, K. E. (1985). Earthworms: Their Ecology and Relation with Soil and Land Use. London: Academic Press. III THE DEVELOPMENT OF PHENOTYPES AND BEHAVIOR

11 The Ontogeny of Phenotypes

H. Frederik Nijhout

A view of phenotypes from a developmental sys- phenotype, and thus refers to those genes that are tems perspective has to be concerned with the in some way related to a phenotype. There are processes that give rise to the phenotype. Varia- actually two sets of genes that need to be consid- tion in the rate, timing, and spatial distribution of ered in the definition of a genotype. We can con- these processes gives rise to variation in the phe- sider either all the genes that affect the variation notype. A distinction is usually drawn between in a character, or all the genes that contribute genotype and phenotype, in which the phenotype to the ontogeny of a character. How one finds all is derivative and the genotype (or a gene) is as- the genes, in either case, is problematic. At one sumed to somehow “control” the characteristics extreme, it could be all the genes that are ex- of phenotypes. Yet for complex traits there is pressed in all the cells that make up the trait, in a usually a weak (and often a variable) correlation lineage going back to the zygote. This is a large between genotype and phenotype. Good exam- number of genes and could conceivably approach ples of such variable association are found in the size of the whole genome. At the other ex- many human genetic diseases where a genetic de- treme, it could be only those genes that have a fect is associated with the disease in only a small major effect on the variation in the trait we are fraction of a population, or where the association interested in (that is, eliminating from considera- exists reliably only in certain families or ethnic tion those genes that have small quantitative groups (Vogelstein and Kinzler 1997; Haines and effects). This is a much smaller number, and, in Pericak-Vance 1998). A developmental systems practice, this is what most people mean when perspective helps to explain, and in some cases to they refer to the “genotype of an organism.” The predict, the degree of association (or lack thereof ) limiting case of this would be a system in which between genetic variation and phenotypic varia- all genes but one have only a single allele (and are tion. The approach I will take in this chapter is to thus invariant). In such a system only the one examine the consequences of interactions within variable gene would constitute the genotype; the developmental systems for the properties of phe- lack of variation in characters other than those notypes and for the correlation between genotype affected by this one gene makes their genetics and phenotype. moot. From a developmental systems perspective it is actually something nearer the first extreme that Genotypes is of interest, because we are interested in the ontogeny and evolution of complex traits (traits Insofar as a discussion of phenotypes makes whose origin and variation depends on many fac- sense only in relation to genotypes (for other- tors). In practice, it is probably not particularly wise we would simply use the term character or useful to consider all relevant genes back to the trait), it is useful to define genotype first. The zygote, because that gets us into an infinite re- usual textbook definition of the genotype as “the gression: The zygote’s development is influenced genetic constitution of an organism” (King and largely by maternal genes, whose regulation de- Stansfield 1985) is not particularly useful because pends on yet additional maternal genes, and so terms like genetic constitution and genetic makeup forth. It might seem that genes expressed early are not well defined. A more useful definition of in the ontogeny of a trait are less relevant to vari- genotype is one that explicitly makes use of what ation of the mature trait than genes expressed we know about the structure of genes. A geno- later, and could therefore be ignored, but it is type is typically considered in the context of a 130 H. Frederik Nijhout

easy to conceive of exceptions to this apparently “activity,” it is sometimes simply the average of sensible viewpoint (for instance, coiling direction the two independent activities of the enzymes of adult snail shells [Limnaea] is determined by a produced by each allele, but more often the joint maternal gene expressed during early embryonic activity is closer to one of them, in which case we cleavage [Sturtevant 1923; Freeman & Lundelius say that one allelic form is dominant over the 1982]). So how are we to determine what portion other to some degree. This is simple monohybrid of the (variable) genes of an organism or species Mendelian genetics, and explaining what hap- constitute the relevant genotype for a given trait? pens at the level of the phenotype in terms of It turns out, as we will see, that for complex traits the characteristics of the genotype is relatively the genes whose variation have the greatest effect unproblematic. on the phenotype differ from one genetic back- A more complex and more typical case is ground to another. In other words, the genes that sickle-cell anemia, a disease “caused” by a known are perceived to affect a trait depends on the con- deﬁciency in a gene for hemoglobin. Two genes text in which that trait is observed. are required to make the components of the adult human hemoglobin molecule: an alpha-globin gene and a beta-globin gene. Hemoglobin has Phenotypes a quaternary structure composed of two alpha- and two beta-globins. A point mutation in the What a phenotype is depends on one’s perspec- code for the sixth amino acid of the beta-globin tive and on the level at which one chooses to do chain is the only genetic defect associated with the analysis. Let’s say that we have a gene for an sickle-cell anemia. As far as we know, this defect enzyme. The simplest level of phenotype is the has no effect at all on the structure or function of protein coded by a gene. At the next level it is the the hemoglobin molecule, as long as the molecule activity of the enzyme, measured as velocity of is in the presence of relatively high oxygen ten- substrate depletion or product accumulation. At sions, and is fully oxygenated. When the mole- yet a higher level, the phenotype is the effect of cule enters an environment with low oxygen ten- the reaction product (or the effect of the absence sion (such as might exist within tissues that have of the substrate). For instance, if the enzyme is a high metabolic rate) the defective molecule may anthocyanidin synthase (which catalyzes the crystalize instead of remaining in a gel-like state reaction from colorless leucocyanidin to purple (so this phenotype is contingent on an environ- cyanidin in morning glories [Holton and Cornish mental factor). Whether or not the hemoglobin 1995]), then we can call the presence of antho- crystalizes depends on the actual tension of oxy- cyanidin synthase the phenotype, or we can call gen, and on whether one or both of the beta purple ﬂower color the phenotype. In all these chains in the hemoglobin are defective. Suppose cases we are looking at different manifestations now that we have a person who is homozygous of the same genetic factor: protein molecule, en- for the genetic defect. Such a person’s hemoglo- zymatic activity, pigment molecule, optical color. bin will crystalize upon deoxygenation. The crys- Now let’s say we have a diploid organism that taline form of hemoglobin is a phenotype, as is is not homozygous for the two alleles for a gene. the deformation of the red blood cell caused by We now have a new phenotype, namely that pro- this crystalization (this is, of course, the pheno- duced by the joint contribution of the two forms type that gave rise to the name of the disease). of the enzyme encoded by the two alleles. In the But there are more phenotypes yet. The sickled chain of manifestations outlined earlier, this new cells may block capillaries and cause tissue necro- phenotype is expressed above the level of the pro- sis (a phenotype), and if this tissue is in the brain, tein molecule. If the phenotype is expressed as it may cause various neurological or behavioral The Ontogeny of Phenotypes 131

symptoms (another phenotype). We could go on and on listing the damage and its symptoms in different tissues as separable phenotypes, because individuals differ in exactly how they manifest the disease. In fact, manifestation of any form of disease is the ultimate phenotype, because whereas some homozygous individual suffer greatly and die in childhood, others live asymptomati- cally to old age. Thus not only can a phenotype (or rather, the phenotypic consequences of genetic variation) be viewed at many different levels, but the actual expression of these phenotypes at any one level is also context-dependent, and is not just a direct consequence of the genetic variation.

A Simple Complex Trait Figure 11.1 Context-dependency can actually be observed in The effect of variation in the “activity” of one enzyme much simpler systems as well. Consider a short on the flux through a metabolic pathway. The relationship becomes progressively more nonlinear as the path- linear metabolic pathway in which the product of way gets longer. Enzymatic variation can come about one biochemical reaction serves as the substrate through mutations, so the horizontal axis can be read for the next reaction, and so forth. In such a sys- as the effect of allelic variation on the genes that code tem the flux through the system (measured for for an enzyme on the activity of that enzyme. Arrows instance, as the rate of accumulation of the final indicate the values of two extreme “alleles” encoding product in the reaction chain) depends in some high (H) and low (L) enzymatic activities, used to con- way on the activities of all the enzymes in the struct figure 11.2. pathway. Metabolic control theory can be used to analyze and predict this dependency (Kacser K is an equilibrium constant, V is a maximal and Burns 1973, 1981; Kacser and Porteous 1987; velocity (V ), M is a Michaelis constant (K ), Keightley and Kacser 1987). Interestingly, if max m and the subscripts refer to the number of the there is genetic variation in only one of these enzyme in the chain. Figure 11.1 illustrates the enzymes, then the degree to which the flux relationship between the activity of any one en- changes depends not only on that enzyme but zyme in such a chain and the flux through the also on how many additional enzymes there are chain. If there is only one enzyme the relation is in the pathway. The dependence of flux on en- linear, but the relationship becomes increasingly zyme activity for a linear chain of n enzymatic hyperbolic as the chain becomes longer. This reactions is given by Kacser and Burns (1981) as means that the sensitivity of flux (a phenotype) to (X :X )K variation in the activity of an enzyme (a lower Flux;———————————————–—–—––—1 n 1n M1!V1+M2K12!V2+M3K13!V3+...+MnK1n!Vn level phenotype) is a systemic property, deter- (11.1) mined by all enzymes in the system. The contribution of each enzyme to the total flux is where X1 and Xn are the concentrations of the ini- described by its sensitivity coe≤cient, which is a tial substrate and final product, respectively, the measure of how much the flux will change given 132 H. Frederik Nijhout

a particular amount of change in the activity of enzymes in the pathway). Therefore, it is the that one enzyme. This coe≤cient can be calcu- ensemble that determines which of the enzymes lated by taking the partial derivative of the flux will be the one with the largest coe≤cient. with respect to the enzyme of interest, and is thus One way to examine how the flux depends on proportional to the tangent at any one point on variation in a single enzyme is by means of a sen- the curves in figure 11.1 (Kacser and Burns 1981). sitivity analysis: examining the effects of varia- The sensitivity coe≤cient of a given enzyme thus tion in one parameter or variable while keeping not only varies with the activity of that enzyme all others constant. The parameters in this system but also varies with the total number of addi- are the various constants that define the kinetics tional enzymes in the pathway. It is therefore a properties of each enzyme: Vmax, Km, and Keq joint property of all the enzymes in the pathway. (see equation 11.1). Each of these constants is One of the most important results of this type determined by the tertiary structure of the en- of metabolic control theory is the summation zyme, which, in turn, is determined by the base property: the magnitudes of the sensitivity co- sequence of the gene that encodes it. In a real e≤cients of all enzymes in a system must sum to population of organisms the genes will be poly- one (Kacser and Burns 1973, 1981). This implies morphic (that is, there will be many alleles), so that the more enzymes there are, the smaller their that at the population level these constants actu- respective coe≤cients will be (assuming all have ally vary among individuals. This means that for approximately the same relative effect on flux). It a sensitivity analysis it is necessary to make a also implies that if not all enzymes have the same decision about the values at which the various sensitivity coe≤cient, only a few can have a large kinetic constants will be held. This could be done coe≤cient. This is because if one or a few en- in a variety of ways, but a simple and heuristical- zymes have a large coe≤cient, then the coe≤- ly instructive method is to assume that there are cients of the rest must be small since they all have only two alleles for each enzyme-encoding gene to sum to unity. The summation property there- in the population, so that in a diploid organism fore indicates that if a system is controlled by we need deal only with three genotypes for each many enzymes, and if all enzymes have an equal gene. We can then hold all genes except for one effect on flux, the sensitivity of the system to vari- either homozygous (AA or aa) or heterozygous ation in any one of them will be very small (if (Aa), and examine the effect of variation of that there are n enzymes in a system and their sensi- one gene on flux through the system. Sensitivity tivity coe≤cients are equal and add to 1, each will analyses examining the effect of variation in Vmax have a sensitivity coe≤cient of 1/n). But if the and Km of one of the enzymes in a linear pathway enzymes do not have an equal effect on flux, then of nine enzymatic steps are illustrated in figure only one or a few of them can have large 11.2. The three graphs in each panel describe the coe≤cients and the rest must have proportionally effects of variation in three different genetic back- small ones, so that all still add up to one. Hence grounds: LL, a background in which the remain- in a system where enzymes do not have equal ing eight genes are homozygous for the allele with effects, one (or a few) will appear to be in control the larger value; SS, a background homozygous of the flux, and the effects of the remainder will be for alleles of the smaller value; LS, a background small and perhaps even undetectable. What is heterozygous for the two alleles. These graphs most important to recognize is that the sensitivi- show that the genetic background is a major ty coe≤cient of an enzyme is not a property of the determinant of the degree to which variation in enzyme itself, but is a systemic property (its value one gene affects the flux through the system. For is determined largely by the denominator in instance, the mean slope of the curves in figure equation 11.1, which contains terms for all the 11.2a shows that variation in the Vmax of an en- The Ontogeny of Phenotypes 133

zyme has only a small effect on flux, and the separation between the curves shows that variation in the genetic background has a much larger effect. Kacser and Burns (1981) showed that dominance is an emergent property of the interaction among enzymes linked in a biochemical pathway. This is because the interaction among enzymes causes the relationship between enzymatic variation and flux to become nonlinear (figure 11.1). Thus, although the two alleles of a gene act additively at the physiological level (additivity means that the physiological effect of each allele in a diploid organism is independent of the effect of the other allele, so that their joint effect is simply the sum of their independent effects; a heterozygote would then have a genetic value exactly halfway between that of the two homozygotes), the interaction among genes causes nonlinear interactions at the level of the phenotype, which is manifested as partial dominance of one allele over the other (figure 11.3). Because all genes acquire a nonlinear relationship with flux, this means that the genetic background in which any one gene acts also becomes nonlinear. The nonlinearity of the genetic background is revealed by the uneven spacing of curves in figure 11.2a. The sensitivity curve for the heterozygous genetic background does not lie Figure 11.2 halfway between the homozygous genetic back- (A) Effect of variation in the value of the Michaelis con- ground curves, as one would expect if the gene in stant (K ) of one enzyme on the flux through a linear m question simply acted additively with the back- biochemical pathway of nine enzymes. (B) The effect, of ground genes. In other words, the epistatic inter- variation in the Vmax of one enzyme in a nine-enzyme pathway. The three curves in each panel represent three action among genes is also nonlinear, at least for genetic backgrounds in which the other eight enzymes variation in Km (but apparently not for variation in the pathway are held constant at genotypes that are in Vmax; figure 11.2a). either homozygous (SS and LL) or heterozygous (LS) for enzymatic values indicated by the arrows in figure 11.1. The magnitude of the effect of enzymatic variation A Developmental System as the Determinant of depends strongly on the genetic background (i.e., on the Phenotype allelic values of all other genes that participate in the ontogeny of the phenotype [flux]). The simplest and most widespread developmental mechanism for pattern formation is a diffusion-gradient threshold mechanism. Diffusion gradients are used throughout embryonic and postembryonic development to specify the location and spatial pattern of developmental events. 134 H. Frederik Nijhout

Figure 11.3 Dominance is an emergent property of nonlinearity in the relationship between genetic variation and phenotypic variation. The curve in this figure represents the function that relates variation in the alleles of one gene (horizontal axis) to the phenotype (vertical axis). The shape of this curve is determined by all the factors that Figure 11.4 interact to produce the phenotype (cf. figures 11.1, 11.5, A diffusion-gradient threshold system. A cell produces and 11.6). If two alleles (A and a) are codominant at the a signaling molecule that diffuses from cell to cell and is genetic level, then the value of the physiological “activ- metabolized over time. Hence progressively farther cells ity” of the heterozygote will be exactly halfway between experience a progressively lower concentration of the that of the two homozygotes. The final phenotype of signal. At a certain point in time the cells “read” the the heterozygote (vertical axis) is, however, closer to concentration of the signal. If the signal is above a one homozygote than the other, so that at the pheno- threshold value the cells express one kind of gene (dark typic level one of the alleles is partially dominant. cells), and if it is below threshold they express a different gene (white cells). The phenotype is the size of the population of cells that is dark. Most processes of pat- A diffusion-gradient threshold mechanism and tern formation during embryonic and postembryonic the phenotype it produces thus constitute a sim- development are regulated by such a gradient threshold ple but realistic developmental system. (Nijhout mechanism. The exact value of this phenotype is deter- and Paulsen 1997; Klingenberg and Nijhout mined by the six parameters shown. 1999). The study of genetic variation and evolution in such a simple system reveals much about properties of developmental systems in general. sitivity analysis of each parameter is shown in We assume that the values of the parameters that figure 11.5. As in the case of the metabolic path- control a diffusion gradient threshold system are way outlined earlier, the relationship between influenced by genes (for simplicity we assume genotypic value of a particular gene and the that alleles at a single gene are responsible for the phenotypic value depends almost entirely on entire genetic variation encountered in each pa- the values of the other genes: the genetic back- rameter). Figure 11.4 shows how six parameters ground. Dominance and epistasis can be seen, can control the value of the phenotype in such a respectively, from the curvature of the relation developmental system. The genetics of the phe- between genetic and phenotypic values and from notype produced can be studied by introducing the spacing of the curves in different genetic back- genetic variation in the parameter values. A sen- grounds. The Ontogeny of Phenotypes 135

Figure 11.5 Effect of genetic variation in the six parameters of a diffusion-gradient threshold mechanisms on the value of the phenotype. As in ﬁgure 11.2, the curves in each panel represent the sensitivity of the phenotype to variation in one parameter when the others are kept constant at a homozygous high, homozygous low, or heterozygous level. Because of the interactive nature of this system, a given genetic value at any one gene can correspond to a broad range of phenotypic values, and a given phenotypic value can be associated with a broad range of genetic values, depending entirely on the values of all other genes in the system. Thus there is no one-to-one correspondence between genetic value and phenotypic value for any of the genes that participate in this system. (After Nijhout and Paulsen 1997.)

Evolution in such a system can be studied sible genotypes, the frequency of each genotype by setting up a population of diffusion-gradient being determined by the relative frequencies of threshold models by computer simulation. Indi- each of the alleles. Random combination of alle- viduals are assumed to be diploid, and each pa- les yields a population with a particular mean rameter is represented by two alleles, one tending phenotype and Hardy-Weinberg frequencies of to produce a high phenotypic value the other a genotypes. We can now study the evolution of low phenotypic value. The frequencies of the two this developmental system in a population ge- alleles in a simulated population can be defined netic context by asking how the allele frequencies (e.g., 10% high alleles and 90% low alleles) and a change under artificial selection on the phenotype. population of individuals can be produced by The e≤cacy of selection is a function of the random combination of the alleles. Six unlinked correlation between genotype and phenotype. By genes, each with two alleles, can produce 729 pos- inspection of figure 11.5 it might seem on first 136 H. Frederik Nijhout

sight that this correlation is very poor: a given type. Another observer, at a different time, would genetic value at any one gene can be associated identify a different set of genes as the ones whose with a broad range of phenotypes, and any phe- variation was most highly correlated with pheno- notype can be associated with an equally broad typic variation and which would thus appear to range of genetic values of every gene. Yet selec- be primarily in control. For instance, in generation on the phenotype leads to an orderly and tion 6 most of the phenotypic variation is due to gradual change in the mean phenotype of the variation in the gene that affects timing, whereas simulated population (figure 11.6a). In each gen- in generation 11 most of the phenotypic variation eration, only individuals with phenotypes larger is due to variation in the gene that affects the than a particular critical value were selected and breakdown rate of the diffusing substance. At then allowed to mate randomly among them- any one time most genes are poorly correlated selves to produce the next generation. Although with the phenotype and thus are neutral to selec- the phenotypic response to selection was orderly, tion on the phenotype. These genes are neverthe- the genetic response was rather more complex less critical contributors to the phenotype, and in (figure 11.6b). Initially only two genes responded a different genetic background one or more of to selection and not until their allele frequencies them could become most highly correlated with had changes considerably did additional genes the phenotype while previously highly correlated begin to respond. At any one time during the genes become nearly neutral. This phenomenon selection process, most of the genetic change was is called pseudoneutrality or contingent neutrality; focused on only one or two genes. The cause of it is not a property of the gene but of the back- this pattern of genetic change is the fact that not ground in which it is expressed. all genes are equally correlated with the phenotype, and that any gene’s correlation with the phenotype depends on its frequency in the popu- The Effect of Environmental Variation on the lation and well as on the frequency of all the Phenotype other genes. The correlations are shown in figure 11.6c. The six genes in this system are not all We have just looked at the effects of genetic vari- equally correlated with the phenotype, and the ation on the phenotype and modeled this effect as correlation of any one gene with the phenotype if environmental variation did not exist. In both changes as selection progresses and allele fre- systems we have examined, the metabolic path- quencies change. One of the most interesting fea- way and the diffusion-gradient threshold mecha- tures of this correlation pattern is that each gene nism, it is actually easy to see how environmental appears to take a turn, as it were, being the most variation could alter the phenotype. The activity highly correlated with the phenotype as selec- of an enzyme, for instance, depends on the tem- tion progresses. Another interesting observation perature, pH, and ionic strength of its envi- is that at most times, the majority of genes are ronment as well as on the presence of chemical very weakly correlated with the phenotype. It ap- inhibitors and synergists. A change in tempera- pears that at any one time only one or a few genes ture can therefore alter the activity of an enzyme are highly correlated with the phenotype while in the same way that a mutation would. Both the most genes are essentially neutral and not seen by equilibrium constants and the reaction constants selection. of enzymatic reactions vary with temperature, An outside observer, chancing upon such a as given by the van’t Hoff and the Arrhenius population of organisms at some arbitrary time equation, respectively. In a multienzyme meta- during selection, would identify only one or two bolic pathway, a change in temperature of course genes as primarily “controlling” the pheno- affects all enzymes simultaneously, but because enzymes differ in their temperature sensitivity, The Ontogeny of Phenotypes 137

Figure 11.6 Response of a diffusion-gradient threshold mechanism to selection on the phenotype. Each of the six parameters was represented by two alleles (a high allele, producing a large phenotypic value, and a low allele, producing a low phenotypic value), whose frequency was initially set at 90 percent high and 10 percent low. An initial population of ten thousand “models” was generated by random recombination of diploids from this gene pool, and their respective phenotypes were calculated. All individuals with phenotypes above the mean were eliminated and the next generation produced by random “mating” of the remaining individuals, and this process was repeated for eighteen generations. (A) The response of the mean phenotype to selection. (B) The response of the allele frequencies for the six parameters to selection. As can be seen, not all parameters responded synchronously. The pattern shown here was identical in several independent runs of the selection, and was not due to stochastic events associated with random mating. (C) The correlation between variation at each parameter and variation of the phenotype. The parameters that respond most rapidly to selection are those that are most highly correlated with the phenotype. Only one or two parameters are highly correlated with the phenotype at any one time. As selection progresses, each parameter in turn becomes “most highly” correlated. Thus each of the parameters could be considered a major gene, a modiﬁer gene, or a nearly neutral gene at different times, depending entirely on the genetic background deﬁned by the other parameters. (After Nijhout and Paulsen 1997.) 138 H. Frederik Nijhout

different enzymes in the pathway will be affected components simultaneously. For instance, a to different degrees. The overall effect of temper- mutation will directly affect the activity of only a ature will therefore be integrated across all en- single enzyme, whereas a change in temperature zymes in a pathway. will directly alter the activities of many enzymes Variation in enzyme activity with temperature simultaneously. Of course, due to the highly in- results in variation in the flux through a metabol- tegrated nature of developmental systems, both ic pathway. Although this flux is a phenotype in mutation and environmental variation can have a its own right, it also affects other phenotypes for host of indirect effects on many parts of the sys- whose development or properties the metabolic tem. The equivalency of genetic and environmen- pathway is necessary. Variation of the pheno- tal variation is demonstrated by the relative ease type with variation in an environmental factor is with which it is possible to produce accurate phe- called a reaction norm (Schlichting and Pigliucci nocopies of known mutations by simple environ- 1998). All phenotypes express reaction norms in mental perturbations (Santamaria 1979; Oyama response to one or more environmental variables. 1981; Mitchell and Petersen 1982). As a rule, a particular correspondence between a In order to produce a predictable phenotype in genotype and a phenotype is defined only under a a variable (varying in time) and diverse (varying restricted set of environmental (and genetic back- in space) environment, the developmental sys- ground) conditions. tem needs to be made robust to this variation. In the more complex diffusion-gradient thresh- Resistance to a variable environment is accom- old system it is likewise easy to see how various plished by homeostatic physiology, which at- environmental factors could affect the pheno- tempts to maintain a fairly constant internal type. Temperature not only affects the kinetics of environment in the face of external variation. A synthesis and breakdown of diffusing substances, diverse environment can be accommodated by but also affects their diffusion coe≤cients. Be- selection for mutations and combinations of alle- cause temperature affects equilibrium constants, les that adjust the dynamics of the developmental it is also likely to affect the value of thresholds. system to produce the target phenotype in each The reaction norm to temperature would be de- particular environment. fined by the joint effects of temperature on all the processes that define the diffusion and threshold system. Other environmental variables such as Can the Findings of Simple Systems Be pH and ionic composition can affect a gradient Generalized to More Complex Phenotypes? threshold system by the same mechanisms they affect enzyme kinetics. As a trait becomes more complex during onto- In general, genetic (mutational) and environ- geny (or evolution) it becomes dependent on, and mental variation have similar effects on the com- therefore sensitive to, an ever greater number and ponents of a developmental system. Both types of diversity of factors, both genetic and nongenetic. variation alter the rates and equilibrium values of Moreover, as our insight becomes more sophis- dynamical processes and therefore both can have ticated and we take more factors into account in the same effect in the final outcome of the our analysis of a trait, the relationship between processes that produce the phenotype. The main any one factor (or determinant) and the pheno- difference between genetic and environmental type becomes increasingly indirect and contingent. variation is that the former can alter the value of Most developmental systems are substantially only a single component of the system (by alter- more complex than the two exemplars discussed ing the properties of a single protein), whereas above. It is therefore useful to consider whether the latter generally alters the properties of many the general features of these simpler systems, The Ontogeny of Phenotypes 139

such as nonlinearity, emergent dominance of alleles, contingent neutrality, and poor correlation between variation at any one gene and the phenotype (in a typical genetic system this would be called variable penetrance), also occur in more complex systems. In addition, we would like to know whether more complex systems have additional properties that emerge from higher-level interactions that are not exhibited by these simpler systems. The emergent properties of the two simple systems are due to two characteristics: (1) nonlinear- Figure 11.7 ity of the relationship between genetic variation Complex relationship between genotype and phenotype (or environmental variation) and phenotypic can arise from nonlinear multifactorial interactions variation, and (2) unequal effects of each of the among the parameters of the developmental system that give rise to the phenotype. The curve in this graph determinants on the phenotype. The interaction illustrates such a (hypothetical) relationship. Three alle- of the nonlinearity with the inequality of effects les and five genotypes of increasing genetic value are makes the strength of the association between shown on the horizontal axis. Due to the nonlinearity a given determinant and the phenotype context of the system, the order of phenotypic values does not dependent, and it changes the absolute as well as correspond to the order of genetic values, and the dom- the relative magnitude of the effect of a given inance among alleles depends on their absolute, not determinant in different genetic backgrounds and their relative, values. As in figure 11.3, the shape of the in different environments. All systems that are relationship is determined by all interacting compo- nonlinear and that have components that inter- nents of the system. Accordingly, mutations at other act nonadditively and with different quantitative genes can change the shape of the curve, and this alters the dominance relationships of the focal gene. effects on the phenotype will exhibit the emergent properties outlined above (dominance, contingent neutrality, and incomplete and variable level of contingency emerges, namely in the dom- penetrance). inance of alleles. Dominance among any two alle- More highly complex systems can deviate from les in such systems depends on exactly where the general characteristics of the simple systems along the curve they occur (figure 11.7). Dom- we have studied in the following ways. More inance relations can change sign in different complex systems can have compensatory or regions of the genetic value axis, or switch from buffering mechanisms (such as feedback loops, overdominance to underdominance. Finally, duplicate pathways, and alternative pathways) more complex developmental systems are typical- that make the phenotype insensitive to variation ly composite, built up of a large number of paral- in one or more of its determinants. In such cases, lel and sequential modules that are themselves variation in those particular determinants would relatively simple (for instance, early embryonic not be expressed as phenotypic variation and development in Drosophila is controlled by a would be essentially neutral to selection. More coupled sequence of diffusion-gradient threshold complex systems can also be more nonlinear, events in which one thresholded gradient de- with U- or S- or N-shaped curves that relate fines the site of expression of a new gene whose genetic (or environmental) variation to pheno- product diffuses and activates a subsequent gene typic variation. In such cases variable penetrance wherever it is above or below some threshold, and contingent neutrality still apply, but a new and so on). If the interactions among the modules 140 H. Frederik Nijhout

have nonlinear components, then new context References dependencies will arise. Freeman, G., and J. W. Lundelius. (1982). The developmental genetics of dextrality and sinistrality in the Conclusion gastropod Limnaea peregra. Wilhelm Roux Archives 191: 69–83. Phenotypes are properties of developmental sys- Haines, J. L., and M. A. Pericak-Vance. (1998). Ap- tems. The rate constants and equilibrium con- proaches to Gene Mapping in Complex Human Diseases. stants of the dynamical processes that define a New York: Wiley-Liss. developmental system are themselves defined by, Holton, T. A., and E. C. Cornish. (1995). Genetics and and can be modified by, both genetic and envi- biochemistry of anthocyanin biosynthesis. Plant Cell 7: ronmental factors. In this view of developmental 1071–1083. systems, the primary role of genes is to supply Kacser, H., and J. A. Burns. (1973). The control of flux. enzymes that catalyze biochemical reactions and Symposia of the Society for Experimental Biology 27: that control the rates of various processes. Allelic 65–104. variation leads to variation in the dynamics of Kacser, H., and J. A. Burns. (1981). The molecular developmental processes, and these lead to varia- basis of dominance. Genetics 97: 639–666. tion in phenotypes. Nonlinearity in the associa- Kacser, H., and J. W. Porteous. (1987). Control of tion between genetic variation and phenotypic metabolism: What do we have to measure? Trends in variation leads to two properties of interacting Biochemical Sciences 12: 5–14. systems. At the level of the alleles of a single gene, Keightley, P. D., and H. Kacser. (1987). Dominance, nonlinearity produces dominance. This means pleiotropy and metabolic structure. Genetics 117: 319– that the effect of the two alleles in a diploid is 329. not simply additive, but that the phenotype asso- King, R. C., and W. D. Stansfield. (1985). A Dictionary ciated with the heterozygote is closer to that of of Genetics. Oxford: Oxford University Press. one of the homozygotes than to the other. At the Klingenberg, C. P., and H. F. Nijhout. (1999). Genetics level of the interacting genes within the devel- of fluctuating asymmetry: A developmental model of opmental system, nonlinearity leads to context- developmental instability. Evolution 53: 358–375. dependent variation in the absolute effect of a Mitchell, H. K., and N. S. Petersen. (1982). Devel- gene on the phenotype. This in turn leads to con- opmental abnormalities in Drosophila induced by heat tingent neutrality of a significant fraction of the shock. Developmental Genetics 3: 91–102. genes whose products play a role in the develop- Nijhout, H. F., and S. M. Paulsen. (1997). Develop- mental system. Due to contingent neutrality it is mental models and polygenic characters. American Naturalist 149: 394–405. possible to detect only a small fraction of the genes that affect a phenotype, and the composi- Oyama, S. (1981). What does the phenocopy copy? Psychological Reports 48: 571–581. tion of this fraction is different in different genetic and environmental backgrounds. A gene that has Santamaria, P. (1979). Heat shock induced pheno- copies of dominant mutants of the bithorax complex in a major effect on a phenotype in one genetic con- D. melanogaster. Molecular and General Genetics 172: text could appear as a modifier gene of small 161–163. effect in a different context, and could not be de- Schlichting, C. D., and M. Pigliucci. (1998). Phenotypic tected at all in yet another context (even though Evolution: A Reaction Norm Perspective. Sunderland, its characteristics and amount of variation do not MA: Sinauer. change). The degree to which genetic variation is Sturtevant, M. H. (1923). Inheritance of direction of associated with phenotypic variation is, there- coiling in Limnaea. Science 58: 269–270. fore, a property not of the genes but of the devel- Vogelstein, B., and K. W. Kinzler. (1997). The Genetic opmental system in which the genes play a role. Basis of Human Cancer. New York: McGraw-Hill. 12 The Development of Ant Colony Behavior

Deborah M. Gordon

Developmental systems theory seeks alternatives I study a population of about three hundred to the idea that an organism’s development is the harvester ant colonies in the desert of southeast- expression of information already in place before ern Arizona. Each year I census all of the col- development began. It is obvious that such alter- onies, which are individually labeled. This census, natives are needed to understand the development now in its fifteenth year, tells me how old each of an ant colony. A colony’s behavior arises from colony is, and that has made it possible to track the relations of ants to each other and to the rest the development of colony behavior, as well as of their environment (Gordon 1999). The col- the demography of the population. ony’s behavior can not be the product of pre- A colony’s behavior changes as it grows older packaged instructions, because there is nowhere and larger. Colonies of any age must perform cer- to locate the package. As the colony grows older, tain tasks, such as building a nest, taking care of the ants die, to be replaced by their younger sis- the juvenile stages (eggs, larvae, and pupae), and ters. The queen lays the eggs but does not direct collecting food. Task allocation is the process the behavior of the ants. Each ant’s behavior is that adjusts the numbers of workers engaged in based on local information. The aggregate of all each task, in response to changes in the environ- ants’ behavior produces the development of the ment and the needs of the colony (Gordon 1996). colony. Task allocation operates differently in young, An ant colony consists of many sterile female small colonies and old, large ones. Old colonies workers and one or more reproductive females or respond in a more stable way to environmental queens. The queen produces new workers. Ant perturbations than do younger ones. In addition, colonies produce offspring colonies, so a colony young and old colonies differ in their relations can be considered as an individual organism: A with neighboring colonies of the same species. colony is born, grows older, reproduces and, Ant colony behavior develops as a result of the when the queen dies and there is no one to pro- ants’ responses to the changing contexts they ex- duce more workers, the colony dies. The behav- perience as the colony grows older. The most ior of the colony develops, year after year, as new obvious change over time in the colony is its size ants are born into the colony and die. This chap- (figure 12.1; Gordon 1992). Colony size thus ter is about the development of colony behavior seems to be the first place to look for explana- in the red harvester ant (Pogonomyrmex barba- tions for age-dependent changes in colony behav- tus), a desert species that eats the seeds, mostly ior. In a large colony with many ants, each ant of grasses, that it collects from the ground. In a experiences a different environment from the one mature harvester ant colony, each summer the that surrounded an ant in the smaller, younger queen produces reproductives, winged queens colony. Colony size affects the pattern of an ant’s and males who fly off to mate with the reproduc- interactions with other ants: more ants can lead tives of other colonies. After mating, the males to a higher rate of contact among ants. Ants die and the newly mated queens found new modify their environment by creating and modi- colonies. A colony may live for fifteen to twenty fying a nest, adjusting the amounts of vegetation, years (Gordon 1991), beginning with a single refuse and chemical traces on the surface of the queen and reaching a stable size of about ten nest mound, and by altering through food collec- thousand workers when the queen, and the tion the distribution of food in the foraging area colony, are five years old. Individual ants live around the nest. More ants modify their environ- only a year. ment more, and in different ways. 142 Deborah M. Gordon

company of a few other workers, perhaps some eggs and larvae. Somehow some ants get out and collect some food, and this food is given to the larvae. Some ants must begin to dig. A year later, by the time the colony has reached its first summer, an ant emerging from its pupal case finds a very different environment. An ant in the one- year-old colony is part of a functioning colony. Several tasks are carried out simultaneously. The processes that shuttle ants from one task to another are in place. There is a nest with a system of tunnels and chambers. In a one-year-old colony, ants come barging out of the nest entrance Figure 12.1 apparently intent on some activity, perhaps head- How colony size, in numbers of workers, depends on colony age. ing straight for a foraging trail or ambling over to the midden pile to move bits of refuse from one place to another. But ants in the first batch A population of colonies thus provides an ex- of workers seem more uncertain (though I have ample of niche construction (sensu Lewontin been able to observe them only in the laboratory). 1983). In harvester ants, this process of niche It might be tempting to believe the first forager construction unfolds over the fifteen- to twenty- to walk out of the nest is “programmed” to col- year lifespan of the colony. To find proximate lect food, bring it back to the nest and feed it to explanations for the ways colony behavior the queen who then produces more and more of changes as the colony grows older, we must ex- these “hard-wired” automata. But there is no rea- amine how the colony modifies its environment son to believe the first ant to come out has some as it grows larger. prior knowledge of what to do, any more than the rest of the ants do. The ant is born in a chamber at the bottom of a straight tunnel. It can go Founding a Colony only around the chamber or up the tunnel. Even- tually it goes up. It gets outside. Being an ant, the The beginning of a colony’s development is the events it encounters, mostly olfactory, elicit re- stage I know least about. Six to eight weeks after sponses that get it eventually to food, and even- the mating flight, the first workers emerge. These tually it picks up some food, and eventually it ants are tiny; the queen fed them as larvae from finds its way back inside. her own fat reserves and they are much smaller What is contained in the notion that “being than later cohorts of workers. These first ants go an ant,” there are some things that the ant is out to forage, with only four to six weeks of for- likely to do? It has an ant’s body, and that offers aging time remaining before the colony’s first a set of possibilities for what it perceives, and winter sends them back inside the nest. By the how it moves, and what it can pick up. Maybe next summer, the colony will have a small nest the joint actions of ants’ bodies are likely to pro- with many tunnels and chambers, a small mound duce chambers with a certain curvature of the above the ground, and five hundred to a thou- walls, because of the ways they move their heads sand ants, and the foragers will begin to explore as they scoop up sand. Being an ant also some- the neighborhood. how establishes a set of possibilities for how ants An ant in the first batch of workers emerges respond to each other. We know that in many from its pupal case into a single tunnel and the The Development of Ant Colony Behavior 143

species of ants, individuals differ in their behav- low probability of survival because there are not ior, especially in how active or mobile they are. enough ants to compensate for each other’s in- For some tasks, in some species, it has been competence or to ensure that ants perform the shown that experience contributes to an ant’s necessary tasks. Maybe it often happens that the task performance. first few foragers out never find their way back; The queen has a minimal role in organizing the probability of getting back to the nest may be colony behavior. She does not direct the behavior low. In larger colonies, ants can use their forag- of the ants in the colony. She has no authority. In ing nestmates as a cue to the direction back to the fact, there is no evidence that any ant ever tells nest, perhaps by following a gradient in the den- another what to do. So it would be absurd to sity of nestmates. But in a very small colony, a imagine that somehow the queen carries inside forager has a small chance of encountering a her all of the instructions that later produce the nestmate on the way home. behavior of the mature colony of ten thousand The accumulation of mistakes similar to those ants. A queen’s life, after her brief, early excur- of the first foragers may prevent small colonies sion outside, consists of eating and laying eggs. In from successfully performing tasks inside the the laboratory, the queen of a large colony usu- nest. In large colonies, some ants dig new cham- ally has some workers nearby. She rarely moves bers and others carry out the soil. The ants that around much, and often seems to be standing scrape soil from the wall of the chamber might doing nothing. Sometimes workers feed her, or just put it down on the floor. Other ants walk by, groom her, or pile up eggs as she lays them. and eventually some ant picks up a bit of soil and Winged reproductives are produced three to six takes it somewhere else, maybe toward the weeks before the day of the mating flight, so a entrance, maybe not. Eventually soil builds up queen will have spent only a tiny portion of her near the entrance where ants pick it up and carry life in her parent colony and will have little prior it out. In a very small colony, there may not be experience of life in a mature colony. She is out- enough ants to ensure that on average, the soil side of a nest for only a day or two to mate and will get out. Perhaps in some tiny new colonies, get to the place where she will dig a new nest. new chambers do not get built when they are Most newly mated queens die before they ever needed. Small colonies may die because the dig a nest; many are eaten by birds and lizards. If chances of getting things done are low. But as a a newly mated queen does manage to start a nest, colony grows larger, eventually there comes a she digs furiously for the first two or three days, point when it has su≤cient numbers of ants creating a single tunnel not much wider than her working in a su≤ciently modified environment own body, about eighteen inches deep. She never that on average, the colony accomplishes its tasks comes out of the nest again, except if the colony often enough to survive. moves to a new nest. Apparently many new colonies do not survive, because there are many more nests of founding Colony Growth queens on the site each year (five hundred to one thousand) than there are one-year-old colonies Colonies grow quickly for the first four or five the following year (twenty to fifty) (Gordon and years (figure 12.1). If a colony manages to survive Kulig 1996). I do not know if the queens die to be two years old, it will probably last out the before they can produce any workers, or if the next fifteen years (Gordon and Kulig 1998). A first batch of workers does not provide enough three- to four-year old colony, with four thou- food to support the queen through the winter. sand to six thousand ants, is in the steepest por- Perhaps the small initial group of workers has a tion of the colony’s growth curve (Gordon 1992). 144 Deborah M. Gordon

By five years of age, a colony can reach its mature the vegetation bit by bit. During the weeks of size of ten thousand ants and begin to reproduce. rainy weather that come each summer, the young This is a stable size; the colony ceases to grow, colonies work on creating a nest entrance that although, because ants live a year, all of its ants will be raised above the level of the ground, which must be replaced each year. Colony growth may prevents water from pouring into the nest en- be limited by the queen’s capacity to lay eggs; to trance. Using tiny twigs they make a thatched maintain a size of ten thousand ants she must lay collar around the nest entrance, and then cover it ten thousand eggs a year. A mature colony has an with dirt. established nest mound and a set of three to eight When a colony is three years old it begins habitual foraging directions, of which it chooses its period of most rapid growth, in numbers of about two to five each day. It begins to repro- ants. By this time most colonies have created a duce, sending winged reproductives to the annual small nest mound. They have begun to cover the mating flight (Gordon 1992). It may remain in mound with tiny pebbles, which they bring in this state for another fifteen years. from the surrounding area. Any bushes remain- Colonies vary in size. Some never seem to ing on the mound are dead, bare branches which reach the size of the largest colonies (ten thou- eventually break off. There is usually a discrete sand to twelve thousand ants), and some never pile or two of refuse, called the midden, with the reproduce. Many of these apparently stunted col- husks of the seeds the ants are currently milling. onies are in very crowded neighborhoods where Ants at work on the nest mound carry out sand competition for food may be especially intense. from construction inside the nest, or sort and pile I suspect that such colonies do not reproduce the midden. because they are not large enough, perhaps be- When we dig up colonies, we find that there is cause they do not have enough workers to collect a cone-shaped mass of chambers, with the top of su≤cient reserves of food. However, there are the cone corresponding to the nest mound on the also colonies that appear to be large but do not surface, and the length of the cone about the reproduce. We counted reproductives as they same as the diameter of the mound. A three-year- leave for the mating flight in two years and in old colony’s mound might be about half a meter both years, about a third of the colonies on the wide, with a correspondingly deep cone of cham- site that were of reproductive age did not send bers; the chambers of a mature colony’s nest may out any reproductives at all (Gordon and Wagner extend down one meter. Somewhere off the bot- 1997; Wagner and Gordon 1999). If the colony tom of the cone is a single tunnel leading down reaches two years of age it is likely to survive for another meter or even two, to a chamber where many years, but its development from ages two to we find the queen and brood when we excavate a five may determine whether it will reproduce. colony. These chambers can be as deep in a three- year-old colony as an older one. Probably the queen and brood do not usually stay so deep Development of the Nest underground. It is clear the brood is not usually this deep, because on warm days it is easily found The nest work of a small colony is different from in chambers just underneath the top surface of that of a larger one, because the one-year-old the mound. In the laboratory, ants bring the colony is still making its nest, while the older col- brood to heated boxes, and the queen is usually ony is maintaining an existing nest. To construct near the smaller brood, eggs and small larvae. a nest the ants have to kill any bushes growing Perhaps in the field, the queen moves along with where the nest is to be. They accomplish this by the brood as it is carried along a temperature gra- destroying the roots underground, and above- dient by the workers. ground by climbing into the bush and clipping off The Development of Ant Colony Behavior 145

In excavated nests we find some chambers 50 percent, works outside the nest. Though the packed with stored seeds, and some with refuse. total numbers in the nest change with colony age Colonies seem to use chambers and then build as shown in figure 12.1, so that a six-year-old col- more as they are needed. ony may be ten times as large as a one-year-old On the surface of the mound, the nest entrance one, the numbers active outside the nest in a large has a smooth runway leading into it because colony are rarely more than two times those in a the larger grains of dirt have been moved aside, small one. leaving only the smallest ones. I can see a few cen- In the larger colony, ants in undisturbed col- timeters into the nest with a fiber optics micro- onies are unlikely to switch tasks from one day to scope. Inside the nest entrance of a three-year-old the next. This suggests that an older colony has a colony, the walls of the tunnels appear to be larger number of ants available to perform exte- rough. In a colony of any age, there is a chamber rior tasks than does a smaller one. In a small just inside the nest entrance. In a three-year-old colony, an ant performing nest maintenance one colony, that chamber usually divides into just day may be shunted into foraging the next. two tunnels. Inside the nest of a colony five years Perhaps in a large colony, an ant tends to be sur- or older, the walls of the chambers and tunnels rounded by more ants doing its own task, so seem smooth. The chamber nearest the nest there is a small probability it will encounter an entrance may branch into three or more main opportunity to do a different task, or the stimulus tunnels. provided by ants engaged in another task. In a nest of any age, most chambers seem to When the colony’s environment changes, so do have at least two openings, but some have more. the numbers of ants engaged in the relevant task. I have never been able to trace the complicated If food appears, more ants forage. If the nest is topology of an entire nest. This topology is im- disturbed, more ants take up nest maintenance portant, though, because the structure of the nest and repair. These shifts lead to shifts in other, probably influences the flow of ants, and thus the unrelated activities. I found in perturbation ex- rate at which ants interact as they come in and periments that foraging, nest maintenance work, out of the nest. and patrolling are all related in this way: a shift in the numbers engaged in one task, caused by a change in the environment relevant to that task, Task Allocation leads to a change in the numbers engaged in the other tasks as well. These shifts arise from two Task allocation of a mature colony, five years or ways that ants of one task respond to changes in older, differs from that of a small, young one in the numbers performing another task: first, ants several ways (Gordon 1987, 1989a). switch tasks, and second, ants may decide to be- In a large colony, only about 25 percent of the come active, or remain inactive. colony works outside the nest, and large numbers These adjustments in the distribution of work- remain inactive inside. These inactive ants may ers performing colony tasks, differ in young and function as reserves that would emerge from the old colonies. The behavior of an older colony is nest if needed in some situation. Such a situation more homeostatic, and more stable, than that of has not occurred in the seventeen summers I have a younger one. The more a young colony is dis- observed these ants, but though seventeen years turbed, the more it changes. However, the more is a long time for a person it is utterly insignifi- an older colony is disturbed, the more its behav- cant on the evolutionary scale for an organism ior seems to resemble that of an undisturbed col- with a five-year generation time. In a young colony. I do not know what process accounts for ony, a larger proportion of the colony, perhaps this result. Perhaps in a larger colony, each ant’s 146 Deborah M. Gordon

behavior is somehow more buffered by the be- released colony, and surprisingly, they retreated havior of other ants. even more. A very young colony’s foraging area The finding that small and large colonies differ after the neighbor was released, was even smaller in behavior suggests that the rate of interaction than it had been before the experiment began. It an ant experiences may influence its task. In an was as though the sudden return of an absent older, larger colony, an ant has more opportuni- neighbor was a stronger deterrent than that neigh- ties to interact than in a smaller one. If an ant’s bor had been before it disappeared. behavior is influenced by interaction rate, then Colonies that are three to four years old, at the even if ants act according to the same rules in steep part of their growth curve, are more per- small and large colonies, the outcome will differ. sistent in encounters with neighbors than younger If an ant is likely to perform task X when it meets or older colonies. When a colony were enclosed, workers of task X at a certain rate, then large its three- to four-year-old neighbors, like neigh- colonies may have more ants performing task X bors of other ages, began to use the foraging area simply because each ant has more task X ants to of the enclosed colony. But when the enclosed meet. Such rules, of course, might involve pos- colonies were released, the three- to four-year- itive or negative feedback. An ant’s behavior is olds, unlike smaller or larger colonies, did not clearly subject to environmental influences as well leave their newly acquired foraging area. They as the influence of its interactions with others, but returned to the site day after day, although this interactions seem to play a role in task allocation. often involved fighting with the newly released For example, in laboratory studies we found that colonies. the probability an ant does midden work is cor- Colonies of five years or more seem to be related with the rate at which it encounters mid- much more staid in their relations with neighbors den workers (Gordon and Mehdiabadi 1999). We than the quickly growing three- to four-year-olds. also find that ants of different task groups differ When colonies were enclosed, older neighbors in cuticular hydrocarbons, so that in the course entered their foraging areas. When the enclosed of a brief antennal contact an ant can determine colonies were released, the older neighbors sim- the task of the ant it meets. We are currently in- ply retreated, foraging elsewhere much as they vestigating how interaction rate contributes to had before the enclosed colony disappeared. task allocation, and we hope that this work will Older colonies adjust their foraging trails so as elucidate why old colonies appear to be more sta- to avoid those of neighbors. Over the fifteen to ble in behavior than younger ones. twenty years of its life, a colony settles into relations with its neighbors. Ants distinguish nestmates from all other ants by a colony-specific Relations with Neighbors odor. At least in harvester ants, workers also distinguish ants of neighboring colonies from ants Very small colonies differ from larger ones in of more distant, stranger colonies, presumably their relations with their neighbors. I did some using the same colony-specific odor (Gordon experiments in which I enclosed some colonies 1989b). and observed the reactions of their neighbors Neighbor relations may change as colonies (Gordon 1992). When a very young colony had grow older because growth rates change. A three- its older, larger neighbor enclosed, its foraging year-old colony has four thousand workers to trails would shift toward the foraging area of the collect and process the food to make six thousand absent neighbor. Neighbors of all ages reacted to ants for the following year. By contrast, a six- the absence of an enclosed colony in this way. year-old colony has ten thousand workers to col- When the enclosed colony was released, the very lect and process the food to make a colony of the young colonies left the foraging area of the newly The Development of Ant Colony Behavior 147

same size the following year. The larvae consume It seems plausible that many of the changes most of the food of the colony; adult ants eat in behavior that we have observed arise from much less. So in a young, growing colony, there changes in colony size, which in turn lead to are more hungry larvae per forager than in an changes in the environment of the colony. Age- older one that is larger but of stable size. In labo- dependent changes in colony behavior seem to be ratory colonies there seems to be an increase in a consequence of the relation between a colony’s numbers foraging in the days following the ap- growth and its modification of its environment. pearance of larvae from eggs. This is the only This relation is complicated and not yet well un- direct evidence I have that the presence of larvae derstood. We do not know what determines the affects the behavior of foragers, but it seems clear rate of colony growth. The availability of food, this relation must be important. When there are how well the colony collects and processes the no larvae, none are fed; when there are larvae, food, the physiological condition of the queen, they are fed; somehow the behavior of foragers competitive pressure from neighbors, all are must be linked to that of the ants that feed larvae, likely to influence colony growth. Colony size, in and somehow that must be linked to the numbers turn, determines how ants interact with each of larvae present. Parts of this chain are under- other, how many ants are available to collect and stood for some social insect species, though we process food, and how a colony fares in competi- know little about the details in this species of har- tion with its neighbors. We know that all of these vester ant. Neighbor relations in the field are con- factors are related to colony size, though we do sistent with the speculation that a forager in a not yet know exactly how. growing three- to four-year-old colony is under One set of important questions about colony more pressure to obtain food than one in an older development concerns the first few months of colony of stable size, because the ratio of brood a colony’s life. What do the first workers do? to foragers is higher in the growing colony. Why is colony mortality so high at this stage? The relation between numbers of larvae to feed Careful observation of very young colonies, both and the behavior of foragers is probably medi- in the laboratory and in the field, would be very ated in some way by the state of the colony’s informative. A second set of questions concerns supply of stored food. Amount of stored food the relation between the physical structure of probably depends on colony size. I have exca- the nest, the flow of ants in and out of the vated colonies of known age to count the ants, nest, and the rate at which ants interact. Here but have not been able to measure food stores both theoretical and empirical work is needed. accurately. My impression, though, is that older, Mathematical models of the complicated three- larger colonies have larger food stores. In one dimensional movement of ants through tunnels colony of twelve thousand ants we found more would help to guide empirical research. A third than a liter of stored seeds. set of questions concerns the differences in task allocation in young and old colonies. Why is the behavior of older, larger colonies more stable Conclusion and homeostatic than that of smaller, younger ones? Again, both theoretical and empirical work Little is known about the development of social is needed. Models are needed that propose sim- insect colonies, because there are few species in ple, plausible rules at the individual level, which which individually labeled colonies have been fol- might lead to the outcomes we observe in real lowed over time. The work described here shows colonies. Such rules determine how an ant’s in- how much we have left to learn about the devel- teractions, with its environment and with other opment of the behavior of a harvester ant colony. ants, affects its decision about which task to per- 148 Deborah M. Gordon

form. Differences between theoretical and empir- Gordon, D. M. (1992). How colony growth affects for- ical results show the gaps in our understanding, ager intrusion in neighboring harvester ant colonies. and point the way to further empirical work. A Behavioral Ecology and Sociobiology 31: 417–427. final set of questions concerns the relation be- Gordon, D. M. (1996). The organization of work in tween food and colony growth. We need to un- social insect colonies. Nature 380: 121–124. derstand why some colonies grow faster than Gordon, D. M. (1999). Ants at Work: How an Insect others. A colony’s neighborhood influences its Society Is Organized. New York: Free Press. growth because neighbors compete for food; thus Gordon, D. M., and A. W. Kulig. (1996). Founding, we need to know how food influences colony foraging and fighting: Colony size and the spatial dis- growth to understand how neighbors affect col- tribution of harvester ant nests. Ecology 77: 2393–2409. ony development. Gordon, D. M., and A. Kulig. (1998). The effect of The development of colony behavior is the neighboring colonies on mortality in harvester ants. process that relates changes in colony size to Journal of Animal Ecology 67: 141–148. changes in the colony’s environment. Each ant Gordon, D. M., and N. Mehdiabadi. (1999). Encounter has the capacity to participate in this process, but rate and task allocation in harvester ants. Behavioral Ecology and Sociobiology 45: 370–377. that capacity is not separate from the process. An ant operates in the context of the colony. An ant Gordon, D. M., and D. Wagner. (1997). Neighborhood density and reproductive potential in harvester ants. alone could not function as an elemental unit of Oecologia 109: 556–560. colony behavior, because there would be no col- Lewontin, R. C. (1983). Gene, organism and environ- ony behavior within which to act. As the colony ment. In D. S. Bendall (Ed.), Evolution from Molecules grows, the ants encounter new circumstances. To to Men, pp. 273–285. Cambridge: Cambridge Univer- our knowledge, an ant in a young colony re- sity Press. sponds to a particular situation in much the same Wagner, D., and D. M. Gordon. (1999). Colony age, way as an ant in an old one. But an ant in a young neighborhood density and reproductive potential in har- colony finds itself in different situations from vester ants. Oecologia 119: 175–182. an ant in an old one. An older colony is larger than a young one, and its environment has been modified by past cohorts of ants. To explain colony development we need to know how the behavior of mature colonies emerges from the changing relations of individual ants and their environment.

References

Gordon, D. M. (1987). Group-level dynamics in harvester ants: Young colonies and the role of patrolling. Animal Behaviour 35: 833–843. Gordon, D. M. (1989a). Dynamics of task switching in harvester ants. Animal Behaviour 38: 194–204. Gordon, D. M. (1989b). Ants distinguish neighbours from strangers. Oecologia 81: 198–200. Gordon, D. M. (1991). Behavioral ﬂexibility and the foraging ecology of seed-eating ants. American Natur- alist 138: 379–411. 13 Behavioral Development and Darwinian Evolution

Patrick Bateson

The One True Cause “We have found a particular behavioral difference between individuals which is associated with The effectiveness of education, the role of parents a particular genetic difference, all other things be- in shaping the characters of their children, the ing equal.” The media and the public might start causes of violence and crime, and the roots of to get the message if plain language like this were personal unhappiness are self-evidently matters used routinely. of huge importance. And, like so many other fun- Fortunately, it no longer seems obscure to damental issues about human existence, they all many others to refer to developmental processes relate to behavioral development. The catalog as systems. Susan Oyama’s witty and eloquent continues. Do bad experiences in early life have a writing (Oyama 1985) has inspired many others lasting effect? Is intelligence in the genes? Can (see the many excellent contributions in Lerner adults change their attitudes and behavior? When [1998]). I like to think that this change in thinking faced with such questions, many people want may reﬂect commonplace experience. From an simple answers. They want to know what really early age many people are exposed to computer makes the difference. Explanations in terms of games, in which outcomes depend on a combina- combinations of conditions is perceived as wooly, tion of conditions. Children playing such games obscurantist, and running counter to the success- meet, for instance, in the dungeon of the dark ful analytical programs of science. castle a dragon that can only be killed with a The search for simple environmental origins, special sword that had to picked up on the top which had wide appeal in the mid-twentieth cen- of the crystal mountain; in doing so, they have tury, has been partly superseded by an equally begun to accustom themselves to the contextual skewed belief in the overriding importance of conditional character of the real world. The lin- genes. If pressed, scientists may concede that ear thinking of a previous generation, with every their talk of genes “for” shyness, maternal behav- event having a single cause, is slowly being re- ior, promiscuity, verbal ability, criminality, or placed by an understanding of coordinated pro- whatever, is merely a shorthand. They may none- cess. That, at least, is my optimistic view. In this theless try to legitimize the language of genes chapter I shall consider why systems approaches “for” behavior, by pointing to seemingly straight- are essential to an understanding of behavioral forward examples like the genes for eye color. development. I shall also argue that they can be Nonetheless, the notion of genes “for” behavior successfully married to a Darwinian approach to undoubtedly corrupts understanding. evolution and to current utility, often viewed as A single developmental ingredient, such as a antagonistic to the developmental systems think- gene or a particular form of experience, might ing (Ingold 1986; Kauffman 1993). produce an effect on behavior, but this certainly does not mean that it is the only thing that matters. Even in the case of eye color, the notion that How Much Nature, How Much Nurture? the relevant gene is the cause is misconceived, because all the other genetic and environmental The importance of both genes and environment ingredients that are just as necessary for the de- to the development of all animals, including velopment of eye color remain the same for all humans, is obvious enough. This is true even individuals. A more honest translation of the for apparently simple physical characteristics, “gene for” terminology would be something like: let alone complex psychological variables. Take shortsightedness, for example. Myopia runs in 150 Patrick Bateson

families but is segregated from other characteris- conditions, the genetic strain to which a pig be- tics, suggesting that it is genetically inherited. But longs will predict its adult body size better than it is also affected by the individual’s experience. other variables such as the number of piglets in a Both a parental history of myopia and, to a lesser sow’s litter. If the animal in question is a cow and extent, the experience of spending prolonged pe- the breeder is interested in maximizing its milk riods studying close-up objects will predispose a yield, then knowing that milk yield is highly her- child to become shortsighted (Zadnik 1997). itable in a particular strain of cows under stan- Different styles of doing science are brought dard rearing conditions is important. into play when dealing with a case like the de- But behind the deceptively plausible ratios lurk velopment of myopia, in which different types of some fundamental problems. For a start, the her- factor affect the outcome. Those who have an itability of any given characteristic is not a fixed aversion to systems thinking sometimes like to and absolute quantity—tempted though many present the purely statistical approach as the only scientists have been to believe otherwise. Its value scientific solution to the age-old problem of the depends on a number of variable factors, such relative contributions of the different factors. The as the particular population of individuals that question becomes: “How much of the variation has been sampled. For instance, if heights are between individuals in a given character is due to measured only among people from a≥uent back- differences in their genes, and how much is due to grounds, then the total variation in height will be differences in their environments?” The suggested much smaller than if the sample also includes answer, satisfying old-style linear thinking, was people who are small because they have been un- provided by a measure called “heritability.” The dernourished. The heritability of height will con- meaning of heritability is best illustrated with sequently be larger in a population of exclusively an uncontroversial characteristic such as height, well-nourished people than it would be among which is clearly influenced by both the indi- people drawn from a wider range of environ- vidual’s family background (supposedly genetic ments. Conversely, if the heritability of height influences) and nutrition (environmental influ- is based on a population with relatively similar ences). The variation between individuals in genes—say, native Icelanders—then the figure height that is attributable to variation in their will be lower than if the population is genetically genes may be expressed as a proportion of the more heterogeneous; for example, if it includes total variation within the population sampled. both Icelanders and African pygmies. Thus, at- This index is known as the heritability ratio. The tempts to measure the relative contributions of higher the figure, which can vary between 0 and genes and environment to a particular character- 1.0, the greater the contribution of genetic varia- istic are highly dependent on who is measured tion to individual variation in that characteristic. and in what conditions. So, if people differed in height solely because they Another problem with heritability is that it differed in their genes, the heritability of height says nothing about the ways in which genes and would be 1.0; if, on the other hand, variation in environment contribute to the biological and height arose entirely from individual differences psychological cooking processes of development in environmental factors such as nutrition, then (Bateson and Martin 1999). This point becomes the heritability would be 0. obvious when considering the heritability of a Calculating a single number to describe the rel- characteristic such as “walking on two legs.” Hu- ative contribution of genes and environment has mans walk on fewer than two legs only as a result obvious attractions. Estimates of heritability are of environmental influences such as war wounds, of undoubted value to animal breeders, for ex- car accidents, disease, or exposure to teratogenic ample. Given a standard set of environmental toxins before birth. In other words, all the varia- Behavioral Development and Darwinian Evolution 151

tion within the human population results from conditions) was the only one that mattered. But environmental influences, and consequently the it was not that simple. In the third type of envi- heritability of “walking on two legs” is zero. And ronment, where the rearing conditions were inter- yet walking on two legs is clearly a fundamental mediate in complexity, rats from the two strains property of being human, and is one of the more differed markedly in their ability to navigate the obvious differences between humans and other maze. These genetic differences only manifested great apes such as chimpanzees or gorillas. It ob- themselves behaviorally in this sort of environ- viously depends heavily on genes, despite having ment. Varying both the genetic background and a heritability of zero. Low heritability clearly the environment revealed a statistical interaction does not mean that development is unaffected by between the two influences.1 genes. An overall estimate of heritability has no If a population of individuals is sampled and meaning in a case such as this, because the effects the results show that one behavior pattern has of the genes and the environment do not simply a higher heritability than another, this merely add together to produce the combined result. The indicates that the two behavior patterns have de- effects of a particular set of genes depend criti- veloped in different ways. It does not mean that cally on the environment in which they are ex- genes play a more important role in the devel- pressed, while the effects of a particular sort of opment of behavior with the higher heritability. environment depend on the individual’s genes. Important environmental influences might have Even in animal breeding programs which use her- been relatively constant at the stage in devel- itability estimates to practical advantage, care is opment when the more heritable behavior pat- still needed. If breeders wish to export a particu- tern would have been most strongly affected by lar genetic strain of cows that yield a lot of milk, experience. they would be wise to check that the strain will Yet another serious weakness with heritability continue to give high milk yields under the differ- estimates is that they rest on the extraordinary ent environmental conditions of another country. assumption that genetic and environmental in- Many cases are known where a strain that per- fluences are independent of one another and do forms well on a particular measure in one envi- not interact. The calculation of heritability as- ronment does poorly in another, while a different sumes that the genetic and environmental contri- strain performs better in the second environment butions can simply be added together to obtain than in the first. the total variation. In many cases this assumption A further point about the mutual actions of is clearly wrong. For example, in one study of individual on the environment and environment rats the animals’ genetic background and their on the individual is brought out nicely by the rearing conditions were both varied; rats from example of myopia already briefly mentioned. two genetically inbred strains were each reared in Nesse and Williams (1994) point out that the one of three environments, differing in their rich- growth of the cornea is affected by the sharpness ness and complexity (Cooper and Zubek 1958). of the image on the retina. Objects close to the The rats’ ability to find their way through a maze eye, such as books, cause the cornea to grow so was measured later in their lives. Rats from both that the image of print is less fuzzy. Individual genetic strains performed equally badly in the differences in the feedback mechanism mean that maze if they had been reared in a poor environ- some people in a modern environment respond ment (a bare cage) and equally well if they had more to the experience of reading than others and been reared in a rich environment filled with toys consequently become more shortsighted. In an and objects. Taken by themselves, these results environment empty of books such people would implied that the environmental factor (rearing not be shortsighted. 152 Patrick Bateson

ceptors. If the incubation temperature is above Alternative Lives 30≥C, a different set of genes is activated, producing female hormones and receptors instead Striking examples of interaction are found (Crews 1996). It so happens that in alligators the throughout the animal kingdom when genetically sex determination works the other way around, identical individuals develop in totally different such that eggs incubated at higher temperatures ways, depending on environmental cues they reproduce males. (In humans and other mammals, ceived when they were young. After a fire on the by contrast, the sex of each individual is generally high grassland plains of East Africa, for example, determined genetically at conception; if it inherits the young grasshoppers are black instead of be- only one X sex chromosome, it becomes male.) ing the normal pale yellowish-green. Something Each grasshopper and turtle starts life with the has switched the course of their development capacity to play two distinctly different develop- onto a different track. The grasshopper’s color mental tunes—green or black, male or female. A makes a big difference to the risk that it will be particular feature of the environment then selects spotted and eaten by a bird, and the scorched which of those tunes the individual will play dur- grassland may remain black for many months ing its life. And once committed, the individual after a fire. So matching its body color to the cannot switch to the other tune. Once black as blackened background is important for its sur- an adult, the grasshopper cannot subsequently vival. The developmental mechanism for making change its color to green, just as a male turtle can- this switch in body color is automatic and de- not transform itself into a female. pends on the amount of light reflected from the The broad pattern of an individual’s social and ground (Rowell 1971). If the young grasshoppers sexual behavior may also be determined early in are placed on black paper they are black when life, with the individual developing along one of they molt to the next stage. But if they are placed two or more qualitatively different tracks. Many on pale paper the molting grasshoppers are the examples are found in the animal kingdom. The normal green color. The grasshoppers actively caste of a female social insect is determined by select habitats with colors that match their own. her nutrition early in life. The main egg producer If the color of the background changes they can of an ant colony, the queen, is part of a teeming also change their color at the next molt to match nest in which some of her sisters care for her off- the background, but they are committed to a spring, others forage, yet others clean or mend color once they reach adulthood. the nest, and finally other sisters specialize in Turtles, crocodiles, and some other reptiles guarding it (Wilson 1971). Locusts may or may commit themselves early in life to developing not become migratory, depending on crowding; along one of two different developmental tracks when the numbers living in a given area build up, and like grasshoppers, they do so in response their offspring develop musculature and behav- to a feature of their environment (Janzen and ior suitable for long flights and then the whole Paukstis 1991). Each individual starts life with swarm moves off (Pener and Yerushalmi 1998). the capacity to become either a male or a female. Vole pups born in the autumn have much thicker The outcome depends on environmental temper- coats than those born in spring; the cue to pro- ature during the middle third of embryonic de- duce a thicker coat is provided by the mother be- velopment (Yntema and Mrosovsky 1982). If the fore birth. The value of preparing in this way for eggs from which they hatch are buried in sand colder weather is obvious (Lee and Zucker 1988). below 30≥C, the young turtles become males. If, The sexual behavior of some primates can also however, the eggs are incubated at above 30 ≥C develop along two or more distinctly different they become females. Temperatures below 30 ≥C tracks. An adult male gelada baboon, for exam- activate genes responsible for the production of ple, will typically defend and breed with a harem male sex hormones and male sex hormone re- Behavioral Development and Darwinian Evolution 153

of females. After a relatively brief but active re- Each of us started life with the capacity to live productive life, he is displaced by another male many different lives, but each of us lives only one. and never breeds again. To position himself so In one sense individual humans are obviously that he can acquire and defend a harem, the male bathed in the values of their own particular cul- must grow rapidly at puberty. He develops the ture and become committed by their early experi- distinctive golden mane of a male in his prime ence to behaving in one of many possible ways. and becomes almost twice the size of the females. Differences in early linguistic experience, for ex- However, when many such males are present in ample, have obvious and long-lasting effects. By the social group, an adolescent male may adopt a the end of a typical high school education, a distinctly different style of reproductive behavior. young American will probably know about fifty He does not develop a mane or undergo a growth thousand different words (Pinker 1994). The spurt. Instead, he remains similar in appearance words are different from those used by a Russian and size to the females. These small males hang of the same age. In general, individual humans around the big males’ harems, sneakily mating imbibe the particular characteristics of their cul- with a female when the harem-holder is not pay- ture by learning (often unwittingly) from older ing attention. Because the small, sneaky male people. When environmental conditions select a never has to fight for females, he is likely to have particular developmental route in animals, the a longer, if less intense, reproductive life. If he mechanisms involved are likely to be different; lasts long enough he may even do better in terms learning may not enter into the picture at all. of siring offspring than a male who pursues the Even so, is it possible that some aspects of human alternative route of growing large and holding a development are triggered by the environment, harem. These two different modes of breeding as though the individual were a jukebox? Was behavior represent two distinctly different devel- each of us conceived with the capacity to develop opmental routes, and each male baboon must along a number of different tracks each of which commit himself to one or other of them before is preadapted to the circumstances in which the puberty (Dunbar 1984). individual finds itself? And does the environment All of these examples illustrate a surprising select the particular developmental track that aspect of development that has intriguing impli- each of us follows? cations for humans. In each case, the individual A series of studies led by the epidemiologist animal starts its life with the capacity to develop David Barker, which assessed people across their in a number of distinctly different ways. Like a entire lifespan from birth to death, has lent jukebox,2 the individual has the potential to play strength to the suggestion that human develop- a number of different developmental tunes. But ment may also involve environmental cues that during the course of its life it plays only one tune. prepare the individual for a particular sort of en- The particular developmental tune it does play is vironment (Barker 1998). The work was based in selected by a feature of the environment in which part on a large sample of men born in the English the individual is growing up—whether it be the county of Hertfordshire between 1911 and 1930. color of the ground, the temperature of the sand, Those men who had had the lowest body weights the type of food, or the presence of other males. at birth and at one year of age were most likely to Furthermore, the particular tune that is selected die from cardiovascular disease later in life. The from the developmental jukebox is adapted to the heaviest babies faced a subsequent risk of dying conditions in which it is played. from cardiovascular disease that was only half Is it the case that people, like grasshoppers or the average for the group as a whole, whereas the baboons, are conceived with the capacity to play risk for the smallest babies was 50 percent above a number of qualitatively different developmental average (in other words, three times greater than tunes—in other words, to live alternative lives? that for the largest babies). Individuals who had 154 Patrick Bateson

been small babies were also more likely to suffer into problems if instead they find themselves from diseases such as diabetes and stroke in growing up in an a≥uent industrialized society to adulthood. which they are poorly adapted. That, at least, is How could a link have arisen between an indi- the hypothesis. vidual’s birth weight and his physical health de- People who grow up in impoverished condi- cades later? The evidence pointed to a connection tions tend to have a smaller body size, a lower with the mother’s nutritional state: women with metabolic rate, and a reduced level of behav- poor diets during pregnancy had smaller pla- ioral activity (Waterlow 1990). These responses centas, and forty years later their offspring had to early deprivation are generally regarded as higher blood pressure (a risk factor for cardio- pathological—just three of the many damaging vascular disease and stroke). But the links with consequences of poverty. But they could also be maternal nutrition went much further back than viewed as part of a package of characteristics that pregnancy. Measurements of the mothers’ pel- are appropriate to the conditions in which the in- vises revealed that those who had a flat, bony dividual grows up—in other words, adaptations pelvis tended to give birth to small babies. These to an environment that is chronically short on small babies, after they had grown up, were much food, rather than merely the pathological by- more likely as adults to die from stroke. The im- products of a bad diet. Having a lower metabolic plication was that poor nutrition during their rate, reduced activity, and a smaller body all help mother’s childhood affected the growth of her to reduce energy expenditure, which can be cru- pelvis which, in turn, curtailed the growth of her cial when food is usually in short supply. offspring during pregnancy, which, in turn, in- Now this conjecture might well be regarded creased her offspring’s risk of stroke and cardio- as offensive. It could be seen as encouraging the vascular disease in adulthood. rich to look complacently at their impoverished Poor maternal physique and health are as- fellow human beings, by arguing that all is for sociated with reduced fetal growth, with con- the best in this best of all possible worlds (as sequences for the offspring’s later health. The Voltaire’s Doctor Pangloss would have had it). question arises, then, as to whether these con- Merely to assert that every human develops the nections make sense in adaptive terms. Could it body size, physiology, biochemistry, and behav- be that, in bad conditions, the pregnant woman ior that is best suited to their station in life would unwittingly signals to her unborn baby that the indeed be banal. The point, however, is not that environment which her child is about enter is the rich and the poor have the same quality of likely to be harsh? (Remember that we are think- life. Rather, it is that, if environmental conditions ing here about what might have been happening are bad and likely to remain bad, individuals ex- tens of thousands of years ago as these mecha- hibit adaptive developmental responses to those nisms were evolving in ancestral humans.) And conditions. To put it simply, they are designed to perhaps this weather forecast from the mother’s make the best of a bad job. body results in her baby being born with adapta- Of course, many of the long-term effects on tions, such as a small body and a modified me- health of a low birth weight may simply be by- tabolism, that help it to cope with a shortage of products of the social and economic conditions food. This hypothetical set of adaptations has that stunted growth in the first place. Ignorance been called the “thrifty phenotype” (Hales and and shortage of money make the prevention and Barker 1992; Hales, Desai, and Ozanne 1997). treatment of disease more di≤cult; overcrowd- And perhaps these individuals with a thrifty phe- ing, bad working conditions, and poverty pro- notype, having small bodies and specialized me- duce psychological stress and increase the risk of tabolisms adapted to cope with meager diets, run infection. People with little money have poorer Behavioral Development and Darwinian Evolution 155

diets, and adverse social or physical factors that the improvements in nutrition have occurred foster depression and hopelessness increase the more recently, the rate of increase in height risks of disease (Wilkinson 1996). In industrial- lagged behind Britain but has been at almost ized nations the poor and unemployed have more three times the rates found in Britain in the last illnesses and die sooner than the a≥uent. But so- few decades. In the United States the trend to- cial and economic conditions do not account for ward ever taller offspring in successive genera- everything, because the connections between low tions, which started earlier than in Britain, has birth weight and subsequent health are still found leveled off in recent years. These findings suggest among babies born in a≥uent homes (Barker an upper bound on the effect of nutrition on hu- 1998). man height. Whether or not the thrifty phenotype hypothesis proves to be correct, everybody agrees that environmental conditions early in development The Demise of Heritability have a significant impact on many other aspects of human biology, including size. People are get- The examples of condition-dependent develop- ting bigger (Eveleth and Tanner 1990). For de- ment do not pose any problems for evolutionary cades now, the average height of men and women biologists, even though they should give pause to in industrialized countries has been steadily in- those who search for universals. From an adap- creasing. Although some of the height differences tationist standpoint, the development of a pheno- between people are due to genetic differences, the type appropriate to the circumstances in which general trend for average height to increase is al- the animal finds itself makes a great deal of sense. most certainly due primarily to improvements in Nevertheless, the striking ways in which environ- nutrition and, to a lesser extent, health. Hence, mental factors can trigger one of a set of alterna- successive generations of the same family have tive responses pose serious di≤culties for those grown taller despite having a similar genetic behavior geneticists who seek to partition varia- makeup. tion into genetic and environmental components. The environmental improvements that have led And worse is to come. The conventional analyti- to this general increase in height have affected cal method that partitions behavioral variation males and females somewhat differently. Men into genetic and environmental components may have been growing taller faster than women. For be misleading in a different way. The two major example, the average height of men in Britain has contributors to variation may not combine even been increasing at a rate of just over one centime- in nonlinear fashion to produce their overall ef- ter every ten years, whereas the average height of fect. For example, the performances of adopted women has been increasing at about one third of children in tests of cognitive ability are related to that rate. In consequence, men are now relatively those of their adopting parents and their bio- bigger than women than they were a century ago logical parents. Commonly in such studies, both (Huh, Power, and Rodgers 1991). types of parents have independent effects on the While the gap between the sexes has been wid- children (Mackintosh 1998). The effects of the ening, the average difference in height between genes (provided by the biological parents) and social classes has remained roughly constant, the effects of the environment (provided by the with men from a≥uent professional homes being adopting parents) seem to add together. In the nearly 2 cm taller than men from manual back- case of IQ scores, for example, each factor ac- grounds, and women from professional homes counts for about 10 percent of the variation in the being 1.6 cm taller. In other countries the trends children’s scores. have been somewhat different. In Russia, where A common-sense view of what happens is that initially the cognitive abilities of the child are 156 Patrick Bateson

most strongly affected by its biological parents, sus nurture, or genes versus environment, gives but that later in development they are increas- only a feeble insight into the processes. The best ingly affected by the experiences the child has had that can be said of the nature/nurture split is that with its adopting parents. However, the quality it provides a framework for uncovering a few of of the exchanges between the adopting parents the genetic and environmental ingredients that and the child will depend on the match between generate differences between people. At worst, it their characteristics. A potentially able child who satisfies a demand for simplicity in ways that are is adopted by dull people might be less stimulated fundamentally misleading. and more frustrated than if he or she had been Any scientific investigation of the origins of adopted by lively, intelligent people. Conversely, human behavioral differences eventually arrives adopting parents who are disappointed by the at a conclusion that most nonscientists would less able child might provide a less supportive probably have reached after only a few seconds’ environment than those whose expectations are thought. Genes and the environment both mat- satisfied by the responsiveness of an able child. ter. The more subtle question about how much Here again, the difference between the child and each of them matters defies an easy answer; no its adopting parents probably matters, but this simple formula can solve that conundrum. This time in the reverse direction. One study, for ex- then raises the need for a systems approach. ample, found that the bigger the absolute difference in IQ between the biological and adopting parents, the more the child was adversely affected Metaphors and Models (Bateson 1987a). The difference between the parents and child accounted for as much of the vari- The common image of a genetic blueprint for ation in the children as the direct influences of the behavior fails because it is too static, too sugges- biological and adopting parents. The conse- tive that adult organisms are merely expanded quences of the relationships between adopting versions of the fertilized egg. In reality, develop- parents and the children were not revealed by a ing organisms are dynamic systems that play an simplistic analysis that assumes that what went active role in their own development. Even when in is directly related to what comes out. The ap- a particular gene or a particular experience is propriate analysis was not carried out until a known to have a powerful effect on the develop- plausible question was asked about the nature of ment of behavior, biology has an uncanny way of the developmental process. I do not know why finding alternative routes. If the normal develop- this should have happened but suspect the beauty mental pathway to a particular form of adult be- of the statistical procedure called Analysis of havior is impassable, another way may often be Variance seemed to offer a su≤cient explanation. found. The individual may be able, through its The very language of “accounting for variation,” behavior, to match its environment to suit its common enough in statistics books, seemed to own characteristics. At the same time, playful preempt thought about what might be happening activity increases the range of available choices during development. I will return to this point and, at its most creative, enables the individual when I consider the types of model that may be to control the environment in ways that would used fruitfully in order to understand develop- otherwise not be possible. mental process. A low-tech cooking metaphor serves to shift Attempts to partition phenotypic variation in the focus onto the multicausal and conditional behavioral characteristics may yield answers of nature of development (Bateson and Martin a kind, but little sense of what happens as each 1999). Using butter instead of margarine may individual grows up. The language of nature ver- make a cake taste different when all the other ingredients and cooking methods remain un- Behavioral Development and Darwinian Evolution 157

changed. But if other combinations of ingredients man 1993). Many mathematical techniques, such or other cooking methods are used, the distinc- as catastrophe theory and “chaos,” have been de- tive difference between a cake made with butter veloped to deal analytically with the complexities and a cake made with margarine may vanish. of dynamical systems. For all that, it is question- Similarly, a baked cake cannot readily be disag- able whether the descriptive use of mathemat- gregated into its original raw ingredients and the ics brings with it any explanatory power. Much various cooking processes, any more than a be- more promising are those approaches that bind havior pattern or a psychological characteristic evidence across different levels of analysis. can be disaggregated into its genetic and environ- Richard Dawkins wrote: “If a computer is mental influences and the developmental pro- doing something clever and life-like, say play- cesses that gave rise to it. ing chess, and we ask how it is doing it, we do To use a different metaphor, development is not want to hear about transistors, we accept not like a fixed musical score that specifies exactly them. . . . We need software explanations of be- how the performance starts, proceeds, and ends. havior. I do not mean that animals necessarily It is more like a form of jazz in which musicians work like computers. They may be very different. improvise and elaborate their musical ideas, But just as the lowest level of explanation is not building on what the others have just done. As always the most appropriate for a computer, new themes emerge, the performance acquires a no more is it for an animal. Animals and com- life of its own, and may end up in a place none puters are both so complex that something on could have anticipated at the outset. Yet it the level of software explanation must be appro- emerges from within a fixed set of rules and the priate for both of them” (Dawkins 1976a). Rules constraints imposed by the musical instruments. like “store a representation of the input and It is clear, then, that because of the system then compare output against it” have been very in which they are embedded, no simple corre- important in obtaining understanding of song- spondence is found between individual genes learning (Marler 1976). Feedback mechanisms and particular behavior patterns or psychological with moving setpoints can explain the dynamic characteristics. Genes store information coding but ordered way in which developmental pro- for the amino acid sequences of proteins; that is cesses can home in on a particular adult endpoint all.3 They do not code for parts of the nervous (Bateson 1976; Chalmers 1987). The conditional system and they certainly do not code for partic- IF ... THEN rules are useful in understanding ular behavior patterns. Any one aspect of behav- alternative modes of development (Caro and ior is influenced by many genes, each of which Bateson 1986). The competitive exclusion rule ex- may have a big or a small effect. Conversely, any plains the dynamics of sensitive periods (Bateson one of many genes can have a major disruptive 1987b). In their different ways, all these ideas effect on a particular aspect of behavior. A dis- when applied to development at the right level connected wire can cause a car to break down, may provide some of the useful ways to reduce but this does not mean that the wire by itself is the complexity. While they may show us how the responsible for making the car move. Without a job might be done, are we wholly satisfied? strong set of binding ideas, it is not easy to think Careful work on developmental processes such about all aspects of the various strands of evi- as imprinting and song-learning in birds have led dence, which often seem to point in opposite di- to models that take account of both behavior and rections. Some theorists have argued that the the underlying neural mechanisms. In the area in seemingly simple and orderly characteristics of which I have worked extensively, the results of development (such as they are) are generated by imprinting experiments depended greatly on the dynamic processes of great complexity (Kauff- conditions that were used (Bateson 1966). 158 Patrick Bateson

When many factors affect the outcome of a de- The first step involves detection of features in velopmental process, what should be done about a stimulus presented to a young bird. Aspects it? The chances are that all the different influences of the stimulus which the bird is predisposed will not add together, and, if they do not, small to find attractive are picked out by particular de- changes in certain factors may sometimes make tectors at this level of processing. The second big differences to the outcome and large changes step involves comparison between what has al- in others will have no effect whatsoever. For a ready been experienced and the current input. Of long while it was thought that, in the case of filial course, before imprinting has taken place, no imprinting, movement was critical to the young comparison is involved. Once it has occurred, bird. A breakthrough came when a reanalysis recognition of what is familiar and what is novel of old data led to the conclusion that features in is crucial. Finally, the third stage involves control the jungle fowl, the ancestral form of the domes- of the various motor patterns involved in execut- tic fowl, were particularly attractive to chicks ing filial behavior. In the case of the tame bird, (Horn and McCabe 1984). The preference for the assumption is that a representation is formed, aspects of naturalistic stimuli had been missed but this representation has no way of gaining ac- because, under laboratory conditions, the detec- cess to the executive system. tors take longer to develop in chicks than do the The model was formalized (Bateson 1991). ones driven by an ambulance light. Even though Parameters described as “features of the organ- the predisposition is less specific than at first it ism” referred to properties of the real animal’s seemed (Johnson and Horn 1988), it looks as nervous system that also have to develop. These though such a head-neck feature detector, along features may or may not be affected by relatively with others responding to movement, color and small changes in the conditions of development contrast, feed into those bits of the brain that or by variation in genotype. If and when they establish representations of the object to which can be estimated, they should not be treated the bird has been socialized (Horn 1991). as though they were similar to gravitational Imprinting involves another type of plasticity constants. When external conditions affect the in the nervous system, namely connecting up the parameters that are regarded as features of the representation of the familiar object exclusive- organism, the developmental effect would be rel- ly to the system controlling filial behavior and, atively nonspecific, an issue that has been fre- much later in development, to the one controlling quently considered in the literature on behavioral sexual behavior (Bateson 1981). One compelling development (Bateson 1976; Lehrman 1970). strand of evidence is the result of taming. When a With Gabriel Horn, I took the formal model of bird is well-imprinted and then exposed to an- imprinting and developed a neural net (Bateson other object, at first the bird withdraws showing and Horn 1994). We tried to ensure that most of every sign of great alarm. By degrees this alarm the subprocesses are as plausible in neural terms habituates and the bird becomes tame. However, as current knowledge allows and that the whole tame birds do not express any social behavior to- system has the behavioral structure of an intact ward the object, which is by now very familiar. animal. The model exhibits a well-known feature They evidently recognize it, but that is all. It of behavioral development seen in animals, tend- seems, therefore, that at least two stages are in- ing to settle into familiar habits, while also able to volved in imprinting. One involves recognition build with increasing elaboration on the basis of and one involves the control of social behavior previous perceptual experience. The preemptive by the representation of the familiar object. From effect of experience means that other stimuli are this consideration emerged a three-stage model less able to produce change unless exposure to involving analysis, recognition, and execution. those novel stimuli is forced and prolonged. This Behavioral Development and Darwinian Evolution 159

aspect of the model simulates the closure of the plorations, surprises are frequently encountered, sensitive period in development, which is known which then propose experiments with real ani- to be dependent on experience, at least in part. mals or reexamination of data not previously The greater the activating value of the features in analyzed. Therefore, from the point of future the experienced stimuli, the more quickly the sen- empirical research, the modeling approach can sitive period comes to an end. suggest proﬁtable new lines of inquiry. The pre- The model also provides a ready explanation dictions may be false, but they are worth testing for evidence that is di≤cult to explain on the just because the assumptions are rooted in psy- basis of a purely additive approach to devel- chological and biological reality. Above all, such opment (Bolhuis and Honey 1994, 1998; Honey a systems approach shows how the old either/or and Bolhuis 1997). When the maternal call of oppositions applied to behavior simply evapor- the domestic hen accompanies the presentation ate when knowledge starts to advance. It bears of a visual stimulus, the domestic chick is more directly on the general arguments about the in- responsive and develops a stronger preference terplay between features of the organism that de- for the visual stimulus. However, if the auditory veloped prior to the stage of interest and features stimulus is played in the absence of the visual of the environment. Finally, it illustrates the in- stimulus before presentation of the compound terplay that is needed between theory and prac- stimulus, the preference for the visual stimulus is tice. If the theories of process are anchored in the weaker. From the standpoint of animal learning, real world, they reduce complexity and offer a an even more striking result is that if the auditory genuine sense of understanding of how real sys- stimulus is played on its own after the compound, tems generate the seemingly elaborate things that the preference for the visual stimulus is also we observe. However, they do more than that. weaker than when the post-compound exposure They make functional sense. was omitted. In terms of the Bateson and Horn model, the playing of the auditory stimulus on its own weakens the link between the analysis mod- Darwinian Approaches to Development ules processing the features of the visual system and the recognition system. This is because the I have long argued for the beneﬁts of using a Dar- downstream modules are active when the up- winian standpoint when approaching the prob- stream modules are inactive. lems of behavioral development (Bateson 1982, The case for some degree of formalization is 1987a, 1988, 1995, 1984, 1986). Many biologists that an explicit working model brings with it want to be told why the job needs to be done. In mental discipline and may expose weaknesses in a other words they want to be given a functional verbal argument that are all too easily missed. It account or think about a characteristic in terms can also serve several other valuable functions. It of design. Robert Hinde noted how much ethol- can show how the dynamics of development do ogy has been helped by considering the adap- not mean that the processes are so complicated tiveness of a behavior pattern, even when the that they are beyond comprehension. When crit- primary concern is with mechanism (Hinde ics say, “You make the whole process of devel- 1982). Why is it advantageous to bring in the opment sound so complicated,” it is possible to functional argument? Animals live in a complex reply, “No, the explanatory devices are really world and most biologists would argue that the rather simple.” Importantly, they enable com- degree of match between their behavior and the binations of conditions to be explored system- conditions in which they live makes a big differ- atically, which would take a long time with live ence to their survival and reproductive success. animals. Furthermore, in the process of such ex- Those who do not like Darwinian arguments about apparent natural design being brought into 160 Patrick Bateson

discussions of development4 see no value in them. Gri≤ths 1999). It is possible, I believe, to resolve If grasshoppers are green when the grass is green the conflict between the developmental systems and black when the savanna has been blackened approach, which takes into account many in- by fire, the obvious benefit to the individual of fluences, and the evolutionary selfish genes ap- minimizing risk from predation is irrelevant to proach, which seems to suggest only one cause. the question of how they come to match the color However, we need to be very clear about the of their background. The appropriate develop- di≤culties of translating one language in to the mental questions are about how the process other and the attendant confusions that arise works. They are right, but they miss the point. when translation occurs carelessly. The design point frames the mechanistic argu- Dawkins suggested that the way to understand ment. It leads to studies that might otherwise not evolution is not in terms of the needs of species, have been done. To give one example from devel- groups or individuals, but in terms of the needs of opment, the functional approach provides a way genes (Dawkins 1976b). Genes recombine in each of thinking about the difference in timing be- generation to form temporary federations. The tween sexual imprinting and filial imprinting. The alliance forms an individual organism. By repro- suggestion is that an animal should not tune its ducing, individuals serve to perpetuate the genes reference point for mating preferences too early which in the next generation recombine in some in development lest it obtains information about other kind of alliance. Genes are selfishly intent the juvenile appearance of its siblings that could on replicating themselves by the best possible not be used effectively when the time comes to means. Dawkins was clearly and deliberately choose a mate. On the other side it must not tune using a heuristic device when he attributed mo- its mating preferences too late in development tives to genes. He obviously did not think that after the family group has broken up and it is gene have intentions. It is easier for most of us to likely to be exposed to non-kin. Indeed, the evi- get our minds round a problem when we can dence suggests that birds delay sexual imprint- think of a complex system in terms of the way ing until their siblings have molted into adult they strive to reach a specific end state. This is not plumage (Bateson 1979). In this way the bird is only true in biology. A great nineteenth-century able use its experience of close kin so that it physicist, William Rowan Hamilton, formulated chooses a mate that is genetically a bit different a general and widely accepted teleological princi- from itself, but not too different. If the two types ple for use in mechanics. It is a powerful way of of imprinting are treated as part of the same gen- thinking about systems, the behavior of which is eral process, the difference seems to be of no im- determined by many factors. portance and is quickly forgotten. However, with The language of intentions can, of course, be attention focused on the problem, we can attempt played many ways. When the ambient environ- to analyze the mechanisms responsible for the mental temperature during development is cru- difference in timing. The point is, then, that the cial for the expression of a particular phenotype, functional approach can stimulate research on changes in temperature by a few degrees may lead the processes of development. to a startling evolutionary change. It may lead to The heuristic point about the mentally enabl- extinction—such as is predicted for turtles whose ing role of functional explanations also relates sex depends on temperature early in development also to the seeming conflict between selfish gene and after global warming would all end up as fe- language and the system theories arising from males. Would not such cases give as much status studies of development—a discussion that runs as to a necessary temperature value as to a necessary a theme throughout this book and other recent gene? The temperature value is also required for treatments of biological philosophy (Sterelny and the expression of a particular set of phenotypes— Behavioral Development and Darwinian Evolution 161

a balanced sex ratio in turtles before global in the causal language of population biolog- warming. It is also stable (within limits) from ists even though the users are ostensibly talking one generation to the next. It may even be trans- about the same issue. mitted from one generation to the next if the sur- For population geneticists, a genetic difference vival machine makes a nest for its offspring. The is identified by means of a biochemical, physio- bird is the nest’s way of making another nest logical, structural, or behavioral difference be- (Bateson 1978). tween organisms (after other potential sources of Dawkins’s riposte to my tease was that nest difference have been excluded by appropriate material doesn’t have the permanence of DNA procedures). Dawkins suggested that his move (Dawkins 1978). Later he developed the point backward and forward between the language of (Dawkins 1982). Nests do not have the causal gene intentions and the more orthodox language significance of genes. “There is a causal arrow of genetic differences was acceptable because they going from gene to bird, but none in the reverse are simply alternative ways of describing the direction. A changed gene may perpetuate itself same thing. To make his point, he described per- better than its unmutated allele. A changed nest ception of the Necker cube. The front edges of will do no such thing unless, of course, the change the line drawing of the cube suddenly flip to the is due to a changed gene, in which case it is the back as we look at them. Each perceived image of gene that is perpetuated, not the nest” (p. 98). In the cube is as real as the other. Dawkins suggests a recent essay, R. D. Gray (in press) has elegantly that, in similar ways, the teleological and mecha- questioned this claim. He points how variation in nistic images of evolution translate backward and the mode of nest construction could plausibly be forward into the other. However, it doesn’t make passed on by extragenetic as well as genetic any sense to attribute motives to a comparison, means. as may be illustrated by a parable. In concluding his discussion of alternative tele- Let us consider the spread of a new brand of ologies and my concern that he was giving too biscuit in supermarkets from the perspective of much status to the gene as programmer, Dawkins the recipe. While shoppers may compare biscuits (1982) wrote amusingly: “As is so often the case, and buy one brand, it is the recipe used for mak- an apparent disagreement turns out to be due to ing desirable biscuits that survives and spreads in mutual misunderstanding. I thought Bateson was the long run. Therefore, the word in the recipe denying proper respect to the Immortal Replica- that makes the biscuit successful is selfish, be- tor. Bateson thought that I was denying proper cause it serves to perpetuate itself. So far so good. respect to the Great Nexus of complex causal fac- But my story loses all coherence if I conclude tors interacting in development” (p. 99). Reflect- by adding that the difference in the wording of ing on this debate many years later, I think that recipes used for making successful and less suc- more was at issue than different emphases and in- cessful is selfish. terests. Dawkins response showed how easily we It is worth developing the biscuits in a super- get snarled up in the language. The details of the market parable just a little further. What shop- evolutionary mechanism involving small changes pers really do in supermarkets is to select a word in DNA had been mixed up with the intentions in a faithfully reproduced recipe. Really? Words that were rhetorically attributed to genes. While in recipes as the units of the shoppers’ selection? Dawkins was quite justified, in my view, in writ- It is an odd idea, since shoppers respond among ing about genes’ intentions, he was wrong when other things to the outcome of the cooking pro- he treated this language as readily translatable cess which gave rise to the biscuits. For very sim- into the language of what genes actually do. The ilar reasons many biologists, myself included, language of selfish genes does not easily translate disliked the idea of genes being treated as the 162 Patrick Bateson

units of selection in Darwinian terms. After all, lings or cousins is once again the province of Darwin had used his metaphor of “natural se- DST. The conditions shared by related entities lection” because he had been impressed by the will commonly consist of genes but will include ways in which plant and animal breeders selected all those nongenetic factors that DST thinking the characteristics they sought to perpetuate. Of has brought back on stage. The value of making course, the gene selection idea grew up because it the three-way distinction is that it separates out made sense of those cases in which the conse- the first and third aspects of the Darwinian mech- quences of an act favored the survival of geneti- anism, which had been run together into the sin- cally related individuals rather than the actor. gle concept of “replicator.” Dawkins (1982) helpfully advanced the argument A confusion of teleological and causal ex- when he drew a distinction between “replicators” planation led critics and supporters alike to a and “vehicles.”5 non sequitur. Proponents argue that if genes are To separate further the many different strands usefully regarded as selfish, it follows that they of thought that have been run together in the uniquely bring into being the phenotypic char- units of selection debate, it is worth making a acters of the whole animal. Opponents argue that three-way distinction that mirrors in part the if it is valuable to treat development as a process mechanism Darwin introduced to explain the involving feedback and many mutual actions evolution of adaptations. Darwin suggested that between different agents, then the selfish gene three components of the mechanism are essential. language is inappropriate. Both arguments are First, competing biological entities must differ in mistaken, in my view. The developmental sys- their characteristics. (The entities have usually tems approach is not in conflict with the selfish been individual organisms in conventional neo- gene approach to evolution. Darwinian thinking, but they may be parts of The evolutionary process does not require a organisms or assemblies of organisms.) Second, simple correspondence between genes and adap- the entities must survive and/or reproduce them- tive behavior. Darwinian evolution operates on selves with differing degrees of success. Third, the individuals that have developed within a partic- entities must be more likely to share characteris- ular set of conditions. If those conditions are sta- tics with their descendants or relatives than with ble for many generations, then the changes that unrelated entities. matter will arise from segregation of factors that The developmental processes creating the phe- give rise to individual differences. Individuals notypes of the entities have been the major con- vary; some survive and reproduce more success- cern of DST and, indeed, of this chapter. Darwin fully than others because they possess a crucial had nothing to say about how phenotypic varia- characteristic; and close relatives are more likely tion is generated, and yet its existence is central to share that characteristic than unrelated in- to his evolutionary mechanism. The modern em- dividuals. Apparent design is produced, even phasis on developmental process has served to when it is at the end of the long and complicated put empirical flesh on notions of what generates process of development. But the environment the variation in the raw material for differential does not cease to be important for evolution just survival and reproductive success. Once the phe- because it remains constant. Change the envi- notypes are created, DST has not much to say ronment and the outcome of an individual’s - about differential survival and differential re- development may be utterly different. Indeed, if productive success. However, the necessary con- an individual does not inherit its parents’ envi- ditions for re-creating the characteristics of the ronment along with their genes, it may not be successful entity in the next generation or for gen- well adapted to the conditions in which it now erating similar characteristics in kin such as sib- finds itself. Behavioral Development and Darwinian Evolution 163

prefer to argue that simplicity and regularity may Conclusion be found in the developmental processes that give rise to unique individuals. Confidence in that Old styles of thought die hard. Even the much- conclusion comes not from general principles but used distinction between sex and gender has a from the careful analysis of particular cases. The strong whiff of the old genes/environment oppo- essence of development—change coupled with sition. “Sex” is biology; but “gender” is part of continuity—starts to make sense. It becomes pos- culture, the acquired behavior deemed appropri- sible to understand how the individual is so re- ate to the social role of that sex. Any fair-minded sponsive to events at one stage and so unaffected person listening to setpiece debates about behav- by them at another. ioral development is likely to be left irritated by Order underlies even those learning processes the claims and counterclaims. And well-meaning that make individuals different from each other. attempts to break out of the nature/nurture Knowing something of the underlying regulari- straitjacket have often resulted in an obscure and ties in development does bring an understanding bewildering portrayal of development as a proof what happens to the child as it grows up. The cess of impenetrable complexity (what Salman ways in which learning is structured, for instance, Rushdie once described in another context as affect how the child makes use of environmental a P2C2E—a Process Too Complicated To Ex- contingencies and how the child classifies percep- plain). Indeed, development seemed so unfath- tual experience. Yet predicting precisely how an omably complex to eighteenth-century biologists individual child will develop in the future from that they believed that it must depend on super- knowledge of the developmental rules for learn- natural guidance. ing is no easier than predicting the course of a The processes involved in behavioral develop- chess game. The rules influence the course of ment do indeed look forbiddingly complicated on a life, but they do not determine it. Like chess the surface. Some would argue that it is worse un- players, children are active agents. They influ- derneath and that such order as is found is gener- ence their environment and are in turn affected ated by dynamical systems of great complexity by what they have done. Furthermore, children’s (Kauffman 1993). One approach to development responses to new conditions will, like chess play- has been to suggest that everything interacts with ers’ responses, be refined or embellished as they everything else. A challenge to such thinking is gather experience. Sometimes normal develop- the evidence for the segregation of characteristics ment of a particular ability requires input from in closely related individuals. How is it that char- the environment at a particular time; what hap- acteristics such as a big nose or a retiring disposi- pens next depends on the character of that input. tion skip a generation? How is it that siblings are The upshot is that, despite their underlying regu- so different from each other at birth? How is larities, developmental processes seldom proceed rapid artificial selection for behavioral character- in straight lines. Big changes in the environment istics such as tameness possible? None of the evi- may have no effect whatsoever, whereas some dence that leads to such questions implies any small changes have big effects. The only way to simple correspondence between an inherited unravel this is to understand the developmental factor and the development of a phenotypic char- processes. acteristic. But the evidence does imply that frac- tionation and independent inheritance of some of the factors necessary for development happens all Acknowledgments the time. In contrast to an agenda that easily renders de- I am very grateful to Susan Oyama and Russell velopment a process too complicate to explain, I Gray for the trouble they took over earlier ver- 164 Patrick Bateson

sions of this chapter. They are good friends as Bateson, P. P. G. (1966). The characteristics and con- well as good editors. text of imprinting. Biological Reviews 4l: 177–220. Bateson, P. P. G. (1976a). Speciﬁcity and the origins of behavior. Advances in the Study of Behavior 6: l–20. Notes Bateson, P. P. G. (1976b). Rules and reciprocity in behavioural development. In P. P. G. Bateson and R. A. 1. Susan Oyama considers the various meanings of “in- Hinde (Eds.), Growing Points in Ethology, pp. 40l–42l. teraction” in her chapter. In its purely statistical sense, Cambridge: Cambridge University Press. “interaction” implies that two or more factors have an effect that would not be deduced from knowledge of Bateson, P. (1978). Review of the Selﬁsh Gene by what each does when the other is kept constant. For ex- Richard Dawkins. Animal Behaviour 26: 316–318. ample, when Factor A is present at its highest level with Bateson, P. (1979). How do sensitive periods arise and Factor B at its lowest level, an outcome measure might what are they for? Animal Behaviour 27: 470–486. have a high value, but when Factor B is also at its high- Bateson, P. (1981). Control of sensitivity to the envi- est level, the outcome measure might have a low value. ronment during development. In K. Immelmann, G. W. Confusion begins when interaction is used for the pro- Barlow, L. Petrinovich, and M. Main (Eds.), Behavioral cess that might give rise to such a result. Development, pp. 432–453. Cambridge: Cambridge Uni- 2. The editors did not like my jukebox metaphor be- versity Press. cause a record is put into the machine ready-made, Bateson, P. (1982). Behavioural development and evo- whereas an environmentally triggered alternative mode lutionary processes. In King’s College Sociobiology of development involves all those mutual actions be- Group (Ed.), Current Problems in Sociobiology, pp. l33– tween the developing individual and its environment l5l. Cambridge: Cambridge University Press. which is the theme of so much of this book. I agree with Bateson, P. P. G. (1984). Genes, evolution and learning. them and understand their discomfort. Nevertheless, I In P. Marler and H. S. Terrace (Eds.), The Biology of hope they will forgive me for persisting in using the Learning, pp. 75–88. Berlin: Springer-Verlag. metaphor despite its misleading aspects and disturb- Bateson, P. P. G. (1986). Functional approaches to be- ingly preformationist character. I persist because the havioural development. In J. G. Else and P. C. Lee image does embody the idea that the various alternative (Eds.), Primate Ontogeny, Cognition and Social Behav- ways in which an individual might develop are coherent iour, pp. 183–192. Cambridge: Cambridge University and the phenotype produced is appropriate to the cir- Press. cumstances in which it develops (except when the forecast is incorrect!). Bateson, P. (1987a). Biological approaches to the study of behavioural development. International Journal of 3. Some would argue that even at the level of base pairs Behavioral Development 10: 1–22. and amino acids, the coding metaphor is inappropriate. Neumann-Held discusses the matter in her chapter. See Bateson, P. (1987b). Imprinting as a process of compet- also Sarkar (1996). itive exclusion. In J. P. Rauschecker and P. Marler (Eds.), Imprinting and Cortical Plasticity, pp. 151–168. 4. Tim Ingold would probably include himself in this New York: Wiley. group. Bateson, P. (1988). The active role of behaviour in evo- 5. Gri≤ths and Gray (1997) point out that the replica- lution. In M.-W. Ho and S. Fox (Eds.), Process and tors are not necessarily genetic, and Gray (in press) feels Metaphors in Evolution, pp. 191–207. Chichester: that “vehicle” is too passive a concept to represent the Wiley. organism’s active engagement with its environment. Bateson, P. (1991). Are there principles of behavioural development? In P. Bateson (Ed.), The Development References and Integration of Behaviour, pp. 19–39. Cambridge: Cambridge University Press. Barker, D. J. P. (1998). Mothers, Babies and Health in Bateson, P. (1996). Design for a life. In D. Magnusson Later Life. (2d ed.). Edinburgh: Churchill Livingstone. (Ed.), Individual Development over the Lifespan: Biolog- Behavioral Development and Darwinian Evolution 165

ical and Psychosocial Perspectives, pp. 1–20. Cam- Political Perspectives. Cambridge: Cambridge Univer- bridge: Cambridge University Press. sity Press. Bateson, P., and G. Horn. (1994). Imprinting and rec- Gri≤ths, P. E., and R. D. Gray. (1997). Replicator II— ognition memory: a neural net model. Animal Behav- judgement day. Biology and Philosophy 12: 471–492. iour 48: 695–715. Hales, C. N., and D. J. P. Barker. (1992). Type 2 (non- Bateson, P., and P. Martin. (1999). Design for a Life: insulin-dependent) diabetes mellitus: The thrifty pheno- How Behaviour Develops. London: Cape. type hypothesis. Diabetologia 35: 595–601. Bolhuis, J. J., and R. C. Honey. (1994). Within-event Hales, C. N., M. Desai, and S. E. Ozanne. (1997). The learning during filial imprinting. Journal of Exper- thrifty phenotype hypothesis: How does it look after 5 imental Psychology: Animal Behavior Processes 20: years? Diabetic Medicine 14: 189–195. 240–248. Hinde, R. A. (1982). Ethology. Oxford: Oxford Univer- Bolhuis, J. J., and R. C. Honey. (1998). Imprinting, sity Press. learning and development: From behaviour to brain Honey, R. C., and J. J. Bolhuis. (1997). Imprinting, and back. Trends in Neurosciences 21: 306–311. conditioning, and within-event learning. Quarterly Jour- Caro, T. M., and P. Bateson. (1986). Organisation and nal of Experimental Psychology. Section B Comparative ontogeny of alternative tactics. Animal Behaviour 34: and Physiological Psychology 50: 97–110. 1483–1499. Horn, G. (1991). Cerebral function and behaviour in- Chalmers, N. R. (1987). Developmental pathways in vestigated through a study of filial imprinting. In P. behaviour. Animal Behaviour 35: 659–674. Bateson (Ed.), The Development and Integration of Cooper, R. M., and J. P. Zubek. (1958). Effects of en- Behaviour, pp. 121–148. Cambridge: Cambridge Uni- riched and restricted early environments on the learning versity Press. ability of bright and dull rats. Canadian Journal of Psy- Horn, G., and B. J. McCabe. (1984). Predispositions chology 12: 159–164. and preferences. Effects on imprinting of lesions to the Crews, D. (1996). Temperature-dependent sex determi- chick brain. Animal Behaviour 32: 288–292. nation: The interplay of steroid hormones and temper- Huh, D. L., C. Power, and B. Rodgers. (1991). Secular ature. Zoological Science 13: 1–13. trends in social class and sex differences in adult height. Dawkins, R. (1976a). Hierarchical organisation: A can- International Journal of Epidemiology 20: 1001–1009. didate principle for ethology. In P. P. G. Bateson and Ingold, T. (1986). Evolution and Social Life. Cam- R. A. Hinde (Eds.), Growing Points in Ethology, pp. bridge: Cambridge University Press. 7–54. Cambridge: Cambridge University Press. Janzen, F. J., and G. L. Paukstis. (1991). Environmen- Dawkins, R. (1976b). The Selfish Gene. Oxford: Oxford tal sex determination in reptiles: Ecology, evolution, University Press. and experimental design. Quarterly Review of Biology Dawkins, R. (1978). Replicator selection and the ex- 66: 149–179. tended phenotype. Zeitschrift für Tierpsychologie 47: Kauffman, S. A. (1993). The Origins of Order. New 61–76. York: Oxford University Press. Dawkins, R. (1982). The Extended Phenotype. Oxford: Lee, T. M., and I. Zucker. (1988). Vole infant develop- Freeman. ment is influenced perinatally by maternal photoperi- Dunbar, R. I. M. (1984). Reproductive Decisions. odic history. American Journal of Physiology 255: R831– Princeton, NJ: Princeton University Press. R838. Eveleth, P. B., and J. M. Tanner. (1990). Worldwide Lehrman, D. S. (1970). Semantic and conceptual issues Variation in Human Growth. (2d ed.). Cambridge: in the nature-nurture problem. In L. R. Aronson, E. Cambridge University Press. Tobach, D. S. Lehrman, and J. S. Rosenblatt (Eds.), Development and Evolution of Behavior Gray, R. D. (in press). Selfish genes or developmental , pp. 17–52. San systems? Evolution without interactors and replicators? Francisco: Freeman. In R. Singh, C. Krimbas, J. Beatty, and D. Paul, (Eds.), Lerner, R. M. (1998). Handbook of Child Psychology Thinking about Evolution: Historical, Philosophical, and vol 1. (5th ed.). New York: Wiley. 166 Patrick Bateson

Mackintosh, N. J. (1998). IQ and Human Intelligence. Oxford: Oxford University Press. Marler, P. (1976). Sensory templates in species-speciﬁc behavior. In J. C. Fentress (Ed.), Simpler Networks and Behavior, pp. 314–329. Sunderland, MA: Sinauer. Nesse, R. M., and G. C. Williams. (1994). Evolution and Healing. London: Weidenfeld & Nicolson. Oyama, S. (1985). The Ontogeny of Information: De- velopmental Systems and Evolution. Cambridge: Cam- bridge University Press. (2d rev. ed., Durham, NC: Duke University Press, 2000.) Pener, M. P., and Y. Yerushalmi. (1998). The physiology of locust phase polymorphism: An update. Journal of Insect Physiology 44: 365–377. Pinker, S. (1994). The Language Instinct. London: Penguin. Rowell, C. H. F. (1971). The variable coloration of the Acridoid grasshoppers. Advances in Insect Physiology 8: 145–198. Sarkar, S. (1996). Decoding “coding”: Information and DNA. Bioscience 46: 857–864. Sterelny, K., and P. E. Gri≤ths. (1999). Sex and Death. Chicago: University of Chicago Press. Waterlow, J. C. (1990). Mechanisms of adaptation to low energy intakes. In G. A. Harrison and J. C. Waterlow (Eds.), Diet and Disease in Traditional and Developing Countries, pp. 5–23. Cambridge: Cambridge University Press. Wilkinson, R. G. (1996). Unhealthy Societies: The A≥ictions of Inequality. London: Routledge. Wilson, E. O. (1971). The Insect Societies. Cambridge, MA: Harvard University Press. Yntema, C. L., and N. Mrosovsky. (1982). Critical periods and pivotal temperatures for sexual differentiation in loggerhead sea turtles. Canadian Journal of Zoology 60: 1012–1016. Zadnik, K. (1997). Myopia development in childhood. Optometry and Vision Science 74: 603–608. 14 Parental Care and Development

Peter H. Klopfer

As a naive young student, I was once shocked choices irreversibly, but this is not true for rear- when a revered elder statesman prefaced his re- ing in a nest in woodland (Wecker 1963). One marks on the neural control of locomotion with can only speculate as to why this should be so, the observation that as a dedicated Marxist he but it does indicate that some rearing experiences “obviously” had to believe . . . I had been raised in are more constraining than others. Similarly, the belief that preconceptions were taboo for Klopfer (1965, 1971) demonstrated that duck- those of us dedicated to experimental science, lings imprinted onto models of different colors and I truly thought this to be true. In the half cen- and patterns would not be equally constrained in tury since, quite aside from personal revelations, their later preferences: Some models allowed for the work of deconstructionists has made it clear a greater range of preferences than others, and that few of us can truly divorce our preconcep- others merely served to activate a preference for a tions from the questions we ask or the answers we different model. favor. This is particularly evident when one con- Among humans, Whiting and Childs (1953) siders the enormous literature on parental care have chronicled cultural differences as great or and the question of how indelibly the actions of greater than those that distinguish the aforemen- the parent are impressed upon the young. This is tioned three species of Peromyscus, but questions the issue we will address here. of adaptivity have seemed of less interest or rele- Patterns of parental care vary enormously vance here. Instead, discussion has centered upon both between and within species (Gubernick and the long-term effects of particular parental rear- Klopfer 1981). Biologists tend to assume these ing styles, that is, upon what in animals was re- differences have some adaptive significance. ferred to as the temporal dimension of adaptivity. Thus, McCabe and Blanchard (1950) detailed the In general three views have prevailed, with a nest-building and parental activities of three fourth recently gaining ascendancy. species of the mouse, Peromyscus (maniculatus, 1. Early fixation. This view emphasizes the truei, and californicus), which show, respectively, effects of primacy, and is often modeled upon the a minimum of parental oversight; some degree ethological concept of imprinting. A metaphor of oversight; and oversight that includes the con- that describes this process is that of softened seal- struction of an elaborate, waterproof home ing wax, which, for a short time only, can receive looked after by both parents. This last species an imprint, whereupon, after hardening, the im- has also fur that is particularly soft and easily print becomes a permanent feature. The exposure wetted by the elements. The differences in their of young precocial birds (viz., those that are mo- parental activities are also reflected in the manner torically able to leave the nest upon hatching, as in which they utilize other resources in their en- opposed to altricial) to a moving object shortly vironment, and these latter differences keep the after hatching, which object then becomes the three species ecologically distinct—that is, they preferred companion or even subject for later prevent their competitively excluding one an- sexual attachments, was popularized by Konrad other (cf. Hutchinson 1951). “Adaptiveness” has Lorenz (1937), though rather more recently a temporal dimension, too, and biologists have Gottlieb (1971) corrected many of the miscon- also taken note of the fact that the ease with ceptions that Lorenz’s descriptions engendered. which whatever influence exercised by parental Gottlieb pointed out the many qualifications and care upon subsequent behavior can be overcome specifications entailed in imprinting, as well as also varies. The effects of deermice being reared other attributes that make it very different from in a pasture nest influence subsequent habitat the simple and permanent process the imprinting 168 Peter H. Klopfer

metaphor suggests. Furthermore, long-term ef- and habits of the old. Kagan, in a study some fects of primary experiences, which may suggest decades ago (1984) described the withdrawn, shy, imprinting, can often be the result of complex but almost secretive behavior of young, evidently subtle feedback processes. Gunther (1961) de- largely neglected children in the highlands of scribes how a newborn human infant may some- Guatemala. Some even displayed many of the times have its upper lip pushed against its nostrils hallmarks of incipient autism. Returning some when it first nurses, the consequence of a mis- years later, Kagan found these same children to match between the shape of the maternal breast have become social, friendly, and self-reliant and the infant’s mouth. When this occurs, the youths, who displayed no signs of early neglect. baby quickly stiffens and seems to thrust itself Somehow they had become transformed. The away from its mother. After a few repetitions of sudden onset of adult micturition patterns in this behavior, many women, seemingly rejected male puppies, once androgenic hormones reach a by their newborn, develop defensive reactions, certain level, is another example, as are the dra- and, in the end, abandon efforts to nurse, feeling matic changes in gender seen in some Dominican themselves rejected and useless. This temporally children, who appear to be and are raised as girls limited scenario with long-term consequences and who, at puberty, suddenly develop androgen may resemble imprinting, though here the effect sensitivity and become functional young men is to create an aversion rather than an attach- (Imperato-McGinley et al. 1974). ment. However, the similarity is clear when one 4. Regulation by interaction: in this example, notes that in domestic goats, lambs, and calves, the focus is not on the developing infant, but an animal that has first suckled for a few minutes rather on the interactive unit, parent/child, which after its birth can be taught to drink from a pail is examined within the context of both biological only with di≤culty (and usually only by allowing and social/environmental constraints. It can be it first to suckle on a finger or other object that is viewed as similar (though I emphasize the sim- then immersed in the pail). In short, the suckling ilarity is a superficial one!) to a goal-seeking - response is initially attached in a fashion suggest- automaton that can adjust its reactions to the ing classical imprinting, though Gunther’s exam- external environment so as to remain on course. ple reveals the important role of sensory feedback Ashby (1958) describes a mechanical “turtle” in maintaining (or terminating) the attachment. that will leave its well-lit hutch (wherein its bat- 2. Gradual shaping. The appropriate meta- teries are recharged) to wander about a room phor here is a windswept tree on a barren coast, whose obstacles it “learns” to avoid. As it “tires” which, in time, displays in the angles of its (viz., its battery runs down), it becomes increas- branches the effects of continual shear. Or, as ingly photopositive and thus returns to its hutch the adage runs, as the twig is bent, so grows the for a “rest,” only to reemerge and continue its ex- limb. There is much observational support for plorations when recharged. A biological example this view, though it ignores occasional, and is provided by Chisholm’s study of Navajo in- major, discontinuities due to the sudden onset of fancy, which is described later. hormone production, which may be seasonally In nonhuman species, where comparative controlled (as in many songbirds) or develop- studies are feasible, the different models of de- mentally triggered (as in many mammals). velopment are at least sometimes susceptible to 3. Metamorphosis. The model here is the hol- being examined from a functional, or adapta- ometabolous insect that spends one part of its life tionist, point of view. Thus, it has been argued as a crawling larva or caterpillar, then pupates (Klopfer 1959) that imprinted preferences are to only to emerge in an altogether different, winged be expected in motorically precocial species and form, with little to suggest in its new life the styles not in their altricial near-kin. If a young duckling, Parental Care and Development 169

it is argued, does not very quickly learn to remain velopment and that also provides a useful heuris- near its sibs and its parents, it will not likely sur- tic. This is the notion that what must be observed vive. A young songbird, on the other hand, phys- and described is the interaction between parent ically unable to disperse from its nest, can take and offspring, and that it is the quality of the time to learn the hallmarks of its parents. Simi- interaction that plays a determining role in the larly, surface-nesting waterfowl can rely upon outcome, not any single learning mechanism. In visual cues when they learn the characteristics of what follows, I describe some earlier work of their kind, while hole-nesting species might be ex- my own on the compensatory responses of goat pected to rely more upon auditory features. Some mothers to their impaired young, work that was of these speculations are both plausible and true, undertaken with an altogether different objective though as Gottlieb (1971) has shown, imprinting in mind, but which, I think, does illustrate the is not the simple phenomenon that Lorenz origi- kinds of interactions that are important. Then, I nally described (cf., Hess 1958). In studies of hu- review a study of Navajo infancy by Chisholm mans, the issue of adaptation at this level is rarely (1983), which provides an especially persuasive raised, and for good reason: Few investigators human example of the role of interaction between of cultural differences in development and child- parent and offspring. rearing would deny that there is more than one Some animals are more equal than others: way to skin a cat. A culture’s rearing myths and Even in groups of highly inbred creatures, at least practices are necessarily of a piece with other fea- in species with some degree of social organiza- tures of that culture, and the evolutionary aspects tion, certain individuals emerge as dominant or of adaptation are rarely if ever relevant (but see as leaders, initiating movements, taking the ini- Chisholm 1983, for an example of adaptational tiative in territorial defense, or displacing their thinking). It is rather a question of how and why conspecifics. Small herds of goats display this early experience associated with different parent- phenomenon clearly: Certain animals have prior- ing styles shapes later behavior. ity of access to localized sources of food, yet With respect to studies that seek to determine others are at the head of the herd when it moves the applicability of any particular model to any to new terrains. Individual differences that reflect given sample, it is embarrassingly evident that the potential for being, in one or another sense, a the investigator’s conceptual frame largely de- “leader” are evident early in the life of the kid: termines the outcome. Even so astute an inves- The most vigorous of a group of month-old kids tigator as Kagan has at different stages of his life almost invariably becomes the dominant animal. interpreted the same body of data as indicating Thus, we asked, how come leaders to their posts, genetic determination of certain traits (such as focusing our attention on a group of about three shyness), or as indicating the salience of early dozen toggenburg goats (Klopfer and Klopfer influences, or as suggesting remarkable resilience 1974). and abilities for compensatory responses in the Manning (1965) and Benzer (1973), among young child (Kagan 1984, 1998; Kagan, Reznick, many others, had previously documented differ- and Snidman 1988). Whorf’s warning that the ences in temperament among animals reared language of our questions shapes our perceptions identically. Benzer’s work on the genetics of be- is particularly pertinent when one examines the havior in Drosophila, in fact, was allegedly the literature on human development and the role of result of his noting the dramatic differences in parental influences (Whorf 1950). the temperaments of his two daughters (Weiner Increasingly, however, one particular model 1999). The differences he observed in the fruitflies is emerging that seems to allow assimilation of were heritable, and presumably linked to genetic much of the accumulated knowledge about de- differences, but epigenetic influences were not 170 Peter H. Klopfer

ruled out. What might these developmental influ- 1977). As expected, the weaker kids, up to a ences have been? point, benefited from the reduced maternal atten- Among goats, twins are the commonest litter tion, getting to the udder and nursing sooner size, at least among multiparous females, and than their more vigorous mate who was being triplets occur frequently, too. However, birth more intently licked (and knocked down) by the weights are highly variable within a litter, even mother. However, where the degree of a kid’s in those cases where the twins are evidently weakness exceeded a threshold, which could be monozygotic. We noted that the heavier kid of specified in the number of minutes required to re- any litter displayed a greater vigor than its lighter sume standing, the feedback failed and the more siblings, vigor being measured in terms of the vigorous kid did nurse first and come to domi- time taken to first stand unsupported and to nate the milk supply. commence udder-seeking behavior. The mother If the mother’s ability to monitor her kids’ attends most to the more vigorous of her kids, ir- behavior was impaired, which we achieved by ap- respective of birth order. This has the paradoxi- plying blindfolds and by pharmacologically re- cal effect of slowing this kids’ progress to the teat. ducing olfactory acuity, the feedback also broke The mother’s attention entails licking and nuz- down. When impaired, mothers stood passively zling the kid, often with su≤cient force that the while their kids sought the udder, and the more kid is knocked off its feet, must rest for a time, vigorous kids invariably reached the udder first and then once again must struggle up and find its and were thus in a position to monopolize the balance. Thus, it often happens that the initially milk supply. As we concluded at the time, mother weaker sibling, lacking interference in its progress love is not blind: The degree to which offspring to the teats, is able to initiate nursing first. The share in the maternal bounty is very much the re- result is that initial differences in strength and sult of an interaction between mother and young. weight are muted, and the weaker animals are not The ultimate status of the neonatal kid is not necessarily relegated to the status of subordinate. specified at birth, nor is there any reason to doubt These results do not answer the question of that subsequent interactions would be equally how a dominant or leadership position develops, effective in modifying behavior. but they suggest the absence of an invariant rela- Navajo infants, it has long been claimed, are tionship between prenatal (including, perhaps, quieter and less irritable than infants of Anglo hereditary) influences and subsequent status. parents (see Chisholm 1983, from whom all of However, the mother’s moderating influence was the following is taken). Depending upon which of seemingly mechanical: she merely licked the more the developmental metaphors one finds more ap- active kid more, and the deleterious effect on pealing, a variety of putative explanations for this its progress to the source of nutrients was inciden- difference can be found. For example, Navajo tal. Suppose that initial differences in vigor were infants are traditionally placed in a cradleboard, su≤ciently great as to suggest that the weaker kid which severely restricts their movements. The had a reduced chance of survival. Or, suppose duration of confinement does vary considerably, that resources to sustain the mother were in short from only a few to twenty-three hours in a day, supply, so that rearing two kids rather than just and may terminate soon after the first year of life one was unlikely. Would the mother modify her or continue for as long as two years. If one as- responses in an adaptive fashion? sumes this confinement to constitute a pertur- To examine the role of the mother in modulat- bation in the normal course of development, ing the effectiveness of her kids’ nursing attempts, equivalent, in one sense, to the administration of we first used drugs to enhance the difference tranquilizing drugs to the kids described above, in the vigor of twin kids (Klopfer and Klopfer one might expect certain effects with long-term Parental Care and Development 171

consequences for the developmental pathway. crawling infant would likely be subject to fre- Chisholm’s study did, in fact, identify some im- quent disciplining for scattering the nuts, while a mediate consequences of the confinement. The cribbed infant would merely be staring into the cradleboarded children, and this was true for sky, both situations less socially rewarding than Anglo babies placed in cradleboards as well, were the continuous face to face communication the less active and less easily aroused. Mothers and cradleboard allows. cradleboarded infants were also less responsive to Chisholm concludes that it is not the individual one another. He concluded (p. 216) that “the im- features of particular developmental practices of mediate effects of the cradleboard on infant be- a culture that are important for development— havior and mother-infant interaction are clear the cradleboard has advantages, but so does its and a substantial corpus of theory suggests that absence, and both practices have short-term these immediate effects should last. The question effects on the behavior and temperament of the now becomes why don’t these clear immediate child. What is significant is the entire parent- effects of the cradleboard last?” His exhaustive ing process, particularly those regularities in the series of tests produces no sign of a cradleboard interaction of infant and parent which assure effect at later ages, just months after cradleboard- predictable outcomes: the infant struggles, its ing ends. confinement ceases. The Navajo child reared in Chisholm suggests three responses to the crad- conventional Western fashion will, by the time leboard, which may serve as the regulating fac- the two are walking, not be distinguishable. tors that assure a stable developmental outcome. The patterns that assure constancy in the out- First, the child itself is allowed to determine when comes may differ from one culture to the other. it is to be removed. Despite the reduced activity But, it is not the culture, or the genetic substrate, and arousal state of the cradleboarded infant, it except in a very general way, nor the rearing can still fret and cry, and Navajo mothers are practices that define the child, but the manner of unanimous in saying that such behavior is the sig- their interaction. This both limits the possible nal for removing the child from its confinement. variance and provides a buffer against pertur- Furthermore, if it resists being replaced, the re- bations that might have serious or long-term con- sistance is generally honored. Thus, despite hav- sequences. Kagan, attacking the allure of infant ing its movements restricted, the infant, especially determinism, states it thus: as it grows older, is able to develop a sense of control over its condition. [It] does not mean that the events of the first two years are without any force. It only means that a fearful, A second potential corrective lies in the fact quiet, tense two-year old who has had an uncertain that directly upon the infant’s removal from the environment remains malleable should benevolent cradleboard the mother and infant engage in a changes occur, and a laughing, securely attached, smart particularly intense period of affectionate interac- two-year-old is not protected from angst should her life tion, which Chisholm likens to the catching up in turn harsh. Both science and autobiography a≤rm that growth experienced by children who have been ill a capacity for change is as essential to human develop- for a time. ment as it is to the evolution of new species. The events Finally, the cradleboard assures a degree of of the opening years do start an infant down a particu- conflict-free contact that a cribbed or crawling lar path, but it is a path with an extraordinarily large infant might not experience. The cradleboard is number of intersections. (Kagan 1998: 149–150) generally placed vertically at a height that allows The conclusion is somewhat paradoxical. A direct eye-to-eye contact between the mother and predictable pattern of parenting is necessary for the child, whether the mother is standing at a normal development of the infant, but differ- loom or seated on the ground shelling nuts. A ent patterns may produce similar outcomes, and 172 Peter H. Klopfer

none of these seem to have effects that are irre- velopment can still lead to similar outcomes, just versible in later life. Another conclusion must be as very minor differences can also produce major that the relationship between early rearing and disparities (viz., the Gunther studies of suckling, later behavior is unique to each species, for it is alluded to above). As Schrödinger (1951) wrote, evident that for other animals different rearing long before the structure of DNA had been de- patterns may not have similar outcomes. This scribed, the determinacy in the development of limits greatly our ability to extrapolate from living organisms requires their feeding upon the greylag geese, to chimpanzees, to humankind order in their environment. Where that environ- (Klopfer 1996). Not all species are subject to such ment becomes randomly variable, determinacy a variety of social and physical environments as must decline, but where it is ordered and pre- are Homo sapiens, so neither should selective dictable, outcomes can be constant even while de- forces promote or sustain the degree of postin- velopment trajectories vary. fant malleability as humans display. Lumsden and Wilson (1981) propose that selection acts to favor particular epigenetic rules. These are regu- References larities in developmental processes that assure a Ashby, W. (1958). An Introduction to Cybernetics. particular phenotype will always appear in a par- London: Chapman and Hall. ticular environment. Those rules, however, derive Benzer, S. (1973). Genetic dissection of behavior. Sci- their power not merely from the fact that the entific American 229: 24–37. products of certain genes are predictable, but Chisholm, J. S. (1983). Navaho Infancy: An Ethological equally from the fact that the genes are expressed Study of Child Development. New York: Aldine. in an ordered and predictable environment, intra- Gottlieb, G. (1971). Development of Species Identity in and extracellular, physical, and social. In the Birds: An Inquiry into the Prenatal Determinants of Per- case of the Dominican children alluded to earlier, ception. Chicago: University of Chicago Press. the genes that are involved in the development Gubernick, D. J., and P. H. Klopfer. (1981). Parental of male morphology are stymied by a cellular Care in Mammals. New York: Plenum. environment (presumably influenced by other Gunther, M. (1961). Infant behavior at the breast. In B. genes) that renders tissues insensitive to andro- M. Foss (Ed.), Determinants of Infant Behavior, pp. gens. When this sensitivity abates, development 37–44. London: Methuen. switches to the male track. One can readily sur- Hess, E. (1958). “Imprinting” in animals. Scientific mise that in most U.S. communities, a child un- American 198: 81–90. dergoing such a metamorphosis, female until age Hutchinson, G. E. (1951). Copepodology for the eleven or twelve, then transforming into a boy, ornithologist. Ecology 32: 571–577. would suffer severe psychosocial trauma, lead- Imperato-McGinley, J. L., L. Guerrero, T. Gautier, ing to a variety of behavioral pathologies. In the and R. E. Peterson. (1974). Steroid 5-alpha reductase Dominican communities accustomed to such deficiency in man: An inherited form of male pseudo- children, their condition is part and parcel of an hermaphroditism. Science 186: 1213–1215. expected scheme and evidently occasions no Kagan, J. (1984). The Nature of the Child. New York: untoward consequences. Again, it is the exis- Basic Books. tence of the ordered social system that allows for Kagan, J. (1998). Three Seductive Ideas. Cambridge, a stable, predictable outcome, even with the vari- MA: Harvard University Press. ations in the developmental sequence. Kagan, J., J. S. Reznick, and N. Snidman. (1988). In humans, the dimensions of the social envi- Biological bases of childhood shyness. Science 240: ronment are multitudinous. Thus, it should come 167–171. as no surprise that many different patterns of de- Klopfer, P. H. (1959). An analysis of learning in young anatidae. Ecology 40: 90–102. Parental Care and Development 173

Klopfer, P. H. (1965). Imprinting: A reassessment. Sci- ence 147: 302–303. Klopfer, P. H. (1971). Imprinting: Determining its perceptual basis in ducklings. Journal of Comparative and Physiological Psychology 75: 378–385. Klopfer, P. H. (1996). Mother-young attachments: On the use of animal models. American Scientist 84: 319– 321. Klopfer, P. H., and M. S. Klopfer. (1974). How come leaders to their posts? American Scientist 61: 560–564. Klopfer, P. H., and M. S. Klopfer. (1977). Compensa- tory responses of goat mothers to their impaired young. Animal Behavior 25: 286–291. Lorenz, K. (1937). The companion in the bird’s world. The Auk 54: 245–273. Lumsden, C. J., and E. O. Wilson. (1981). Genes, Mind and Culture. Cambridge, MA: Harvard University Press. Manning, A. (1965). Drosophila and the evolution of behavior. In J. D. Carthy and C. L. Duddington (Eds.), Viewpoints in Biology 4, pp. 125–169. Reading, MA: Butterworths. McCabe, T. T., and B. D. Blanchard. (1950). Three Species of Peromyscus. Santa Barbara, CA: Rood Associates. Schrödinger, E. (1951). What Is Life? New York: Cam- bridge University Press. Wecker, S. C. (1963). The role of early experience in habitat selection by the prairie deermouse, Peromyscus maniculatus bairdi. Ecological Monographs 33: 307–325. Weiner, J. (1999). Lord of the ﬂies. New Yorker, April 5, pp. 44–51. Whiting, J. M., and J. L. Childs. (1953). Child Training and Personality. New Haven: Yale University Press. Whorf, B. L. (1950). Language, Thought, and Reality. Cambridge, MA: MIT Press.

IV RETHINKING DEVELOPMENT AND EVOLUTION

15 Terms in Tension: What Do You Do When All the Good Words Are Taken?

Susan Oyama

Anyone attempting to articulate a developing a partial account of my own reasons for adopting position is likely to feel immobilized from time and sometimes revising (perversely, some may to time by the very conceptual and historical think) these terms. Misapprehensions occur no baggage that makes “good words” good—by the matter how precise one tries to be, and by sug- rich connectivity that gives them their power. gesting the conceptual locus from which they are One needs to recognize important differences and made, mistakings can be informative in them- commonalities with other traditions without selves. Though what follows is partly about recti- compromising the project at hand. Never easy, fying or preventing misreadings, then, I hope it such articulation becomes more risky still if one will also give readers a sense of the ongoing dis- seeks to make contact with other disciplines, for agreements and alliances out of which the devel- each has its own theoretical wrangles. It often opmental systems perspective has been, and is seems that the good words have already been now being, interactively constructed. In addition, taken, not to mention retaken and mistaken. One I would like to impart a sense of the potential of this book’s aims is to present some of the work scope of these interlocking ideas: how they pro- that clusters around the notion of the develop- vide a conceptual framework within which phe- mental system, but another is precisely this kind nomena can be brought together in ways that are of exploratory contact with allied approaches. both consonant and novel. In this chapter I investigate the concepts of interaction and system in the context of a view of development and evolution as constructive pro- Interaction cesses. These concepts are uniﬁed by an enlarged understanding of inheritance (Oyama 1982, 1985), Philip Kitcher (forthcoming) begins his contrib- and, like notes in a chord, they must be under- ution to a festschrift for Richard Lewontin with stood in relation to each other. They must also be a quotation from my Ontogeny of Information viewed in relation to the problems that the notion (Oyama 1985: 26–27): of the developmental system is meant to resolve. “But wait,” the exasperated reader cries, “everyone Words like this are untidy to work with, but, as I nowadays knows that development is a matter of inter- implied earlier, much of their resonance derives action. You’re beating a dead horse.” from their historical complexity. If we are careful, we can draw on these resources, heightening I reply, “I would like nothing better than to stop beat- some harmonics and damping down the rest. The ing him, but every time I think I am free of him he kicks me and does rude things to the intellectual and political alternative is neologism. This last should not be environment. He seems to be a phantom horse with a rejected in principle, but in the end we cannot thousand incarnations, and he gets more and more sub- make anything pristinely new. Nor is it clear why tle each time around. . . . What we need here, to switch we would want to. Whether we use good words metaphors in midstream, is the stake-in-the-heart or shun them, we must confront their existing move, and the heart is the notion that some inﬂuences senses. are more equal than others....” In sampling the possibilities and di≤culties of words like interaction, system, and construction, I In my book, the passage in question follows a dis- do not offer an etymology or systematic history, cussion of the ways in which DNA is given a spe- much less an all-purpose method for dealing with cial role in accounts of ontogenesis. The causal semantic dilemmas, but something more homely: privileging in such accounts is not a matter of saying that only genes are needed for develop- 178 Susan Oyama

ment (no one says this). It is the attribution of ceived. Although the dialectical biologists and I special directive, formative, or informative power have tended to focus on different authors and to genes—in short, the treating of some causes as issues,1 Kitcher is not wrong to see us as having more equal than others. This is often done de- similar theoretical preferences, even if they reject spite the fact that the analyses themselves show a the term and I do not. variety of elements playing analogous roles in Lewontin, Rose, and Kamin (1984) say that in- regulation, differentiation, and variation. Equal teractionism is “the beginning of wisdom,” but recognition, I argued, was not being rendered for add that one way it errs is in treating organism equal work, however the analyst seemed to be and environment as separate and independent measuring “work” at the time: The rules kept rather than as interpenetrating and interdepen- changing. To reject a special executive role for the dent (p. 268). It is an ism they object to, however, DNA is not to deny that all sorts of distinctions and many of the ways in which they use interac- can be made among factors and among the ways tion as a “civilian” word rather than as a special they impinge on development. Nor is it to deny term are very much in accord with the ways I do. the possibility of prediction or of making various They say of the constituent parts of a whole, for kinds of qualitative and quantitative statements instance, that “the interaction of these units in about particular sets of data. It is to say that the construction of the wholes generates com- many of the classical ways of describing biologi- plexities,” or “[O]rganisms alter the external cal processes are ill conceived, including many of world as they interact with it” (p. 11). There is those that claim to be paying appropriate atten- in these writings an appreciation of mutually tion to a variety of factors. As the allusion to deﬁning and mutually inﬂuencing organisms Orwell’s (1946) Animal Farm suggests, claims of and environments: of interactive emergence. Out- equality in biological causation can be given the comes tend to be attributed to a distributed set of lie by actual practice, just as they can in everyday participants, whose actions and effects are inter- political life. dependent, rather than to actors neatly separated Kitcher (forthcoming) believes that the stan- from that on which they operate. dard “interactionist Credo” signals a kind of This systemic view of causality also character- causal egalitarianism. He defends that interac- izes the developmental systems perspective (or tionism against a number of attackers, including, DST, for theory, broadly construed), which in it turns out, me. His charge that dialectical biol- addition stresses the dependence of both invari- ogy is hostile to interactionism is perhaps under- ant and variant features on the same kinds of standable; Levins and Lewontin (1985) explicitly complex sequences of interactions, rather than reject the term, as do Lewontin, Rose, and depicting some as principally organized from the Kamin (1984). The reason for including develop- inside and others from the outside. Both stability mental systems theorists in his treatment is less and change, similarity as well as variety, must immediately obvious; still, even though I have then be explained in the same general terms. As a always considered myself an interactionist of a way of illuminating the meaning of interaction- certain type, I have also been critical of what ism in DST, I am going to comment on several I tend to call “standard” or “conventional” or aspects of Kitcher’s critique. These remarks will “traditional” interactionism (in order to distin- not include a broad defense of the other writers guish it from what I now call constructivist in- he discusses, though I will refer to them from teractionism; see also B. H. Smith’s [1997: xxi] time to time. (See Gray, forthcoming, and Gri≤- related “constructivist-interactionist account of ths 1999, for overlapping responses.) The dialec- belief”). The question, then, is how standard in- tical biologists and I appear to have attended to teractionism and its alternatives are to be con- different aspects of current usage, and the situa- Terms in Tension 179

tion is too disordered to allow pronouncements and the nature of the causal processes produc- on what “interactionists” as a group really say ing those outcomes. (In the second part of this and do (for there is no “group” as such, and chapter, in fact, we encounter the opposite worry, Kitcher does not name names). Nevertheless, I that the developmental systems view posits more will permit myself some observations. regularity than is reasonable. Such attributions Readers of Vaulting Ambition (1985) will rec- reflect just the well-entrenched assumptions ognize the stance Kitcher assumes in the chapter about contrasting developmental explanations in question, as a reasonable, politically sensitive that are being rejected; see also Wimsatt [1986].) person who readily admits that people sometimes One of my own complaints about the various go beyond the evidence and who seeks to rectify conventional interactionisms is that their for- such mishaps without offending good sense. He mulations are often presented as resolutions of defends standard interactionism against its well- nature/nurture problems, when in fact they meaning but misguided critics, viewing some of accept the basic dichotomies from which those our concerns with sympathy but fearing that we problems grow (see Johnston 1987; Oyama 1985, work against ourselves by yearning for a better 2000b). Someone might say, for instance, that the response to genetic determinism, in the form of nature/nurture opposition is nonsensical because an undefined and obscure “trans-interactionist” some things don’t just mature, but require in- view (apparently his term). According to him, teraction. Here it is evident that interaction, far any problems in the literature flow from careless from challenging the concept of internally driven use of basically sound methods and concepts, not maturation, assumes it. The fatal weakness of op- from more fundamental di≤culties: Genetic de- positional stances that stress either genetic or en- terminism can be adequately dealt with using the vironmental effects, in fact, is that no matter how instruments at hand. useful their examples may be in disputing par- Briefly described, DST’s constructivist interac- ticular deterministic claims, they miss the larger tionism can resemble a number of other views, chance by accepting too much. real or imagined. It is not, however, to be con- In light of all this, I contend that Kitcher mis- fused with the bloodless interactionism that boils represents at least a significant segment of stan- down to “the environment is important too” dard interactionism, even as he misrepresents the (e.g., Buss et al. 1999; for contrast see Caporael’s positions of its critics. Though his own views are [1997] concept of repeated assembly). Versions not always laid out fully enough in the paper like Kitcher’s, which explicitly allow genetic and under discussion to allow a judgment, moreover, environmental variables to be analyzed in similar it is possible to argue that he himself is not a stan- ways, do not automatically qualify either; recall dard interactionist at all. I will look first at his that much privileging takes place despite such discussion of norms of reaction. One of his more analyses. We do not claim that statistical inter- disconcerting claims is that DST and dialectical actions will be found in every data set.2 Nor is biology consider them to be incoherent. Then I constructivist interactionism opposed in principle turn to the separation of causal factors and to the possibility of developmental stability or causal democracy. regularity. Although it is frequently most easily demonstrated by altering conditions to alter out- Norms of Reaction comes, that is, it should not be mistaken for some sort of environmental determinism or an insis- A norm of reaction shows the set of phenotypes tence that developmental plasticity is the rule. resulting when a given genotype develops in var- How could it be? Part of the point is to distin- ious (or more sweepingly, all possible or feasible) guish between issues of regularity of outcomes environments. At the same time that Kitcher 180 Susan Oyama

claims that DST and dialectical biology consider people tend to think of the body as genetically norms of reaction nonsensical, he acknowledges caused; the analytic logic, however, is the same Lewontin’s influential writings on the utility of for both.) Because Kitcher’s analysis of genetic such graphs. Gray (1992, forthcoming) and van determination may tacitly rely on the existence of der Weele (1999; see chapter 24) have also found an alternative genotype with a different norm of them useful. None of us, to my knowledge, has reaction, furthermore, we can add to my coun- said the idea of norm of reaction is itself ill terexample an alternative developmental environ- formed. We might well argue with some of the ment, perhaps one that yields some genetically ways it is used, however (see, for instance, Oyama correlated phenotypic variation. (It could be that 2000a, where I discuss privileging in statistical in this second environment, organisms lacking representations, including norms of reaction). the organ in question can survive.) Would we Kitcher says that one way to understand ge- now want to say that the morphological feature netic determinism, “the idea that genetic causes in the first case is completely environmentally de- take priority” (forthcoming), is as a claim of a flat termined because it appears with all of the geno- norm of reaction. I doubt that this ism can be so types (all surviving organisms have just one of the neatly pinned to particular predictions, but it parts in question)? And in the second case, does allows him to start with a convenient represen- the environment have less causal e≤cacy, and the tational device that makes genetic determinism a genes proportionately more? matter of degree. A flat norm of reaction is then Like Lewontin, Kitcher evinces considerable one extreme on a continuum. Notice that this mistrust of heritability studies, especially the choice bypasses some questions that immediately drawing of conclusions about causation from come up in a DST framework: What does the them. Yet, given Kitcher’s affection for the shape of a norm of reaction have to do with the “model of causal analysis that looks at the effects idea that “genetic causes take priority”? As I of a single genotype across varying environ- argue later, priority can be given in a variety of ments,” I am intrigued by his dismissal, for this ways. When Kitcher speaks of norms of reaction sounds like much of what goes on in investi- revealing causal contributions, I would ask, con- gations of heritability. (The “single genotype” tributions to what? What is being quantified would really be some set of organisms judged to here? If the genotype in question is being placed be genetically similar, possibly only at a single in every environment consistent with survival, locus, by a criterion that may involve no investi- say, then many, many other factors are invariant gation of the genetic material itself.) Heritable as well. Despite the fact that these are by defini- (genotype-correlated) variance can be quantified tion needed for development, Kitcher attributes by comparing the developmental products of var- causal e≤cacy only to the genes. Under what ious pairings of genotypes and environments in conception of developmental processes are genes an analysis of variance (ANOVA). It is possible more important in traits that emerge reliably to view the set of phenotypic outcomes for any in many environments than in those that vary particular genotype as a mini-norm of reaction: across them? Is the degree of environmental “de- the phenotypes that result when organisms with termination” proportionately greater if a genetic the “same” genotype are raised in varying devel- mutation brings about no phenotypic change?3 opmental environments. It is partial, of course, Imagine a morphological feature—having just but every actual norm of reaction shows only a one of a particular body part, perhaps—for small sample of possible conditions, and we are which there is no known genetically correlated typically warned against extrapolating beyond variation consistent with viability in some fixed them. The entire ANOVA matrix can then be developmental environment. (The point may be seen as an array of such norms of reaction (suc- sharper with body parts than behavior, because cessive rows, say), each one showing the average Terms in Tension 181

outcomes for individuals of the same genotype in The distinction between accounting for a differ- one of the environments studied. ence among phenomena and accounting for the Faced with a set of results that shows genetic phenomena themselves is significant. It is this main effects, one can associate differences in out- distinction (and more generally, the difference come with differences in genotype. Some account between population and developmental analysis) for similarities in outcome by shared genes, but that lies behind the warning that is so familiar in my view this is suspect, though it is the ex- to those who followed the race and IQ contro- planatory strategy Kitcher seems to prefer. He at- versies of the 1980s: not to confuse accounting tributes the “genetically determined,” uniform for variation among organisms with accounting developmental outcomes in a flat norm of reac- for the ontogeny and nature of the organisms tion to the genes that do not vary (are held con- themselves. It is one thing to speak of contri- stant, shared, identical, and so forth, albeit across butions to patterns of similarities and differ- a wider range of conditions than can be encom- ences among organisms under particular circum- passed by an ANOVA—or any other analytic stances. ANOVA and norms of reaction are technique I am aware of ). He appears to do with among the methods that allow this, and if part of norms of reaction, that is, what behavior geneti- Kitcher’s point is that we should attend to the cists often do with heritability—explain similar- range of conditions sampled, it is a point often ities by shared genes.4 made and well taken. It is another matter, how- The problem lies less with the analytic tech- ever, to say that some aspects of us are made nique per se than with the preanalytical assump- more by (more programmed by, more completely tions that guide its use, and thus the meanings inscribed in, more informed by) the genes than that are attached to the results. Kitcher may be others. The understandings that support what we taking objections to the latter as outright rejec- might call genetic essentialism is one of my con- tions of the norm of reaction as a tool. In DST, a cerns here, and they are not addressed by single factor might under certain circumstances Kitcher’s critique. This is one reason I consider be “determinative” in the sense of being a good the widespread usage of genetically determined to predictor across some range of cases, but a good indicate a phenotype with a flat norm of reaction deal of mischief comes from confusing this with to be so terribly unfortunate. A related di≤culty the more common causal meanings; they come is that the phrase thus used marks a lack of vari- together when genetically determined is used for ation, rather than the systematic variation that I organisms or characteristics rather than for pat- take to be the proper statistical use of determine. terns of outcomes. The question of developmen- The latter, stricter, usage is in fact frequently tal fixity, one sense of which is a flat norm of invoked by behavior geneticists and others who reaction, is often a significant ingredient of bio- bemoan the mishandling of the language of de- logical or genetic determinism, but it is not the termination. All this makes it very hard to hold only one. The “priority” of genetic causes can on to the distinction between differences and also be understood in terms of an underlying things.5 It does, however, fit the idea that there meaning, truth, or essence, perhaps pressing for are two theoretically separable kinds of explana- “expression.” Certain subtle and often more con- tions for organisms and their features, an idea sequential ways of prioritizing genes are therefore DST disputes. not touched by discussions that concentrate only on what I call issues of incidence—frequency, Separation of Causal Factors predictability, inevitability. My concern is thus not to endorse some meth- According to Kitcher, dialectical biology and ods and condemn others, but to ask what con- DST reject interactionists’ claim that the genes clusions are being drawn when a method is used. can be separated out as causal factors. There may 182 Susan Oyama

be something to this, but our objection is not press) rightly says that such organismic “determi- what Kitcher thinks.6 He attributes to Lewon- nation” of the relevant environment is consistent tin a belief that I have never seen expressed by with “the attempt to draw norms of reaction that that author or by any of the other people under identify the causal contributions.” I know no one discussion—indeed, cannot imagine any of us who has suggested otherwise, although the mean- holding: that something cannot be a “causal facing of “causal contributions” may be a point of tor” in the development of a phenotype if that disagreement. I would say, and I suspect Kitcher thing depends on the phenotype. Recall the would concur, that the genes cannot be separated causal language of the quotation from Levins out as prime movers: autonomous imparters of and Lewontin (1984: 100), about organisms alter- form and function, carrying in their very molecu- ing the world as they are altered by it. Unless lar structure representations of their potential or one simply stipulates that causes must be inde- proper developmental outcomes. Like any in- pendent of the things they affect,7 a chemical teractant, they become causes only with respect would be a “causal factor” affecting an organ- to other participants in an interaction, and their ism’s health even if it only becomes toxic when effects are similarly system-dependent, the ubiq- the organism’s digestive juices act on it. I believe uitous language of genetic privilege notwith- that such reciprocal dependency is one of the standing. The prevalence of such privileging was points of dialectical biology’s organism-environ- one of the early motivators for my own work; ment interpenetration. It is certainly one of the these notions are not easily addressed with the points of my concept of constructivist interac- analytic tools we have been discussing, but they tion: that this is the way these causal relations do yield to a different kind of approach. work, interdependently and interactively—not In DST constructivist interaction, interde- that they are not causal relations at all. So the pendence, and reciprocal contingency counterclaim is hardly that genetic effects on organisms act one-way causal stories, reminding us that cannot be identified, but that the genes have their although we may use our methods of random- effects by being affected by other factors—by ization and control to “isolate” the effects of a their cellular environments, if you will—and factor, we do so precisely because our factors are these often include the very processes they influ- abstracted from a dense causal complex. Statis- ence. The impact of gene products, furthermore, tical control, furthermore, hardly eliminates the tends to vary with other conditions. Starting an causal roles of stabilized factors. It simply ex- account with genetic transcription, and treating cludes them from the analysis. It is the neglect of the DNA as an “independent variable” that “ini- mutual influence, and the unprincipled separat- tiates” an interesting cascade of events, leads only ing out of genes as controlling, instructive agents, too easily to obliterating from the causal land- that DST resists, not conventional analyses of scape the events and conditions that brought that contributing factors or the possibility of causal transcription about. influence, whatever views one may have of cau- Kitcher then interprets interdependence as sality itself. This brings us to the issue of parity. causal relevance rather than actual physical impact, as when only certain aspects of a bird’s sur- Causal Democracy and Parity roundings are relevant to it, or, in the example just given, when a chemical is toxic to one organ- DST makes extensive use of parity of reasoning. ism but not another (see also Godfrey-Smith Descriptions and explanations of development 1996). In this sense, what counts as the bird’s en- are often asymmetric: The logic that is used vironment depends on what/how it is and what it (when there is a logic at all) to characterize cer- does, a matter to which we will return. Kitcher (in tain factors as informing, coding, controlling, Terms in Tension 183

and so forth, could be, but typically is not, ap- tions to a wider range of entities and factors than plied to other factors that play demonstrably is usual, to see whether conventional practice comparable roles. In contrast, DST includes as has prematurely or arbitrarily restricted its cate- full-fledged interactants many factors that are gories. The payoff is a broad reconceptualization generally left in the background (for examples of development, heredity, and evolution. and discussion of the pragmatics of research, Referring to an exchange with DSTers see van der Weele 1999). When Kitcher blandly Gri≤ths and Gray over the “genes for” locution preaches “causal democracy,” then, I am fasci- (Gri≤ths and Gray 1994, 1997; Sterelny and nated. Not only is most standard interactionism Kitcher 1988; Sterelny, Smith, and Dickison shot through with asymmetries, but the notions 1996), Kitcher concedes that his favored strategy, of causal symmetry, or parity, which do have of fixing genotype and varying environment to a democratic ring, inform the very concept of a see if a phenotype is genetically determined, can developmental system. Thus my pleasure that be reversed, so the environment is fixed while the Kitcher seems to like parity was somewhat, shall genotypes are varied. Once one accepts the logic, we say, muted. (Remember that he presents his he admits, this technique can reveal an environ- democracy as a tenet of the reasonable inter- mental factor “for” a phenotype, just as the re- actionism that needs to be defended against well- verse operation can be said to show genes “for” intentioned but fuzzy-headed folks like me.8) phenotypes. The exercise is useful because it ex- Confusion is diminished when one realizes that plicitly plays out the “gene for” logic, and so DST’s parity is quite different from Kitcher’s de- demonstrates its limitations. I would be very sur- mocracy. In fact, he says that DST cannot adopt prised if his standard interactionists, whoever his democracy principle, because the principle is they are, embraced this little bit of parity—let based on separating (but not prioritizing) causal alone its wider developmental and evolutionary factors (personal communication, May 1999). implications, which Kitcher does not discuss. The question is what it means to separate factors. Those phantoms of the allegedly dead horse of It appears that Kitcher has in mind more or less the nature/nurture dichotomy haunt the litera- what behavior geneticists do when they insist on ture, dragging behind them clanking chains of the legitimacy of separating genetic from envi- invidious distinctions among classes of causal ronmental causes (even though they believe this is “equals”—between those that define potential incompatible with “interactionism”; recall note and those that just select outcome, say—distinc- 6), and the framework is a population genetic tions that both require the dichotomy and per- one. petuate it. Hence my wish for just the kind of In DST, in any case, systematic even-handed- far-reaching reworking of perspective that DST ness is not a matter of outlawing particular meth- is about, and that Kitcher thinks is unnecessary. ods but of detecting and avoiding unprincipled Being several weary miles further down the privileging. If this is what you mean by “control” road than I was when I wrote the passage that or “inheritance,” does it characterize this other serves as Kitcher’s epigraph, I am now less likely case, too, even though you use another term for to talk about swift and lethal stakes in the heart. it? Notice that parity is not a matter of pro- As I suggested, a great deal turns on conceptions nouncing all factors to be “equal” or “the same” of causality. Many scientists avoid looking too in the sense that they cannot be distinguished. closely at such matters, and with good reason: Because there is an infinite number of ways one The philosophical literature in this area is ba- can compare things, this would be absurd. Parity roque. That a sophisticated and in some ways is not identity; it is consistency with respect to a sympathetic reader should misconstrue DST as criterion. DST applies certain criteria and defini- Kitcher does highlights the importance of attend- 184 Susan Oyama

ing to the conceptual frame that surrounds nar- ing on the political importance of the matters rower analytical matters. Having remarked in he tackled in that book, including some of the Vaulting Ambition on the vagueness of the “ge- deeply held assumptions about biological nature netic bases” that loomed so large in the socio- that dialectical biology and DST challenge. Pre- biology debates, he states that the controversy sumably he was also commenting on the lame- was really about evidence (1985: 7–8), and in the ness of formulaic egalitarian talk in the face of paper under discussion he counsels us to turn strictly stratified school systems. He says (forth- from “quibbling about the proper definition of coming), “[N]o interactionist denies that many genetic determinism” to investigating norms of causal factors are involved in development (that, reaction. I, on the other hand, doubt that such after all, is the point of interactionism).” I suggest evidence will do the trick without closer attention that this kind of thinly “liberal,”9 many-factors- to how “genetic bases” and “determination” are are-involved interactionism is not enough, either. used, and what conclusions they are thought to If, like many of the standard interactionists I justify. Such attention might well shed light on have written about over the years (Oyama 1982, what he calls the “fragility of our representations 1985, 2000b), he believes that biology and culture of the non-genetic causal factors.” (In Vaulting are “transmitted” via different “channels,” and Ambition, in fact, he does some useful quibbling that they “interact” in ways that allow their rel- of his own.) ative contributions to the constitution of or- I share Kitcher’s conviction that entirely too ganisms to be quantified, then we have serious many pronouncements are made about the rela- differences that have not been captured in his cri- tive importance of genes and environments in tique. If, on other hand, his vision is really of this or that trait (as opposed to careful investi- interdependent, symmetrically treated causal in- gation of particulars). Cavalier extrapolation is fluences, any one(s) of which may serve to dis- no substitute for case-by-case research, and I tinguish organisms in some study but none of am hardly against such work. I do think that re- which is a mastermind bearing representations of search alone will not dispel the problems under traits or organisms, or is even more responsible discussion. After all, the standard interactionist for making the organism, and all of which help travesties have flourished in a rich soil of findings make up an organism’s inheritance, then he is as on various degrees and kinds of susceptibility to atypical an “interactionist” as I am. various classes of developmental contingencies. Often they are attempts to accommodate developmental variation within quite retrograde System frameworks. Kitcher misses the myriad explanatory asymmetries maintained in standard interac- The emphasis on distributed causality, the emer- tionism, and he does not follow the evolutionary gence of form and function in the interaction implications of the more globally symmetrical of heterogeneous internal and external causal approach implied by the “environmental feature influences on various scales, the causal inter- for” analysis. It would be interesting to know dependencies and lack of absolute distinctions whether in his democracy some causes would still between causes and effects—such ideas hardly end up being more equal than others. originate with DST (or dialectical biology), but Kitcher began Vaulting Ambition with an anec- are associated with more general notions of sys- dote about a cousin who failed a major sorting tems. The constructivist interactionism I have exam in the English educational system and was been limning is in fact a quite “systemsy” one, given a bicycle as consolation. In saying, “[A] which brings me to my second good word. bicycle is not enough” (1985: 11), he was remark- This section is organized by several questions, in contrast to the first one, much of which was di- Terms in Tension 185

rected at single paper (albeit one that raised world governed by “self-regulating feedback sys- several pertinent questions). Must system imply tems”; diagrams of energy flow through ecosys- tightly predictable self-regulation, or can it serve tems showed clearly defined circuits, eventually broader purposes? Is it consistent with DST’s manipulable from the outside (Taylor and Blum stress on interaction? How does it fit with the idea 1991: 277). Communities became steady-state of construction? Effective exchange within and devices. across disciplines requires attention to such ques- This accords with popular visions of nature’s tions. Serious terminological and conceptual mis- balanced harmony. Ecological anthropologist matches with ecology, for instance, could present Andrew Vayda (1996: 9–15) warns against sys- di≤culties, for treating organisms and their (de- tems approaches in ecology, charging that no- velopmentally and evolutionarily effective) envi- tions of self-regulation divert attention from the ronments together suggests the fundamental need to establish particular causal linkages. He significance of ecological concepts (see chapters laments analysts’ habit of treating social pro- 9, 10, 19, and 22). cesses and ecosystems as predefined, with established boundaries, rather than allowing them to Systems: Regular by Definition? “grow through being studied,” as their various connections are explored (see also Taylor and Ecology’s pivotal term, ecosystem, has the word García-Barrios 1995). in question built right into it. At the beginning of Vayda is thus skeptical about DST’s use of the twentieth century, Frederic Clements treated terms like system and process (personal commu- plant communities as individuals with life cycles. nication, June 1997). Yet my own misgivings He took experimental physiology as his model: about the internalism of traditional treatments of Ecological succession was a community’s on- organismic development have fixed on the same togeny (Hagen 1988). In Joel Hagen’s (1992) sorts of problems. For him, systems thinking is history of the discipline, the fortunes of the eco- the culprit, whereas for me it has been part of the system concept and the organism metaphor are remedy. closely entwined. Organismic development, as Earlier, with regard to dialectical biology, I ob- usual, was considered autonomous, regular, goal- served that terminological differences may par- directed, internally driven. Arthur Tansley, the tially obscure considerable accord about overall plant ecologist who originated the term ecosys- approach, and perhaps this is the case here. But tem in the 1930s, complained that Clements took there is more. The privileging of wholes over a useful analogy too literally, seeing succession as parts that Vayda and I mistrust is an aspect of the maturation to a single climax state. Most ecolo- reifying holism that Lewontin, Rose, and Kamin gists later chose ecosystems over organismic (1984, chap. 10) regard as an inadequate response analogies, but Hagen (1992: 48) argues that the to reductionism: uniform parts subservient to community-as-cybernetic-system idea was de- wholes with minds of their own. Hagen (1992: 98) rived from the organism metaphor and incorpo- describes the tendency of particular species and rated aspects of it. populations to “evaporate” once ecosystems are Organism and machine metaphors coexisted in seen in terms of energy or information flow. these early ecologists’ work (Hagen 1992). Both When higher levels are privileged over lower Hagen and Peter Taylor (1988), however, report ones, a possible consequence is exaggerated a shift toward more formal cybernetic models confidence that the expected outcomes will ap- after World War II, when many scientists were pear. Another is inattention to the actual interac- drawn to information theory. Inspired by war- tions themselves, and to differences among the time engineering successes, they envisioned a interactants. 186 Susan Oyama

Systems: Emblematic of Unpredictability? predictability echo both the tensions Hagen writes about and the nature/nurture debates from Must system be constitutively contaminated with which my own investigations grew. Ecological this brand of holism?10 Environmentalists are as succession, international politics, and organismic likely to emphasize the fragility of ecosystems life cycles give varying impressions of orderliness, as their stability, to the point that Donella depending partly on the levels at which they are Meadows (1988: 16) speaks of the dismayed pa- observed and the indicators that are used, so to ralysis that can come from contemplating “Awful some extent the diverse emphases of these theo- Interconnections.” Systemic linkages can pro- rists are easily explained. The point, which rests duce cascades of catastrophes. Some might see in part on such variety, is that the concept of sys- such cascades as intrusions into an otherwise pre- tem has more possibilities than are typically ex- dictable natural order, but Hagen (1992: 27) ploited in any one literature, and should not be reports that even the staunchest believers in unduly circumscribed. predictable succession allowed for destabiliz- The notion of a developmental system can do ing events. He thinks scholars have missed the useful work only if it is construed broadly enough. heterogeneity and flexibility of early ecologists’ It cannot prejudge the issue (issues, really) of reg- ideas, maintaining that a constant tension be- ularity, but requires, by the stark expedient of tween stability, integration, and balance on the blocking the exits, that developmental questions one hand, and change and unpredictability on the be answered by looking at developmental pro- other, runs through Darwin and the entire his- cesses and products, not at what labels they bear. tory of ecology (1992, chap. 4). To keep the questions of regulation and variation Robert Jervis (1997) uses systemic contin- appropriately open, a quite lean sense of system is gencies, nonlinearities, and multiple and indirect called for: a collection of interdependencies that effects to show how often, and how startlingly, can, under certain conditions, be so regular as to predictions of social relations can be wrong. He appear to be imbued with goal-seeking agency, comments on the dangers of ceteris paribus but that is not defined by such regularity. This clauses (p. 76), criticizing those who treat systems minimal sense can capture the varied interactions as entities with wills, capable of subordinating and mutual influences of such networks without their constituent parts. Change and variation assuming their degree of integration and self- on the domestic level, he asserts, can influence regulation. It is minimal, that is, not because it national and international events: The variable allows only a few factors or only simple interac- characteristics of the parts can be important (pp. tions—on the contrary—but because it does not, 103–104). Commanding a richly varied store of as a matter of definition, involve specific kinds of examples, Jervis covers many other significant relations among constituents or among levels, or aspects of complex systems, such as historicity specific kinds of outcomes (but see chapter 17). and path dependence (pp. 155–156), as well as This kind of mobile complexity is not easily visi- research and intervention strategies (pp. 73–74, ble in the cogwheels and electrical circuits that 282–294). The surprises and uncertainties he Taylor and Blum’s (1991) postwar ecologists de- describes are not pathological irruptions; they vised. It is more easily seen in the looser sense of follow from the multileveled complexity of the system I advocate, though, as well as in Taylor’s systems themselves. own work (chapter 22), in which regularity of various kinds is possible but is not bought at the Open Systems, Open Questions cost of heterogeneity and historicity. Early systems theorist Ludwig von Bertalanffy The point is not that systems “really are” unpre- (1967: 66–69) at one point describes a system sim- dictable.11 The alternatives of regularity and un- ply as “a complex of components in interaction.” Terms in Tension 187

The mechanistically stimulus-response cyberne- skin. Tansley’s critique was mentioned earlier tic system is one subtype; but the attributes of (Hagen 1992: 78–86). He did not question inter- what Bertalanffy calls general (or open) systems nalist understandings of ontogeny, only their im- are easier to connect with the developmental sys- moderate application in ecology. Before settling tems approach. They include “dynamic inter- on ecosystem, he even considered quasi-organism. action between many variables” rather than the From the present point of view two things are simpler “circular causality” of feedback regula- noteworthy about his account: (1) ecosystems tion; an emphasis on exchanges of matter rather were identified and somewhat artificially isolated than information; and the ability to self-organize, by the scientist confronting a nature of overlap- a kind of self-making of which a cybernetic sys- ping parts and indistinct boundaries (compare tem is incapable, but which is indispensable if we with Vayda on predefined systems, cited earlier), are to talk about developmental construction. It and (2) they included abiotic components. Ray- is this last, the possibility of growth and differ- mond Lindeman, a young colleague of G. Evelyn entiation, that has attracted those studying the Hutchinson, developed this idea with his mentor ontogenies of actual organisms. a decade later. How, Lindeman asked, could one As noted earlier, Lewontin, Rose, and Kamin separate the living from the nonliving, given the (1984: 279) believe that early systems theorists fact that the former disintegrated into molecules, were occasionally guilty of wooly holism. This whereas abiotic materials were taken up and may be the case. I am not attempting to recover moved through food chains? Biogeochemistry an ur-systems theory, but to show that those blurred the boundaries. writings may be roomier than is sometimes Ecosystem modeling relied, however, on single thought. Developmental biologists have tended currencies of energy or information. This is fi- to be impressed by regularity, but there is more nally not as congenial to the present project as than one way to explain it. What does it mean for the boundary-breaching itself, which sits nicely ecological succession to be (like) development? with the inclusion in a developmental system of Organismic maturation is the model, but this can not only abiotic resources but also other organ- be conceived in noninternalist terms (Oyama isms. As Johnston (chapter 2) shows, certain psy- 1982), and ecological succession has been in- chologists and biologists saw that an account of voked precisely to challenge the notion of pro- species-typical development must be extended grammed development. Yrjö Haila (1999) argues beyond the confines of the body. Bertalannfy’s that ecological succession and seasonal changes (1967: 69) mention of the heterogeneity of inter- should be seen as developmental, but confirms acting components, along with his considerably (personal communication, December 1998) that earlier declaration that “determination factors” this must not be misconstrued. Development it- could be found both inside and outside the egg self must be understood in a more open and con- (1962: 68–70), also suggest historical precedent tingent manner than usual. To put it crudely, for a more ample notion of system.12 while a social theorist or ecologist might have There is irony here. The concept that some find asked whether societies or sand dunes were “like useful in undoing the internalism of traditional organisms,” and meaning by this, were they self- notions of development has, for others, its own contained, self-guiding, and self-maintaining, internalist aura (just as the interactionism dis- certain developmentalists with a systems bent cussed in my first section seems externalist to eventually asked whether organisms were “like some). It is instructive to consider the contrast to organisms” in that sense. system in each case. In ecology it was something One of the more precise questions nested in rather more “individualistic”—if not an unthe- this last one is whether an adequate explanatory orized collection of creatures and features, then complex is to be found within the organism’s at least a focus on particular populations and 188 Susan Oyama

species (Hagen 1992, chaps. 2, 5). For develop- and external factors capture the interdependence mental systems thinkers the contrast is with an of these phenomena. opposition between internal control and external Development has conventionally been ex- input. The kind of system that has served us plained by internalist models and evolution, involves heterogeneous, interdependent causal largely by externalist (especially selectionist) factors both inside and outside the skin; the pos- ones. But developmental constancy is no less a sibility of more or less (sometimes much more or product of (systemic) interaction than is varia- much less) orderly processes without a preformed tion. In like manner, lability, unpredictability, plan; and the emergence of structure and func- and variability is no less (interactively) systemic tion from specific causal interactions among very than is constancy. I have argued for a parallel specific conditions—in short, a very “interac- synthesis of evolutionary insides and outsides tiony” sort of system. The collisions and cooper- that I take to be in line with the idea of organ- ations of varied and distributed causal factors ism-environment interpenetration (Levins and should invite, even demand scrutiny, rather than Lewontin 1985). Both developmental and evolu- being easily black-boxed (with the justification tionary construction are historical comings into that all the “information” is inside the pack- being, by concrete events and activity on a vari- age). The role of the investigator in defining the ety of scales: changes in the dimensions, material system is crucial, however; one doesn’t glance out constituents and modes of functioning of an or- the window and see one of these organism- ganism and its worlds. The emphasis is on tem- environment complexes, shrink-wrapped and porality and physicality, partly to counteract the ready to go. disembodied, formal quality of the language of programs, algorithms and gene pools (Hendriks- Interactive Construction in Systems Jansen 1996; Varela, Thompson, and Rosch 1991). The notion of construction is inseparable from It should go without saying that there is no the readings of interaction and system being pre- constructor, but I have learned I must say it: The sented here. There is a recurrent tendency to gene does not build organisms in some special associate predictability and constancy mainly centrally controlled way that other interactants with insides, and (certain kinds of ) change and do not. But then, the organism does not make variation with outsides. And yet Lehrman (chap- (most of ) its own environment, either, though it ter 3), Gottlieb (1997; see also chapter 4), and does select and alter its surroundings. Nor does other developmentalists describe exchanges with “the environment” make organisms or adapta- the external environment that are crucial to tions over ontogenetic or phylogenetic time. The the ontogeny of characters usually considered conjoining of construction to interaction and sys- “innate”—not as acquired additions to a “biolog- tems is meant to work against this persistent de- ical base,” but as aspects of the growth of the sire to meet the maker. organism itself. And as Lewontin (chapter 6) Also to be headed off is any reading of con- and other evolutionists have argued, develop- struction deriving from the very dichotomies it mental, behavioral, and other “internal” factors denies. This is not construction as opposed to (that is, characteristics of species, and thus in maturation. It is not socially constructed as op- many treatments internal with respect to the evo- posed to biological or real or natural or essential lutionary environment) contribute to the natural or universal. Rather, it is a reminder that every selection usually said to shape populations from aspect of an organism, rare or boringly predict- the outside. In neither case does the mere (stan- able, must develop, from some always incom- dard interactionist?) inclusion of both internal pletely characterized and changing (over both Terms in Tension 189

developmental and evolutionary time) complex tension between unpredictability and stability). of interactants. Interaction, at least the reciprocally constraining, It is said in DST that the developmentally rele- mutually constructive interactionism of a devel- vant environment depends on the organism. This opmental system, can keep a complex from melt- means that whether, and how, any aspect of the ing into an undifferentiated whole or closing in surround is involved in producing an organism is on itself and becoming an inscrutable black box. a function of that organism’s characteristics and It is a reminder of concrete events, of organisms its activity. But the operative features of the sur- in real worlds, cells in real tissues, groups in real roundings are not just “preregistered,” as Horst societies. System, at least the open, loosely de- Hendriks-Jansen (1996) puts it in a discussion of fined (or better, always to-be-specified) systems I situated robotics, there in some straightforward have described, can help us see that the interac- way, but are identified by reference to the organ- tions are connected, though not always in neat ism and may emerge only through its activity. He spatial bundles. Hence my interactiony systems, describes a device that follows walls without my systemsy interactions. Together with construc- being instructed specifically to do so. It navigates tion, they offer a different way of dealing with by “landmarks” that are not actual features of developmental and evolutionary stability and the surroundings but are based on correlations change. between what their sensors pick up and their own It is not a bad thing that these words remain movements (p. 135). Likening the landmarks to richly complicated and variously read. They can Gibsonian invariants, which are produced as an no more be made clear once and for all than a animal moves, Hendriks-Jansen points out that developmental system can be pinned down to a they exist neither in the head nor in the sur- fixed location or frozen in time, and its elements roundings. They are constructed during an in- enumerated once and for all. We can, and in- teractional history (p. 9), and “the emergent evitably will, keep talking about them, but more phenenomena open up possibilities for behavior to the point, using them. What can be accom- that did not exist prior to their emergence” (p. plished with distinctions, comparisons, and lists 30).13 This is what interactive construction looks of disclaimers is in the end limited, though I have like in a living system, too. It is this kind of mu- done my best. In the long run what will be per- tual definition of influences that lies behind suasive, or not, is the systematic deployment of DST’s and dialectical biology’s talk of the ulti- this vocabulary, so familiar in some ways (com- mate impossibility of separating causes, and it is posed as it is of rich, powerful, messily good not captured by Kitcher’s analytic “democracy.” words) and so subversively odd in others, to pose questions, analyze, interpret, distinguish, and tie together. A sampling of such efforts is collected in Terms in Tension this volume. Their appeal will depend on their ability to produce that combination of novelty, One could say that in DST interaction and system coherence, and utility that recommends a con- are linked in a kind of dynamic tension, a tension ceptual framework to the interested scholar. that is at once a source of their utility and their trickiness. I am, with some seriousness, adopting a constructivist-interactionist-systems attitude Acknowledgments toward these terms, for words, too, are always in a context of language and practice, each with a I thank Yrjö Haila, Philip Kitcher, Barbara history. Any of my good words alone threatens to Smith, Peter Taylor, and Rasmus Winther for slip toward precisely the implications one wants comments on earlier drafts of this chapter. A to keep at bay (recall Hagen’s account of the skeletal version of this material was presented at 190 Susan Oyama

the July 1997 meetings of the International is easy to forget because of the habit, encouraged by Society for History, Philosophy and Social nature/nurture oppositions, of treating “the environ- Studies of Biology, in Seattle, Washington. The ment” as unitary and thus capable of being manipu- session was entitled “Conceptualizing Develop- lated or fixed as a whole. One might argue that, given the selectivity of factors mental Processes (DST 3).” and levels in an ANOVA, Kitcher would reserve the language of determination for outcomes in all possible Notes (or “almost all,” etc.) environments. But then it could virtually never be used, and we would be left with the 1. Lewontin, Rose, and Kamin (1984: 272) mention question of how to treat results of analyses that could Popper, Lorenz, Campbell, and Piaget; their remarks actually be done, as opposed to ones that could not. on, and dissatisfaction with, interactionism should be Practically feasible ones produce the partitionings of read in light of this. My own sources included the tra- variance that are typically taken to show degrees of ge- dition of animal behavior studies discussed by Gottlieb netic and environmental determination, but different and by Johnston (see chapters 2 and 4), as well as cer- studies will produce different partitionings, and I know tain trends in developmental biology and developmen- of no way of placing them on the overarching contin- tal psychology (see discussions in Oyama 1999, 2000b). uum Kitcher seems to have in mind, precisely because I mentioned Piaget in my book because of his emphasis this continuum presupposes knowledge of all the perti- on constructive activity, and because of his willingness nent environments. to take a developmental approach to certain species- 5. The causal issues are daunting, so it is di≤cult to common cognitive features, not just those that vary bring diverse treatments into proper confrontation. Be- among individuals. When it comes to his tendency to havior geneticists frequently insist on a strict language describe the environment as setting evolutionary and of populations and differences, not individuals and cognitive problems for organisms (and, perhaps, his similarities. At other times, however, they are quite insistence on the necessity of logicomathematical content to talk of inherent gene-based natures or of knowledge), I suspect the dialectical biologists and I similarities among individuals being due to shared have similar objections. genes (Plomin 1986: 7–9; Rowe 1994: 3; see Oyama 2. Because statistical interactions involve nonadditive 1988 for comments). variance, I find the phrase “nonadditive interactions” Sober (1988: 317–318) argues for more distinctions: unhelpful, even when used to make very much the sorts between something’s being a cause and its making a dif- of points I make here (as in Gray, forthcoming, and ference, and between its “relative contribution” and the Sterelny and Gri≤ths 1999). Pat Bateson, in a number difference it makes. The genes, he says, can have “zero of personal communications over the years, has ob- magnitude” in the case of a person who would have had jected to interaction itself on similar grounds: that it im- the same height if she had had different genes, so that plies the statistical meaning. I would have been inclined “genes can be a cause of height, even if they are judged to adopt his alternative, interplay, if only it lent itself to to have zero magnitude.” This is a kind of converse of adjectival and adverbial constructions: I would be hard the flat norm of reaction, which for Kitcher shows put to talk about an interplayive relationship! genetic determination (100% causal magnitude?). 3. The spelling out of the connection between causal Perhaps some such set of distinctions informs Kitcher’s priority and such representations will require some at- presentation. tention to quite basic conceptions of causality, with 6. It is amusing to juxtapose his allegedly interactionist special attention to the relations between individual goal of separating the causal roles of genes and en- and population analyses. It may well be, as Godfrey- vironments with behavior geneticist Robert Plomin’s Smith suggests (chapter 20), that Kitcher’s particular testy reference to the “mistaken interactionist notion views on causality explain some of these striking that the separate effects of heredity and environment incongruencies. cannot be analyzed” (1986: 7). 4. The reason this is suspect is related to the point 7. Kitcher does not do this explicitly, and my impres- made earlier: Many aspects of the developmental envi- sion is that his (1989) more formal treatment of causal- ronment are also shared, even as others vary. This ity does not require it. Terms in Tension 191

8. There is a gap here. What he actually says is that Bertalanffy, L. von. (1967). Robots, Men and Minds: standard interactionists should be democratic. His dis- Psychology in the Modern World. New York: George cussion, however, implies quite strongly that they are. Braziller. In the restricted sense of being willing to attribute some Buss, D. M., M. G. Haselton, T. K. Shackelford, A. L. role to both genetic and environmental factors in cer- Bleske, and J. C. Wakefield. (1999). Interactionism, tain analyses, they sometimes are, but the question is flexibility, and inferences about the past. American Psy- how such roles are interpreted. chologist 54: 443–445. 9. I use this language in response to Kitcher’s appro- Caporael, L. R. (1997). The evolution of truly social priation of democracy. I do not, however, endorse the cognition. Personality and Social Psychology Review, 1: political terminology or wish to adopt it more gener- 276–298. ally. The rhetorical gestures it licenses are likely to sow Godfrey-Smith, P. (1996). Complexity and the Function complication without conceptual gain. of Mind in Nature. Cambridge: Cambridge University 10. Hagen (1992: 137) suggests that holism aided in the Press. formation of a new specialty and stimulated research; Gottlieb, G. (1997). Synthesizing Nature-Nurture: such things can help at one time and become a hin- Prenatal Roots of Instinctive Behavior. Mahwah, NJ: drance at another. Lawrence Erlbaum Associates. 11. Oddly enough, Schaffner (1998) associates DST’s Gray, R. D. (1988). Metaphors and methods: Be- “developmental emergentism” with unpredictability. havioural ecology, panbiogeography and the evolving See Gri≤ths and Knight (1998). synthesis. In M. W. Ho and S. W. Fox (Eds.), Evolu- 12. In this last passage he seems to be resisting a privi- tionary Processes and Metaphors, pp. 209–242. London: leging of internal factors over external ones. The term Wiley. determination factor is from Roux, who located these Gray, R. D. (1992). Death of the gene: Developmental formative influences inside the egg, distinguishing them systems strike back. In P. E. Gri≤ths (Ed.), Trees of from mere “factors of realization,” which could be ex- Life: Essays in Philosophy of Biology, pp. 165–209. ternal and which supplied matter and energy. These Dordrecht: Kluwer Academic. very familiar kinds of arguments are grist for parity analysis. Gray, R. D. (forthcoming). Selfish genes or developmental systems? Evolution without replicators and ve- 13. The robot was designed by Maya Mataric. In situ- hicles. In R. Singh, C. Krimbas, J. Beatty, and D. Paul ated robotics the emphasis is on bottom-up emergence (Eds.), Thinking about Evolution: Historical, Philosophi- of behavior from a few simple low-level rules, involving cal, and Political Perspectives. Cambridge: Cambridge actions like advancing and turning. Though these are University Press. behavioral examples, developmental interactions are similar. In Bateson’s mismatches between children and Gri≤ths, P. E. (1999). The fearless vampire conserva- adoptive parents (chapter 13), the significant variable is tor. In C. Rehmann-Sutter and E. M. Neumann-Held Genes in Development. Rereading the Molecular not a characteristic of parents, children, or “environ- (Eds.), Paradigm. ment,” but rather a relationship that emerges as they in- Manuscript submitted for publication. teract with, and react to, one another. Peter Klopfer Gri≤ths, P. E., and R. D. Gray. (1994). Developmental (chapter 14) gives other examples. systems and evolutionary explanation. Journal of Phi- losophy, 91: 277–304. Gri≤ths, P. E., and R. D. Gray. (1997). Replicator II: References Judgement day. Biology and Philosophy 12(4): 471–492. Gri≤ths, P. E., and R. D. Knight. (1998). What is the Bateson, P. P. G. (1988). The active role of behaviour in developmentalist challenge? Philosophy of Science 65: evolution. In M. W. Ho and S. W. Fox (Eds.), Evolu- 253–258. tionary Processes and Metaphors, pp. 191–207. London: Wiley. Hagen, J. B. (1988). Organism and environment: Frederic Clements’s vision of a unified physiologi- Bertalanffy, L. von. (1962). Modern Theories of Devel- cal ecology. In R. Rainger, K. R. Benson, and J. opment: An Introduction to Theoretical Biology. New Maienschein (Eds.), The American Development of York: Harper & Row. 192 Susan Oyama

Biology, pp. 257–280. Philadelphia: University of Penn- chology. [Reciprocal book reviews and author’s replies.] sylvania Press. Developmental Psychobiology 21: 97–100, 101–105. Hagen, J. B. (1992). An Entangled Bank: The Origins of Oyama, S. (1992). Pensare d’evoluzione. L’integrazione Ecosystem Ecology. New Brunswick, NJ: Rutgers Uni- del contesto nell’ontogenesi, nella filogenesi, nella cog- versity Press. nizione (Thinking about evolution: Integrating the Haila, Y. (1999). “Biodiversity” and the nature/culture context in ontogeny, phylogeny and cognition). In M. divide: Conflicting tendencies. Biodiversity and Conser- Ceruti (Ed.), Evoluzione e Cognizione. L’Epistemologia vation 8: 165–181. Genetica di Jean Piaget e le Prospettive del Costruttivis- mo, Hendriks-Jansen, H. (1996). Catching Ourselves in the pp. 47–60. Bergamo, Italy: Lubrina Editore. Act. Cambridge, MA: MIT Press. Oyama, S. (1999). Locating development, locating de- Jervis, R. (1997). System Effects: Complexity in velopmental systems. In E. K. Scholnick, K. Nelson, S. Conceptual De- Political and Social Life. Princeton, NJ: Princeton Uni- A. Gelman, and P. H. Miller (Eds.), velopment: Piaget’s Legacy, versity Press. pp. 185–208. Hillsdale, NJ: Lawrence Erlbaum. Johnston, T. D. (1987). The persistence of dichotomies in the study of behavioral development. Developmental Oyama, S. (2000a). Causal democracy and causal con- Philosophy of Science Review 7: 149–182. tributions in DST. 67 (proceedings): 332–347. Kitcher, P. (1985). Vaulting Ambition: Sociobiology and Evolution’s Eye: A Systems View of the Quest for Human Nature. Cambridge, MA: MIT Oyama, S. (2000b). the Biology-Culture Divide. Press. Durham, NC: Duke Uni- versity Press. Kitcher, P. (1989). Explanatory unification and the Development, Genetics, and Psychol- causal structure of the world. Minnesota Studies in the Plomin, R. (1986). ogy Philosophy of Science 13: 410–505. . Hillsdale, NJ: Lawrence Erlbaum Associates. The Limits of Family Influence: Kitcher, P. (forthcoming). Battling the undead: How Rowe, D. C. (1994). Genes, Experience, and Behavior. (and how not) to resist genetic determinism. In R. New York: Guilford Singh, C. Krimbas, J. Beatty, and D. Paul (Eds.), Press. Thinking about Evolution: Historical, Philosophical and Schaffner, K. F. (1998). Genes, behavior, and develop- Political Perspectives. Cambridge: Cambridge Univer- mental emergentism: One process, indivisible? Philoso- sity Press. phy of Science 65: 209–252. Levins, R., and R. Lewontin. (1985). The Dialectical Smith, B. H. (1997). Belief and Resistance: Dynamics of Biologist. Cambridge, MA: Harvard University Press. Contemporary Intellectual Controversy. Cambridge, Lewontin, R. C., S. Rose, and L. J. Kamin. (1984). Not MA: Harvard University Press. in Our Genes. New York: Pantheon. Sober, E. (1988). Apportioning causal responsibility. Journal of Philosophy Meadows, D. (1988). World interconnectedness also 85: 303–318. works in our favor. Annals of Earth 6(1): 16. Sterelny, K., and P. Gri≤ths. (1999). Sex and Death: An Introduction to Philosophy of Biology Orwell, G. (1946). Animal Farm. New York: Harcourt . Chicago: Univer- Brace. sity of Chicago Press. Oyama, S. (1982). A reformulation of the concept of Sterelny, K., and P. S. Kitcher. (1988). The return of Journal of Philosophy maturation. In P. P. G. Bateson and P. H. Klopfer the gene. 85: 339–361. (Eds.), Perspectives in Ethology vol. 5, pp. 101–131. Sterelny, K., K. C. Smith, and M. Dickison. (1996). New York: Plenum. The extended replicator. Biology and Philosophy 11: Oyama, S. (1985). The Ontogeny of Information: Devel- 377–403. opmental Systems and Evolution. Cambridge: Cam- Taylor, P. J. (1988). Technocratic optimism, H. T. bridge University Press. (2d rev. ed., Durham, NC: Odum, and the partial transformation of ecological Duke University Press, 2000.) metaphor after World War II. Journal of the History of Biology Oyama, S. (1988). Reply to Robert Plomin’s review of 21: 213–244. The Ontogeny of Information, and Populations and phe- Taylor, P. J. (1995). Building on construction: An notypes: A review of Development, Genetics, and Psy- exploration of heterogeneous constructionism, using Terms in Tension 193

an analogy from psychology and a sketch from socio- economic modeling. Perspectives on Science 3(1): 66– 98. Taylor, P. J., and A. S. Blum. (1991). Ecosystems as circuits: Diagrams and the limits of physical analogies. Biology & Philosophy 6: 275–294. Taylor, P. J., and R. García-Barrios. (1995). The social analysis of ecological change: From systems to intersecting processes. Social Science Information 34(1): 5– 30. van der Weele, C. (1999). Images of Development: Envi- ronmental Causes in Ontogeny. Albany: State Univer- sity of New York Press. Varela, F. J., E. Thompson, and E. Rosch. (1991). The Embodied Mind. Cambridge, MA: MIT Press. Vayda, A. P. (1996). Methods and Explanations in the Study of Human Actions and Their Environmental Ef- fects. Jakarta: CIFOR/WWF. Wimsatt, W. C. (1986). Developmental constraints, generative entrenchment, and the innate-acquired distinction. In W. Bechtel (Ed.), Integrating Scientiﬁc Dis- ciplines, pp. 185–208. Dordrecht: Martinus-Nijhoff.

16 Darwinism and Developmental Systems

Paul E. Gri≤ths and Russell D. Gray

Open almost any biology textbook and you will the range of phenomena that can be given find the following definition: Evolution is change adaptive-historical explanations. in gene frequency.1 This definition reflects the conventional view of natural selection and the conventional view of heredity. Natural selection Inheritance occurs because individuals vary, some of these variations are linked to differences in fitness, and What does an organism inherit? Certainly more some of those variants are heritable (Lewontin than the nuclear DNA. A viable egg cell must 1970). Because variants that are not heritable contain a variety of membranes, both for its own cannot play a role in natural selection, and be- viability as a cell and to act as templates for the cause the mechanism of inheritance is presumed assembly of proteins synthesized from the DNA to be genetic, evolution is defined as change in into new membrane. A eukaryote cell must con- gene frequencies.2 tain a number of organelles, such as mitochon- Developmental systems theory (DST) is a rad- dria, with their own distinctive DNA. But the full ical challenge to dichotomous accounts of de- variety of the contents of the cell is only now velopment—accounts centered on a dichotomy being uncovered. For normal gene transcription between genes on the one hand and every other to occur, DNA must be accompanied by the ele- causal factor on the other. Proponents of DST ments of the chromatin marking system. For argue that the empirical differences between the normal differentiation of the embryo, initial cyto- role of DNA and that of other developmental re- plasmic chemical gradients must be set up within sources do not justify the metaphysical distinc- the cell. The essential role of still further parts of tions currently built upon them. In particular, the package, such as microtubule organizing cen- any sense in which genes code for phenotypic ters, is becoming apparent. But unpacking the traits or program development or contain de- inherited resources in the cell is not the end of un- velopmental information can be equally well ap- packing inheritance. In multicellular organisms plied to other factors required for development the parental generation typically contributes ex- (Gray 1992, 2001; Gri≤ths and Gray 1994a; tracellular resources. An ant in a brood cell is ex- Johnston 1987). In this paper we explore the imposed to a variety of chemical influences that lead plications of this “parity thesis” (Gri≤ths and it to develop as a worker, a queen or a soldier. Knight 1998) for the orthodox picture of evolu- A termite inherits a population of gut endosym- tion by natural selection. It is generally accepted bionts by coprophagy. In viviparous organisms that the neo-Darwinian synthesis marginalized the environment of the womb provides not only developmental biology (Depew and Weber 1995; nutrition but also a range of stimulation essential Gilbert, Opitz, and Raff 1996). Attempts to refor the normal development of the nervous sys- introduce developmental considerations have tem (for examples, see chapter 4 this volume). often been framed as attacks on (neo-) Darwin- This stimulation continues after birth. The effects ism (Goodwin 1984). We think this is a mistake. of severe deprivation of conspecific stimulation in Here we explore how the traditional Darwinian infants has been documented in many tragic “ex- concepts of inheritance, selection, adaptation, periments” (Harlow and Harlow 1962; Money and lineage can be reworked from a developmen- 1992). Nor are these effects confined to animals. tal systems perspective. Rather than diminish Many eucalypt species have seeds that cannot the power of natural selection, in the spirit of germinate until they have been scorched by a Darwin’s original insight this approach expands bushfire. To increase the frequency of bushfires 196 Paul E. Gri≤ths and Russell D. Gray

to the point where this system works reliably, nomena discussed, and their evolutionary sig- local populations of eucalyptus trees must create nificance recognized, in John Maynard Smith’s forests scattered with resinous litter and hung own work (Maynard Smith and Szathmáry with bark ribbons. These are carried aloft by the 1995). The concept of inheritance is used to ex- updraft as blazing torches and spread the fire to plain the stability of biological form from one new areas. Even after the resources created by the generation to the next. In line with this theoreti- population as a whole are added in, a range of cal role, developmental systems theory applies other factors must be present before the sum of the concept of inheritance to any resource that is the available resources adds up to a viable pack- reliably present in successive generations, and is age. Development frequently requires gravity or part of the explanation of why each generation sunlight or, for a hermit crab, a supply of dis- resembles the last. This seems to us a principled carded shells from other species. These factors definition of inheritance. It allows us to assess the are unaffected by the activities of past genera- evolutionary potential of various forms of inher- tions of the species that rely on them. Neverthe- itance, rather than immediately excluding every- less, the organism must position itself so that thing but genes and a few fashionable extras. these factors interact with it and play their usual One way to conceptualize the role of extra- role in development. While the evolving lineage genetic inheritance is as a number of separate cannot make these resources, it can still make (“parallel”) channels for the transmission of de- them part of its developmental system. velopmental information. The most traditional It is uncontroversial to describe all these re- multiple channel model has two systems of hered- sources as playing a role in development. But it is ity: genes and culture. In recent years more bio- highly controversial to say that these same re- logically sophisticated models, with their roots sources are “inherited.” With the exception of in actual empirical work on inheritance, have genes, and more recently the chromatin marking emerged (see chapters 9, 10, and 23 of this vol- system, their roles are not supposed to extend ume). Multiple channel models are an effective to the intergenerational processes of evolution. way to draw attention to the phenomena over- Nongenetic factors, it is generally supposed, do looked by a purely genocentric account of hered- not have the capacity for replication through ity. However, we have strong reservations about many generations, and lack the potential to pro- multiple inheritance or “extended replicator” duce the kind of variation upon which natural (Sterelny, Dickison, and Smith 1996) models. We selection can act: “The special status of genetic believe that it is both more biologically realistic factors is deserved for one reason only: genetic and, in the long run, more productive to think of factors replicate themselves, blemishes and all, the life cycle being reconstructed by a system of but non-genetic factors do not” (Dawkins 1982: resources. Let’s start with issues of biological re- 99). Or, more bluntly: “Differences due to na- alism. So-called channels are not generally in- ture are likely to be inherited whereas those due dependent of one another. Many “channels” are to nurture are not; evolutionary changes are so strongly intertwined that they cannot affect de- changes in nature, not nurture” (Maynard Smith velopment unless other channels develop in a way 2000). that is automatically “epistatic” with the first The continued popularity of this argument is channel. The chromatin marking system, often puzzling. Many nongenetic resources are reliably described as a “parallel” inheritance mechanism, passed on across the generations. Variations in modifies the pattern of gene expression. It can be these resources can be passed on, causing changes useful to treat DNA sequences and chromatin in the life cycle of the next generation. It is still marks separately in some modeling exercises, but more puzzling to find many of these very phe- this is an idealization akin to leaving out linkage Darwinism and Developmental Systems 197

in a genetic model. It should not be built into the dition makes use of it (see, for example, chapter 4 basic way we conceive the system. Furthermore, of this volume). However, describing this work the developmental system as we conceive it in- in the way we have just done keeps the context- cludes not only the “channels” of the other for- dependence of causation in development in the mulations, but also developmental resources center of the stage. One of the main motivations which are not easily represented as “channels” or of developmental systems theory is to draw at- “replicators.” It is hard to think of germination tention to fact that developmental causes do not in eucalypts as a character transmitted via the have their effects in isolation, but as part of a bushfire channel, but it would be necessary to wider system of causes. Causation in develop- think this way to make a multiple channel model ment is thus intrinsically likely to be context- complete. The extended replicator theory handles dependent. The very idea of “developmental this case by treating it as the replication of selfish information” runs some risk of disguising this bushfires which use eucalypt trees to achieve their fact. Susan Oyama (1985) points out that once in- goals. Whether or not that is adequate, the ex- formation is localized in, for example, a sequence tended replicator theory also has to model the of DNA, it is all too easy to forget that the devel- standing features of the physical world which opmental effect of this sequence is a function of form part of most developmental systems. Sun- context. The same DNA sequence in a different light, gravity, mineral concentrations in the local time or place might convey quite different infor- soil, and many other factors must be present if mation. “Information” here is being used in the “channels” are to convey and “replicators” to statistical sense, that of correlation between de- replicate. There are a number of ways in which velopmental input and developmental outcome.3 evolving lineages can ensure the inheritance of But the associations of the vernacular concept of these factors. These range from highly active information are often present when the statistical methods, such as habitat and host imprinting, to concept is applied. In the vernacular sense, infor- entirely passive methods such as the biogeo- mation is “intentional”: it is the meaning con- graphic association between a lineage and a ferred on a symbol or a message by its creator. region. It is clear that evolving lineages can do This meaning can be misinterpreted by the recip- better or worse than one another because of ient of a message, but the meaning of the message fitness differences caused by these developmental is not thereby changed. Hence thinking of devel- factors. As we have argued elsewhere (Gri≤ths opmental causation as the expression of informa- and Gray 1997), the idea that developmental sys- tion carries the association that the significance tems can be reduced to a collection of indepen- of a cause is independent of the context in which dent replicators is either inadequate or has to it acts. The idea of dual (or multiple) inheritance recognize relationships to persistent features of systems runs a similar risk of pushing context the environment as an addition to the cast of dependency into the background. Consider, for replicators. The “selfish standing-biogeographic- example, the methylation inheritance system association-with-a-low-rainfall-region” is un- (Jablonka and Szathmáry 1995). The devel- likely to appeal. opmental significance of a methylation pattern We would also argue that a developmental depends on the gene whose transcription it modi- systems conceptualization is more heuristically fies, and on much else. It is, of course, possible to valuable than a multiple channel or multiple rep- identify predictive relationships between patterns licator model. Holding most of the developmen- of methylation and developmental outcomes. tal system constant in order to tease out the roles However, the idea that these developmental out- of a single factor is a valuable technique, and comes are transmitted down the methylation in- much research in the developmental systems tra- heritance channel obscures the way in which the 198 Paul E. Gri≤ths and Russell D. Gray

relationship between methylation pattern and aphid hosts to utilize what would otherwise be outcome depend on what is happening in numer- nutritionally unsuitable host plants. Aphids that ous “other” channels. Eva Jablonka recognizes have been treated with antibiotics to eliminate this di≤culty when she says that the different the bacteria are stunted in growth, reproductively inheritance systems cannot be treated as au- sterile, and die prematurely. A lineage that in- tonomous. However, the very idea of separate herits bacteria is clearly at an advantage over one systems suggests autonomy and it would be de- that does not. Once again there is variation (line- sirable to find a formulation that avoids this. ages with either different Buchnera bacteria or without Buchnera), these differences confer differences in fitness, and they are heritable. All biolo- Natural Selection gists would recognize the first case as an example of natural selection in action, but they would Armed with a thoroughly epigenetic view of de- probably balk at categorizing the aphid/bacteria velopment and an expanded view of inheritance, system in the same way. Yet why should these let us now turn to the concept of natural selec- cases be treated differently when both meet the tion. In principle, there seems no reason why this three criteria for natural selection? concept should not be decoupled from gene- An obvious response would be to claim that if centered theories of development and evolution. there is selection in this case, then it can be re- After all, Darwin developed the theory of natural duced to selection of genetic differences. Aphids selection prior to the mechanisms of inheritance with genes for passing on their endosymbionts being discovered. The three requirements for have evolved by outcompeting aphids with genes natural selection (variation, fitness differences, for not passing on endosymbionts. However, it is heritability) are agnostic about the details of in- possible to have differential reproduction of the heritance. In Daniel Lehrman’s classic phrase, aphid/bacteria system without any genetic differ- “Nature selects for outcomes” (Lehrman 1970: ence between the two lineages involved. An aphid 28). The developmental routes by which fitness lineage that loses its bacteria will produce off- differences are produced do not matter as long as spring without bacteria. These offspring remains they reliably reoccur. genetically identical to the lineages with which Consider the following two cases: Newcomb they compete, but have a lower expected repro- et al. (1997) found that a single nucleotide change ductive output. A naturally occurring instance in blowflies can change the amino acid at an ac- of this sort of selectively relevant non-genetic tive site of an enzyme (carboxylesterase). This variation is found in the North American fire change produced a qualitatively different enzyme ant Solenopsis invicta (Keller and Ross 1993). (organophosphorous hydrolase), which con- Colonies containing large, monogynous queens ferred resistance against certain insecticides. This and colonies containing small, polygynous case fulfills the three requirements for natural se- queens were shown to have no significant genetic lection. There are phenotypic differences in in- differences. Differences between queens are insecticide resistance, these differences are likely to duced by the type of colony in which they have produce differences in fitness, and these differ- been raised, as shown by cross-fostering exper- ences are heritable. Moran and Baumann (1994) iments. Exposure of eggs from either type of col- discuss a similar, fascinating example of evolu- ony to the pheremonal “culture” of a polygynous tion in action. Certain aphid species reliably pass colony produces small queens who found poly- on their endosymbiotic Buchnera bacteria from gynous colonies, leading to more small queens, the maternal symbiont mass to either the eggs or and so forth. Exposure of eggs from either type of developing embryo. The bacteria enable their Darwinism and Developmental Systems 199

colony to the pheremonal “culture” of a monog- ples of extragenetic inheritance discussed in this ynous colony produces large queens who found book do not fit this dichotomy (e.g., chromatin monogynous colonies, leading to more large marking systems, chemical traces from the ma- queens, and so forth. What appears to happen ternal diet passed on via fetal olfactory condi- here is that a “mutation” in a nongenetic element tioning or in maternal milk, the inheritance of gut of the developmental matrix can induce a new symbionts, and the inheritance of fire ant colony self-replicating variant of the system which may type). A division of these cases into those that are differ in fitness from the original. “sort of genetic” and those that are “sort of cul- The moral that proponents of developmen- tural” will be largely arbitrary. A somewhat more tal systems theory draw from the comparison of insightful response to the challenge of expanded these cases is that the power of selective explana- inheritance has been outlined by Kim Sterelny tions need not be limited to genetic changes. The (chapter 23). Following Richard Dawkins’s well- range of phenomena that can be given selec- worn track up Mt. Improbable (Dawkins 1996), tive explanation should be expanded to include Sterelny argues that only cumulative selection differences dependent upon chromatin marking can produce complex adaptive structures. systems (chapter 9), prions (Lindquist 1997; Sterelny then outlines some of the requirements Lansbury 1997), dietary cues in maternal milk, an inheritance system would need to make cu- cultural traditions and ecological inheritance mulative selection possible (e.g., a large range of (Gray 1992; chapter 10). Selection for differences possible phenotypes, longevity, high fidelity repli- in one of these heritable developmental resources cation, vertical transmission, and developmen- is likely to have consequences for other aspects of tal modularity). According to Sterelny, extended the developmental system. Whitehead (1998) has forms of inheritance like cultural traditions and recently argued that cultural selection has led to ecological inheritance are unlikely to satisfy these genetic changes in this way. He observed that in requirements; they are inheritance systems, but species of whales with matrilineal social systems not highly evolvable inheritance systems. In con- mitochondrial DNA diversity is ten times lower trast, however, he argues that symbiont transmis- than in those with nonmatrilineal social systems. sion might score quite highly on his criteria for He suggested that differences in maternally trans- evolvability. mitted cultural traits, such as vocalizations and While there is much that we agree with in feeding methods, have conferred a su≤cient ad- Sterelny’s analysis, there are two important con- vantage to lead to the spread of some maternal ceptual differences between our views. First, lineages, and thus their mtDNA. The mtDNA Sterelny adopts a particularly Dawkinsian view that exists today remains because it hitchhiked of what is important in evolution (i.e., cumulative along with the cultural traits that were selected selection leading to adaptation). While we do for. not deny that cumulative selection is an impor- At this point orthodox gene-centered biolo- tant part of evolution, there is a lot more to be ex- gists might concede that natural selection can plained than just this. Our Darwinian mission is, indeed be generalized to cover cases of expanded after all, to explain the diversity of life—the myr- inheritance. Having made this concession they iad fascinating changes in shape, size, physiology, might then attempt to minimize its significance. behavior, and ecology. Extended forms of inher- Genic selectionists, for example, might be itance can play important roles in evolution with- tempted to reduce cases of expanded inheritance out providing the heritable basis for cumulative to a dual inheritance model—genes and their cul- selection. For example, Maynard Smith and his tural equivalent (memes). But many of the exam- collaborator Eörs Szathmáry (1995) argue that 200 Paul E. Gri≤ths and Russell D. Gray

epigenetic inheritance has played a crucial role possible phenotypes, longevity, fidelity, vertical in some of the major evolutionary transitions. transmission, and developmental modularity). Sterelny is sympathetic to this view. He notes that We will now address each of these criteria in turn. symbiosis might be of considerable significance in the generation of evolutionary novelty. After Natural Selection and the Range of Possible all, the eukaryotic cell is probably an example Phenotypes of frozen symbiosis. Not only might expanded forms of inheritance play an important role in the Maynard Smith and Szathmáry have introduced generation of evolutionary novelty they could a distinction between limited and unlimited sys- also significantly alter the dynamics of evolution- tems of heredity (Maynard Smith and Szathmáry ary change. Pal and Miklos (1999) recently mod- 1995; Szathmáry and Maynard Smith 1997). eled the impact of epigenetic inheritance, such They argue that most nongenetic inheritance sys- as chromatin marking, on the evolutionary tra- tems can only mutate between a limited number jectory of a population through an adaptive land- of states. In contrast, they note that the genome scape. Their results suggest that this expanded and language both have recursive, hierarchical inheritance can facilitate transitions from subop- structures, and hence an indefinite number of timal to higher peaks, thus creating more effective possible heritable states. This unlimited range of evolutionary dynamics than would be possible combinatorial possibilities enables microevolu- under strict genes-only conceptions. Expanded tionary change and cumulative selection to take forms of inheritance may also be the cause of place. These points are all perfectly legitimate, reproductive isolation and hence of speciation. but from a developmental systems perspective the Parasitic Wolbachia bacteria infect up to 20 per- significance of unlimited inheritance should not cent of insect species, and it has been suggested be oversold for three reasons. that this cytoplasmically inherited microorgan- First, the unlimited nature of an inheritance ism may be a major cause of speciation in insects. system is a property of the developmental system In many species, individuals infected with one as a whole, not only of the resource in which we strain of the bacteria cannot successfully fertilize find the recursive structure. The vast coding po- individuals infected with another strain. Some tential of genes, language and perhaps phero- clearly separate species also become interfertile mones is created by the way in which combina- when “cured” of their Wolbachia infections tions of these factors “mean something” to the (Vines 1999). In all these ways, epigenetic inheri- rest of the developmental system. Asking if a tance can be a major factor in evolution. system is limited or unlimited holds the current Our second point of departure from Sterelny’s developmental system fixed, and asks what can position is that we do not divide expanded inher- be achieved by ringing the changes on one of the itance into separate inheritance systems or multi- existing developmental resources. But the lesson ple replicators. As we argued above, these factors of the evolutionary transitions—the introduction are physically and functionally linked. The effects of whole new levels of biological order, such as of differences in these factors are likely to be con- multicellularity—is that evolution can change de- text sensitive, and should be seen as part of a sys- velopmental systems so as to massively expand tem of causes rather than as separate information the possible significance of existing developmen- channels or replicators. Putting these conceptual tal resources. A base-pair substitution in a multi- differences aside, there is still a great deal that can cellular organism has potentials that it lacked in be gained by examining Sterelny’s criteria for an a unicellular ancestor. If the substitution occurs evolvable inheritance system (i.e., the range of in a regulatory gene it could mean a new body Darwinism and Developmental Systems 201

plan. The role of systems of “limited heredity” in terns. Thus, in the words of Boyd and Richerson these evolutionary transitions is considerable, as (1996), culture is common but cultural evolution Maynard Smith and Szathmáry make clear. is rare. Second, from a selectionist viewpoint the com- Although this is certainly a possibility, evolu- binatorial richness of an inheritance system must tionary biologists know remarkably little about be measured in terms of the number of different the longevity of cultural traditions in animals and phenotypic effects, not just the number of com- the degree to which they have involved cumula- binations of components (chapter 23). If the rest tive changes. Research on these questions, while of the developmental system were such that exciting, is in its infancy. For example, Whiten et the indefinitely many base-pair combinations of al. (1999) have documented substantial cultural DNA collapsed into only a few developmental variation in tool use, grooming, and courtship outcomes, then for all its combinatorial structure behaviors between populations of wild chim- DNA would not be an unlimited heredity system. panzees (Pan troglodytes). However, there is little It not hard to imagine cellular machinery with evidence that these variations have persisted for this result because the existing genetic code is long periods of time, nor that they are a product substantially redundant in just this way. Several of cumulative cultural change. It is possible that codons produce the same amino acid. sensitive periods and social and ecological scaf- A third and final reason not to place too much folding may facilitate reliable cultural inheri- emphasis on the limited/unlimited distinction tance. One example is the reliability of human is that it treats genetic and extragenetic inheri- linguistic inheritance. If the controversial claim tance separately. From a developmental systems that human language relationships are congruent perspective these sources of heritable variation with our evolutionary history is correct, then lin- should viewed as acting together. Adding one guistic inheritance has tracked genetic divergence form of inheritance to another causes a multipli- for as long as 200,000 years (Cavalli-Sforza et al. cation of evolutionary possibilities, not just an 1988; Penny, Watson, and Steel 1993). In non- addition to them. The greater the range of pos- humans we know much less. However, a study sibilities the more scope there is for cumulative on cha≤nches on Atlantic islands by Lynch and selection and microevolution. Baker (1986) found substantial congruence between a tree based on morphology and a tree Natural Selection, Longevity, and Fidelity constructed from song syllables, indicating a common evolutionary history going back one Another strategy that could be used to minimize to two million years. Whitehead’s (1998) study the evolutionary significance of expanded forms on the possible effect of cultural selection on of inheritance is to argue that they lack su≤cient mtDNA diversity in whales suggests that cultural longevity and high fidelity replication to be the inheritance has exhibited considerable longevity targets of cumulative selection. Sterelny (chap- and fidelity in the species with matrilineal social ter 23) claims that this is likely to be the case systems. The results of his computer simulations with cultural inheritance. Following Tomasello, indicate that, assuming a 10 percent reproductive Kruger, and Ratner (1993) he argues that high advantage is culturally transmitted down mater- fidelity cultural inheritance requires genuine im- nal lineages, it would take more than two hun- itative learning, and this is rare outside humans. dred generations to produce the observed tenfold Other forms of social learning, like local enhance- reduction in mtDNA diversity. ment and emulation,4 are unlikely to result in While the cases both for and against cultural faithful reproduction of the same motor pat- inheritance are limited by a lack of evidence, the 202 Paul E. Gri≤ths and Russell D. Gray

same cannot be said for some other forms of ex- populations. Linguistic inheritance may operate tended inheritance. Perhaps the most impressive like this (Gray and Jordan 2000). The inheritance involves the aphid/bacteria symbiosis analyzed of species-speciﬁc strains of Wolbachia bacteria by Moran and Baumann (1994) and discussed in some species is also extremely unreliable for in- earlier. Phylogenetic trees of the bacteria and dividuals while being highly reliable for the their aphid hosts are perfectly congruent. This species as a whole. Ecological and cultural inher- suggests that speciation of the aphids has lead to itance could then play an important evolutionary speciation of the bacteria—they have cospeci- role by providing the basis for higher level trait ated. Molecular and fossil evidence suggests that groups (see Wilson 1997 and the discussion on this association is incredibly ancient—between lineages that follows). In fact, a group selective 160 and 280 million years. The Buchnera bacteria explanation has been offered for the individually are thus a vital and remarkably reliably inherited unreliable nature of transmission in Wolbachia developmental resource. (Vines 1999: 47). However, as Sterelny notes, the conditions that allow group selection are quite Vertical Transmission restrictive. An alternative possibility is that although changes in ecological and cultural in- The lack of strict vertical (parent to offspring) inheritance might not provide the heritable basis heritance is another potential problem for the for cumulative selection, they could play impor- ability of certain forms of extended inheritance tant evolutionary roles both in opening up new to lead to cumulative selection. Again Sterelny sets of adaptive possibilities and by facilitating (chapter 23) has discussed the problem with con- the dynamics of evolutionary change. It should siderable insight: also be emphasized that the problem of diffuse horizontal inheritance does not apply to all forms Transmission [of ecological and cultural differences] is of expanded inheritance. Sterelny notes that cyto- not vertical. Indeed, it is not even individual. It is dif- plasmic factors, symbiont systems, and ant nest fuse. Groups of trees engineer their soil structures or a ﬁre-prone understory; individual trees do not make types are all likely to be inherited in a highly ver- their microenvironments for themselves and their tical fashion. descendants. In most cases groups of animals make warrens, trackways, track-and-bowl systems, beaver Modularity and One Reason Why Selectionists lodges, termite mounds and other structures ultimately Cannot Ignore Development taken over by the next generation. If this is transmission at all, it is diffuse and development is holistic. (p. 344). At the end of his classic paper on adaptation, Lewontin (1978) notes that adaptive evolution It is likely that even quite small departures requires quasi-independence. By quasi-indepen- from 100 percent vertical transmission could un- dence he means that selection must be able to act dermine the evolutionary coherence of some on a trait without causing deleterious changes in forms of extended inheritance with the rest of other aspects of the organism. If all the features the developmental system. For example, in of an organism were so closely developmentally Whitehead’s (1998) simulation of the effects of integrated that quasi-independent variation did cultural selection on mtDNA diversity in whales, not exist, then “organisms as we know them if horizontal transmission was much greater than could not exist because adaptive evolution would 0.5 percent, then there was little reduction in ge- have been impossible” (Lewontin 1978: 169). netic diversity. One possibility is that ecological This means that we must add a caveat to and cultural inheritance, while diffuse and hori- Lehrman’s slogan “Nature selects for outcomes” zontal at the individual level, might be vertically (Lehrman 1970: 28). Development does matter. inherited in larger units such as families and local Darwinism and Developmental Systems 203

The reliable reoccurrence of an advantageous sion or cultural traditions will be any less mod- variant is not enough. The developmental route ular in their developmental consequences than which produces the variation must be quasi- genetic factors, and thus extended forms of in- independent if it is to be the basis for cumulative heritance cannot be brushed off as lacking an selection. A common finding in artificial selection essential requirement for cumulative selection. experiments is that although many traits initially Third, because extended inheritance must be respond rapidly to selection, the response will taken seriously rather than brushed aside, the often slow and reach a plateau (Lerner 1970). range of phenomena that can and should be One interpretation of these results is that devel- given selectionist explanations is considerably in- opmental links mean that many traits can only be creased. This is something that we as Darwinian changed to a certain degree without having dele- biologists should be excited about. DST expands terious effects on other aspects of the phenotype. the scope of Darwinian explanation, and that Further directional change would thus require is exactly the general conclusion of this section some kind of developmental reorganization. on natural selection. From a DST perspective There has been considerable recent interest in there is lots more work to be done—there are ex- the extent to which the organization of devel- citing new questions that we have, at best, only opment really is modular. For example, Halder, partial answers to. As students of cumulative Callerts, and Gehring (1995) demonstrated the selection, we really need to know the extent to modularity of eye formation in Drosophila by which extended forms of inheritance fulfill the re- successfully inducing eyes on the antennae, quirements of longevity, fidelity, and vertical in- wings, and legs of Drosophila. The targeted mis- heritance, and we really need to investigate the expression of the “eyeless” gene produced struc- modularity and evolvability of developmental tures in these unusual locations that contained a systems. cornea, bristles, and photoreceptors and were responsive to light. Wagner and Altenberg (1996) suggest that directional selection might act on de- Adaptation and Niche Construction velopmental systems to reduce pleiotropic effects between characters with different functions, In three seminal papers Richard Lewontin criti- thereby enhancing the modularity and evolvabil- cized the metaphors that have traditionally been ity of these developmental systems. They spec- used to represent the process of adaptation by ulate that there should be evolutionary trends natural selection (Lewontin 1982; Lewontin towards increased modularity. Brandon (1999) 1983a; Lewontin 1983b). The metaphorical con- goes so far as to suggest that developmental mod- ception that Lewontin criticized is the so-called ules are the units of selection. The study of devel- lock and key model of adaptation. Adaptations opmental modularity is still in its infancy and the are solutions (keys) to the problems posed by the extent of modularity far from resolved. However, environment (locks). Organisms are said to be from a DST perspective the exact extent of mod- adapted to their ways of life because they were ularity is not a pivotal concern. DST is agnostic made to fit those ways of life. In place of the on this question (Sterelny in press). Instead, we traditional metaphor of adaptation as “fit” would emphasize three general implications that Lewontin suggested a metaphor of construction. arise from the modularity issue. First, analyzing Organisms and their ecological niches are co- development itself is the key to understanding the constructing and codefining. Organisms both ability of selection of act in a cumulative manner. physically shape their environments and deter- Second, there is no reason to think that extended mine which factors in the external environment forms of inheritance such as symbiont transmis- are relevant to their evolution, thus assembling 204 Paul E. Gri≤ths and Russell D. Gray

such factors into what biologists describe as a sense becomes still more tenuous, however, once niche. Organisms are adapted to their ways of it is recognized that occupied landscapes owe life because organisms and their way of life were much of their physical structure to the activities made for (and by) each other. Lewontin also of the organisms that occupy them. In this tenu- revised the popular metaphor of a “fitness ous sense there were niches for species requiring landscape.” In this image, populations occupy a high rainfall in the Amazon basin before the rugged landscape with many fitness peaks and biota which make it a high-rainfall region had evolve by always trying to walk uphill. But evolved. So a region of space and time contains because organisms construct their niches, the not only the niches that can be defined using landscape is actually much like the surface of a its existing features, but all those that could be trampoline. As organisms climb the hills they defined using the features induced by the action change the shape of the landscape. Lewontin’s of all the species that could evolve so as to make metaphor of construction is not merely a new a niche in that region! way to describe the same evolutionary process. It Sterelny and Gri≤ths have argued that the is the public face of a substantially revised model concept of a “vacant niche” makes sense only in of the actual process of natural selection, re- an ecosystem that has already been structured defining the causal relationships which ecology by a collection of organisms which are part of it and evolutionary biology must seek to model. (Sterelny and Gri≤ths 1999). It may then be pos- Lewontin’s ideas challenge one of the central sible to determine that an organism of a specified elements of contemporary neo-Darwinism, the type not present in that ecosystem could maintain idea that the source of the selective pressures that itself were it introduced. The move from this idea explain adaptive evolution can be sought in a rel- to the idea of an unoccupied landscape with a de- atively independent science of ecology.5 Darwin terminate niche structure is an illicit idealization. himself was well aware of the reciprocal influence It idealizes away from precisely those factors that of organism and environment, and conducted create the possibility of identifying vacant niches. pioneering studies on the role of earthworms in Similarly, the idea that knowing the “shape” of the formation of soils (Darwin 1881). Despite the vacant niches allows us to predict how or- this, many presentations of Darwinism treat the ganism will evolve to fit them ignores the fact environment as a source of fixed problems which that different organisms will construct different every organism must solve or die. Environments niches. Some eucalypt species can establish and for which there is no fossil or other direct evi- sustain “islands” of dry sclerophyll forest in rain- dence can be reconstructed by “reverse engineer- forest by facilitating bushfires. Once this process ing” the organisms that those environments is understood it is possible to identify a vacant shaped to fit themselves. Some even suggest that niche for these species in many other landscapes, evolutionary research can proceed by first identi- despite the fact that they would not even germi- fying the niches in an environment and then pre- nate if simply planted there (Mount 1964). We dicting how organisms will evolve to fill them— would argue that the idea of a vacant niche for so-called adaptive thinking.6 But as Lewontin has E. delegatensis in all non-Australian rainforests pointed out, there are indefinitely many overlap- is simply perverse in comparison to the idea of ping niches in an unoccupied physical landscape. niche construction. Until the organisms that occupy the niches are The most detailed attempt to develop the new specified, the concept of the niche is completely metaphor of construction is that of F. J. Odling- unhelpful. Of course, there is a sense in which Smee and his collaborators (for an overview, see every possible niche that an organism could con- chapter 10 of this volume). The current promi- struct in an area of space and time “exists.” This nence of the term niche construction is due to this Darwinism and Developmental Systems 205

Table 16.1 Three pictures of the dynamical equations for evolution Traditional Neo-Darwinism Lewontin’s Constructionism Odling-Smee’s General Coevolution dO!dt;f (O, E) dO!dt;f (O, E) dOpop!dt;f (Opop, Epop) dE!dt;g(E) dE!dt;g(O, E) dEpop!dt;g(Opop, Epop)

d (Opop, Epop)!dt;h(Opop, E)

E, Environment; O, organism; Epop, organism-referent environment of a population; Opop, population of organisms. These variables are related by functions f, g, h. See text for explanation. group. The first two columns in table 16.1 give ual organism and its developmental environment the traditional neo-Darwinian model of adapta- as the organism’s life cycle unfolds. By introduction as “fit” and the model of adaptation as con- ing these indices, Odling-Smee is making explicit struction as these two models are described by what was already implicit in the explanation of Lewontin (1982, 1983a). In the traditional pic- Lewontin’s equations given in the previous para- ture, change in organisms over time is a function graph—the term O in those equations refers to of the state of the organism and its environment populations of organisms, not to some individual at each the previous instant. The environment organism. Earlier versions of DST (e.g., Oyama acts on the existing state of organisms by select- 1985; Gray 1992) and some of Lewontin’s writing from the pool of variation those individuals ings are sympathetic to Odling-Smee’s idea that best fitted to the environment. The environment there is a significant parallelism between the way itself changes over time too, but as the bottom populations of organisms and their environments equation shows, these changes are not a function reciprocally influence one another and way in of what organisms are doing at each previous in- which individual organisms and their develop- stant. In Lewontin’s alternative picture, shown in mental environments do so. But this is not the the center column of table 16.1, organisms and place to give this idea the attention it deserves. their environments play reciprocal roles in each The second way in which Odling-Smee’s treat- other’s change. Change in the environment over ment differs from Lewontin’s is that he is con- time is a function of the state at each previous in- cerned not to represent the organism-environ- stant of both the environment and the organisms ment system as a closed system, as the equations evolving in that environment. in the center column would seem to imply. The right-hand column of table 16.1 shows Although the eucalypt-bushfire relationship, for Odling-Smee’s model of evolution as the co- example, is one of mutual construction, the construction of organism and environment. changes in this system over time are externally Odling-Smee’s general coevolutionary model dif- driven by the progressive drying of the Australian fers from Lewontin’s in two ways (Odling-Smee continental climate. Organisms feel the impact of 1988). First, Odling-Smee hoped to generate a changes in the environment in the traditional common framework in which to represent both sense of that term—their total biotic and abiotic development and evolution. This explains why surroundings—but they experience these impacts the terms Epop and Opop occur in the equations in via the environment as it appears in relation table 16.1. Evolution is a process in which pop- to them, and thus different lineages experience ulations and their environments co-construct one “the same changes” quite differently. Odling-Smee another over time. If the terms were Ei and Oi, tries to respect this situation by assigning sepa- then in Odling-Smee’s notation the equations rate roles to the environment of a particular line- would describe the co-construction of an individ- age of organisms and what he calls the “universal 206 Paul E. Gri≤ths and Russell D. Gray

physical environment.” The former, organism- a habitat association). It is the organism-referent referent description of the environment is the description of the environment—the ecological source of evolutionary pressures on that organ- environment—which captures the aspects of the ism, and the organism is the source of niche- environment that are relevant to the organism constructing forces on that environment. The and defines what counts as a “change.” latter, the universal physical environment, is a The developmental systems model of evolution source of exogenous change in the organism’s (Gray 1992, 1997; Gri≤ths and Gray 1994a, 1997) environment. can be clarified and improved by the insights of Robert Brandon’s theory of the role of the en- Odling-Smee and his collaborators. In particular, vironment in adaptation is a useful complement the insight that exogenous factors can affect the to Odling-Smee’s ideas. Brandon distinguishes availability of developmental resources has not three different senses of “environment” (Brandon been su≤ciently stressed in previous presenta- 1990, 1992). His “external environment” corre- tions. There remains, however, one major differ- sponds to Odling-Smee’s universal, physical envi- ence between DST and work on niche construc- ronment. All organisms in a particular region of tion up to and including the present time. Niche space and time share an external environment. construction is still a fundamentally dichotomous The “ecological environment” must be described account of evolution (and, indeed, of devel- with reference to a particular evolving lineage. It opment). There are two systems of heredity— consists of those environmental parameters genetic inheritance and environmental inheri- whose value affects the reproductive output of tance. There are, correspondingly, two causal members of the lineage. Finally, the “selective en- processes in evolution—natural selection of the vironment” is that part of the ecological environ- organism by the niche and construction of the ment which differentially affects the reproductive niche by the organism. The niche-construction output of variant forms in the evolving lineage. It model could be modified to take account of re- is this last which contains the sources of adaptive cent work on narrow epigenetic inheritance, with evolutionary pressures on the lineage. Brandon a category like “intracellular inheritance” taking has used these ideas in the context of his own ex- the place of genetic inheritance. This, however, ploration of organism-environment coevolution would seem merely to substitute one rigid bound- (Brandon and Antonovics 1996). Organisms ary for another. A central theme of the DST modify the selective and ecological environments research tradition has been that distinctions be- in numerous ways. All these can potentially tween classes of developmental resource should influence their evolution. Only some of these be fluid and justified by particular research inter- modifications of the selective and ecological envi- ests, rather than built into the basic framework of ronments also constitute modifications of the ex- biological thought. Fundamentally, the unit of ternal environment, but whether they do or not is both development and evolution is the develop- unimportant when determining their role in the mental system, the entire matrix of interactants future evolution of the organism-environment involved in a life cycle. The developmental sys- system. This point, which was touched on earlier, tem is not two things, but one, albeit one that is one reason why the simple notion of (external) it can be divided up in many ways for different environment is inadequate. Many changes in the theoretical purposes. Hence we would inter- external environment do not constitute change pret niche-construction models “tactically,” as a from the point of view of the organism and, con- method for rendering tractable some aspects of versely, an organism can transform its environ- evolution. We would not interpret them “strate- ment without actually changing the universal gically” as a fundamental representation of the physical environment (for example, by changing nature of the evolutionary process. This response Darwinism and Developmental Systems 207

is closely related to the comments about “multi- there are different types of organism in a single ple channel” models of inheritance given above. evolving population, each reproducing its own The DST model of evolution can be repre- differences (polymorphism) or because there are sented in such a way as to make it directly com- variations in the developmental matrix from one parable with the models in table 16.1. We can generation to the next (facultative development). aptly represent the developmental system with For example, there are tall and short human fam- the symbol Œ. We retain Odling-Smee’s insight ilies and heights also vary from one generation to that evolutionary change in organism-environ- another due to nutrition. These features are han- ment systems is often exogenously driven by dled in the same way as in characterizations of using E to represent the universal physical (exter- “the” phenotype of an evolving lineage (Gri≤ths nal) environment. We end up with the equation: and Gray 1997). The resultant description of the idealized developmental system of a particular ! ; dŒpop dt f (Œpop, E) lineage at some stage in its evolution is highly Evolution is change in the nature of popula- self-contained. Because the focus is on how the tions of developmental systems. This change is complete life cycle is achieved, everything needed driven both endogenously, by the modification for that life cycle is assumed to be present. So by each generation of developmental systems of everything that impinges on the process is an ele- the resources inherited by future generations, and ment of the system itself. It is this that creates the exogenously, by modifications of these resources impression that all change in the system must be by factors outside the developmental system.7 endogenously driven and creates the apparent puzzle about the source of selection pressures. Fitness and Adaptation The puzzle is only apparent because to think about evolution we need to switch from describ- This representation of developmental systems ing the developmental system characteristic of an evolution allows us to answer a persistent objec- evolving lineage at a time to describing an evolv- tion to DST. Since we claim that there is no dis- ing population of individual developmental sys- tinction between organism and environment, tems. We need to look at the causes of variation, where do evolutionary pressures on the develop- as well as how the characteristics of the lineage mental system come from?8 What causes adapta- are reliably reconstructed. Hence we need to look tion? To give a clear answer we must go back to at the causes of idiosyncratic development in par- the definition of the developmental system given ticular individuals. These causes lie “outside” the in Gri≤ths and Gray (1994a). The developmental description we have constructed of the typical de- system of an individual organism contains all the velopmental system of the lineage. A population unique events that are responsible for individual of individual developmental systems will exhibit differences, deformities, and so forth. Just as a variation and differential reproduction for a traditional model of evolution abstracts away number of reasons. Parental life cycles may fail to from the unique features of individual pheno- generate the full system of resources required to types, developmental systems theory must ab- reconstruct the life cycle. Resources generated by stract away from these features in order to tackle the activities of an entire population (such as evolutionary questions. In evolutionary terms the bushfires in eucalypt forest) may also be scarce, developmental system contains all those features or patchily distributed, so that some individuals which reliably recur in each generation and which lack an important element of their developmental help to reconstruct the normal life cycle of the system. Finally, persistent resources—those de- evolving lineage. Of course, many species have velopmental factors whose abundance is inde- more than one normal life cycle, either because pendent of the activities of the lineage—may be 208 Paul E. Gri≤ths and Russell D. Gray

scarce or patchy and so some individuals may be discarded shells of other species. A dearth of unable to reestablish the relationship to these re- shells would be an exogenous cause of selective sources that is part of their life cycle. The external pressure on the lineage. Variants with a beneficial environment (E) can impinge on developmental set of behaviors or a beneficial habitat associ- systems by any of these routes. But this does not ation that allowed them to continue to reliably mean that we can go back to thinking of evolu- reestablish their relationships to shells would be tion as a response to the demands of the external favored by selection. Shells will typically be an in- environment. The effect of changes in the exter- dependently persistent resource, and the case in nal environment on the evolution of a lineage can which an independently persistent developmental be understood only when those changes are de- resource acts as a limiting resource has obvious scribed in terms of how they change the organ- resonance with traditional ideas of selection of ism-referent environment (Epop). “Changes” in the organism by an independent environment. parameters of the external environment that are But, to fictionalize the example slightly, suppose developmentally equivalent are not changes from the crab life cycle includes disturbing the sand the point of view of the evolving system. People in such a way as to expose a greater supply of in different regions of Britain experience substan- discarded shells. That would make shells a pop- tially different quantities of dissolved limestone ulation-generated resource, but they might still in their drinking water, but this is generally of no act as a limiting resource. Or suppose a lineage ecological significance. Conversely, apparently evolves behaviors that allow crabs to bequeath trivial changes may seem momentous when de- shells to their offspring when they themselves scribed in terms of a particular developmental seek a larger home. Shells would then be paren- system. Far smaller changes in the concentration tally generated, but exogenous change in the of lead from one region to another would have availability of shells might still leave some off- momentous consequences. This is, of course, the spring without them, just as a shortage of a trace point already made by Lewontin, Odling-Smee’s, element in the parental diet may lead to a birth and Brandon’s work and by the concepts of eco- defect in a viviparous species. logical environment and organism-referent de- One idea that really is missing from this picture scription of the environment. is the notion that the external (universal physical) environment poses definite problems that line- So far we have concentrated on how failures ages must seek to solve. Instead, the lineage helps of development can lead to evolutionarily sig- to define what the problems are. A dearth of nificant variation. But positive innovations are shells is a feature of the ecological environment of possible as well. An individual difference in the a hermit crab and a problem for the hermit crab, system of developmental resources may allow but it is completely invisible to a blue-swimmer some individuals to cope better when both are de- crab. The number of discarded shells per square prived of some developmental resource because meter is a feature of the external environment of of exogenous change. Alternatively, an individual both species, but it is only a feature of the eco- difference may simply alter the life cycle in such logical environment of one of them. So it is true a way that it gives rise to a greater number of that the developmental systems treatment of evo- descendants. The source of novelty can be a mu- lution does not incorporate Darwin’s original, in- tation in any of the developmental resources— tuitive idea of fitness as a measure of the match parentally generated, population generated, or between an organism and an independent en- independently persistent. To make this discussion vironment (e.g., Darwin 1859/1964: 472). But more concrete, imagine a typical population of this is a feature which the developmental sys- hermit crabs. A key component of the devel- tems treatment shares with conventional neo- opmental system in this lineage is a succession of Darwinism. Adaptation is no longer defined in- Darwinism and Developmental Systems 209

tuitively, as the sort of organism/environment causal explanation of evolutionary success. relationship that a natural theologian would see Fitness in general, however, does not correspond as a sign of God’s beneficent plan. Darwin set out to any single physical property (Rosenberg 1978). to explain the fact that the biological world is full The only general account of fitness describes its of adaptations in this sense, but as so often hap- role as a parameter in population dynamic equa- pens in science, the phenomenon to be explained tions. It is clear that this orthodox account of got redefined in the process of explaining it. In fitness applies equally well to the developmental modern usage, an adaptation is whatever results systems theory. There is no puzzle about how de- from natural selection, even when what results velopmental systems that incorporate the whole is intuitively perverse and ine≤cient. Gould and range of resources that reconstruct the life cycle Lewontin once described a mutation in a bird could come to vary in their success in recon- which doubles clutch size when the population is structing themselves and be selected on that at the limits of the carrying capacity of the envi- basis. ronment. The mutation sweeps to fixation with a consequent doubling of chick mortality. They described this as a case of selection operating with- Individuals, Lineages, and the Units of Evolution out producing adaptation (Gould and Lewontin 1978). They were using “adaptation” in its origi- A coherent theory of evolution requires an ac- nal sense and consequently swimming against the curate conception of its fundamental units. Ac- neo-Darwinian tide. To most of their critics it cording to DST an evolutionary individual is one seemed obvious that this trait was a paradigmatic cycle of an complete developmental process—a adaptation. The first example of this shift in the life cycle. We have shown that natural selection concept of adaptation is, of course, Darwin’s can act on populations of developmental systems own idea of sexual selection. In at least some of and give rise to adaptation, but in doing so we his moods Darwin still saw this as a separate have assumed that developmental systems are the force that could act in opposition to natural sort of things that can be counted, that they have selection. That suggests that he was limiting the clear boundaries and that they do not overlap so idea of natural selection to processes that pro- much that they cannot be distinguished from one duce adaptations in the sense of William Paley’s another. We now turn to justifying this assump- “contrivances”—features that suggest the world tion. Developmental systems include much that was designed by a beneficent creator. But it has is outside the skin of the traditional phenotype. been a long time since sexual selection seemed This raises the question of where one develop- anything other than one kind of natural selection. mental system and one life cycle ends and the We hope it is now clear how DST can explain next begins There is an enormous amount of adaptation, in the modern sense of that term. cyclical structure in most biological lineages. As Change over time in the developmental system well as the life cycles associated with traditional of a lineage is driven by the differing capacity physiological individuals there are “repeated as- of variant developmental systems to reconstruct semblies” (Caporael 1995) within a single indi- themselves, or, in a word, differential fitness. vidual, such as cells or morphological parts like What is fitness? In contemporary evolutionary the leaves of a tree. There are also repeated as- theory fitness is a measure of the capacity of a semblies of whole individual organisms, such as unit of evolution to reproduce itself (Mills and the characteristic mother/child dyad or the Beatty 1979). Fitness differences are caused by ephemeral dyad formed of a buyer and seller in a physical and behavioral differences between the market. It has recently been suggested that re- individuals in the population. So fitness can be peated assemblies of human individuals like these translated on a case-by-case basis into a detailed can themselves be units of evolution (Wilson and 210 Paul E. Gri≤ths and Russell D. Gray

Sober 1994). In previous publications we have mental resources that go to make up the develop- tried to identify what makes a repeated assem- mental system of an organism. Nevertheless, uni- bly a developmental system in its own right, as cellular eukaryotes are regarded as individuals, opposed to a part of such a system or an aggre- not collections of individuals. In contrast, the gate of several different systems (see especially ants and acacia trees of the ant/acacia symbiosis Gri≤ths and Gray 1997). Our focus has been on are normally regarded as two separate evolution- identifying developmental systems that could be ary lineages. We suggested that the distinction evolutionary individuals. We have also been con- between one and many turns on whether one ele- cerned to examine the extent to which DST can ment of the symbiosis can give rise to new cycles support a hierarchical model of natural selection. of itself that are not coupled to the other mem- On the one hand, there is no reason why natural bers of the symbiosis in the characteristic way. selection should not operate at different levels of The meaning of “cannot” here is that the free- biological organization (Brandon 1988; Sober living form is very distant in the space of biologi- and Lewontin 1982). On the other hand, not cal possibility. There are strong barriers to the every repeated assembly is the focus of a selection components of the eukaryotic cell evolving back process. So we need criteria to identify which de- to free-living forms. The barriers for the ant and velopmental systems count as evolutionary indi- the acacia tree are weaker. We defended the viduals. While we see some merit in our previous vagueness of this answer by arguing that there is suggestions, we have learnt a great deal from the no clear line on the continuum between strong work of David Sloan Wilson and Elliott Sober on symbioses, facultatively colonial organisms like trait-group selection and the concept of a super- slime-molds and obligate colonial organisms organism, and also from Kim Sterelny’s work on like sponges and metazoans at which individ- higher level selection (Sober and Wilson 1994, uality springs into existence. We suggested that 1998; Sterelny 1996; Wilson 1997; Wilson and Maynard-Smith and Szathmáry’s concept of Sober 1994). In this section we revise our previ- “contingent irreversibility” of an evolutionary ous account of the individual in the light of this development in a lineage came close to the con- work. cept of biological (im)possibility we required In earlier work we suggested that an evolu- (Maynard Smith and Szathmáry 1995). Our tionary individual was distinguished from a col- proposal was intended to assist in deﬁning an lection of such individuals by the strength of the evolutionary lineage, as well as an evolutionary evolutionary association between its compo- individual. If two life cycles become coupled in nents: “We argue that the eukaryotic cell should a way that is contingently irreversible, then the be seen as a single life-cycle because its con- evolutionary lineages of which they are represen- stituents are obligate symbionts and there are tatives have also merged. strong barriers to their evolving back to free- We now see these ideas as strongly convergent living forms. Strongly obligate symbioses like this with Wilson and Sober’s idea of “shared evolu- one should be regarded as a single evolutionary tionary fate” (Wilson and Sober 1994). In our lineage” (Gri≤ths and Gray 1997; 478). The 1997 paper we wrote: “Two lineages whose evo- inspiration for this idea was the now almost uni- lutionary fates were previously separable (though versal acceptance that symbiotic lineages some- interacting) are now inseparably bound together” times merge into a single lineage, as in the origins (Gri≤ths and Gray 1997: 478). However, we had of eukaryotic cells (Margulis 1970). The descen- not really assimilated Wilson and Sober’s pro- dants of these symbiotic organisms, the cell nu- posal, one of whose advantages is the substance it cleus and the cell organelles, are replicated with gains from its relationship to Wilson’s concept of different periodicities and have different patterns trait-group selection (Wilson 1983). A trait group of inheritance, just like the various develop- is a set of organisms relative to which some adap- Darwinism and Developmental Systems 211

tation is, in economic terms, a public good. The individuals. What matters is that both the beavers that share a lodge form a trait group with cleaner-fish example and more standard cases of respect to dam-building adaptations because it is reciprocal altruism are examples of population not possible for one beaver to increase its fitness structured evolution and trait group selection. In by dam building without increasing the fitness of the case of the fish, it is the patchy distribution of its lodge mates. The water will be 1 cm deeper for cleaners on the reef that creates a correlation everyone. Trait group selection can occur when between not eating cleaners and getting cleaned. there is a correlation between having an adapta- In reciprocal altruism, the source of population tion and being part of the relevant trait group. structure is refusal to associate with those who If the beavers that share a dam are closely re- do not dispense benefits. There are many other lated, shared descent will produce such a cor- possibilities. Richard Dawkins has described a relation—a mechanism long recognized as kin bizarre fictional example. He imagines some peo- selection. Another long-accepted evolutionary ple with green beards who are disposed to help mechanism—reciprocal altruism—produces the anyone else with a green beard (Dawkins 1976: same effect. If individuals differentially associate 96–97). Given suitable cost-benefit ratios, indi- with those who return their favors, then there will viduals with this trait complex could outcompete be a correlation between giving and receiving those who neither dispense nor receive benefits. favors. The advantage of the trait group-selection Dawkins uses this example to demonstrate that it model is that it brings out the underlying unity of is not kinship that is important in kin-selection the various models for the evolution of coopera- models, but shared genes. He postulates that his tion and reveals the whole landscape, only a few green-bearded people are unrelated but share a peaks of which had previously poked up through pleiotropic gene that confers the trait complex the fog of the group-selection debate. upon them. However, the example would work The general phenomenon of which kin selec- just as well if different green-bearded individuals tion and reciprocal altruism are special cases is had many different genes that produced the same population structured evolution. Organisms do effects. The different genes could all impair the not interact with equal probability with every same normal biochemical pathway—many “ge- other member of the population. Population netic diseases” work this way. Dawkins’s exam- structure creates opportunities for trait group ple would also work if the trait complex were selection. The most obvious form of population produced by cultural inheritance—a subculture structure is geographic structure. Robert Triver’s keen on hair dye and cooperation. The particular famous paper on the evolution of reciprocal al- mechanism of inheritance is irrelevant. What the truism gives as an example a symbiotic relation- hypothetical green-beard example actually de- ship between small cleaner fish and the large monstrates is not that kin selection is a special predators whose parasites these fish remove case of gene selection, but that it is a special case (Trivers 1971). The larger fish do not eat the of trait-group selection. It is not necessary that small cleaners after they have finished cleaning, individuals who selectively benefit one another be as a simple game-theoretic model would suggest. related or that they share a gene. The essential Trivers argues that this is because the large fish feature needed to get models like this to work is find it hard to locate cleaners. They solve this population structure. Something has to create a problem by returning to the same cleaning site statistical association between dispensing benefits each time. By eating the cleaners they would re- and associating with other individuals who dis- duce their probability of being cleaned next time. pense benefits. Trivers treats this as a case of reciprocal altruism, Wilson and Sober argue that trait groups are but it is more common to define reciprocal al- units of evolution. More specifically, trait groups truism as requiring recognition and memory of are interactors sensu David Hull (Hull 1980). 212 Paul E. Gri≤ths and Russell D. Gray

That is to say, it makes sense to assign fitnesses roles are played by different physical objects— to trait groups and to track the evolution of traditionally one or more genes and one or more adaptations due to the differential reproduction phenotypic traits. Proponents of the replicator/ of their associated replicators. In one respect, interactor framework believe that identifying the however, a trait group is very different from the occupants of these two roles is essential to under- kinds of evolutionary individuals we tried to standing how natural selection is operating in a define in previous work. Each trait potentially domain. We think the replicator/interactor defines a different trait group. The beavers in a framework has fatal flaws as a representation dam are not a trait group with respect to foraging of evolution, embodying as it does the gene/ behaviors. In humans, each cooperative behavior environment dichotomy rejected by developmen- may define a different trait group. Paying for tal systems theory (Gri≤ths and Gray 1994a, your shout in the bar does not benefit the same 1994b). But be that as it may, the arguments in group as helping with the housework. This has favor of the framework have as much application caused Dawkins to deny that there is any room in the case of trait group selection as elsewhere. for the replicator/interactor distinction in the In this sense, Dawkins is wrong and there are in- selection processes described by Wilson and teractors in the selection processes Wilson and Sober (Dawkins 1994). According to Dawkins, Sober describe. “vehicles” (interactors) only exist in a special sub- The concept of an interactor has a double life, set of selection processes where the fate of a col- however. It also serves to generalize the notion of lection of replicators is strongly linked by their an organism. Sometimes, when we look for the joint investment in a complex, physiologically interactor in a potential domain for natural selec- interdependent collection of adaptations like or- tion we are looking, not for something that can ganism. In other selection processes, there are play the abstract role of interactor in selection only individual replicators and their impact on theory, but for something that corresponds to the their own replication prospects. This criticism organism. In earlier group selection models a seems to us both right and wrong, and to point to spatially cohesive local population of a species— the need to clearly distinguish the concept of an a deme—was assumed to correspond to the or- interactor as it figures in the replicator/interactor ganism. In that case the analogy with paradigm distinction from the interactor as a generalization cases of individual selection seemed clear enough. of the concept of the organism. If we concentrate The new trait-group selection models show that on the replicator/interactor framework for think- the interactor role can be fulfilled in such a piece- ing about natural selection then Dawkins’s meal way that nothing at this new, higher level of criticism is clearly misguided. Replicator and selection corresponds to the organism in para- interactor are two aspects of the process of selec- digm cases. In this sense, Dawkins was right that tion. In classic cases of gene selection, such as in some trait-group selection models there is no meiotic drive, the replicator and the interactor interactor. Sterelny and Gri≤ths have argued are the same physical object—a stretch of DNA. that trait-group selection sometimes gives rise Nevertheless, this object is playing two separate to functionally organized units with many roles. The fact that it plays the replicator role is adapted traits and sometimes does not. The old supposed to explain the stable replication of form term superorganism is a useful one to replace the from one generation to the next. The fact that the ambiguous interactor for these higher-level evolu- same stretch of DNA plays the interactor role tionary units. explains the selection of those replicators What makes a group into a superorganism? associated with the most e≤cient interactors. In At an intuitive level, an ant nest is a much more paradigm cases of phenotypic evolution, the two convincing superorganism that a lodge full of Darwinism and Developmental Systems 213

beavers. There are a number of features that seem some of the very different mechanisms that are to underlie this intuition, such as the functional used to bind the interests of cell lineages together differentiation of parts and the dependence of in plants and fungi (Buss 1987). In bee nests, the parts on the whole for their viability. Sterelny queen marks her eggs with a pheromone that in- and Gri≤ths argue that what is fundamental to a hibits workers from eating them. Eggs laid by superorganism is that very many traits of its or- workers are eaten by other workers, so the only ganisms are selected with respect to a single trait realistic way for workers to bring about the re- group. That trait group is the superorganism construction of their life cycle is via the larger, (Sterelny and Gri≤ths 1999: 172–177; see also colony life cycle (Ratnieks and Visscher 1989). Wilson 1997). The ants in a nest and the cells in The worker bee is reduced to a part of a larger a human body have a shared fate not just with cycle as effectively as the cell of a metazoan body respect to one part of their activities, but with is by segregation of the germline or an equivalent respect to all of them. A liver cell does not have mechanism. some adaptations with respect to the whole body The idea that evolutionary individuals are trait and other with respect to the liver alone. This is groups thus converges on our older idea that an because the only way the liver cell can reproduce individual is a life cycle whose components can- itself is via the success of the whole organism. not reconstruct themselves when decoupled from Similarly, the only way an ant can contribute to the larger cycle. Much trait-group selection does its own reproduction is via the success of the nest not give rise to new levels of individuality, but as a whole. The general phenomenon that we see only to transitory interactors. However, when in these cases is the existence of evolved features new features evolve, presumably by trait-group that suppress competition between the compo- selection, which link together the members of nent parts of the superorganism. The best known a group in a way that is contingently irreversible, of these is the segregation of the germline. In a new kind of individual emerges. At this point, most animals a particular cell lineage is physio- the population structure on which trait-group logically isolated relatively early in development selection depends is no longer just a cause of as the source of all future gametes. This means trait-group selection, but an effect of trait-group that other cell lines that do not contribute to suc- selection. An individual is a system in which the cessful functioning of the whole organism are, parts form a trait group with respect to most fu- like cancer cell lineages, doomed to extinction. ture evolutionary processes. This account of the Segregation of the germ line is an important evolution of individuality can actually explain mechanism. Organisms that do not have this fea- why the distinction between a colony of organ- ture, such as slime-molds, rapidly lose their mul- isms or a symbiotic association and an individual ticellularity when selection for it is relaxed and organism is not a sharp one. The mechanisms become once again a population of free-living that bind the trait group together can be more or individuals. One species of aphid seems to segre- less effective. They may also keep the evolution- gate a proportion of its inherited endosymbionts ary interests of the same group aligned across a as the exclusive source of founder populations wider or narrower range of traits. The metazoan of endosymbionts for its offspring, presumably organism and the unicellular eukaryotic cell are in order to keep the remaining endosymbionts clearly individuals. Jellyﬁsh, lichens, eusocial in- at work (Frank 1996). However, it is easy to sect colonies, and the ant/acacia symbiosis are overstate the importance of this particular mech- less clearly so. Each of these has a life cycle and anism. Plants typically do not have germline a developmental system that feed into its devel- segregation, so it cannot be a prerequisite for opment. But in most of these latter cases, it is complex multicellular life. Leo Buss has explored possible to describe evolutionary pressures with 214 Paul E. Gri≤ths and Russell D. Gray

respect to which the smaller life cycles nested The benefits of this reconceptualization of within the larger cycle do not form a trait group. evolution in DST terms are considerable. The more forced and implausible these scenarios, Phenomena that are marginalized in current the less theoretical role there is for a description genocentric conceptions of evolution, like ex- in which these cycles are treated as independent panded inheritance, niche construction and de- and not as parts. velopmental organization, are placed center stage. Here are some suggestions for the kind of research questions we think DST encourages. Conclusion 1. Treat all claims about instincts, genetic pro- The aim of this chapter is quite simple. The fear grams and other black boxes as potential re- that DST will lead to wholesale rejection of cur- search questions for developmental analysis, that rent evolutionary theory is not well founded. is, how does this trait actually develop, what re- DST is not anti-Darwinian, nor does it render the sources does its reliable development depend basic explanatory terms of evolutionary theory upon, are there many developmental routes to incoherent. To the contrary, DST expands the this outcome or only one, over what range of scope and power of adaptive/historical explana- parameters is this developmental outcome station. The core Darwinian concepts of inheri- ble, how does the “environment” change as a tance, natural selection, adaptation, individual, function of initial development differences that and lineage can be productively reworked in DST produce this trait? terms as follows: 2. Study expanded forms of inheritance. Con- • Developmental system—the interactants and duct studies to investigate the longevity and processes that produce a life cycle. fidelity of extended inheritance. Are there physiological and developmental mechanisms that • Evolutionary developmental system—the inter- enable these forms of inheritance to be vertically actants and processes that produce those devel- reproduced down lineages with fidelity and opmental outcomes that are reliably reproduced longevity? If extragenetic inheritance is an adap- in a lineage. tive developmental resource, then developmental • Inheritance—the reliable reproduction of de- systems that reliably and e≤ciently pass on that velopmental resources down lineages. resource would be at an advantage. Test adap- • Natural selection—the differential reproductive hypotheses about extended inheritance using tion of heritable variants of developmental comparative methods—for example, is species systems due to relative improvements in their diversity greater or rates of evolution higher in functioning. lineages with certain forms of extragenetic inheritance? Develop mathematical models of the im- • Adaptation—the product of natural selection. pact of different types of extragenetic inheritance • Individual—the most inclusive sequence of de- and their coevolution with genetic change. How velopmental events (life cycle) such that the tightly coupled do different developmental re- smaller repeated cycles nested within it form a sources have to be to influence each other’s evo- trait group with respect to most plausible evolu- lutionary dynamics? What role have expanded tionary scenarios. forms of inheritance played in major evolution- • Lineage—a causally connected sequence of ary transitions? similar individual life cycles. 3. Study “niche construction.” Conduct field ex- • Evolution—change over time in the composi- periments to assess the fitness impact of niche tion of populations of developmental systems. construction. Develop models to investigate the Darwinism and Developmental Systems 215

evolutionary consequences of the ways in which disputed. In G. C. Williams’s influential view, when we organisms select and modify their environments. measure change in gene frequencies we are getting at Test predictions from these models using com- the heart of the process of evolution. It is genes that parative methods. compete and are selected (Williams 1966). In contrast, phenotypic or hierarchical views of evolution accept 4. Investigate the extent and functional basis of that evolution can be represented by change in gene fre- developmental modularity. How modular is de- quencies, but locate the processes of competition and velopment? Are functionally linked traits also selection at higher levels of biological organization. linked together in development? Are there evolu- 3. In the mathematical theory of information tionary trends in the degree of modularity? Does (Shannon and Weaver 1949) and its relatives (Dretske the modularity only hold over a restricted range 1981), a signal sender conveys information to a receiver of parameters? Can these parameters be changed when the state of the receiver is correlated with the state by selection? Are extended forms of inheritance of the sender. The conditions under which this correla- more or less modular than genetic inheritance in tion exists constitute the “channel” between sender and receiver. Changes in the channel affect which state of their developmental consequences? the receiver corresponds to which state of the sender. Neo-Darwinism was the result of the union of The information conveyed by a particular state of the Darwin’s theory of natural selection with a par- receiver is as much a function of the channel, the context, as it is of the sender. ticular view of heredity. The new view of heredity transformed Darwin’s vision and gave rise to a 4. In local enhancement the model directs the subject’s attention to salient features of environment and the wide range of research questions. In rejecting the subject then develops the appropriate behavior through narrowly gene-centered view of heredity and individual trial-and-error learning. In emulation the bringing developmental processes back into our subject learns about cause and effect relations rather account of evolution, we are not rejecting the the- than the behavior itself. ory of natural selection but are attempting to 5. The traditional neo-Darwinian explanation of ad- unite it with the developmental systems account aptation is an “externalist” explanation. Explanation of heredity and thus to reveal new and promising flows from the environment to the organism and not research agendas. vice versa. For a good discussion of “externalist,” “internalist,” and “constructionist” explanatory strategies in general, see Godfrey-Smith (1996). Acknowledgments 6. For example, Dennett (1995) and Pinker (1997). The idea is popular in the evolutionary psychology move- We would like to thank Kendall Clements, Fiona ment (Barkow, Cosmides, and Tooby 1992). For an Jordan, Susan Oyama, Kim Sterelny, and Debbie attempt to integrate the idea that each trait has its own Waldron for comments on drafts of this chapter. selection of relevant environmental parameters into evolutionary psychology, see Irons (1998). 7. Susan Oyama (personal communication) is less Notes happy than we are with the idea of the external or universal physical environment. The next section makes 1. Here is a random selection: “Evolution 1. Process by clear that our formulation actually preserves the DST which organisms come to differ from generation to gen- insight that a full description of the developmental eration. 2. Change in the gene pool of a population systems of an individual organism, as opposed to a de- from generation to generation” (Arms and Camp 1987: scription of the typical developmental system of a pop- 1121). “Evolution is the result of accumulated changes ulation, will include all the causal factors that influence in the composition of the gene pool” (Curtis and Barnes that individual’s development. Oyama is legitimately 1989: 989). concerned that the need for a concept of endogenous 2. Although the definition of evolution as change in sources of change in evolution will create the impres- gene frequencies is widely accepted, its significance is sion that there is a need for sources of change outside 216 Paul E. Gri≤ths and Russell D. Gray

the developmental system in the individual case. That data. Proceedings of the National Academy of Sciences would be exactly the sort of dichotomous account of 85: 6002–6006. development that DST seeks to avoid, albeit with the Curtis, H., and N. S. Barnes. (1989). Biology. (5th ed.). boundary in an unusual place. New York: Worth Publishers. 8. We are not aware of any published version of this Darwin, C. (1859/1964). On the Origin of Species: A criticism, but it was ﬁrst suggested to Gri≤ths in con- Facsimile of the First Edition. Cambridge, MA: versation by Lindley Darden in 1994 and has also been Harvard University Press. raised by Alexander Rosenberg (personal communica- Darwin, C. (1881). The Formation of Vegetable Mould, tion). Through the Action of Worms, with Observations on Their Habits. London: Murray. Dawkins, R. (1982). The Extended Phenotype. Oxford: References Freeman. Arms, K., and P. S. Camp. (1987). Biology. (3d ed.). Dawkins, R. (1994). Burying the vehicle. Behavioral and New York: CBS College Publishing. Brain Sciences 17: 616–617. Barkow, J. H., L. Cosmides, and J. Tooby (Eds.) Dawkins, R. (1996). Climbing Mount Improbable. (1992). The Adapted Mind: Evolutionary Psychology London: Viking. and the Generation of Culture. Oxford: Oxford Univer- Dennett, D. C. (1995). Darwin’s Dangerous Idea. New sity Press. York: Simon and Schuster. Boyd, R., and P. J. Richerson. (1996). Why culture is Depew, D. J., and B. H. Weber. (1995). Darwinism common, but cultural evolution is rare. Proceedings of Evolving: Systems Dynamics and the Genealogy of Nat- the British Academy 88: 77–93. ural Selection. Cambridge, MA: Bradford Books/MIT Brandon, R. (1988). The levels of selection: A hierarchy Press. of interactors. In H. Plotkin (Ed.), The Role of Behavior Dretske, F. (1981). Knowledge and the Flow of Informa- in Evolution, pp. 51–71. Cambridge, MA: MIT Press. tion. Oxford: Blackwells. Brandon, R. (1990). Adaptation and Environment. Frank, S. A. (1996). Host control of symbiont trans- Princeton, NJ: Princeton University Press. mission: The separation of symbionts into germ and Brandon, R. N. (1992). Environment. In E. Fox Keller soma. American Naturalist 148: 1113–1124. and E. A. Lloyd (Eds.), Keywords in Evolutionary Gilbert, S. F., J. M. Opitz, and R. A. Raff. (1996). Biology, pp. 81–86. Cambridge, MA: Harvard Univer- Resynthesizing evolutionary and developmental biol- sity Press. ogy. Developmental Biology 173: 357–372. Brandon, R. (1999). The units of selection revisited: Godfrey-Smith, P. (1996). Complexity and the Function The modules of selection. Biology and Philosophy 14: of Mind in Nature. Cambridge: Cambridge University 167–180. Press. Brandon, R., and J. Antonovics. (1996). The coevolu- Goodwin, B. C. (1984). Changing from an evolutionary tion of organism and environment. In R. Brandon to a generative paradigm in biology. In J. W. Pollard (Ed.), Concepts and Methods in Evolutionary Biol- (Ed.), Evolutionary Theory: Paths into the Future, pp. ogy, pp. 161–178. Cambridge: Cambridge University 99–120. New York: Wiley. Press. Gould, S. J., and R. Lewontin. (1978). The spandrels of Buss, L. (1987). The Evolution of Individuality. San Marco and the Panglossian paradigm: A critique of Princeton, NJ: Princeton University Press. the adaptationist programme. Proceedings of the Royal Caporael, L. (1995). Sociality: Coordinating bodies, Society of London 205: 581–598. minds and groups. Psycoloquy. http://www.princeton. Gray, R. D. (1992). Death of the gene: Developmental edu/pub/harnad/Psycoloquy/1995.volume.6. systems strike back. In P. E. Gri≤ths (Ed.), Trees of Cavalli-Sforza, L. L., A. Piazza, P. Menozzi, and J. Life, pp. 165–210. Dordrecht: Kluwer. Mountain. (1988). Reconstruction of human evolution: Gray, R. D. (1997). “In the belly of the monster”: Bringing together genetic, archaeological and linguistic Feminism, developmental systems, and evolutionary Darwinism and Developmental Systems 217

explanations. In P. A. Gowaty (Ed.), Evolutionary Lehrman, D. S. (1970). Semantic and conceptual issues Biology and Feminism, pp. 385–413. New York: in the nature-nurture problem. In L. R. Aronson, E. Chapman and Hall. Tobach, D. S. Lehrman, and J. S. Rosenblatt (Eds.), Gray, R. D. (2001). Selfish genes or developmental sys- Development and Evolution of Behavior: Essays in Mem- tems? Evolution without interactors and replicators? In ory of T. C. Schneirla, pp. 17–52. San Francisco: R. Singh, C. Krimbas, J. Beatty, and D. Paul (Eds.), Freeman. Thinking about Evolution: Historical, Philosophical, and Lerner, I. M. (1970). Genetic Homeostasis. New York: Political Perspectives, pp. 184–207. Cambridge: Cam- Dover Publications. bridge University Press. Lewontin, R. (1970). The units of selection. Annual Gray, R. D., and F. M. Jordan. (2000). Language trees Review of Ecology & Systematics 1: 1–14. support the express train sequence of Austronesian ex- Lewontin, R. C. (1978). Adaptation. Scientific pansion. Nature 405: 1052–1055. American 239 (September): 156–169. Gri≤ths, P. E., and R. D. Gray. (1994a). Developmental Lewontin, R. C. (1982). Organism & environment. In systems and evolutionary explanation. Journal of Phil- H. Plotkin (Ed.), Learning, Development, Culture, pp. osophy 91(6): 277–304. 151–170. New York: John Wiley. Gri≤ths, P. E., and R. D. Gray. (1994b). Replicators Lewontin, R. C. (1983a). Gene, organism and environ- and vehicles—or developmental systems? Behavioral ment. In D. S. Bendall (Ed.), Evolution: From Molecules and Brain Sciences 17: 623–624. to Men, pp. 273–285. Cambridge: Cambridge Univer- Gri≤ths, P. E., and R. D. Gray. (1997). Replicator II: sity Press. Judgement day. Biology and Philosophy 12(4): 471– 492. Lewontin, R. C. (1983b). The organism as the subject Gri≤ths, P. E., and R. D. Knight. (1998). What is the and object of evolution. Scientia 118: 65–82. developmentalist challenge? Philosophy of Science 65(2): Lindquist, S. (1997). Mad cows meet psi-chotic yeast: 253–258. the expansion of the prion hypothesis. Cell 89: 495. Halder, G. P., P. Callerts, and W. J. Gehring. (1995). Lynch, A., and A. J. Baker. (1986). Congruence of mor- Induction of ectopic eyes by targeted expression of the phological and cultural evolution in Atlantic island eyeless gene in Drosophila. Science 267: 1788–1792. cha≤nch populations. Canadian Journal of Zoology 64: Harlow, H. F., and M. K. Harlow. (1962). Social dep- 1576–1580. rivation in monkeys. Scientific American 207(5): 136– Maynard Smith, J. (2000). The concept of information 146. in biology. Philosophy of Science 67: 177–194. Hull, D. L. (1980). Individuality and selection. Annual Maynard Smith, J., and E. Szathmáry. (1995). The Review of Ecology and Systematics 11: 311–332. Major Transitions in Evolution. Oxford: W. H. Irons, W. (1998). Adaptively relevant environments Freeman. versus the environment of evolutionary adaptedness. Mills, S., and J. Beatty. (1979). The propensity inter- Evolutionary Anthropology 6: 194–204. pretation of fitness. Philosophy of Science 46: 263–286. Jablonka, E., and E. Szathmáry. (1995). The evolution Money, J. (1992). The Kaspar Hauser Syndrome of of information storage and heredity. TREE 10(5): 206– “Psychosocial Dwarfism”: Deficient Statural, Intellec- 211. tual, and Social Growth Induced by Child Abuse. Buffalo, Johnston, T. D. (1987). The persistence of dichotomies NY: Prometheus Books. in the study of behavioural development. Developmen- Moran, N., and P. Baumann. (1994). Phylogenetics of tal Review 7: 149–182. cytoplasmically inherited microorganisms of arthro- Keller, L., and K. G. Ross. (1993). Phenotypic plas- pods. Trends in Ecology and Evolution 9: 15–20. ticity and “cultural transmission” of alternative social Mount, A. B. (1964). The interdependence of the organisations in the fire ant Solenopsis invicta. Be- eucalpyts and forest fires in southern Australia. Austra- havioural Ecology and Sociobiology 33: 121–129. lian Forestry 28: 166–172. Lansbury, P (1997). Yeast Prions: Inheritance by Newcomb, R. D., P. M. Campbell, D. L. Ollis, E. seeded protein polymerisations? Current Biology 7: Cheah, R. J. Russell, and J. G. Oakeshott. (1997). A R617. single amino acid substitution converts a carboxy- 218 Paul E. Gri≤ths and Russell D. Gray

lesterase to an organophosphorous hydrolase and con- leading to life. Journal of Theoretical Biology 187: fers insecticide resistance on a blowfly. Proceedings of 555–571. the National Academy of Science 94: 7464–7468. Tomasello, M., A. C. Kruger, and H. H. Ratner. Odling-Smee, F. J. (1988). Niche-constructing pheno- (1993). Cultural learning. Behavioral and Brain Sciences types. In H. C. Plotkin (Ed.), The Role of Behavior in 16(3): 495–552. Evolution, pp. 73–132. Cambridge, MA: MIT Press. Trivers, R. L. (1971). The evolution of reciprocal altru- Oyama, S. (1985). The Ontogeny of Information: De- ism. Quarterly Review of Biology 46(4): 35–57. velopmental Systems and Evolution. Cambridge: Cam- Vines, G. (1999). Gendercide. New Scientist 158: 44–47. bridge University Press. (2d rev. ed., Durham, NC: Wagner, G. P., and L. Altenberg. (1996). Complex Duke University Press, 2000.) adaptations and the evolution of evolvability. Evolution Pal, C., and I. Miklos. (1999). Epigenetic inheritance, 50: 967–976. genetic assimilation and speciation. Journal of Theore- Whitehead, H. (1998). Cultural selection and genetic tical Biology 200: 19–37. diversity in matrilineal whales. Science 282: 1708–1711. Penny, D., E. E. Watson, and M. A. Steel. (1993). Trees Whiten, A., J. Goodall, W. C. McGrew, T. Nishida, from languages and genes are very similar. Systematic V. Reynolds, Y. Sugiyama, C. E. G. Tutin, R. W. Biology 42: 382–384. Wrangham, and C. Boesch. (1999). Culture in chim- Pinker, S. (1997). How the Mind Works. New York: panzees. Nature 399: 682–685. Allen Lane. Williams, G. C. (1966). Adaptation and Natural Ratnieks, F. L. W., and P. K. Visscher. (1989). Worker Selection. Princeton, NJ: Princeton University Press. policing in honeybees. Nature 342: 796–797. Wilson, D. S. (1997). Biological communities as func- Rosenberg, A. (1978). The supervenience of biological tionally organized units. Ecology 78: 2018–2024. concepts. Philosophy of Science 45: 368–386. Wilson, D. S., and E. Sober. (1994). Reintroducing Shannon, C. E., and W. Weaver. (1949). The Mathe- group selection to the human behavioral sciences. matical Theory of Communication. Urbana: University Behavioral and Brain Sciences 17: 585–608. of Illinois Press. Sober, E., and R. C. Lewontin. (1982). Artifact, cause & genic selection. Philosophy of Science 49: 157–180. Sober, E., and D. S. Wilson. (1994). A critical review of philosophical work on the units of selection problem. Philosophy of Science 61(4): 534–555. Sober, E., and D. S. Wilson. (1998). Unto Others: The Evolution and Psychology of Unselfish Behavior. Cam- bridge, MA: Harvard University Press. Sterelny, K. (1996). Explanatory pluralism in evolutionary biology. Biology and Philosophy 11(2): 193–214. Sterelny, K. (in press). Development, evolution and adaptation. Philosophy of Science. Sterelny, K., M. Dickison, and K. Smith. (1996). The extended replicator. Biology and Philosophy 11(3): 377– 403. Sterelny, K., and P. E. Gri≤ths. (1999). Sex and Death: An Introduction to the Philosophy of Biology. Chicago: University of Chicago Press. Szathmáry, E., and J. Maynard Smith. (1997). From replicators to reproducers: The first major transition Generative Entrenchment and the Developmental Systems Approach to 17 Evolutionary Processes

William C. Wimsatt

Developmental Systems Theory as a developmental and evolutionarily significant fac- Methodological Theory tors that can be modeled, generate predictions, and yield new understandings. If so, GE provides DST is not a theory in the sense of a specific model that an important engine for evolutionary processes produces predictions to be tested against rival models. on the DST perspective—or perhaps a transmis- Instead DST is a general theoretical perspective on sion coupling developmental and evolutionary development and evolution at the same level of gener- processes to the same effect. In what is becoming ality as genic seletionism. (Gray 1999) a tradition for DST, this also involves recogniz- DST is a broad research perspective rather than a ing what we already do and know in a new light. specific model. Hence to succeed as a research How can generative entrenchment—a source program it must generate or entrain other theory, of developmental constraints—be an engine? that is, specific models, data gathering, simula- Blades in a water turbine constrain the flow of tions, and experiments. We all have to eat, and water and provide motive power—acting as an theories or research perspectives without specific engine even though they are not the source of the issue will not be pursued indefinitely. For recon- energy—by transforming the potential energy or ceptualizing relations, and reanalyzing data, par- rectilinear motion of the water stream into the ticularly to compromise facile “nothing-but” angular momentum and torque of the turbine style claims about where “ultimate” explanations shaft. They are levers, one of the five “fundamen- 2 lie (somehow always microlevel and “genetic”), tal machines” of high school physics. Similarly, DST has done a lot already. (See Oyama’s 1985 GE affects and constrains which kinds of varia- groundbreaking critiques of traditional nativist tion get incorporated in processes of evolutionary theories, and Gri≤ths and Gray’s 1994 particu- change and elaboration. Things that look like larly compact and illuminating presentation of constraints on processes on a shorter time scale DST.) We will all benefit from these reconceptu- can look like engines of that process in a larger alizations. But we should work for more. context or on longer time scales. The main effects I have long felt my ideas consilient with DST. of GE constraints are to control or modulate the Perhaps this is self-selection; it is certainly also relative rates of different kinds of change pro- due to early extended associations with Richard cesses (Schank and Wimsatt 1988; Wimsatt and Lewontin and Richard Levins. (See Wimsatt Schank 1988). That is not only an engine, but 1986 on development and nativism, and Wimsatt potentially a cybernetic engine—a potential 1980b, 1981b for early analyses supporting source of control structures. But unlike the pic- Gray’s attacks on gene selectionism and reduc- ture painted by gene-selectionists, the loci of tionism.) Nonetheless, I had tended not to count control here are distributed across system and myself as a DST theorist because I sought more environment, and up and down the levels of specific models.1 Now I see from Gray’s paper organization. that I am a DST theorist after all—at least if DST Neo-Darwinism ignored development in “the is a family resemblance concept. (Some others in major synthesis” bringing population genetics to this book are more “specific” than I!) Here I will the theoretical core of evolutionary biology. In try to bridge the gap between theoretical perspec- the last two decades this has emerged as a serious tive and specific models. Generative entrench- error. Perhaps at mid-century there was not the ment (or GE) reflects factors of great generality work (in developmental genetics and other work having natural resonance with DST—a class of on development from neurophysiology to animal 220 William C. Wimsatt

ethology and social psychology) to force a pro- (by one traditionally more reductionistic) reveals ductive synthesis. But times have changed. “Evo- his own surprise at the developmental complex- devo” is one of the hottest research areas in both ities he found in the supposedly simple and de- evolutionary biology and development. (Raff terminate nematode soil worm Caenorhabditis 1996 is a synthetic work by probably the major elegans. We must deny the total hegemony of the spokesman of this new field.) But we must be genetic perspective, but at the same time we must careful lest this come to be seen as just claiming not be afraid to use some of its products. the larger playing field that genes deserve.3 Progress and balance require that the other relevant fields continue to develop in analyzing the Darwin’s Principles Embodied: The Evolution and causal e≤cacy of processes at their levels of or- Entrenchment of Generative Structures ganization. An extended multilevel population genetics surely must be part of an adequate pic- Where should we start with our top-down theo- ture, but this will not su≤ce. As DST urges, the rizing to put developmental genetics in context? whole ecologically embodied and embedded life Why not with the only other general biological cycle, or for multiple levels, interdigitated life theory we have—evolution. Any evolving sys- courses or cycles—Caporael’s (1997) “repeated tems must meet what Lewontin (1970) has called assemblies”— must be treated as relevant units of “Darwin’s principles.” they must: 4 analysis. DST must play an active role to leave 1. have descendants that differ in their properties its mark in evo-devo and elsewhere as it con- (variation), tinues to develop. We need more “top-down” con- 2. some of which are heritable (heritable varia- ceptual and theoretical work to properly situate tion), and the significance of the emerging results. I have long argued against “nothing-but” style genetic 3. have varying causal tendencies to have descen- reductionism.5 We also need room to study the dants (heritable variation in fitness). same systems from the top down. Both ap- These three principles—all met simultane- proaches are legitimate. ously for the same processes or properties of the But this is not and should not be a one-way same entities—are widely acknowledged as core street. DST advocates should also be prepared to requirements for evolution to occur. Any popula- learn from the bottom up: we will learn as much tion of entities meeting all three of these condi- in new evidence and ideas from developmen- tions will undergo an evolutionary process, and tal genetics and more broadly from evo-devo as none which fail to meet even one of them can do we have to give to it. We have all benefited from so.6 They might be thought of as the logical or careful analyses at the genetic level in the work of conceptual conditions for an evolutionary process Jablonka (see chapter 9), Moss (1992, 1998; see because they key into the fundamental require- chapter 8), Keller (1995), Neumann-Held (1998; ments of evolution by natural selection. see chapter 7), and back to work by Lewontin But there are two further conditions of great (1983) in criticism of the genes as self-activating generality. These conditions recognize develop- and self-expressing. These works are as much ment’s central role in the evolutionary process. I about development as about embodiment. know of no interesting evolutionary process Griesemer’s move (1999a, 2000) to reconceptual- whatsoever (physical or conceptual) that does ize and replace replicator-talk with developmen- not meet them.7 With these DST becomes not just tally embodied “reproducers” promises notions an additional perspective on evolution, but fun- of inheritance more friendly to DST. Schaffner’s damental to it.8 The same entities which meet the (1998) rich exploration of developmental genetics first three conditions must also be: Generative Entrenchment and the DST Approach 221

4. structures which are generated over time so much like entropy, with dynamical changes in the they have a developmental history (generativity), system tending to increase genericity (Schank and and Wimsatt 1988). These properties are physically 5. some elements that have larger or more perva- robust just as entropy increase is on the statist- sive effects than others in that production (differ- ical form of the second law of thermodynamics: ential entrenchment). it allows reversibility and reductions of entropy in principle but de facto guarantees irreversibil- Then different elements in the structures char- ity and increasing entropy if the system is not acteristically have downstream effects of different already at an entropic maximum! magnitudes. The generative entrenchment (GE) of 2. Evolutionarily generic. Schank and I (1988) an element is the magnitude of those effects in suggested that Kauffman’s treatment of generic that generation or life cycle. Elements with larger traits as monadic properties of the system was degrees of GE are generators. This is a degree too narrow. We proposed a second complemen- property. The GE of an element in an evolution- tary class of generic properties that are invariant ary unit has multiple deep consequences for its across different selection regimes. An example of evolutionary fate and character, and that of sys- this is discussed further below: The fitness loss tems impinging on it. I return to this later. due to loss of a deeply GE’d trait is profound, But first, why are these deep principles, and and, because it affects so many things is almost what kinds of principles are they? Exceptions are context-independent. There is almost nothing logically possible, so they cannot be logical prin- you can do to help! This situation will therefore ciples. And we can imagine physical systems be almost universally selected against. If physi- which fail to meet them, so they are physically cally generic means that a property is realized for possible, though, as we will see, rather unlikely. nearly all contexts, evolutionarily generic means But there are weaker but still very strong kinds of that it is selected for (or against) in nearly all principles producing universal or near-universal contexts. behavior. I characterize them and then raise some qualifications: 3. Rate-dominant generic. Simon’s (1962) argument for the evolution of hierarchical organiza- 1. Physically generic. Kauffman (1993) urges tion via the formation of (nearly decomposable) us to build theory for the emergence of complex- stable subassemblies suggest that hierarchically ity around properties which are stochastically organized complex structures will evolve much inevitable (or stochastically self-organizing)— faster than other alternatives. And as evolution- motivated by his belief that there are things ary biologists know, those that get there first get too complex to be maintained by selection (or to eat the others as they arrive! So the evolution- by selection alone).9 Consider an ensemble of ary systems that we see should be hierarchical systems defined by a set of constructional con- and nearly decomposable.10 These could also be straints but free to vary in other defined direc- thought of as deep design principles of “meta- tions. A generic property for that ensemble is one engineering” (Dennett 1995). characterizing virtually all systems in it. Thus (1) 4. Symmetry-breaking generic: those properties a system picked at random from the ensemble of interest emerge or self-organize through self- will probably have the property. Further, dynam- amplifying deviations from conditions of homo- ical processes which change systems from one geneity or symmetry. Most discussed cases are not ensemble state to another will probably (2) con- principles but examples of pattern formation, like serve generic properties and (3) produce generic that of the famous Zhabotinski-Belusov reagent, properties in systems lacking them. If degree of in which spiral colored patterns of activity genericity is quantifiable, it will behave very 222 William C. Wimsatt

emerge from local density fluctuations in the explaining the laws of embryology. (Darwin 1859: reactants, and expand to cover and entrain the 13–14) whole system. I will argue that condition (5)— Darwin considered this a “rule governing differential entrenchment—is such a principle. inheritance” though it would not appear as such Finally, there are results derivable from some (indeed, not at all!) in any genetics text in the combination of the above with possibly broad post-Mendelian era.13 It was not about combina- initial conditions and other assumptions. Ap- torial heredity, but about characters in relation to parent conflicts among these kinds of principles development. If we apply this rule in sequence to may arise when the conditions for deriving one of peculiarities (and to normalities!) appearing them are violated, or when expectations derived throughout the life course (not thereby implying from application of them to simpler systems are a repeated trajectory), then if the organism has overgeneralized. offspring, repeated application of this rule for the different characters predicts that there will be (at So How Do Our Two New Principles Stand? least roughly), a life cycle—thereby implying a repeated trajectory. Any physical process (including reproduction) The “laws of embryology” Darwin mentions takes time, so part of (4) is trivially satisfied. But are presumably those of von Baer—in post- is any temporally extended process a develop- Darwinian guise, that earlier developmental mental process?11 Surely not. Lineages are not traits tend to be more evolutionarily conserva- sequences of arbitrary and dissimilar develop- tive, so more taxonomically widespread, than mental trajectories: Developmental structures later ones (Gould 1977). These would produce or processes have characteristic patterns or life life cycles with less recognizable variation at ear- cycles. This is very important—enough to be lier developmental stages than at later ones—not added as a separate condition if we had to. But exactly similar (for developmental as well as for we don’t: it already follows from the rest of (1) hereditary reasons), but easily recognizable as life thru (5), via the evolutionary conservatism of cycles. Von Baer’s laws are a straightforward pre- entrenched elements of the life cycle. That there diction of the simplest models of the action of are recognizable life cycles—though not necessar- generative entrenchment in evolution (Wimsatt ily all of their details12—is a consequence of gener- 1986). So condition (4) is established as central, ative entrenchment—(and thus of 5 in the context unavoidable, and having an honorable history. of 1–4). We get them “for free”! Not only traits Condition (5) is equally fundamental—and but their pattern of appearance is heritable, and inevitable, though not widely recognized as such. may itself be treated as a trait. And phenotypic It is inevitably satisfied for heterogeneous struc- characters, which depend for their form of ex- tures: Try to imagine a machine or system whose pression upon a prior relatively specific sequence breakdowns are equally severe for any kind of events, or more generally, on developmental of failure, in any part, under all conditions.14 timing, depend substantially upon it. (There aren’t any!) Any differentiated system— Darwin recognized age-specific patterning as biological, cognitive, or cultural—exhibits vari- “a rule governing inheritance” in the first chapter ous degrees of generative entrenchment among of The Origin of Species: its parts and activities. Condition (5) is essentially a generic property (sensu Kauffman 1993) of dif- A much more important rule . . . is that, at whatever ferentiated systems. Almost all possible systems period of life a peculiarity first appears, it tends to of a given kind will meet it. But there is more. If appear in the offspring at a corresponding age. . . . I you could find and begin with a system violating believe this rule to be of the highest importance in (5), so everything had effects of the same magni- Generative Entrenchment and the DST Approach 223

tude, mutational fluctuations and selection pro- evolutionary change—stated in a way that fits cesses acting on it are self-amplifying, pushing also exaptive or macroevolutionary change. evolution toward systems which increasingly sat- This is a self-amplifying process. Mutation and isfy it. So (5) is doubly guaranteed both as a selection should act over long evolutionary periods generic property of mechanisms, and as a evolu- to break symmetrical or homogeneous distribu- tionary generic property of arbitrary selection tions of functions and selection coe≤cients, so that regimes acting on structures which (for whatever systems with equal or nearly equal selection co- improbable reasons) fail to meet it. These sur- e≤cients for their parts would be driven by selec- prising points were first made in Wimsatt and tion to greater variability in the distribution of the Schank 1988. The argument for the latter goes as parts’ selection coe≤cients.15 Stochastic fluctua- follows: tions in selection coe≤cients due to environmen- Evolution is opportunistic—it uses what is tal fluctuations (ecological drift) can also break already there, with new additions to modify exist- symmetries. ing or to make new adaptive systems. (1) Con- So, in summary, why add conditions (4) and sider a system in which all elements make equal (5) to the widely accepted Darwin’s principles? contributions, and thus engender equal fitness Because any nontrivial physical or conceptual sys- losses if they are knocked out. (2) Assume they all tem satisfying (1–3) will do so via causal (pheno- have equal probability per unit time of sustaining typic) structures satisfying (4) and (5). So any such a decrement. (3) Assume also a given prob- evolutionary systems will satisfy (4) and (5). They ability per unit time that any element (gene, will thus have a development, and if they can aspect of the phenotype, behavior, or phenotype- reproduce and pass on their set of generators, will environment relational structure) will acquire a have a heredity. This order is expository, not dependent adaptation—one adaptive only in the causal: without a minimally reliable heredity, presence of that factor. (When genetic, these are they cannot evolve a complex developmental called “modifier genes”.) (4) Elements that stay phenotype, but developmental architecture can around longer have a higher probability of ac- increase the e≤cacy and reliability of hereditary quiring such adaptations, and a higher expected transmission. Heredity and development thus number of such acquisitions. bootstrap each other, as emerging genotype and Assumptions 1–3 are conditions of the ideal- phenotype, through evolution. This provides a ization setting up the symmetry argument. (4) symmetric view for the nature and origin of life, is entailed by (3) and the laws of probability. requiring only that generators retain their gener- Continuing: (5) Elements that acquire dependent ative powers under a su≤cient fraction of acces- adaptations have, in consequence, greater gener- sible small changes in their structure. Then they ative entrenchment. (6) Degradational mutations will also show phenotypic variations which (by in them or in genes affecting them would now the logic of Darwin’s argument) yield fitness dif- cause greater fitness losses than before they ferences, natural selection, and evolution. acquired their dependent adaptations. (7) Ana- lytical results of population genetic theory (Wright 1931) show that they then have increased Genes or Generators? chances of staying around longer, and indeed, the effect is nonlinear. By a reapplication of (4), they “Darwin’s principles” never mention genes. have also increased their chances of acquiring Neither do the new additions. This expanded list still other dependent adaptations—“downstream of conditions gives heredity and development with- modifiers.” This basic process is the developmen- out ever introducing the usual replicator/interactor tal expression of what happens in much micro- distinction, widely adopted since Hull (1980) as the appropriate generalization of the genotype- 224 William C. Wimsatt

phenotype distinction for the informational gen- commonly, where autocatalysis is such a distrib- eralization of “genes.” It is too early to tell uted, complex, and diffuse process (with identities whether we can eliminate it in all contexts, but it or similarities determined more by context than seems quite clear that here Griesemer’s account by content), that there seems no point to try to of reproduction fits more naturally. Griesemer track compact lineages through it. Paradoxically, (1999, 2000) roots a conception of heredity with- if I am right, the DST approach may provide in in an account of evolution as a materially con- that case (in at least some ways) a simpler and tinuous lineage of developments to facilitate the more compact account than a memetic one—as if central reintegration of development into our genetics and development were changing places accounts of the evolutionary process.16 Genes are (Wimsatt 1999a). (How does a scientific theory agents in these stories (in biology) not the privi- make a copy of itself? Very indirectly!) Econ- leged bearers of information, but coactors with omist Kenneth Boulding used to quip that “a car (the rest of) the developing phenotype and its is just an organism with an exceedingly compli- environmental resources as bearers of relation- cated sex life.” Like a technological virus, it takes ally embodied information.17 over a complex social structure and redirects the I proposed in 1981b (see also Callebaut 1993: resources of a large fraction of it to reproduce 425–429) that generative entrenchment could more of its own kind.19 Indeed, our economic sys- be used to individuate genes in terms of their tem has fostered an environment—a “culture heterocatalytic role by their phenotypic activity, dish” in which the invention, mutation, and ex- rather than talk of copying (an abstraction of pansion of such cultural viruses is encouraged their autocatalytic role) as replicator accounts try until, many environmentalists would say, it has to do. A “heterocatalytic” account of (biological) assumed cancerous proportions (see also Sperber genes consilient with this is nicely elaborated by 1995). Eva Neumann-Held (1998). Sterelny, Smith, and Dickison (1996) present a rich and insightful friendly critique of DST, but in doing so defend The Consequences of Differential Generative a notion of “replicator” far broader than most Entrenchment (and closer both to Griesemer’s contrasting “reproducer” concept, and to a heterocatalytic If a structure meeting (1–5) is even minimally ad- one)—one that if fleshed out is not ultimately at apted to its environment or task, then modifi- odds with DST. There is no reason however why cations of more deeply GE’d elements will have the heterocatalytic roles picked out by generative higher probabilities both of being maladaptive, entrenchment criteria must correspond to well- and of being more seriously maladaptive. These delineated genes—the context dependence and become more extreme—in the simplest models, distributed character of the associated activities exponentially so—either for larger structures may be so great as to make that seem unwise (see (e.g., if they grow by adding elements down- chapter 7). stream), or as one looks to more deeply en- Heterocatalytic gene-like things picked out by trenched elements in a given structure. Either the GE criterion in biology would include some change increases the degree of “lock in” of 20 but not all (traditional) genes, some things which entrenched elements. are not genes, and most often, heterogeneous Selection acts on the structure as a whole, so complexes of both.18 In some domains (like cul- parts of an adaptive structure are inevitably tural evolution) gene-like things may be picked coadapted to each other, and to different compo- out by GE criteria where there is arguably noth- nents of the environment. Larger changes in this ing picked out by autocatalytic criteria, or more structure will have to meet more design constraints. Fewer changes can do so. Ever larger Generative Entrenchment and the DST Approach 225

changes have a rapidly decreasing chance of acquired distinction (Wimsatt 1986, 1999b). being adaptive. Mutations in deeply GE’d ele- Turner (1991) employed generative entrenchment ments will have large and diverse effects, so are to analyze the distinction between literal and much more likely to be severely disadvantageous figurative meaning. Gri≤th (1996) uses GE in or lethal. Simple analytical models (Wimsatt a related way to suggest an analysis for natural 1986) and more realistic structures and simu- kinds. It has powerful applications to cultural lations (Rasmussen 1987; Schank and Wimsatt evolution in general, and scientific change in par- 1988; Wimsatt and Schank 1988) all show that ticular (see Callebaut 1993: 331–334, 378–383, system elements with greater GE tend to be much 425–429; Griesemer and Wimsatt 1989; Wimsatt more evolutionarily conservative. Changes accu- 1995). Entities with larger generative roles are mulate elsewhere while these deeper features ap- more foundational (in role and properties), more pear relatively “frozen” over evolutionary time. likely to persist to be observed and, (for some This is the basis of von Baer’s “law”—revealed fraction of them) to grow. If they are more robust in our models as a probabilistic generalization. than alternatives, they are more likely to have While things can appear early without being been there from the beginning, and to apparently GE’d, and or appear late and be GE’d, there are have an almost unconditional necessity. strong stochastic associations between earliness Returning to biology, Arthur (1997) provides in development and increasing probabilities and the broadest current review but does not cover degrees of generative entrenchment. everything. In developmental genetics, Rasmus- We have ways for modulating or weakening sen (1987) explicitly drew on my 1986 to predict generative entrenchment for some purposes so we the broad architecture of the developmental pro- can (occasionally) make deeper modifications gram of Drosophila melanogaster from effects of and get away with it (Wimsatt 1987). Crucial for genetic mutants and data from comparative phy- cultural evolution, most of these ways do not logeny. Even emerging before most of the HOX apply or apply only in weaker forms for biol- “explosion” (which yield a strikingly rich GE ogy.21 For this and other reasons (notably the structure), the vast majority of its predictions horizontal transmission of practices and para- remain sound. Schank and I have applied GE sites22)—cultural evolution commonly proceeds models to problems in biological evolution and much faster than biological evolution.23 development ranging from the architecture of These exceptions aside, one can predict which gene control networks (1988), and the role of parts of such structures are more likely to be pre- modularity in development (1998), to the evolu- served, or to change—and over broader time tion of complexity (Wimsatt and Schank 1988). scales, their relative rates of change—in terms of The generative form of adaptive structures their generative entrenchment. What kinds of affects the foci and relative rates of their evolu- structures? They could be quite diverse: proposition. Evolution acts on that form. So one should tions in a generated network of inferences; laws be able to trace feedbacks from developmental or consequences in a scientific theory; experimen- pattern to evolutionary trajectory and back tal procedures or pieces of material technology; again, identifying pivot points where relative sta- structures or behaviors in a developing pheno- sis or elaboration can cause major changes in type; relationships between organism and niche evolutionary direction. (Generative entrench- resources; cultural institutions or norms in a soci- ment as an engine!) These are second-order ety; or the dynamical structures—biological and effects. They give GE theories greater explana- cognitive—driving cognitive development. GE tory power than one might at first suppose. necessitates and provides a reconceptualiza- Campbell’s (1974) “vicarious selectors” have tion and replacement of the traditional innate/ demonstrated abilities to create or facilitate new 226 William C. Wimsatt

effectively autonomous higher level dynamics.24 can massively change evolutionary history.26 It Thus perception (a vicarious selector) plus mate- seems plausible—indeed (as I will argue) almost choice can create runaway sexual selection pro- inescapable—that a successive layered patch- cesses leading natural selection in new directions work of contingencies has affected not only the (Todd and Miller 1998). Cultural evolution detailed organic designs we see, and variations (Boyd and Richerson 1985; Wimsatt 1999a) is among conspecific organisms, but also much similarly open-ended—partly for similar reasons. deeper things—the very configuration and defini- These processes build upon and interact richly tion of the possible design space, and the regions with each other. Partial products of GE, they in it they occupy. Deep accidents from the distant provide both new opportunities for its action and past not only define the constraints of current occasions for its use as a tool of analysis. Applied (so-called) optimizations, but constraints on to the evolution of cognitive and cultural sys- these constraints, and so on, moving backward tems, generative entrenchment extends evolu- through a history of the deposition of exaptive tionary epistemology with new predictions and dependencies (Gould and Vrba 1982) which explanations. Differences among these processes become framing principles for the design of suc- in different areas affect how the models must be cessively acquired and modified adaptations. developed and applied and the kinds of results GE provides an explanation, perhaps the only expected. These differences, and the character, possible explanation, for how and why this is consequences of, and interactions among these possible. In reproducing heritable systems, GE elaborating second-order effects will provide and selection may provide su≤cient conditions many areas for further development of theory for the incorporation and growing importance and models. over time of contingency, and of history, in the The biological theory of evolution is not prop- explanation of form. Not only that: Because erly seen as an eliminative reductive one, but as older contingencies become more general and an integrative theory articulating a diversity of more entrenched, we have an explanation other theories about biological structures and through GE for the broader and deeper (contin- processes. So DST must be capable of articulat- gent) regularities of nature—usually not for the ing the insights and assessing the consequences of details of the particular regularities, but for why the processes, forces, groups, and objects studied there should be any such regularities.27 In this GE in the various biological and human sciences. is like “chaotic” dynamics—there can be sensitive Generative dependency structures and differen- dependence on initial conditions, but unlike that tial entrenchment are found widely there, and case, the interest of GE arises only after the im- should therefore provide a widely useful tool in pact of these conditions has become relatively their analysis. entrenched, by which time further small changes lose their capacity to affect those aspects of the outcome. Generative Entrenchment, Satisficing Design, and Because of the dominant role of population Historical Contingency genetics in evolutionary biology, stochastic genetic events and processes—point mutations, History matters to evolution. It is not too far tandem duplications, inversions, segregation wrong to say that everything interesting about events, independent assortments, and other re- adaptation is a product of selection for improve- combinations in inheritance—are the most comments in design, or of history, or their interac- monly cited sources of contingency in evolu- 25 tion. Gould (e.g., 1989) has emphasized the role tion. Equally important are chance ecological of contingency in evolutionary processes, arguing events: meetings leading to matings, migrations, that minor unrelated “accidents” or “incidents” symbioses, exclusions, parasitisms, and preda- Generative Entrenchment and the DST Approach 227

tions. As George Williams (1966) quipped, “To a you approach an adaptive peak and the longer plankton, a great blue whale is an act of God.” you remain there, the less obvious is the path you Better design as a plankton will not help if it is in took to get there—or at least so it would seem. the wrong place at the wrong time. The only systematic means of escape from the But accidental or contingent events do not by fluctuations of fleeting optima over time is themselves make history. They need not leave dis- through generative entrenchment via an impor- tinguishable historical traces much later. Most do tant property of the latter: The more deeply gen- not. They may be: eratively entrenched something becomes, the 1. Not heritable, or more context-independent and the larger is the 2. even if locally heritable, are averaged out fitness loss if it is disturbed. Deeply generatively (or erased) in changing drift processes in space entrenched things become really established—the and time—lost in multiple intersecting entropic deepest become functional necessities. This says in processes. another way that deeply GE’d things will be pre- 3. Damped out via “noise tolerant” design in served.28 If the fitness decrement from their loss phenotype or genotype. Biological processes are gets larger and more invariant, so do they. paradigmatically noise tolerant at all levels: syn- To mark history, an event must cause cascades onymies in the genetic code, diploidy, alternative of dependent events that affect evolution. Some metabolic pathways for many critical functions, contingent events are massive, immediately up through developmental canalization, growth marking diverse biotic and geophysical processes, allometries, bilaterally symmetric organs, and like the “large body” impacts recorded at the macroscopic regulatory features of individual K-T boundary extinguishing the dinosaurs and physiology, a hierarchy of learning and gener- most other species, giving small mammals a alization processes, to various mechanisms in chance. No surprise: such a massive cause should social groups, breeding populations, ecosystems, have had far-reaching effects. and trophic levels. Multiple realizeability, so But most evolutionary “contingencies” start favored by philosophers of psychology, has its small—the luck of the draw for George Wil- primary significance as noise tolerance to evolu- liams’s plankton or single-base mutations ini- tionists. Noise tolerance produces neutrality or tiating selective cascades of layered exaptations near-neutrality among evolutionary variants, with divergent consequences. Oxygen produc- allowing drift among them and increased prob- tion—metabolic byproduct in ancient plants— ability of loss of any particular variant. (Most presumably started small, but hardly any contin- “neutral” mutations are not intrinsically so, but gency has had broader or greater consequences owe their neutrality to this class of mechanisms.) for evolution, by spreading as these plants suc- So things captured here feed to (2). Even if neu- ceeded, and becoming a much larger process. As trality doesn’t lead to loss, it may lead us to miss this atmospheric poison rose in concentration, the variation—or to decide not to notice it (as oxygen was initially adapted to, then became uti- unimportant) in our accounts. lized almost universally throughout the animal 4. Optimization (or selection) also erases his- kingdom, driving an energetically richer metab- tory on evolutionary time scales if the optima do olism which now depends on it. This is prob- not stand still. Usually they do not. Over time, ably the best possible single case for Levins and changing adaptations to changing circumstances Lewontin’s (1983) argument that organisms con- gradually erode and reconfigure everything that is struct their environments as well as conversely. changeable (this point is robust, and arguable in Small contingencies that leave a mark in evo- different ways: contrast Lewontin’s 1966 “capri- lutionary history do so by becoming larger— cious evolution” with Van Valen’s 1973 “Red amplified by recurrent processes—for organisms, Queen”). But if the optima were static, the closer by reproduction.29 228 William C. Wimsatt

Cascading sequential dependencies also oc- way, it is probably not a bad first approximation cur in each individual during development— to say: generative entrenchment is why history commonly, but not exclusively through de- matters! scent processes in cellular reproduction.30 Not all So these systematic effects—the structure of of these are “contingencies.” “Generic” events dependencies in development—can also have (Kauffman 1993) or forces reflecting laws of systematic evolutionary consequences. Devel- nature (e.g., gravity or surface tension) can opmental processes are the source of these become GE’d—deeply utilized or “presumed” in patterns—for why contingencies can persist. So the design of adaptive structures (Wimsatt 1986; taking developmental systems seriously simulta- Schank and Wimsatt 1988). D’Arcy Thompson neously explains two of the most striking features (1917) saw in the rich similarity of animate and of organic life: inanimate natural forms the direct action of laws First, generative entrenchment explains the fre- of nature, and thought he could dispense with quent deep similarities in organic architecture of selection. But in most cases on the organic side he varying degrees of generality roughly mapping was seeing the interplay of selection with proper- phylogeny onto causal depth in developmental ties of matter and deep entrenchment—the gener- process in the life cycle of organisms. A cluster of ation of what Gri≤ths and Gray (1994) would related heuristic principles utilizing GE play classify as highly heritable nongenetic resources. major roles in inferences from development to Thus it is an error to suppose that generative phylogeny, and conversely. Haeckel’s “Ontogeny entrenchment and self-organization must be recapitulates phylogeny” is dead, but these sim- alternative competing explanations for deep reg- ilar inferential principles are alive and well, ularities of form. They are inextricably interdigi- though not unrestricted (Raff 1996; Wimsatt tated (Wimsatt 1986). 1998; see Rasmussen 1987 for some of the princi- But GE can also happen to arbitrary contin- ples and how they work; Nelson 1998 considers gencies, rendering them more or less context- issues of GE in phylogenetic inference in detail). dependent adaptive necessities—all the more Second, GE explains why some features that striking for their seeming arbitrariness. These seem bizarre, ine≤cient, or only arbitrary should reflect an evolutionary history of contingencies, show so much persistence, so much evolutionary of exaptation layered upon exaptation, a history inertia after chance or selection have kept them unique to, characteristic of, and divergent in dif- for a time, and other things have come to depend ferent lineages. This is the architecture of adap- upon them. It does not explain why they occur in tation. This creation of layered dependencies the first place. These marks of contingency are produces GE. Laws of nature, and generic prop- more quickly lost—never visible (as in the quan- erties of ensembles of systems can be expected tum transitions through which an ionization to show through a lot of noise—developmental might lead to a base substitution), or local and noise and evolutionary noise. But the vast major- fleeting (the predator from which it zigged ity of arbitrary small contingencies vanish with- instead of zagged and got away), or not so local out a trace, through one of the four mechanisms and not so fleeting but still not saved (changing given earlier. selective optima). So GE is essentially the only way that such An important part of seeing how GE can act as smaller contingencies have a reasonable chance an explanation here is to see that differential GE to be preserved over long stretches of macroevo- is a generic feature of phenotypes—of life cycles, lution. Drift alone will not su≤ce. We see evolu- and repeated assemblies. Differential GE is law- tion as a contingent process primarily because of like in character—both in its widespread applica- generative entrenchment. Or to put it another bility, and in the near inevitability of its conse- Generative Entrenchment and the DST Approach 229

quences. As something gets more deeply GE’d, ters’ rights) by population genetics. Some of the selection forces opposing changes in it become— most powerful explanations in evolutionary the- stochastically, but inevitably—more uncondi- ory arise from observations or assumptions tionally negative, more context independent. This about reproductive rates, body size, and genera- is because of the embeddedness of GE’d parts, tion times of different species. Fitness is closely the richness of their connections. The diverging connected with reproductive rates, and evolu- consequences lead to less localized (more holistic) tionary rates with generation time, and genera- effects. And things with very different degrees of tion times with body size. On our time scale we GE in the same repeated assembly can reliably can confidently predict that insects will rapidly and robustly—even if not certainly—be expected evolve pesticide resistance, bacteria will generate to change at different rates. That is a crisp pre- antibiotic resistance even faster, that there will diction. In these important ways any holism be new influenza viruses each year that threaten attributable to DST is not totally symmetric, not humans (even those who had it last year), that uninformative, not uninvestigateable, not experi- large fluctuations in prey numbers will render mentally intractible. comparably sized predators at least locally People have pointed to the importance of con- extinct. (Predictions of this sort are strikingly tingency in evolution, argued for it, presupposed robust and independent of detail for such com- it, but not tried to explain it—not as a property of plex phenomena: unlike the hypothetical and individual cases (left- versus right-coiling shells) usually falsified predictions of idealized models.) or periods (the great loss of Cambrian Bauplans) Because population genetic models abstract —but as a generic property of systems that can away from phenotypic process—body size, re- fix and build up richly textured tapestries of ac- productive rate, and generation time must enter cidents. GE does so. It does not explain their by assumption. This escapes notice because such origin, or early preservation. But the longer the information is close to our experience, and we contingencies persist, becoming more deeply do not derive it from our models. (Even a rocket generatively entrenched, and commonly, more scientist untutored in biology would know that widely distributed among descendant species, the flies reproduce like hell and that there are lots of larger a mark they have left on evolution, and the them.) So flies have lots of mutations, lots of larger the probability they will continue to leave generations, and lots of offspring for every one of marks. GE is also a sound and robust expla- each that we can muster. So they can always nation for why evolution should be so much run faster than we can—evolutionarily speaking. marked by contingency—and for why the con- These properties are close to and flow naturally tingencies persist. Unlike the leveling and ho- from developmental models. (Body size, repro- mogenizing processes attributed to the second ductive rate, and generation time are all related law of thermodynamics, evolution seems to leave phenotypic properties, and show strong correla- an ever more complex and filigreed history over tions across species which are relatively insen- time. Generative entrenchment is the primary sitive to all kinds of empirical details (see Bonner explanation for this difference. Note that in this 1965, chap. 2). Those who say that they don’t sense, evolution can be cumulative without being need DST do not realize how much they have progressive. filched from it already or from common knowledge of developmental processes. To see what kinds of things we might get from A Parting Shot Promising Later Broadsides development, let’s pick an area that has undergone massive recent elaborations without benefit It is time to reclaim something for DST that has of developmental insights. This is cultural evolu- been long and proudly held (but only with squat- 230 William C. Wimsatt

tion viewed through the lens of the selfish meme. from that distance, estimate their deceleration rates. In The closest meme theorists come to metazoan terms of their cornering coe≤cients, estimate how fast development (or any sort of multiunit complex- they will be able to go before spinning out on a 100- ity) is vague talk of “meme-complexes.” These yard radius track. Finally, assuming that they are all made of iron, run into a concrete wall at 60 MPH., and references are so little elaborated that one cannot all of their kinetic energy is turned into heat, how much tell whether meme-complexes are supposed to hotter do they get? How much should their front correspond to multilocus genotypes, develop- bumpers collapse? List any assumptions you make as mental complexity, ecosystems, all of the above, you go along. For cases in which the published answers or none of the above. One of many problems are different from your calculated ones (essentially all with “memetics” (if there is any such beast) is of them), can you give plausible reasons for the dif- that in focussing on autocatalytic ideas of self- ferences? You may need additional information for replication, intuitions from development have some problems, so tell me what it is and how you would been almost totally left out. find out. How will you evaluate your sources? Which I will not provide such a theory here (though I problems can’t be solved with plausibly available information at all? Invent a problem of your own of compa- am working on one). I sketch a simple model of rable richness utilizing Newton’s laws (automotive cultural transmission, (Wimsatt 1999a; Callebaut design has many others) and solve it or tell why you 1993: 425–429, figure 9.2). Assume we are de- couldn’t for as many parts as possible. For two weeks veloping individuals who acquire cultural traits, from today, design a lab to investigate and illustrate to practices, knowledge systems, and skills over others a related phenomenon that you found particu- time, embedding them in and relating them to larly interesting, and think of what assignments you other generative things we have acquired. Usu- would give them to take home. ally (save for very simple things) we pass on these This made-up time slice has predecessors and generative structures not by passing on the whole descendants in the same lineage, starting with developed thing (the adult phenotype as it were), arithmetic and the lever, and ranging up to parti- in one bolus, but by passing on only a few gener- cle physics and chaotic dynamics, and is loaded ators, (heterocatalytically characterized gene-like with sequential dependencies for necessary skills. things). We provide—as individuals, and through Pick the “gene-like” generators as Newton’s laws our institutions—the resources to interact with in this simple story of a developing condensed- what we have already passed on, acquiring fur- matter physicist (there are many other generators ther generators, to develop a mature understand- as well). It is clear how many and varied the con- ing and working knowledge in the recipients. So ceptual, material, and social resources these stu- imagine a just-so story greatly truncated, collaps- dents will have to draw upon to complete the ing into a stage what would be months of sequen- assignment, and how much richer their knowl- tial assignments, a high school physics teacher edge of physics will be than the mere statement of gives his already substantially prepared students these laws, even though many textbooks act as if the following extended assignment: all of the information “followed from” by the So copy down Newton’s three laws, and work in groups axioms. Consider how much development and to solve the simple problems next to them on the board. integration of the resources is required for each You may ask me any questions. I’ll help, but I won’t tell successful answer, explained anomaly, or analy- you the answers. Then go home and buy an annual car sis of why no answer is possible. If this seems to magazine for this year—one with all of the specifica- have parallels with the gene-selectionist story, it tions in it. In terms of the horsepower and weights for should. But let me make some predictions about each, estimate the ratios of the 0–60 times for all new cultural evolution from a DST perspective— cars, and what proportion of the ideal values their actual acceleration times are. Given their stopping times turning only on reproductive rate and generation time, and not significantly utilizing the conserva- Generative Entrenchment and the DST Approach 231

tive properties of generative entrenchment, which from high school and undergrad tutors through has particularly rich application in this case assistant professors to emeriti. (Or brown belts (Wimsatt 1987): through third-degree black belts running Tae Kwon Do clubs with members of different 1. Things that are not GE’d are paradigmatic degrees of sophistication, and skill-ranked tour- “naked memes” and should be capable of the naments.) Using a series of immature individuals highest horizontal transmission rates consistent to transmit simpler parts before they could trans- with the channel characteristics. These are cases mit the mature theory or practice can substan- which r-selective meme theorists seem to like the tially increase the reproductive rate (a “master” most. (They also reflect—in themselves—the multiplies his presence with younger assistants— least structure, and are likely to most anger social see the workshops of major Renaissance artists) scientists when offered as examples of “culture”.) and shorten the effective generation time by 2. The more complex the cultural trait, the more finding simpler tasks through which an individ- slowly it should spread (it is more complex to ual can produce useful labor. In addition, ap- learn and to teach, and there may be increased prentices learn by teaching, and any teacher is complexity in interfacing it with other elements likely to remain more interested if they are opti- of the culture.) mally challenged, teaching not too far below (or 3. If the trait is entirely new to the culture (so above!) their competency level. (See Csikszent- that no one knows much of what they need to mihalyi 1990, on “flow.”) So more are likely to master it), there should also be a longer lag before complete the course of instruction, more usefully, it can spread (because it takes time to teach and and with greater competence. to learn) and In addition, having more individuals who know parts of the complex practice introduces 4. it will be learned by a smaller number of indi- redundancy—not just as multiple copies might, viduals (who specialize for it) unless it becomes but because they are intelligent agents who “socially required.” together may be able to reconstruct, correct, or 5. All of these will tend to slow down the repro- improve pieces that they would not have other- ductive rate and decrease the fraction of the pop- wise as mere passive redundant copiers. Thus are ulation picking it up. The chance of “falling off born scientific communities, craft guilds, and the track” will increase even for those intending priesthoods. to complete it. Is this too far from biology? Notice that we These are heavy reproductive burdens for com- have just reinvented population growth with plex cultural traits. How can we escape these lim- overlapping generations and social structure— itations? Consider a complex cultural trait or for memes. (With an important difference, however: The memotype transmitted by immature related set of traits as a conceptual phenotype. If parents is also less “mature.” That’s why “popu- “cell lineages” in the phenotype differ in length, lation memetics” would be such a mess: We don’t and gene-like generators (transmissible genera- just track genes or genotypes regularly assembled tive parts) can be placed on shorter lineages, then from pairs of parents, but all of the developmen- transmission can be accomplished before the phe- tal stages of phenotypes—variously assembled in notype achieves full maturity—(defined as being different proportions from variable numbers of able to transmit the generators for the full com- “parents.”) For all that, the teaching of younger plex theory or set of practices). So then it is pos- apprentices parallels the “helping” behavior of sible to transmit generators to produce immature brown jays in Costa Rica. (Lawton and Guindon or partial phenotypes earlier, and one can have 1981; Lawton and Lawton 1985, 1986; Williams, conceptual overlapping generations, ranging 232 William C. Wimsatt

Lawton, and Lawton 1994). The brown jays who of the “quantitative” sciences. By contrast, prin- are older siblings capable of building nests and ciple (7) reflects the situation for first-language laying eggs do not make very good nests and do acquisition, and for second-language acquisition not have the complexities of parental care down if it is su≤ciently similar to the first, and begun too well, so their expected yield of successfully su≤ciently early. And, if so, the self that gets fledged offspring is not much above zero. But if organized should not be just the language ability they stay at their parents’ nest they learn how “on but the cluster of competencies through which the job” (those who have helped for a year do it emerges. These probably include a variety of better than those of the same age who didn’t) cognitive, conative, social, affective, and motor (Lawton 1985), while their parents raise more skills. In this sense, language ability is prob- related young than they would be able to other- ably no more self-organizing than genes are self- wise. Helpers sometimes “cheat” by laying an egg replicating. (It is a common reductionistic bias to of their own in their parents’ nest—getting a “free draw the boundaries of the system in the wrong ride”. But maybe that happens in cultural evo- place—always too small; see Wimsatt 1980b, lution too—like borrowing the family car for a 1997b.) Finally principle (8) is an observation date! that follows when the organization of any pro- I leave this speculative example to suggest duced system increases in complexity and is more general principles of adaptive design for structured or maintained at least partially along cultural evolution which come out of a develop- functional lines. And that is—taken as a whole— mental perspective. many more parts of a functionally organized structure are means than ends (Wimsatt 1997b). 6. If a cultural trait has a structure which is This is likely to be the most controversial thing GE’d, complex, and important, more elaborate I’ve said. But to make sense of it, we need to scaffolding should evolve to facilitate learning it. remember that activities can become autotelic And if the scaffolding can be used to learn other through learning and exposure—we can acquire things, the scaffolding in turn may become more tastes defining things as good-in-themselves. And entrenched than the original. (This latter is exap- that does not mean that they may not nonetheless tive elaboration.) also serve cultural ends. 7. Evolution will structure the organization of Finally, are any of the predictions of this sec- developmental programs (dispositions of re- tion surprising? Perhaps not. Most are already sources) so that assembly of GE structures will familiar. We had other ways of getting them become (and look) self-organizing, and self- already, or they were parts of our embodied com- maintaining reliably across the range of envi- mon sense—save perhaps for the last four. ronments normally encountered. (So GE and (We’ve had examples of all of them staring us in self-organization interact richly; see chapter 18.) the face!) What is more surprising is where the 8. Judgments of “importance” for most cultural predictions came from. And these will not be the traits will derive in large part from their genera- last. tive entrenchment in the production of other cultural traits. Acknowledgments Principle (6) roughly reflects the situation for learning mathematics, physics, or other complex Some of this paper is elaborated from my 1999a “acquired” traits. And mathematics—though a and 1999b. In thinking about developmental sys- di≤cult and GE’d subject matter in its own tems, I have benefited from long collaborations right—becomes scaffolding for all kinds of devel- with Jim Griesemer and Jeff Schank, more recent opment, learning and empowering of theory in all discussions with Julio Tuma and Jason Roberts, Generative Entrenchment and the DST Approach 233

and timely feedback from Susan Oyama and Paul of “the selfish gene,” though it would be foolish not to Gri≤ths. Schank’s work has been particularly use the rich structure arising from genetic specifications pivotal in developing many of these ideas into of the problem as well, when that is genuinely avail- richer and more testable forms. able. (Most commonly it is not.) So called multilocus traits are multidimensionally environmental too! An adequate genetic specification without genomic and Notes environmental context-dependencies seems (just contingently?) never to be at hand. 1. See Glassman and Wimsatt (1984); Rasmussen 9. Before (1993), e.g., in (1985), Kauffman presented (1987); Schank and Wimsatt (1988, 1998); Wimsatt and self-organization and selection as alternative causes. Schank (1989); Wimsatt (1981: 178–181, 1986, 1999a, In 1993, he adopted a weaker view, with selection 1999b); and Callebaut (1993: 331–334, 378–382, 385– and self-organization as complements (the parenthetic 386, 425–427). qualifications)—a view I have urged in print since 1986 2. Indeed, a closer look at any internal combustion (and in conversation much earlier). Kauffman had been engine would show nothing different: structures yield- treating genericity as a dichotomous variable, suggesting patterns of constrained motion which, with the re- ing the former. But it is a degree property (Schank and lease of imported sources of energy, are converted into Wimsatt 1988) so the latter seems inevitable. the ability to do work. 10. I argued (1974) that it is not that simple. Simon’s 3. This would be the wrong lesson to draw from argument ignores coevolution of subassemblies and Schaffner (1998). Another careful exploration of this must be transformed to utilize development. See also territory is Sterelny, Smith, and Dickison (1996). But Lewontin’s “quasi-independence” (additivity or near- both are more gene-centered in flavor than I prefer. decomposability for small or moderate changes in a 4. Sterelny, Smith, and Dickison (1996) see this thicket system that is globally epistatic), and “evolvability” of overlapping boundaries as a problem, which they (Lewontin 1978; Wimsatt 1981b; Dawkins 1989; apparently think is special to DST. I disagree. Some Kauffman 1993; Dennett 1995; Schank and Wimsatt kinds of systems, and problems require recognizing and 1998; Brandon 1999). dealing with multiple boundaries simultaneously, and 11. I do not give a su≤cient analysis of a developmen- multiple relevant ways of decomposing larger systems tal process; see Griesemer (2000a). into parts. See Wimsatt (1974, 1994) for analysis and 12. Some of these details are of crucial importance McHarg (1971) for lovely examples. though—such as that metazoans go through a single- 5. See Wimsatt (1976b, 1980b, 1981b, 1994, 1997a, cell stage. Without this, GE structures could look radi- 2001). The best recent review of this enormous litera- cally different. So there is still a lot of work to do to get ture is Sarkar (1998). to life-cycles as we know them. But there is some hope; 6. For qualifications see Brandon (1990: 6–9). See also see Grosberg and Strathmann (1998). Campbell’s (1974) “Blind variation and selective reten- 13. Darwin opens (p. 13) by saying that “the laws gov- tion.” And substitute “number of” for “causal ten- erning inheritance are quite unknown,” but then goes dencies to have” in (3), and drift is also covered. on to list a number of regularities variously described as 7. Focusing on replication as copying abstracts away “rules” or “facts” about inheritance. They appear not from both of these processes. The focus is on the in- to be described as laws because (1) they have exceptions formation transmitted, and not on the process which or (2) may be described in terms of tendencies, or (3) the Griesemer calls “reproduction” (2000a, 2000b). conditions under which they work is unclear. 8. Arguments from and about fundamentality are 14. This is a kind of unavoidable axiom of design almost inevitably misused. We must learn to deal with anchored in a combination of physically and evolution- complex systems simultaneously through a variety of arily generic principles which I will discuss elsewhere. theoretical perspectives and at various levels of organi- 15. This argument works only from a symmetric start- zation—usually without fundamentally privileging any ing point—and depending on conditions and the result- of them (Wimsatt 1994). But these five conditions give ing organization of the phenotype, for some distance a richer and more powerful view of evolution than that around it. It does not show that for any distribution of 234 William C. Wimsatt

selection coe≤cients over functional elements, “the rich while others are not—synonymous code variants for will get richer, and the poor, poorer.” Indeed, su≤cient- the same protein, and any pathways which are not lim- ly extreme distributions could be expected to get less iting in a redundant network. These are both discussed extreme. even in Wimsatt (1986). But there are many other pos- 16. Griesemer and I (1989) argue that misinterpreta- sibilities. Michael Wade and I will review them with tion of E. B. Wilson’s diagram of Weismannism earlier particular reference to Raff’s hourglass in the near this century contributed to the emergence of the mod- future. ern genetic reductionist interpretation of evolutionary 22. Hull urges that from the practice’s (or parasite’s) theory. point of view, the transmission is vertical not horizon- 17. Griesemer’s story (and those of Moss and tal (so that differences between cultural and biological Neumann-Held) are all richer than my proposal inheritance have been overstated). While correct, trans- (1981b) to define genes in terms of their heterocatalytic mission is still “horizontal” relative to the host’s function. Griesemer sees material transmission across biological descent. This situation arises with multiple generations as a crucial feature of reproduction—which intersecting channels or lineages of heredity. (We will may differentiate it from cultural transmission, with tend to treat lineages of the larger units as vertical, and further interesting consequences. He also discusses the the smaller horizontal, especially if the smaller things distorted representations of the biology induced by the also replicate faster.) “informational” or “replicator” conception of the gene. 23. But biology often outruns culture, even if our biol- 18. In cases where either genes or environmental fac- ogy does not. Bacteria may change properties (their tors alone were picked out, it would usually be because virulence) affecting their fitness and ours in days or explanatory contexts seemed to call for a focal object or weeks in the course of an epidemic—much faster than process—which would never constitute the total cause. insects evolve pesticide resistance, in turn notoriously 19. From notes of Boulding’s lecture at Max Black’s faster than our response. Not only can biology be fast, seminar in the Humanities, Science, and Technology but culture can be slow: Boyd and Richerson note program at Cornell University, 1974–75. (1985: 60) that Olduwan and Aechulian toolmaking practices were transmitted essentially unchanged for 20. “Lock in” is Brian Arthur’s (1994) term for the tens of thousands of generations—through major envi- same process; in economic models of developing tech- ronmental changes and migrations. nologies, city formation, and cases where adoptions of an action or technology reduces the cost of doing so to 24. Conditions for “dynamical autonomy” of higher- others—generating other adoptions. His simple models level phenomena consistent with reductive explanation are in effect microevolutionary frequency-dependent of those same phenomena are provided in Wimsatt selection models. They complement our microevolu- (1981, 1994). tionary simulations and macroevolutionary discussions 25. Whether satisficed or optimized (I favor the former of larger structures with bigger differences in GE. for theoretical and empirical reasons) design is subject 21. “Deep” modifications in biology are a key theoret- to (usually unspecified or underspecified) constraints. ical problem in paleontology and developmental genet- History includes any particularities of initial conditions ics, where GE arguments are commonly made. The in space and time. These could seem like two capacious problem is how to explain “Raff’s hourglass”—the wastebaskets that together capture almost anything. evolutionary large amount of early developmental variation in close- Perhaps, but history imposes additional ly related species of amphibia, echinids, and some other constraints of generational repetition, and size and time kinds of organisms. (Raff’s 1996 famous “direct devel- scales which are used in the argument. And as we will oping” sea urchins are evolutionarily recent but skip an see, GE gives history structure—a life on multiple scales entire and normally important larval stage.) The sim- simultaneously. plest GE models would predict monotonically declining 26. Evolutionary biologists use contingency differently variation at earlier developmental stages, a “cone” from modern metaphysicians, but not unlike Aristotle’s (Arthur 1997). So what is entrenched, what is not, and distinction between necessity versus chance or accident: how does what is not escape? Some early things are The opposition is not between “necessary” or “anal- obviously deeply entrenched—the genetic code, ma- ytic” versus “contingent” (as possibly false), but law- chinery for cell division, and the famous HOX clusters, bound or highly probable versus unlikely or arbitrary. Generative Entrenchment and the DST Approach 235

27. Human history need not turn on the GE of contin- Callebaut, W. (1993). Taking the Naturalistic Turn: gencies, though invocation of cultural norms or institu- How to Do Real Philosophy of Science. Chicago: Uni- tions (which are characteristically GE’d) would bring it versity of Chicago Press. in. Narrative explanations can serve other ends. In con- Campbell, D. T. (1974). Evolutionary epistemology. In texts of fixing responsibility we can pick any event, P. A. Schilpp (Ed.), The Philosophy of Karl Popper vol. large or small, and ask what lead up to it. If small, it 2, pp. 413–463. LaSalle, IL: Open Court. need not lead to or through other events with diverging Caporael, L. (1997). The evolution of truly social cog- consequences. (But if it did we would likely use them to nition: The core configuration model. Personality and provide organizing patterns—likely particular and Social Psychology Review 1: 276–298. dated—for our story.) Csikszentmihalyi, M. (1990). Flow: The Psychology of 28. Preserved, and also (re functional necessities) that Optimal Experience. New York: Harper and Row. our very conception of that kind of system now includes that feature—and returns to Gri≤th’s (1996) use Darwin, C. (1859). The Origin of Species. London: John of GE in his nonessentialist gloss on natural kinds. It Murray. also has obvious implications for teleosemantics. Dawkins, R. (1989). The evolution of evolvability. 29. This applies also to their detection. A found fossil, In C. Langton (Ed.), Artificial Life, pp. 201–220. even the only known instance of its type, was likely of Redwood City, CA: Addison-Wesley. an organism much more common in its time and place. Dennett, D. (1995). Darwin’s Dangerous Idea. New 30. Cellular descent is an important process for GE, York: Simon and Schuster. but only one of many. This was a flaw in Arthur’s ear- Glassmann, R. B., and W. C. Wimsatt. (1984). Evolu- lier presentations of GE-style theories. He drops this tionary advantages and limitations of early plasticity. in his (1998), recognizing, as we have urged—see In R. Almli and S. Finger (Eds.), Early Brain Damage Rasmussen (1987); Schank and Wimsatt (1988)—that vol. 1, pp. 35–58. Orlando: Academic Press. GE works as well for biochemical pathways (see Gould, S. J. (1997). Ontogeny and Phylogeny. Cam- Morowitz 1992) and “horizontal” interactions. Thus bridge, MA: Harvard University Press. circulating hormones produced by localized tissues Gould, S. J. (1989). Wonderful Life: The Burgess Shale spread consequences to other cellular lineages, to the and the Nature of History. New York: W. W. Norton. organism as a whole, and even to the social group Gould, S. J., and E. Vrba. (1982). Exaptation—a miss- (Stern and McClintock 1999). ing term in the science of form. Paleobiology 8: 4–15. Gray, R. (2000). Selfish genes or developmental sys- References tems? Evolution without replicators and vehicles. In R. Singh, C. Krimbas, J. Beatty, and D. Paul (Eds.), Arthur, B. W. (1994). Increasing Returns and Path Thinking about Evolution: Historical, Philosophical, and Dependence in the Economy. Ann Arbor: University of Political Perspectives, pp. 184–207. Cambridge, Cam- Michigan Press. bridge University Press. Arthur, W. (1997). The Origin of Animal Body Plans: A Griesemer, J. R. (2000a). Reproduction and the reduc- Study in Evolutionary Developmental Biology. Cam- tion of genetics. In P. Beurton, R. Falk, and H-J. bridge: Cambridge University Press. Rheinberger (Eds.), The Concept of the Gene in Devel- Bonner, J. T. (1965). Size and Cycle. Princeton, NJ: opment and Evolution, pp. 240–285. Cambridge: Cam- Princeton University Press bridge University Press. Boyd, R., and P. Richerson. (1985). Culture and the Griesemer, J. R. (2000b). Development, culture and the Evolutionary Process. Chicago: University of Chicago units of inheritance. In Proceedings of the Philosophy of Press. Science Association 1998 vol. 2 (Philosophy of Science, 67: S348–S368. Brandon, R. (1990). Adaptation and Environment. Princeton, NJ: Princeton University Press. Griesemer, J. R. (forthcoming). Reproduction in the Evolutionary Process. Brandon, R. (1999). The units of selection revisited: The modules of selection. Biology & Philosophy 14(2): Griesemer, J. R., and W. C. Wimsatt. (1988). Picturing 167–180. Weismannism: A case study in conceptual evolution. In 236 William C. Wimsatt

M. Ruse (Ed.), What Philosophy of Biology Is, pp. 75– ciation 1992 vol. 1, pp. 335–348. East Lansing, MI: The 137. Dordrecht: Martinus-Nijhoff. Philosophy of Science Association. Gri≤ths, P. (1996). Darwinism, process structuralism Moss, L. (1998). What Genes Can’t Do. Doctoral dis- and natural kinds. Proceedings of the Philosophy of sertation, Northwestern University. Science Association 1996 vol. 1 pp. 1–9. Morowitz, H. (1992). The Origins of Cellular Life. New Gri≤ths, P., and R. Gray. (1994). Developmental sys- Haven: Yale University Press. tems and evolutionary explanation. Journal of Philos- Neumann-Held, E. M. (1999). The gene is dead—long ophy 91: 277–304. live the gene! Conceptualizing genes the constructionist Grosberg, R. K., and R. R. Strathmann. (1998). One way. In Peter Koslowski (Ed.), Sociobiology and Bio- cell, two cell, red cell, blue cell: The persistence of a uni- economics—The Theory of Evolution in Biological and cellular stage in multicellular life histories. TREE 13(3): Economic Theory, pp. 105–137. Berlin: Springer. 112–116. Nelson, P. (1998). Common Descent, Generative Entren- Hull, D. L. (1980). Individuality and selection. Annual chment, and the Epistemology of Evolutionary Inference. Review of Ecology and Systematics 11: 311–332. Doctoral dissertation, University of Chicago. Hull, D. L. (1988). Science as a Process. Chicago: Uni- Oyama, S. (1985). The Ontogeny of Information. Cam- versity of Chicago Press. brdige: Cambridge University Press. (2d. rev. ed., Dur- Kauffman, S. A. (1993). The Origins of Order. Oxford: ham, NC: Duke University Press, 2000.) Oxford University Press. Raff, R. A. (1996) The Shape of Life: Genes, Develop- Keller, E. F. (1995). Refiguring Life: Metaphors of ment and the Evolution of Animal Form. Chicago: Uni- Twentieth-Century Biology. New York: Columbia versity of Chicago Press. University Press. Rasmussen, N. (1987). A new model of developmental Lawton, M. F., and C. F. Guindon. (1981). Flock com- constraints as applied to the Drosophila system. Journal position, breeding success and learning in the brown of Theoretical Biology 127(3): 271–301. jay. The Condor 83: 27–33. Sarkar, S. (1998). Genetics and Reductionism. Cam- Lawton, M. F., and R. O. Lawton. (1985). The breed- bridge: Cambridge University Press. ing biology of the brown jay in Monteverde. The Schaffner, K. F. (1998). Genes, behavior and develop- Condor 87: 192–204. mental emergentism: One process, indivisible? Philos- Lawton, M. F., and R. O. Lawton. (1986). Hetero- ophy of Science 65: 209–252. chrony, deferred breeding and the evolution of avian Schank, J. C., and W. C. Wimsatt. (1988). Generative sociality. In R. Johnston (Ed.), Current Ornithology entrenchment and evolution. In A. Fine and P. K. vol. 3, pp. 187–224. New York: Plenum Press. Machamer (Eds.), Proceedings to the Philosophy of Lewontin, R. C. (1966). Is nature probable or capri- Science Association 1986 vol. 2, pp. 33–60. East cious? Bioscience 16: 25–27. Lansing, MI: The Philosophy of Science Association. Lewontin, R. C. (1970). The units of selection. Annual Schank, J. C., and W. C. Wimsatt. (2000). Evolvability: Review of Ecology and Systematics 1: 1–18. adaptation, and modularity. In R. Singh, C. Krimbas, J. Beatty, and D. Paul (Eds.), Thinking about Evolution: Lewontin, R. C. (1978). Adaptation. Scientific Amer- Historical, Philosophical, and Political Perspectives, pp. ican 239(3): 156–169. 322–335. Cambridge: Cambridge University Press. Lewontin, R. C. (1983). The organism as the subject Simon, H. A. (1962). The Architecture of Complexity. and object of evolution. Scientia 118: 63–82. (3d ed. 1996.) Cambridge, MA: MIT Press. Lorenz, K. Z. (1965). Evolution and Modification of Sperber, D. (1995). Explaining Culture. London: Behavior. Chicago: University of Chicago Press. Blackwell. McHarg, I. (1971). Design with Nature. New York: Sterelny, K., K. Smith, and M. Dickison. The extended Natural History Press. replicator. Biology and Philosophy 11: 377–403. Moss, L. 1992. A kernel of truth? On the reality of the Stern, K., and M. K. McClintock. (1998). Regulation genetic program. In D. Hull, A. Fine, and M. Forbes of ovulation by human pheromones. Nature 392: 177– (Eds.), Proceedings of the Philosophy of Science Asso- 179. Generative Entrenchment and the DST Approach 237

Szathmáry, E., and Maynard Smith, J. (1997). From Wimsatt, W. C. (1995). The analytic geometry of genet- replicators to reproducers: The first major transitions ics: The evolution of punnett squares. (Unpublished leading to life. Journal of Theoretical Biology 187: manuscript.) 555–571. Wimsatt, W. C. (1997a) Aggregativity: Reductive Thompson, D. W. (1917). On Growth and Form. Cam- heuristics for finding emergence. In L. Darden (Ed.), bridge: Cambridge University Press. Proceedings of the Philosophy of Science Association Todd, P., and G. Miller. (1998). Biodiversity through 1996 Vol. 2, pp. S372–S384. sexual selection. In C. Langton (Ed.), Artificial Life V. Wimsatt, W. C. (1997b). Functional organization, Cambridge, MA: MIT Press. functional analogy, and functional inference. Evolution Turner, M., (1991). Reading Minds: The Study of and Cognition 3(2): 102–132. English in the Age of Cognitive Science. Princeton, NJ: Wimsatt, W. C. (1998). Simple systems and phylo- Princeton University Press. genetic diversity. Philosophy of Science 65: 267–275. Van Valen, L. (1973). A new evolutionary law. Evolu- Wimsatt, W. C. (1999a). Genes, memes, and cultural tionary Theory 1: 1–30. inheritance. Biology and Philosophy 14: 279–310. Williams, D., M. F. Lawton, and R. O. Lawton. (1994). Wimsatt, W. C. (1999b). Generativity, entrenchment, Population growth, range expansion, and competition evolution, and innateness. In V. Hardcastle (Ed.), in the cooperatively breeding brown jay, Cyanocorax Where Biology Meets Psychology: Philosophical Essays, morio. Animal Behavior 48: 309–322. pp. 139–179. Cambridge, MA: MIT Press. Williams, G. C. (1966). Adaptation and Natural Selec- Wimsatt, W. C. (2001). Re-engineering Philosophy for tion. Princeton, NJ: Princeton University Press. Limited Beings: Piecewise Approximations to Reality. Wimsatt, W. C. (1974). Complexity and organization. Cambridge: Harvard University Press. In K. F. Schaffner and R. S. Cohen (Eds.), Proceedings Wimsatt, W. C., and J. C. Schank. (1988). Two con- of the Philosophy of Science Association 1972 vol. 20, straints on the evolution of complex adaptations and pp. 67–86. Dordrecht: Reidel. the means for their avoidance. In M. Nitecki (Ed.), Wimsatt, W. C. (1976). Reductive explanation: A func- Evolutionary Progress, pp. 231–273. Chicago: Univer- tional account. In A. C. Michalos, C. A. Hooker, G. sity of Chicago Press. Pearce, and R. S. Cohen (Eds.), Proceedings of the Wright, S. (1931). Evolution in Mendelian populations. Philosophy of Science Association 1972 vol. 30, pp. Genetics 16: 97–159. 671–710. Dordrecht: Reidel. Wimsatt, W. C. (1980). Reductionistic research strategies and their biases in the units of selection controversy. In T. Nickles (Ed.), Scientific discovery—Vol. 2: Case Studies, pp. 213–259. Dordrecht: Reidel. Wimsatt, W. C. (1981). Units of selection and the structure of the multi-level genome. In P. D. Asquith and R. N. Giere, (Eds.), Proceedings of the Philosophy of Science Association 1980 vol. 2, pp. 122–183. Lansing, MI: The Philosophy of Science Association. Wimsatt, W. C. (1986). Developmental constraints, generative entrenchment, and the innate-acquired distinction. In W. Bechtel (Ed.), Integrating Scientific Disciplines, pp. 185–208. Dordrecht: Martinus-Nijhoff. Wimsatt, W. C. (1987). Generative entrenchment, scientific change, and the analytic-synthetic distinction. (Unpublished manuscript.) Wimsatt, W. C. (1994). The ontology of complex systems: Levels, perspectives and causal thickets. Canadian Journal of Philosophy Supp. vol. 20, pp. 207–274.

18 Developmental Systems, Darwinian Evolution, and the Unity of Science

Bruce H. Weber and David J. Depew

We take it as a presumption of this chapter that who appreciate the ecological orientation of the we are living at a time when the central problem older movement, would like to see evolutionary in theoretical biology is the integration of devel- theory put back into the developmentalist matrix opmental biology with genetics and evolutionary from which statistical Darwinism broke free in theory (Gilbert, Opitz, and Raff 1996). Even its the twentieth century (Salthe 1993; Webster and founders admit that developmental biology was Goodwin 1996; Ulanowicz 1997; see Depew and set aside for the time being when the Modern Weber 1995 and Gilbert 1991 for an histor- Evolutionary Synthesis was shaped (Mayr and ical discussion of this topic). For their part, Provine 1980).1 Surprisingly, orthodox neo- Darwinians of every persuasion predictably resist Darwinians (who assume that organisms are the these proposals as falsely implying that evolution primary targets and beneficiaries of selection) over phylogenetic time is progressive in some have made little progress on this issue since then. meaningful sense. They are keenly sensitive to the Accordingly, the field has been left to contes- fact that most public discourse about evolution- tations between molecular reductionists, who ary matters shows ill-digested traces of the “old” assume that the problem of development is sim- evolutionary synthesis, many of which are far ply the problem of turning structural genes on from benign in their consequences. and off, and those who identify in one way or Developmental systems theory (henceforth another with the contemporary “developmen- DST) is a welcome intervention in this unhappy talist challenge,” who are confident that what dialectic. Although those associated with it typ- genes do is far from the whole story.2 ically exhibit a marked animus against genocen- A tendency to drive this contestation to ex- tric versions of Darwinism, DST (which should tremes manifests itself whenever proposed revi- be construed as part of a more encompassing sions in received evolutionary theory are brought “developmentalist challenge”) has in recent years into play. At one extreme, molecular reduction- shown that it can present itself as a genuine ists favor “selfish” genes as units and beneficiaries Darwinian research program (Gri≤ths and Gray of Darwinian selection (Dawkins 1989). To do 1994). DST’s advocates begin by arguing that if so, they blur the distinction between molecular we take as seriously as we should the fact that and evolutionary gene concepts. At the other organisms develop we must regard them as epi- extreme, some developmentalists suspect that genetically constructed in each generation from integration between developmental biology and a large array of developmental resources, some evolutionary theory cannot take place until evo- heritable, some not, rather than as read out or lutionary theory has been wrested away not sim- printed out from a causally primary, quasi- ply from reductionistic versions of genetic preformationist genetic program (Oyama 1985). Darwinism, but from the Darwinian tradition as In consequence, genes are regarded by DST as a whole. Authors of the latter stripe are keenly only one of a number of developmental re- aware that the Modern Evolutionary Synthesis sources. Natural selection is construed as able to was not in fact the first evolutionary synthesis, act in principle on variations in any or all of these and so might not be the last. Post-Weismannian resources. The differential retention of these vari- Darwinism, of which the Modern Synthesis is a ants will manifest itself as fitness-enhancing development, broke with the assumed parallelism changes in aspects of the life cycle in a particular between ontogeny, ecology, and phylogeny that environment. was the common coin of evolutionists of all From the point of view of philosophers of biol- stripes at the end of the nineteenth century. Some ogy, among whose major functions is to identify contemporary developmentalists, especially those conceptual biases that retard the course of scien- 240 Bruce H. Weber and David J. Depew

tific work and to recommend new conceptual argument, we offer some putative extensions of frameworks calculated to get things moving DST’s main claims, some of which might well be again, DST is perhaps the most interesting devel- contested by DST’s advocates. We hope that our opment since the nineteen seventies, when philo- remarks will be taken in the purely exploratory sophers of biology intervened in the so-called spirit in which they are offered. units of selection controversy in order to suggest that Darwinism is not formally committed to selection at only one level. In part, that is because DST as a Darwinian Research Program DST itself intervenes in this very controversy in a way that is intended to put an end to it. For Susan Oyama’s The Ontogeny of Information is a DST’s champions infer from the presumptive seminal text for developmental systems theorists causal parity of all developmental resources that (Oyama 1985). In the process of deconstructing the replicator/interactor distinction, on which the the nature/nurture dichotomy, Oyama articulates units of selection debate has been predicated, is ill a view in which each ontogenetic cycle employs a conceived (Gri≤ths and Gray 1994; Gri≤ths and set of largely, but not exclusively, heritable devel- Knight 1998). When phenotypic traits are con- opmental resources. These are reconstructed in strued as developmental resources it can be seen each generation in what Maturana and Varela that they are as much replicators as interactors. call an “autopoietic” process—a self-organizing Conversely, when the full extent of the depen- process that does not rely on a central informa- dence of genes on other developmental resources tion source (Maturana and Varela 1980). DST, comes into view, they lose their supposedly priv- according to Paul Gri≤ths and Russell Gray, ileged status as replicators. emphasizes the self-organizing properties of the In this essay, our aim is to discuss an aspect of system of physical resources which occupy a DST that has not, to the best of our knowledge, stretch of space-time as a result of the activities of been fully explored by its most prominent ad- past generation. Self-organization reconstructs vocates. It is better able, we maintain, to forge the life-cycle (Gri≤ths and Gray 1997). To be closer links with other sciences than the ver- sure, the resources needed to complete each stage sions of Darwinism that it seeks to displace. In of ontogenesis include genes. But other chromo- particular, we believe that, in addition to being somal, cytoplasmic, and metabolic structures able to solve the specific problems within evolu- (some of which are heritable in their own right), tionary theory to which it overtly addresses itself, as well as behavioral, environmental, and social DST allows evolutionary biology to be more sub- factors, play a causal role that should in no case stantively connected to the physical and chemical be subordinated a priori to genes. On this view, sciences below and to the much-contested evolu- functional information is the result of ontogeny tionary behavioral sciences above. In this sense, rather than ontogeny being the result of the DST is able to support reasonable intuitions transmission of information stored in a genetic about the “unity of science” by maintaining a program. Thus it becomes inappropriate to say productive mean between the extremes of auton- that genes “contain” information. omy for evolutionary biology advocated most We may place research programs that flow vociferously by Ernst Mayr (see Mayr 1988, for from this guiding insight within a much longer example) and the reductionist impulses of gen- lived research tradition, namely the epigenetic ocentric Darwinians (see E. O. Wilson 1998)— view of embryogenesis that by the eighteeenth both of which seem to us generally more effective century was vigorously contesting the preforma- as means of freezing existing science into place tionism that had flourished in the seventeenth. than of encouraging new bursts of cutting-edge Following a suggestion of Evelyn Fox Keller, we problem-solving. In the course of making this are even tempted to call Oyama’s proposal, and Developmental Systems and the Unity of Science 241

the DST research programs that it has inspired, Jablonka and Lamb’s Epigenetic Inheritance and “the new epigenesis” (Keller 1995). Reasonable Evolution: The Lamarckian Dimension (Jablonka Darwinians, such as Mayr, have always recog- and Lamb 1995). We will note some of their nized that there are traces of preformationism in results below. post-Weismannian Darwinism. The whole or- The claim that DST can be construed as a ganism may not be antecedently encased within Darwinian research program achieves its greatest the sperm or egg, but the information for pro- salience against the background of renewed inter- ducing the organism is. These preformationist est in nongenetic forms of heritability (Gri≤ths traces are most visible in genocentric forms of and Gray 1994, 1997). Gri≤ths and Gray implic- Darwinism like that of Dawkins, with its stress itly ask why we should believe that the existence on “immortal replicators.” Viewed in the light of of heritable developmental resources other than DST’s strongly epigeneticist approach, however, genes should in any significant way conflict with one can find traces of preformationist rhetoric the theory of natural selection as such, even if even in Mayr, against whose notion of a genetic they do conflict with certain versions of that the- program Oyama protests. In rejecting all prefor- ory. After all, heritability is simply “a measure of mationist elements, DST reverts, in fact if not in how well the state of the parent predicts the state name, to the original meaning of the term epige- of the offspring” (Sterelny and Gri≤ths 1999: 35). nesis, which was lost under the impact of post- If you hold an environment constant enough you Weismannian thinking, when it came to refer to can increase the heritability of traits no matter all processes in forming the phenotype other than what is happening with the genes. Working from genetic transcription and translation. The older this simple recognition, Gri≤ths and Gray do not meaning of epigenesis, which runs from Aristotle criticize standard versions of genetic Darwinism through Harvey to an array of later eighteenth- in the manner of Darwinian “pluralists,” who century biologists, holds that an organism devel- multiply the units and levels at which natural ops through a process in which the proximate selection can work, or those who limit its adap- cause of each step is the total set of interactions at tive power by assigning a large role to phyloge- the immediately preceding state, starting with the netic constraints, pure chance, or mechanisms of procreative act, moving through the differen- nonadaptive natural selection. On the contrary, tiation and articulation of physical and psycho- they appear to be rather strong adaptationists, logical traits, and ending with the initiation of working for the most part at the traditional or- another life cycle. It is to just such a robustly epi- ganismic level. With a little effort, their view can genetic conception that DST adverts. even be read as a defense of orthodox organism- The antiepigenetic notion that genes play a level selection against both genocentric versions causally dominant role in evolution has been of natural selection and the fashionable “expan- encouraged by the view that other aspects of sion” of Darwinism to countenance both higher ontogeny are not in fact heritable. This is, how- and lower units and levels of selection. For the ever, false. Prions, for example—proteinacious suborganismal and superorganismal traits that particles that are responsible for “mad cow” and have led others rather promiscuously to postulate other scrapie diseases—are heritable, but are in the existence of “superorganisms” above and no way coded for by DNA or RNA, as the Nobel “selfish genes” below are, whenever possible, laureate Stanley Prusiner has demonstrated. reconceived as resources for the generation- One may be loath to call a prion a developmen- by-generation construction of organisms. Varia- tal resource. But a comprehensive inventory of tion in these resources drives adaptive natural heritable, nongenetic factors that are indeed de- selection.3 velopmentally relevant, including cytoplasmic This is possible in part because of an ontologi- inheritance, has been collected and assessed in cal dimension of DST. Organisms are viewed by 242 Bruce H. Weber and David J. Depew

DST’s proponents as self-organizing processes a predictable property of certain kinds of self- rather than as discrete, hard entities on which organizing systems, would eventually allow “forces” impinge. Developmental resources often Darwinism and developmentalism to be brought lie beyond the traditional boundary of the onto- into a new unity. We also suggested that the logically hardened organism—in the environ- developmental and self-organizational aspects of ment, for example, and especially in that part of organisms are rooted in the prebiological fact the environment of some species that can be that organisms are autocatalytic chemically dissi- called cultural. In consequence, DST’s proposed pative systems before they are anything else, and extermination of the preformationist elements that autocatalytic chemically dissipative systems latent within post-Weismannian Darwinism sug- are by their very nature self-organizing devel- gests that the centrality of organismic selection opmental systems. What we did not have, how- in traditional Darwinism can be preserved partly ever, was an argument showing how Darwinism because the boundaries of what counts as a should be construed in a way that is sensitive to “part” become productively blurred when onto- the self-organizing developmental nature of liv- geny is viewed as a process. ing systems. And what we especially did not have was an argument showing that the most problematic aspects of contemporary Darwinian Developmental Systems as Autocatalytic theory could be resolved by just such a recon- Dissipative Structures struction. That is what we found by reading Gri≤ths and Gray and rereading Oyama. In sum, When we first read Gri≤ths and Gray’s 1994 what we have learned from DST’s proponents is paper we had the experience Huxley reports him- what follows from the fact that organisms are self as having had when he first encountered self-organizing developmental systems. What we Darwin’s theory: How very stupid of us not to hope to offer them is some sense of why organ- have thought of that. For in Darwinism Evolving: isms are self-organizing developmental systems in Systems Dynamics and the Genealogy of Nat- the first place—and why, in our view, natural ural Selection we had, like Oyama, Gri≤ths, and selection properly so called can arise only in such Gray, insisted that organisms are developmen- systems. tal processes (Depew and Weber 1995). Like Our answers to these questions reflect the fact Oyama, Gri≤ths, and Gray too, we had thought that one of us is a biochemist who has worked on of ontogenesis as self-organizing rather than the evolution of proteins and the origins of life. as causally manipulated by genes. Moreover, Living things, whatever else they may be, are we had expressed regret that the developmen- open systems, pumping matter and energy into tal and self-organizational nature of organisms themselves and dissipating it in degraded form had not yet been well integrated with Darwin- to their environments. As such, organisms are ian theory, leaving the field to antiselectionist what Prigogine called “dissipative structures” developmentalists. (Prigogine 1980). The inherent tendency of dissi- Our own response to this unsatisfactory situa- pative structures to increase, even to maximize, tion was to argue in some historical detail that in their dissipative rate is linked with their ability to its various stages and incarnations Darwinism build better dissipative pathways, in the form of had always benefited from reconceiving natural more e≤cient internal structures, which enable selection in terms of models drawn from various them to make other entities pay their entropic sorts of systems dynamics, and to express our debts. The most effective way of building such hope (based on this weak induction from the structures in chemical systems is by means of past) that new models of natural selection, in autocatalysis. A chemical reaction that produces which that phenomenon comes to be viewed as Developmental Systems and the Unity of Science 243

a substance that can facilitate the production of proliferation. In earth’s early environment, ac- more of the original reactant will show a rapid cordingly, an enormous competitive advantage amplification of the concentration of that sub- would have accrued to entities that could make stance. This is called an autocatalytic cycle. Or- use of templates, however crude, to enhance the ganisms are chock full of autocatalytic cycling— reliability, speed, and redundant pathways of indeed, within the environments to which they their autocatalytic activity. At the dawn of life are deeply coupled, they simply are autocatalytic this probably occurred through an intense coevo- systems of a certain sort. Moreover, not only are lution between proteinoids, which have long been cells highly complex autocatalytic cycles of ener- known spontaneously to form in chemical en- gy use and dissipation, but, in their developmen- vironments such as those that obtained early in tal trajectories, the organisms that are composed earth’s history, and nucleic acids, which may of such cells show precisely the predictable pat- have formed either separately or, alternatively, in terns of bifurcation and differentiation that we the presence of polypeptides within amphiphi- observe in all complex dissipative structures, such lic vesicles. Energy-capture mechanisms would as the Belousov-Zabotinsky reaction. One can have driven polymerization of amino acids and see the traces of such patterns in the developmen- nucleotides in a chemically bounded space of tal trajectories of organisms like slime molds, amphiphiles to produce “generic” proteins and whose pattern of aggregation and dispersal nucleic acids. These generic macromolecules proceeds along, and takes adaptive advantage of, would have had weak, but broad catalytic activ- dynamical paths already marked out by self- ity such that a complex array of chemical reac- organization (Bonner 1993). tions would be possible within such protocells The fact that living things are autocataly- (Weber 1998). tic dissipative systems constitutes, we believe, the An ensemble of protocells so characterized root cause of the fact that organisms are self- can be said to exhibit weakly heritable variation, organizing developmental systems. Accordingly, even in the absence of genetic mechanisms. As although we agree with advocates of the auton- Stuart Kauffman’s simulations have suggested, omy of evolutionary biology that some physical increased coupling of such a system of autocat- and chemical aspects of organisms are more or alyzed reactions would have led to catalytic clo- less irrelevant to their evolutionary histories sure (Kauffman 1993). Catalytic closure means (Burian and Richardson 199l), we do not think that every member of an autocatalytic set has at that all of them are. On the contrary, we think least one of the possible last steps of its formation the chemical and biochemical underpinnings of catalyzed by some other member of the set. There organisms, which in turn rest on the thermo- would have been increased rewards for catalytic dynamical parts of physics, are highly relevant e≤ciency, and thus selection pressure for a better to the evolutionary process. fit between the sequence space of a given macro- The potential relevance of chemical self-organ- molecule and the catalytic task space with which ization and physical dissipation to selective it roughly overlaps. Under selection pressure for processes arises because, in addition to sponta- more e≤cient autocatalytic cycling, a better fit neous differentiation and bifurcation, chemical between the sequence space of a macromolecule autocatalytic cycles themselves already exhibit a and its catalytic function space would have certain sort of selection—not natural selection occurred. That step might have taken place in the for fitness, to be sure, but what we will call following way: A tightening of the interactions “chemical selection” for autocatalytic e≤ciency. between nucleic acids and proteins would have The differential rates at which an autocatalytic led to a division of labor between them, one help- cycle is completed, for example, entail differential ing more reliably to ensure generation by genera- 244 Bruce H. Weber and David J. Depew

tion reconstruction, the other carrying out the all and only entities that exhibit differential reten- bulk of the catalytic tasks. As patterns of increas- tion of heritable variation (Lewontin 1980). But, ingly reliable and robust autocatalysis emerged, a as Lewontin himself admits, natural selection, transition from protocells to true cells, and hence formally defined by the three conditions of varia- to life properly so called, would have ensued. tion, heritability, and differential retention, is not Thus life would have emerged by way of a pro- su≤cient for producing an adaptation. What is cess we call “chemical selection,” a process that, needed in addition is continuity and directional- in its very conception, is inseparably linked to the ity maintained over a good deal of transgenera- self-organizational and differentiating tendencies tional time in an environment made constant of autocatalytic dissipative structures and to the enough (in part through the activities of organ- thermodynamical imperatives that underlie them. isms themselves) to add up to something useful This approach to the origins of life bears on a (Sterelny and Gri≤ths 1999: 31–38). We agree number of contested issues in both theoretical with DST’s advocates, who, perceiving this, biology and the philosophy of biology. For one argue that transgenerational continuity and thing, it constitutes a critique of “magic mole- directionality of the required kind can be assured cule” thinking, according to which life emerged only when ontogeny is construed as the reliable once, and by accident, in the form of a statisti- reconstruction in each generation of a range of cally fluky RNA molecule that thereupon pro- developmental resources. But we are also inclined ceeded to decorate itself out with survival to think that this process itself depends on the machines. Because that sort of thinking lies at self-organizational aspects of ontogenesis, and the basis of “selfish gene” theory, our rejection that these in turn rely substantially on autocat- of “magic molecule” thinking is intended as a alytic processes at the level of matter and energy critique of “selfish gene” theorizing generally. transformation. From this “bottom-up” point of We see the basic fallacy of this approach as its view, in which it is construed as an emergent tendency to push the notion of natural selection phenomenon, natural selection is presumptively down into the very sorts of processes from which adaptive. For adaptive natural selection is what natural selection itself must arise. Thinking fur- realistically accounts for the emergence of the ther along these lines, we noticed that our ac- phenomenon of natural selection in the first count of the emergence of life by way of chemical place; merely formal natural selection, which selection also constitutes a critique of the rather may seem primary from a purely definitional promiscuous, substrate-neutral conception of na- point of view, and only contingently related to tural selection that currently flourishes among adaptation, appears from a more realistic view as Darwinians. We believe that natural selection, a departure from the adaptive norm. properly so called, is not, at least in the first in- It was at this point that our thinking began to stance, an explanatory model, or a universal me- link up more substantively, if also more problem- chanism, or an “algorithm,” as Dennett calls it atically, with that of DST’s advocates. For it (Dennett 1995). Nor do we believe that it is more would seem that we are making a stronger claim or less indifferent to the kinds of materials and than Gri≤ths and Gray. They contend that entities on which it works. On the contrary, nat- developmental systems can be, and indeed are, ural selection is itself an emergent natural phe- produced by natural selection. We contend in nomenon4—a phenomenon that arises in, and addition that anything that comes into being by properly applies only to, the chemically autocat- natural selection, considered as a real phenome- alytic and physically dissipative systems of which non, must be a system that develops. We main- we been speaking (Depew and Weber 1998). tain, that is to say, that only the developmental It is one thing to say, with Lewontin, that nat- systems identified by DST contain a rich enough ural selection, formally considered, ranges over notion of heritability to sustain the kind of selec- Developmental Systems and the Unity of Science 245

tion that can properly be called adaptive natural Our use of the terms “bounded” and “informed” selection, as distinct from chemical-autocatalytic has caused some squinting on the part of DST’s selection below and certain conceptions of cul- proponents, who suspect that talk about the tural selection above. It does not, of course, fol- boundedness of organisms encourages too entita- low from this that natural selection is reducible to tive a view of the ontology of living things, and chemical selection. What follows is that natural hence obscures the processive, constructive view selection presupposes, and is uniquely predicable of ontogeny that DST supports (Gri≤ths, per- of, systems in which chemical selection, as we sonal communication). Similar worries arise in have described it, has already been at work— connection with our use of the term “informed.” and that cultural selection does not depend on For it might suggest the encoding of heritable a separate process working on cultural repli- information in a genetic program. cators (“memes”), but is an integral part of the We are reasonably sure, however, that this coupling between organisms and aspects of their worry can be assuaged by clarifying the point environments. of view from which we take organisms to be This way of situating the argument provides bounded and informed. Organisms are bounded additional support for one of DST’s most salient in contrast to more loosely coupled, prebiotic criticisms of the selfish gene hypothesis. DST’s autocatalytic systems. For the latter are not advocates have noted a strange anomaly in the- su≤ciently buffered from aspects of their envi- ories that treat genes not only as the units over ronment to count as genuinely living systems.5 which the selective “algorithm” ranges, but as the This does not in the least mean or imply that chief beneficiaries of this iterative process. Be- all of the developmental resources organisms use cause on this account they are merely repositories to reconstruct themselves are located within of transgenerational information, genes seem on their monadic boundaries—boundaries that are this view to have no biological function at all erected in the received view of natural selection (Gri≤ths and Knight 1998; Agar 1999)! The less by empirical facts than by an inappropriately absurdity of this becomes especially clear when atomistic conceptual apparatus. Nor does our one views organisms in the way we do—as self- use of the term “informed” imply that genes organizing, autocatalytic dissipative systems. For “contain” information that is supposed to be in this case one can readily enough surmise that semantically packaged within DNA and RNA, the biological function of protein-coding sectors which is then “opened up” and “expressed.” of molecules like RNA and DNA is to stabilize Genes, on our view, perform a stabilizing func- and buffer developmental cycles by enhancing tion, as noted above. This increases the degree of their ability to repeat more reliably in new auto- correlation between one generation and the next. catalytic cycles what has been effective in earlier The information associated with these correla- ones. Genes are, metaphorically at least, catalysts tions is of necessity distributed over the entire of catalyzing reactions. process, not encoded in informational packets Bearing this picture in mind, we defined organ- and read out into matter by way of a ghostly isms not only as autocatalytic dissipative systems, computerlike program. but as “bounded and informed” autocatalytic dis- It is possible that DST’s advocates will greet sipative systems (Weber and Depew 1996). For our proposal to root natural selection, devel- the catalytic effect of genes results in, and is ad- opment, and self-organization in prebiotic pro- aptive for, enhanced reliability in the generation- cesses as imposing unnecessary baggage on them by-generation construction of highly integrated in their efforts to gain a foothold against the re- autocatalytic systems, and hence is responsible ceived view of genetic Darwinism. It is di≤cult for the increased correlation between the traits of to argue at one and the same time for a new view parents and offspring, ancestors and descendants. of the “units” on which natural selection works 246 Bruce H. Weber and David J. Depew

and against the autonomy of biology stance that terns, membranes, and organelles, as well as ex- often accompanies the standard picture, espe- tracellular factors such as the cellular matrix, cially because the latter also serves to mark off developmental and environmental signals, and the turf of evolutionary biologists in the discur- behavior. Taken collectively, Jablonka and sive environment of academic life. Accordingly, it Lamb call these “epigenetic inheritance systems” might be thought that our proposal holds DST’s (EISs). The net effect of EISs is to provide for arguments about the units of selection hostage to natural selection a wider range of heritable vari- quite independent arguments against the autono- ation than that provided by genetic variation mist stance. We certainly hope this is not the case. based on nuclear DNA alone. We cannot, however, repress the passing thought It might be said by proponents of strongly that treating developmental systems as emergent genocentric versions of Darwinism, who assume from autocatalytic dissipative systems by way of that information must be stored, read out, and a linked series of types of selection might actually indeed be the chief beneﬁciary of natural selec- help DST’s advocates make their views about the tion, that these phenomena are too weakly and units of selection controversy more persuasive. transiently heritable to have much of an effect on For when one looks at the topics of boundedness natural selection and evolution. It is just here, and information from the more “bottom-up” however, that one must guard against one’s con- perspective we are advocating, DST will not ap- ceptual framework and ontology getting in the pear merely to be negating the received Dar- way of the facts. Admittedly, such variations winian view by heavily policing the use of terms might not be taken up by natural selection when like informed and bounded. The best defense is natural selection is construed as occurring in sys- a good offense; and we suggest that in this case tems that do not in any marked or essential way the best offense is a more physically and chemi- exhibit strongly nonlinear forms of feedback, cally rooted conception of organisms as develop- including autocatalysis, within a self-organizing mental processes, a view that will reveal clearly developmental system. In the standard view of the biological functions that genes play in rela- natural selection positive feedback is more or less tively bounded and reliably reconstructable absent. Thus one will be tempted to cry out for an developmental systems. In the following section, immortal, self-creating replicator, as Dawkins we will try to support this claim by exploring the does (Dawkins 1989), or for mechanisms that can potential relevance of our view to nongenetic “pump up” the otherwise plodding pace of adap- forms of inheritance. tation, as Dennett does (Dennett 1995). But, as we have argued above, it is precisely in systems that do exhibit self-organization through auto- Epigenetic Inheritance and the Darwinian- catalytic positive feedback that natural selection Lamarckian Divide is at work. That in turn makes it much easier for otherwise weakly heritable variants to be taken We stated above that DST’s ability to present up and ampliﬁed by natural selection. For such itself as a Darwinian theory has been stimulated variants to count in the process of adaptive natu- by empirical recognition that quite concrete ral selection, all that is required is that they be resources other than genes vary and are heritable. reliably reconstructed in cooperation with a host Much of the evidence for these forms of heritable of other interacting resources in each epige- variation has been summarized by Jablonka netic cycle. Accordingly, the potential relevance and Lamb (Jablonka and Lamb 1995; see also to evolutionary theory of epigenetic variation in- Jablonka and Lamb 1998). These include cyto- creases, at least theoretically, when DST is as- plasmic factors, organizing centers for basal sumed as a background theory; and it perhaps bodies and microtubules, DNA methylation pat- increases even more when our proposal to situ- Developmental Systems and the Unity of Science 247

ate DST within a context that highlights the (Flavell and O’Dell 1990). Sano and colleagues self-organizing nature of ontogeny through its have shown that changes of methylation pattern essential connection to chemical autocatalysis in rice result in heritable phenotypic change that is brought into play. remained stable over multiple generations, even We can be more concrete about this. The best- though the relevant DNA sequence was un- documented example of an EIS is DNA methy- changed (Sano et al. 1990). Thus variation in lation, which is one of the mechanisms of methylation patterns can, in principle, be subject chromatin marking. Chromatin marking is im- to the action of natural selection, and can lead portant for cell memory. It ensures that during to evolutionary adaptation by means of various development and differentiation a cell lineage mechanisms, of which Waddington’s genetic as- that has become a kidney cell, for example, con- similation is one of the best known (Wadding- tinues upon somatic reproduction to produce ton 1940). Admittedly, plants, which have late to kidney cells and not some other type of cell. This nonexistent germline segregation, seem to have example serves to remind us that genic reduc- a higher probability of having epigenetic inheri- tionists too often forget that chromosomes them- tance play a direct role in adaptive evolution. selves are not exclusively made up of the base Epigenetic factors in animals are more likely to sequences of DNA. Other molecular structures, affect early development or germ-cell lineages, such as various nucleoproteins and chemical when segregation of a germ line from somatic modifications of the DNA bases like methyla- processes is far less rigid. Among animals with tion, may affect gene expression. Admittedly, the complex behaviors, however, the presumptive control of the pattern of methylation during irrelevance of epigenetic variation can be com- development is not yet well understood. But what pensated for by increased sensitivity to environ- is clear is that such a pattern is preserved during mental signals and by behavioral variations. mitotic division through the action of the methyl The subtitle of Epigenetic Inheritance and transferase enzyme. Methylation has been shown Evolution, in which Jablonka and Lamb report to alter the gene expression of tissue-specific results such as those just discussed, is The genes, stage-specific genes, and silent genes, Lamarckian Dimension. Their use of the term whereas the expression of genes needed for basic Lamarckian is not intended to deny the evolu- metabolism of any cell is not so controlled. Dif- tionary power of natural selection (Jablonka, ferences in DNA methylation have led to differ- personal communication). It suggests instead ent chromatin structures, and in this respect they opposition to forms of Darwinism that attempt can be considered alternative phenotypes of a to assimilate the evolutionary gene concept to the gene. Kermicle has suggested that these alterna- molecular gene concept, thereby fetishizing and tive phenotypes of a particular locus be called exaggerating the causal role of naked DNA and “epialleles” (Kermicle 1978). Epialleles arise dur- RNA in such a way that the heritability of vari- ing normal chromatin alterations that occur dur- ants that do not fit this model, such as epialle- ing somatic differentiation. Although all the les, goes unrecognized or underestimated in its DNA sequences at that locus are identical, there potentially adaptive effects. So, although they are will be a number of epialleles within an individual not necessarily rendered heritable in the stereo- organism and in a population. typical ways that have led to the ill repute of This source of epigenetic variability is ex- Lamarckian evolutionary mechanisms by twen- pressed during development. It has been demon- tieth century Darwinians, it is useful to call at- strated in several cases, moreover, that epigenetic tention to these phenomena by using the term variation can be transmitted to offspring. Flavell Lamarckian. and O’Dell found seven epialleles for wheat high- We suspect, however, that this decision is not molecular-weight glutinin that were heritable rhetorically cost-free. Even such a benign use of 248 Bruce H. Weber and David J. Depew

the term Lamarckian testifies to the power of biology, but between a newly unified biology and genocentric Darwinians to monopolize the term the thermodynamics and chemical kinetics of dis- Darwinism in contemporary discourse, a power sipative structures. that has increased remarkably in recent years This proposal can easily provoke fears of re- through the popularity of works on sociobiology ductionism. Our general response to that fear is and evolutionary psychology. For in this discur- to distinguish between the unification of science sive context, the term Lamarckian is more likely and programs for reductionism. Reductionism to suggest violations of the Central Dogma of failed over and over again as the twentieth cen- Molecular Biology, which requires that infor- tury unfolded, provoking a strong reaction that mation must never run directly from proteins to prevails today in favor of incommensurability, DNA and RNA, than to call attention to in- pluralism, and even “the disunity of science” stances of epigenetic inheritance and natural (Dupre 1993). Yet it remains true that one of the selection that do not quite fit the genocentric most salient characteristics of science is its in- Darwinian stereotype. Jablonka’s and Lamb’s creasing unity. This fact has been ill served by primary purpose is to dispute genetic Darwin- philosophers of science who, by means of the ism’s empirical marginalization of nongenetic deductive-nomological account of explanation, forms of inheritance, and not to argue for too readily universalize the quest within mathe- stronger, more problematic, versions of Lamarc- matical physics for greater and greater reduction kism that puts it into direct opposition with the under a smaller and smaller range of laws. In Central Dogma (although results by Steele, actual fact, science as a whole becomes unified by Lindley, and Blanden that might suggest this an ongoing series of discoveries in which, as we have been reported [Steele, Lindley, and Blanden come to understand a given phenomenon better, 1998]). This distinction becomes a good deal we also come to see more clearly its relation to clearer when it is linked to the strongly epigenet- other phenomena. We wish to go on record, ac- ic view of ontogeny that Jablonka and Lamb cordingly, as supporting the unity of science in clearly assume. Seen in this light, their evidence this wide sense without supporting reductionism for epigenetic inheritance supports DST’s effort or the philosophy of science that gives it un- to expunge from evolutionary theory all traces due importance in actual processes of scientific of preformationism, and so to combine a pre- inquiry. Weismannian conception of epigenesis with a Unfortunately, genetic Darwinism has not yet post-Synthesis reconstruction of Darwinian nat- exhibited this pattern. In its well-motivated ural selection. DST offers a framework for the efforts to steer clear of reductionism, especially notion of epigenetic inheritance that puts these after the rise of molecular biology, the Modern phenomena in the best light. Evolutionary Synthesis has attempted to preserve its autonomy from physics and chemistry by cat- egorically denying the relevance of the physical DST, Developmental Psychology, and the Unity and chemical sciences to specifically biological of Science problems (Mayr 1988). Our support for proposals to reformulate Darwinism in terms of DST’s In this essay, we have suggested that develop- conception of ontogeny, and in turn to refer this mental systems, as DST describes them, can best conception to autocatalytic and dissipative probe viewed as emergent from autocatalytic dissi- cesses, contains an implicit plea to relax this pative systems. On this account, DST’s proposed stance. For our hope is that this conception of reunion of Darwinism with developmental biol- ontogeny will lead to hitherto unsuspected links ogy will lead not only to a closer relationship between biology and certain aspects of chemistry between evolutionary theory and developmental and physics without in the least suggesting that Developmental Systems and the Unity of Science 249

evolutionary biology is reducible to these sciences to see that learned behaviors, and the learning in the sense that biological processes are “nothing mechanisms that support them, are extremely ef- but” lower-level phenomena operating under cer- fective ways of assuring matches between organ- tain initial and boundary conditions. isms and otherwise shifting environments. These It is not only with sciences that on a conven- will, all other things being equal, be favored by tional ordering lie below evolutionary biology, adaptive natural selection. Learning mechanisms however, that we wish to urge a unifying stance, that rebuild behaviors in each generation by but also with some sciences above it. We are espe- using mimetic capacities to better match organ- cially interested in this respect in evolutionary isms to environments are clearly evident among developmental psychology, which is at present primates, including ourselves. One consequence in a state of intense ferment that we suspect can is the pressure we are able to put on other species be clarified by Darwinian versions of DST. For and the diminished pressure they are able to put in treating aspects of the environment as devel- on us. opmental resources, DST allows us to see what By stressing the role of learning capacities and genetic reductionism obscures, namely that fea- the behaviors they generate as developmental tures of cultural environments, rather than being resources, DST provides additional conceptual sites at which selection over memic units takes backing for sometimes overlooked aspects of the place of natural selection, are, in many cases, the orthodox, organocentric view of Darwinism. aspects of the process of organismic natural selec- This approach also sounds a greater note of nat- tion itself. uralism, and so far forth the unity of science. For Organocentric Darwinians like Mayr insist, it tends to break up the long-standing agreement quite correctly, that natural selection is a two- that cultural evolution is “Lamarckian” (and all stage process (Mayr 1988; see Brandon 1985, the speedier for that), while biological evolution, 1990). Even the best genes have no future unless properly so called, is under the control of ad- the organisms carrying them are successful aptive natural selection as measured by chang- enough in dealing with their environments to ing gene frequencies. The distinction between pass them on. Thus what Robert Brandon calls “memic” and “genic” selection is merely a ver- “environmental selection” in each generation sion of the latter understanding. For DST, on the must precede natural selection of genotypes over other hand, culture does not “arise from” nature, multiple generations (Brandon 1990). From this nor is it “constrained” by it. In view of the role account, it is easy to see that, all other things cultural elements can play in the generation-by- being equal, there will exist selection pressure for generation construction of organisms with reli- behavioral innovations that help organisms re- ably heritable traits, culture is part of nature. spond to contingencies in their environments. A relatively uncontroversial way in which Lewontin in particular has insisted that popu- DST-oriented Darwinians challenge standard lations commonly evolve in a fitness-enhancing understandings of the nature/nurture divide is to way by taking advantage of the considerable urge against most versions of “evolutionary psy- variation available in gene pools to evolve adap- chology” that generation-by-generation main- tations that enable organisms to construct and tenance of the genetic structures that facilitate reconstruct their environments so that existing learning should count as producing adaptations genotypes may remain as nearly as possible ad- no less than the selection pressure that was first apted to stabilized conditions (Lewontin 1983). It responsible for the emergence of the trait in ques- is precisely because the environment has been station (Sterelny and Gri≤ths 1999: 219). On this blized by the agency of organisms, in fact, that view, too strong a distinction between the adap- the traits whereby it comes to be so stabilized can tations by which the learning abilities of our be heritable. From here it is not much of a leap hunting and gathering ancestors first arose and 250 Bruce H. Weber and David J. Depew

the subsequent “exaptation” of these capacities view, because this putative phenomenon became for the uses of modern life should be resisted. too closely linked to debates about the general Many of the later uses of earlier capacities are e≤cacy of Waddington’s genetic assimilation. certainly adaptive. But, in view of the power of But it was also because the Modern Evolutionary natural selection to maintain, as well as to select Synthesis likes to think of fit genotypes as being for, genotypes that foster current fitness, there is sifted from the process of phenotypic selection in no reason to refuse to call these adaptive traits each generation. By contrast, the Baldwin effect adaptations, and far less reason to call them asserts that reconstruction of learned behaviors “exaptations.” On this view, the sociobiological at the phenotypic level can occur over multiple notion that our genes “hold us on a leash” more generations before any genotypic change begins or less evaporates. With it too evaporate many of to follow suit (which presumably does not hap- the motives that have provoked Darwinian op- pen at all in a large number of cases). The latter ponents of sociobiological thinking to rely too notion seems to us quite congenial to DST. Be- heavily on strong forms of group selection to cir- cause in the context of contemporary interest in cumvent what wrongly are seen to be the limita- the evolution of language and mind the Baldwin tions of individual-level selection. effect has recently made a big comeback as a We wish to end this essay on a note of query putative evolutionary “driver” or “lifter” among about this issue, posing two questions that sug- thinkers of widely varied stripe (Dennett 1995; gest how open the question of cultural evolution Michel and Moore 1995; Deacon 1997; Hinton still is, as well as our confidence that this disputed and Nowlan 1996), we wish to call upon DST’s topic can be illuminated by using a DST inter- advocates to help clarify the theoretical struc- pretive strategy. ture of Baldwin-type phenomena as well as their The first question has to do with the recent actual effects, if any. When, if reliably recon- revival of talk about the so-called Baldwin effect. structed in each generation, should we begin to First proposed in the 1890s as a way of recon- talk about adaptations? ciling Weismannian neo-Darwinism with some A related question has to do with cases that aspects of Lamarckism, the Baldwin effect gen- call into question the standing assumption that erally asserts that if an organism chances to adaptations produced by natural selection must exhibit a behavior that permits more effective always involve changes in nuclear gene frequen- interaction with its environment, and can pass cies. A case that raises this question has been that behavior along mimetically, then descen- reported in matrilineal species of whales—species dents of such individuals and populations will on whose members live with female relatives and average do better than their competitors, in turn form new groups when females and their young making it more probable that germline factors fission off from larger groups (Whitehead 1998). that promote the behavior, if they independently These species, it seems, have five times less diver- arise, will be more quickly and effectively taken sity in their mitochondrial DNA than humpback up and moved in an adaptive direction (Bald- whales or dolphins, which do not have such a win 1896). Even though, properly interpreted, social structure, while nuclear DNA, the pre- the Baldwin effect does not suggest, or at least sumptive site for standard Darwinian models require, that multigenerational phenotypic conti- of evolutionary change, remains unchanged. To nuity will induce the kind of genetic change nec- explain this, the investigator who has reported essary to support it, this notion was downplayed these results hypothesizes that cultural learning by the makers of the Modern Synthesis until it over many generations—the imparting of songs, became very recessive indeed, killed for the most migration strategies, foraging techniques, or part by the usual method of accepting it in theory baby-sitting tactics—may become adaptive in but denying in practice. In part, that was, in our ways that are evidenced by, but not caused by, Developmental Systems and the Unity of Science 251

a reduction in genetic diversity in mitochron- from molecular genetics, have advanced little beyond drial DNA, which hitchhikes on these changes the state that Morgan found them. (Gottlieb 1997: 98). (Whitehead 1998). Recall that mitochondrial 2. We take the term developmentalist challenge from DNA is passed exclusively through the female Schaffner (1998). Schaffner mentions Gottlieb, Gray, line and is not involved in adaptations except Gri≤ths, Bronfenbreener and Cici, Oyama, and for the biochemical functions carried out by Lewontin among others as partisans of this challenge. The heterogeneity of the views of those listed suggests mitochrondia themselves. Where, then, are the the pluralism of the developmentalist challenge. expected gene frequency changes that underlie 3. DST Darwinians do not object to group selection. this presumptive adaptation? Rather, they tend to offer support for construing group Standard versions of Darwinism can readily selection in the sense articulated and defended by David answer this question by assigning to whales the Sloan Wilson. In effect, they recode Wilson’s notion of so-called Lamarckian, or cultural, adaptive evo- trait groups in an interpretive framework that turns lution that has hitherto been exclusively attrib- such traits into developmental resources of individuals, uted to humans, or at least primates. That is not not ontological markers of entities (see Sterelny and unreasonable considering the communicative Gri≤ths 1999: 160–177). traits of whales and the connective power of their 4. The concept of phenomenon in play here is roughly brains. Assuming that it is a real phenomenon, that of Bogen and Woodward 1988. however, we think it merely dogmatic not to call 5. In this connection, our view departs from the wish of this a case of natural selection, especially because many anti-Darwinian developmentalists to extend the the process presumably goes on through a differ- notion of living system to ecosystems, and indeed to the ential retention of behavioral variants, either at entire earth, as in the Gaia hypothesis. individual or trait-group levels. Its status as a genuine adaptation, built up over generations by References fixation of heritable variation, can, we suspect, be a≤rmed by DST’s approach to natural selection. Agar, N. (1996). Teleology and genes. Biology and But if this is the case, DST’s innovative interven- Philosophy 11: 289–300. tion in the quarrels that have always attended the Baldwin, J. M. (1896). A new factor in evolution. multifaceted and still vital Darwinian tradition American Naturalist 30: 441–451. might have to confront a potential clash between Bogen, J., and J. Woodward. (1988). Saving the phe- its constructivist view of ontogeny and the as- nomena. Philosophical Review 97: 303–352. sumption that adaptive natural selection is Bonner, J. T. (1993). Life Cycles: Reflections of an Evo- always measured by gene frequency changes in lutionary Biologist. Princeton, NJ: Princeton University nuclear DNA. At the very least, that should be Press. interesting. Brandon, R. N. (1985). Adaptive explanations: Are adaptations for the good of replicators or interactors? In D. J. Depew and B. H. Weber (Eds.), Evolution at a Notes Crossroads: The New Biology and the New Philosophy of Science, pp. 81–96. Cambridge, MA: MIT Press. 1. Some think it had already been aside when T. H. Brandon, R. N. (1990). Adaptation and Environment. Morgan separated genetics from embryology (see Princeton, NJ: Princeton University Press. Gilbert, Opitz, and Raff 1996). Gottlieb has a more charitable view of Morgan, in which the separation was Burian, R., and R. Richardson. (1991). Form and order intended tactically and only for a time; “eventually in evolutionary biology: Stuart Kauffman’s transfor- Morgan believed that the gene would be incorporated mation of theoretical biology. Proceeding of the Philo- into the developmental process and its activity seen sophy of Science Association 1990 vol. 2, pp. 357–402. as reciprocally altered thereby” (Gottlieb 1992: 139). East Lansing, MI: Philosophy of Science Association. Gottlieb further comments that the current theories of Dawkins, R. (1989). The Selfish Gene. (2nd ed.) Oxford: differentiation, although informed by a wealth of data Oxford University Press. 252 Bruce H. Weber and David J. Depew

Deacon, T. (1997). The Symbolic Species. New York: Jablonka, E., and M. J. Lamb. (1998). Bridges between W. W. Norton. development and evolution. Biology and Philosophy 13: Dennett, D. C. (1995). Darwin’s Dangerous Idea: 119–245. Evolution and the Meanings of Life. New York: Simon Kauffman, S. A. (1993). The Origins of Order: Self- and Schuster. Organization and Selection in Evolution. New York: Depew, D. J., and B. H. Weber. (1995). Darwinism Oxford University Press. Evolving: Systems Dynamics and the Genealogy of Nat- Keller, E. F. (1995). Reﬁguring Life: Metaphors of ural Selection. Cambridge, MA: MIT Press. Twentieth-Century Biology. New York: Columbia Uni- Depew, D. J., and B. H. Weber. (1998). What does nat- versity Press. ural selection have to be like in order to work with self- Kermicle, J. L. (1978). Imprinting of gene action organization? Cybernetics and Human Knowing 5: 18– in maize endosperm. In D. B. Walsden (Ed.), 31. Maize Breeding and Genetics, pp. 357–371. New York: Dupre, J. (1993). The Disorder of Things: Metaphysical Wiley. Foundations of the Disunity of Science. Cambridge, Lewontin, R. C. (1980). Adattamento. The Encycl- MA: Harvard University Press. opedia Einaudi. Milan: Einaudi. Reprinted as Adap- Flavell, R. B., and M. O’Dell. (1990). Variation and tation in E. Sober (Ed.), (1984), Conceptual Issues in inheritance of cytosine methylation patterns in wheat at Evolutionary Biology: An Anthology, pp. 234–251. high molecular weight glutinen and ribosomal RNA Cambridge, MA: MIT Press. loci. Development 1990 Supplement: 15–20. Lewontin, R. C. (1983). Gene, organism and environ- Gilbert, S. F. (1991). Induction and the origins of devel- ment. In D. S. Bendall (Ed.), Evolution from Molecules opmental genetics. In S. F. Gilbert (Ed.), A Conceptual to Men, pp. 273–285. Cambridge: Cambridge Univer- History of Modern Embryology, pp. 181–206. New sity Press. York: Plenum. Maturana, J., and F. Varela. (1980). Autopoiesis and Gilbert, S. F., J. M. Opitz, and R. A. Raff. (1996). Cognition: The Realization of the Living. Dordrecht: Resynthesizing evolutionary and developmental biol- Reidel. ogy. Developmental Biology 173: 357–372. Mayr, E. (1985). How does biology differ from the Gottlieb, G. (1992). Individual Development and physical sciences? In D. J. Depew and B. H. Weber Evolution: The Genesis of Novel Behavior. New York: (Eds.), Evolution at a Crossroads: The New Biology and Oxford University Press. the New Philosophy of Science, pp. 43–63. Cambridge, Gottlieb, G. (1997). Synthesizing Nature-Nurture: MA: MIT Press. Prenatal Roots of Instinctive Behavior. Mahwah, NJ: Mayr, E. (1988). Toward a New Philosophy of Biology: Lawrence Erlbaum Associates. Observations of an Evolutionist. Cambridge, MA: Har- Gri≤ths, P. E., and R. D. Gray. (1994). Developmental vard University Press. systems and evolutionary explanation. Journal of Philo- Mayr, E., and W. Provine. (1980). Perspectives on the sophy 91: 277–304. Uniﬁcation of Biology. Cambridge, MA: Harvard Uni- Gri≤ths, P. E., and R. D. Gray. (1997). Replicator II: versity Press. Judgement day. Biology and Philosophy 12: 471–492. Michel, G. F., and C. L. Moore. (1995). Developmental Psychology: An Interdisciplinary Science. Gri≤ths, P. E., and R. D. Knight. (1998). What is the Cambridge, developmentalist challenge? Philosophy of Science 65: MA: MIT Press. 253–258. Moss, L. (1992). A kernel of truth? On the reality of Proceedings of the Philosophy of Hinton, G. E., and S. J. Nowlan. (1996). How learning the genetic program. Science Association 1992 can guide evolution. In K. Belew and M. Mitchell vol. 1, pp. 335–348. East (Eds.), Adaptive Individuals in Evolving Populations, pp. Lansing, MI: Philosophy of Science Association. 443–446. Reading, MA: Addison-Wesley. Oyama, S. (1985). The Ontogeny of Information: Devel- opmental Systems and Evolution. Jablonka, E., and M. J. Lamb. (1995). Epigenetic Cambridge: Cam- Inheritance and Evolution: The Lamarckian Dimension. bridge University Press. (2d rev. ed., Durham, NC: Oxford: Oxford University Press. Duke University Press, 2000.) Developmental Systems and the Unity of Science 253

Prigogine, I. (1980). From Being to Becoming: Time and Complexity in the Physical Sciences. San Francisco: W. H. Freeman. Salthe, S. N. (1993). Development in Evolution: Com- plexity and Change in Biology. Cambridge, MA: MIT Press. Sano, H., I. Kamada, S. Yousseﬁan, M. Katsumi, and H. Wabiko. (1990). A single treatment of rice seedlings with 5-azacitidine induces heritable dwarﬁsm and undermethylation of genomic DNA. Molecular and General Genetics 220: 441–447. Schaffner, K. F. (1998). Genes, behavior, and developmental emergentism: One process, indivisible? Philo- sophy of Science 65: 209–252. Steele, E. J., R. A. Lindley, and R. V. Blanden. (1998). Lamarck’s Signature: How Retrogenes are Changing Darwin’s Natural Selection Paradigm. Reading, MA: Perseus Books. Sterelny, K., and P. Gri≤ths. (1999). Sex and Death: An Introduction to the Philosophy of Biology. Chicago: Uni- versity of Chicago Press. Ulanowicz, R. (1997). Ecology, the Ascendent Perspective. New York: Columbia University Press. Waddington, C. H. (1940). Organizers and Genes. Cambridge: Cambridge University Press. Weber, B. H. (1998). Emergence of life and biological selection from the perspective of complex systems dynamics. In G. Van de Vijver, S. N. Salthe and M. Delpos (Eds.), Evolutionary Systems: Biological and Epistemological Perspectives on Selection and Self- Organization, pp. 59–66. Dordrecht: Kluwer Academic Publishers. Weber, B. H., and D. J. Depew. (1996). Natural selection and self-organization: Dynamical models as clues to a new evolutionary synthesis. Biology and Philosophy 11: 33–65. Webster, G., and B. Goodwin. (1996). Form and Trans- formation: Generative and Relational Principles in Biol- ogy. Cambridge: Cambridge University Press. Whitehead, H. (1998). Cultural selection and genetic diversity in matrilineal whales. Science 282: 1708–1711. Wilson, E. O. (1998). Consilience. New York: Alfred A. Knopf.

From Complementarity to Obviation: On Dissolving the Boundaries between 19 Social and Biological Anthropology, Archaeology, and Psychology

Tim Ingold

The roots of what I have to say in this chapter go the entire edifice of Western thought and sci- back to concerns that originally led me, as an un- ence—namely, that between the “two worlds” of dergraduate at Cambridge in the late 1960s, to humanity and nature. For this is what has given take up the study of anthropology. I had just us the overriding academic division of labor be- completed my first year as a student of natural tween those disciplines that deal, on the one science and was profoundly disillusioned. It was hand, with the human mind and its manifold lin- not that I was any less fascinated by the phe- guistic, social, and cultural products, and on the nomena of nature. My disenchantment stemmed other, with the structures and composition of the rather from a dawning realization that the sci- material world. And it also cleaves anthropology entific establishment was so heavily institution- itself into its sociocultural and biophysical divi- alized, internally specialized, and oppressively sions, whose respective practitioners have less to hierarchical that the most one could achieve as a say to one another than they do to colleagues in professional scientist would be to become a very other disciplines on the same side of the academic small cog in a huge juggernaut of an enterprise— fence. Social or cultural anthropologists would one moreover that seemed to have lost touch rather read the work of historians, linguists, phi- both with its sense of social responsibility and losophers, and literary critics; biological or phys- with its original mission to enlarge the scope of ical anthropologists prefer to talk to colleagues in human knowledge, and to have become largely other fields of biology or biomedicine. subservient to the military-industrial complex. I am not content to live with this situation. It Looking around for something else to study, I was, in part, the challenge of closing the gap be- wanted a discipline that would help to reconnect tween the arts and humanities on the one hand, the sense of intellectual adventure associated with and the natural sciences on the other, that drew scientific inquiry with the realities of human ex- me to anthropology in the first place, and I still perience in a world increasingly ravaged by mas- believe that no other discipline is in a better po- sive technological intervention. Anthropology sition to accomplish it. In this article, I present seemed, at the time, to fit the bill. Indeed, the rea- a program for how this might be done. My ar- sons why I took up anthropology then are still gument, in a nutshell, hinges on a distinction the reasons why I continue to study it, though I between two approaches to thinking about the might now express them in rather different terms. relations between those aspects of human exis- I believe that the discipline has a critical contri- tence that have conventionally been parceled out bution to make to the way we understand the between different disciplines (or, in the case of process of human being-in-the-world which is anthropology, subdisciplines) for separate study. badly needed in an intellectual, political, and eco- For convenience, I call these the complementar- nomic climate that has always tended to divorce ity and obviation approaches, respectively. The human affairs from their bearings in the contin- first regards every aspect as a distinct, substantive uum of organic life. component of being. It admits that the study of Since embarking on my studies in anthropol- each component is bound to yield only a partial ogy I have never looked back. I have, however, account, but promises that by putting these ac- often looked from side to side, observing with counts together it should be possible to produce a mounting despair how it has been torn apart by synthetic account of the whole. These syntheses the very divisions I thought it existed to over- are characteristically denoted by such hybrid come. These divisions ultimately seem to derive terms as biosocial, psychocultural, or even biopsy- from a single, master dichotomy that underpins chocultural. The obviation approach, by contrast, 256 Tim Ingold

is intent on doing away with the boundaries by this, however, the philosopher might respond which these components have been distinguished. with the observation that the very possibility of It claims that the human being is not a composite such a description is only open to a creature for entity made up of separable but mutually com- whom being is knowing, one that can so detach plementary parts, such as body, mind, and cul- its consciousness from the tra≤c of its bodily in- ture, but rather a singular locus of creative teractions with the environment as to treat the growth within a continually unfolding field of re- latter as the object of its concern. It is in this tran- lationships. In what follows, I argue for an obvi- scendence over nature, our philosopher might ation approach. isist, that the essence of our humanity resides. Before proceeding further, I should add a note In short, the human being can only appear as a about the terms I use for the different fields of an- naturally selected, empirical object in the eyes thropology. For a start, I do not deal here with of the rationally selecting epistemic subject. the distinction between social and cultural an- This paradox, that accounting for our exis- thropology: I believe this distinction is already tence in nature means taking ourselves out of it, widely regarded as obsolete, and I have no inten- runs like a thread through the entire history of tion of reinstating it. So when I place the word Western thought and science. And it lies at the social before anthropology, I mean it as a short- root of the idea that humans—uniquely among hand for social-cultural. Likewise, I am not con- animals—exist simultaneously in two parallel cerned with the distinction between biological worlds, respectively of nature and society, in the and physical anthropology. To my ear, the latter first as biological individuals (organisms), in the designation has a rather archaic ring, suggesting second as cultural subjects (persons). As organ- a preoccupation with measuring skulls and exca- isms, human beings seem inescapably bound to vating for fossil bones. I prefer the designation the conditions of the natural world. Like other biological, since it suggests a more rounded con- creatures, they are born, grow old and die; they cern with the conditions of human life, both now must eat to live, protect themselves to survive and and in the past. Finally, I shall make no attempt mate to reproduce. But as persons, humans seem to distinguish between archaeology and prehis- to float aloof from this world in multiple realms tory, and will use the first term indiscriminately of discourse and meaning, each constitutive of a to cover both. specific historical consciousness. From this ex- alted position they are said to transform nature, both ideationally through the imposition of Social and Biological Anthropology schemes of symbolic representation and practically through the application of technology, It is notoriously di≤cult to explain, to those new thereby converting it into the object of relations to the subject, what anthropology is all about. among themselves, relations that are taken to What, they might ask, is this being, this anthro- make up the distinct domain of society. pos, from which our discipline takes its name? It Now a complementarity approach would ac- is one thing, it seems, to ask what is a human cept this division between the organism and the being, quite another to ask what is human being. person, and would aim to put together the partial The first question is an empirical one, the second accounts of human life obtainable on each of the is a question of ontology. A modern evolutionary two planes, of nature and society, to produce a biologist, for example, might describe a human complete “biosocial” picture. The obviation ap- being as an individual of a species with a suite of proach, by contrast, would reject the comple- built-in characteristics that owe their origin to a mentarity assumption, that human existence can process of variation under natural selection. To be neatly partitioned into its biophysical and so- From Complementarity to Obviation 257

ciocultural components, not, however, by simply but also how such ways of behaving are chan- collapsing one side of the dichotomy into the neled, evaluated and made meaningful, and the other as in the more extreme forms of sociobiol- persons to whom they are directed categorized, ogy and social constructionism, but by doing in terms of culturally specific, representational away with the dichotomy itself. Whereas an ad- schemata. An obviation approach to the study vocate of complementarity might assert that the of kinship, on the other hand, would begin by human being is not merely a biological organism recognizing that behavioral dispositions are nei- nor merely a social person, but the compound of ther preconstituted genetically nor simply down- one thing plus the other, the obviation approach loaded onto the passively receptive individual asserts that humans are indeed all organism, as from a superior source in society, but are rather indeed they are all person, for in the final analy- formed in and through a process of ontogenetic sis organism and person are one and the same, development within a specific environmental con- and there is nothing mere (that is, residual or in- text. Kinship is about the ways in which others in complete) about either (Ingold 1990: 220). By the the environment contribute—through their pres- same token, this approach would reject the idea ence, their activities and the nurturance they that there is an essence of humanity that sets us provide—to this process.1 Thus, insofar as it con- radically apart from all other creatures whose cerns the growth of the organism-person within a lives are wholly contained within the world of field of ongoing relations, kinship is indissolubly nature, and with it the possibility of a purely ob- biological and social. But the biology pertains to jective account of the human being as a naturally development, not genetics; and the social to the existing, evolved entity. We may of course imag- domain of lived experience rather than its cate- ine ourselves to be suspended in a world of inter- gorical representation. subjective meaning, over and above that of our The contrast between the two approaches may material life, but such imaginings can only be car- be illustrated by way of one other example. Bi- ried on by a being who is already positioned in pedal locomotion, the capacity to walk on two the world and, by virtue of that fact, already feet, is generally assumed to be one of the hall- committed to relations with determinate compo- marks of our species, and as such to form part of nents of the environment. You have to be in a an evolved human nature. Yet as we all know, world to imagine yourself out of it, and it is and as Mauss famously observed in his essay of through this being-in-the-world that you become 1934 on body techniques (Mauss 1979: 97–123), what you are. people in different cultures are brought up to Let me briefly compare the two approaches walk in very different ways. These ways are as they might be applied to one of the classic acquired, or as Mauss put it (p. 102), “there is no fields of anthropological inquiry, namely kinship. ‘natural way’ [of walking] for the adult.” How The complementarity approach would reject would a complementarity approach deal with both the radical sociobiological thesis, that kin- this? It would argue—very much, in fact, as ship can be reduced to a calculus of genetic re- Mauss did—that although the body is innately latedness, and the equally radical humanistic predisposed to walk, it is also educated by a re- alternative, that it is an arbitrary social construct ceived social tradition, transmitted orally or by that bears no relation to genetic connection at all. other means, consisting of certain ideal rules and Rather, it would suggest that for a complete un- conventions that lay down standards of propri- derstanding of human kinship we need to recog- ety, perhaps specific to age, sex or gender, that nize not only how individuals may be innately walkers are enjoined to follow, and in terms of predisposed to behave in certain ways toward which their performance is evaluated and inter- those to whom they have a close genetic link, preted. Thus, while the capacity to walk is a bio- 258 Tim Ingold

logical universal, particular ways of walking are body undergoes processes of growth and decay, expressive of social values. and that as it does so, particular skills, habits, But the obviation approach can readily ﬁnd capacities, and strengths, as well as debilities fault with this argument. For a start, human and weaknesses, are enfolded into its very con- babies are not born walking; rather, the ability to stitution—in its neurology, musculature, even its walk is itself an acquired skill that develops in an anatomy. To adopt a distinction suggested by environment that includes walking caregivers, a Connerton (1989: 72–73), this is a matter of in- range of supporting objects, and a certain terrain. corporation rather than inscription. Thus walk- How, then, can one possibly separate learning to ing, for example, is embodied in the sense of walk from learning to walk in the approved man- being developmentally incorporated through ner of one’s society? Surely, the development of practice and training in an environment. The walking skills is just one aspect of the growth same, indeed, goes for any other practical skill. of the organism-person within a nexus of envi- Having said that, however, I must admit to a ronmental relations, and as such is closely bound growing unease with the fashion for the “body” up with kinship. Walking is certainly biological, in current social anthropology, and indeed with in that it is part of the modus operandi of the the very notion of embodiment. Advocates of the human organism, but it is also social—not be- “paradigm of embodiment,” such as Csordas cause it is expressive of values that somehow (1990), have drawn inspiration from the philoso- reside in an extrasomatic domain of collective phy of Merleau-Ponty (1962) in treating the body representations, but because the walker’s move- as the form in which the human person, qua cul- ments, his or her step, gait and pace, are contin- tural subject, is intentionally present as a being- ually responsive to the movements of others in in-the-world. One of their aims in doing so is to the immediate environment. It is in this kind of break away from the Cartesian bias, still domi- mutual responsiveness or “resonance” (Wikan nant in mainstream psychology, toward treating 1992), not in the subjection of behavior to cate- the body as the executive arm of a disembodied gorical rules, that the essence of sociality resides. mind that, sheltered from direct contact with the external world, is presumed to organize the data Body, Organism, and Development of experience and to be the ultimate source of all meaning and intention. I sympathize with this Clearly, the problem with the complementarity aim, but I am not sure that the best way to over- thesis is that it is unable to offer a coherent ac- come the troublesome mind/body dichotomy is count of ontogenetic development. Human be- by dropping the former term and retaining the ings are supposed to be in part preconstituted latter. It would seem just as legitimate to speak of genetically, in part moulded through the super- enmindment as of embodiment, to emphasize the imposition (through enculturation or socialisa- immanent intentionality of human beings’ en- tion) of ready-made structures. Real humans, gagement with their environment in the course of however, grow in an environment furnished by perception and action. The distance between a the presence and activities of others. It is precisely Merleau-Pontyan phenomenology of the body because the dynamics of development lie at the and what Bateson (1973) christened the “ecology heart of the obviation approach that it is able to of mind” is not as great as might ﬁrst appear. dispense with the biological/social dichotomy. Perhaps this is merely an issue of semantics. And it leads, naturally, to a focus on issues of em- Behind it, however, there lies a more fundamen- bodiment. By this I do not mean that the human tal question. How, if at all, are we to distinguish body should be understood as a site or medium the body from the organism? One answer might for the inscription of social values. I would rather be that the body is a discrete object composed of use the term to stress that throughout life, the organs and tissues, like as not dead or at least From Complementarity to Obviation 259

anesthetized, as it might appear before the sur- those committed to phenomenological or ecolog- geon in the operating theatre, whereas the organ- ical approaches, are markedly reluctant to go this ism is a living being, situated and functioning in far. Their hesitation may be attributed in part to its proper environment. But neither Merleau- the continuing influence of dualistic thinking, but Ponty nor those who have followed his lead mean in part—too—to a certain nervousness about the the body in this sense. They are rather referring implications of the position set out earlier for the to “the living body . . . with feelings, sensations, distinction between culture and biology. perceptions and emotions” (Ots 1994: 116), or These implications are indeed radical. If, as I what is known in German as Leib (as opposed to have suggested, those specific ways of acting, per- Körper). Yet in their determination to treat the ceiving, and knowing that we have been accus- leibly body as the subject of culture, anthropolo- tomed to call cultural are incorporated, in the gists such as Jackson (1989: 119) and Csordas course of ontogenetic development, into the neu- (1990: 5) cannot avoid the implication that there rology, musculature, and anatomy of the human exists some kind of biological residuum that is organism, then they are equally facts of biology. objectively given, independently and in advance Cultural differences, in short, are biological. Now of the cultural process.2 Culture and biology re- of course, it was precisely on the premise that cul- main as far apart as ever, only the body has been tural variation is independent of biology that repositioned: formerly placed with the organism anthropologists could claim to have refuted the on the side of biology, it has now reappeared with raciology of the early decades of this century the person on the side of culture. Hence the body (Wolf 1994). In 1930, no less an authority than as subject is split off from the organism as object, Boas had declared that “any attempt to explain leaving the latter bodiless, reduced to an inchoate cultural form on a purely biological basis is mass of biological potential. The embodiment of doomed to failure” (Boas 1940: 165). From then culture leads to nothing less than the disembodi- on, the biophysical and sociocultural divisions of ment of the organism! anthropology have proceeded along markedly di- It seems to me that the theoretical gains vergent paths. It is no wonder that contemporary brought by the paradigm of embodiment will be social anthropologists should be fearful of going more apparent than real, so long as we fail to back on such a fundamental tenet of disciplinary take one final, and crucial step, which is to recog- integrity. nize that the body is the human organism, and I believe this is an issue that has to be con- that the process of embodiment is one and the fronted. How can we rest secure in the conviction same as the development of that organism in its that raciology has long since been expurgated environment. Once this step is taken, then one or from the discipline, now that the premises on other of the two terms, body and organism, be- which this was done seem increasingly shaky, if comes effectively redundant. Given the choice of not downright incoherent? Evidence of this in- which term to retain, I would opt for the latter, coherence is not hard to come by, for example since it better conveys the sense of organized in the “statement on race” recently endorsed by process, of movement, connectivity, and relation- the International Union of Anthropological and ality, that I take to be fundamental to life. Substi- Ethnological Sciences. Article 10 of this state- tuting life for mind, and organism for body, the ment begins with the well-worn claim that “there notion of a mindful body may be replaced by that is no necessary concordance between biological of living organism, a substitution that has the characteristics and culturally defined groups,” effect both of restoring human beings to their and ends by asserting that “it is not justifiable to proper place within the continuum of organic attribute cultural characteristics to the influence life, and of laying the Cartesian dualism finally to of genetic inheritance” (IUAES 1996: 19–20). rest. Most social anthropologists, however, even What is striking here is the implicit attribu- 260 Tim Ingold

tion of “biological characteristics” to “genetic of this specification takes place over numerous inheritance”—and this despite the recognition, generations, through changes brought about by elsewhere in the document, that “biological dif- natural selection in the frequency of its informa- ferences . . . are strongly influenced by nutrition, tion-bearing elements, the genes. Development way of life and other aspects of the environment” is then understood to be the process whereby the (Article 4). To return to my earlier example: con- genotypic specification, by definition context- sider a culturally specific way of walking. Is this independent, is translated within a particular en- not a property of the organism, the outcome of a vironmental context into the manifest form of the process of development, and hence fully admis- phenotype. In this standard account, the geno- sible as a “biological characteristic”? Despite type is privileged as the locus of organic form, Boas’s strictures, there is nothing wrong with ac- while the environment merely provides the mate- counting for this or any other aspect of cultural rial conditions for its substantive realisation. To form on a “purely biological basis,” so long as be sure, an organism may develop different fea- the biology in question is of development, not tures in changed environments, but these differ- genetics. ences are regarded as no more than alternative Evidently, the real source of the problem is not phenotypic expressions of the same basic design. the identification of the social or cultural with the Only when the design itself changes does evolu- biological, but the assignment of the biological to tion occur. the genetic.3 For it is the latter assumption, which Let me return for a moment to the example still lies unquestioned at the heart of much an- of walking. According to orthodox evolutionary thropological theory as well as in the discipline’s biology, bipedal locomotion is one of a suite of public pronouncements, that forces us to choose anatomical and behavioral characteristics that between treating, say, a locally specific way of have emerged in the course of human evolution. walking either as basically nonbiological or ex- It—or, rather, a program for its development— trasomatic, governed in its bodily execution by a must therefore form part of the species-specific scheme of acquired mental representations, or as genotypic endowment that each one of us re- biological but genetically inherited. The first al- ceives at the point of conception. It is in this sense ternative reinstates the Cartesian antinomies of that human beings are said to be universally mind and body; the second takes us right back to equipped, as part of their evolved makeup, with raciology. Breaking the link between biological an innate capacity to walk on two feet, regardless form and genetic inheritance, however, is easier of how they walk in practice, or of whether they said than done, for this link underpins the entire walk at all—or go everywhere by car! Specific edifice of modern evolutionary theory and justi- ways of walking have not themselves evolved, fies the fundamental precept on which it rests, they are just alternative phenotypic realizations namely that the life history of the individual or- of an evolved, genotypic trait. By the same token, ganism, its ontogenetic development, forms no we should all be genotypically endowed with the part of the evolution of the species to which it capacity to rest for long periods in a squatting belongs.4 position, yet this is something that I (along with fellow Westerners) am quite unable to do, since I The Myth of the Genotype have been brought up in a society where it is normal to sit on chairs. As this example shows, the In brief, what is supposed to evolve is not the or- notion of capacity is almost totally vacuous un- ganism itself or its manifest capabilities of action, less it refers back to the overall set of conditions but rather a formal design specification for the that must be in place, not only in the individual’s organism known as the genotype. The evolution genetic constitution but also in the surrounding From Complementarity to Obviation 261

environment, to make the subsequent develop- tions for their own future development and that ment of the characteristic or capability in ques- of others to which they relate, they figure not tion a realistic possibility (Ingold 1996a). One as passive sites of evolutionary change but as cre- would otherwise have to suppose that human be- ative agents, producers as well as products of ings were genotypically endowed, at the dawn of their own evolution. Third, and most crucially history, with the capacity to do everything that for my present purposes, this applies equally to they ever have done in the past, and ever will do human beings. “Our basic image of human on- in the future—not only walk and squat but also togeny,” as Robertson (1996: 595) insists, swim, ride bicycles, drive cars, fly airplanes, carry “should therefore be that of a lifespan set be- out scientific research, and so on (the list would tween an ascendent and a descendent generation, be endless). linked by the process of begetting and being be- What this means, in general terms, is that the gotten.” Human lives overlap: fashioned within forms and capacities of human and other organ- contexts shaped by the presence and activities of isms are attributable, in the final analysis, not to predecessors, they in turn affect the conditions of genetic inheritance but to the generative poten- development for successors. There is nothing tials of the developmental system (Oyama 1985), strange about this idea; on the contrary it sums that is, the entire system of relations constituted up the process we are used to calling history. So by the presence of the organism in a particular conceived, however, history is not so much a environment. This is not to deny that every or- movement in which, as Maurice Godelier puts it ganism starts life with—among other things—its (1989: 63), human beings “produce society in complement of DNA in the genome. Orthodox order to live,” as one in which, in the course of evolutionary theory has it that this DNA encodes their social lives, they grow one another, estab- the formal design specification. Because, how- lishing by their actions the conditions for each ever, there is no reading of the genetic code that other’s development. But taken in this sense, his- is not itself part of the process of development, it tory is no more than a continuation, into the field is only within the context of the developmental of human relations, of a process that is going on system that we can say what any particular gene throughout the organic world. That process is is for. It follows that there can be no specification one of evolution. The distinction between history of the characteristics of an organism, no design, and evolution is thus dissolved (Ingold 1995a: that is independent of the context of develop- 210–211). ment. The genotype simply does not exist. And so too, in the case of human beings, there is no such thing as “bipedal locomotion” apart from the Anthropology and Archaeology manifold ways in which people actually learn to walk in different communities (Ingold 1995a). Between Evolution and History Now if, as I have argued, organic form is a property not of genes but of developmental sys- This is where prehistoric archaeology comes in. tems, then to account for its evolution we have to Do archaeologists study human history or hu- understand how such systems are constituted and man evolution? So long as the distinction remains reconstituted over time. This conclusion has in place, archaeology seems to fall awkwardly be- three major implications. First, far from being a tween the two stools. An indicator of this predic- tangential offshoot of the evolutionary process, ament is the fact that while there has long been a ontogenesis is the very crucible from which it un- strongly held view in social anthropology that folds. Second, because organisms, through their there is little to distinguish it from the discipline activity, can influence the environmental condi- of history, and despite the rather obvious links 262 Tim Ingold

between history and archaeology (in that both study the lives of people in the past), the majority of social anthropologists insist that their subject has little or nothing to do with archaeology. Of course, this was not always so. The evolutionary anthropologists of the nineteenth century were keen to study allegedly primitive peoples because Figure 19.1 Organic evolution and the history of culture (after it was thought that their present existence could Kroeber 1952: 50). “In this illustration,” Kroeber ex- illuminate the earlier conditions of humankind in plains, “the continuous line denotes the level inorganic; the spheres of social and intellectual life, just as the broken line the evolution of the organic; the line of archaeology could reveal the early stages of ma- dots, the development of civilisation. Height above the terial culture. But the subsequent rejection of this base is degree of advancement.” A marks “the begining kind of progressive evolutionism broke the link of time on this earth,” B “the first human precursor,” C between social anthropology and archaeology, “primitive man,” and D “the present moment.” and at the same time ruled out of order any suggestion that humans might be more or less cultural, or that they might be further along or lag that they would rather not have to think about. If behind in the course of history. human history has a point of origin, what could So far as most contemporary social anthropol- it mean to have been living close to that point, or ogists are concerned, living beings either inhabit even at the crucial moment of transition itself ? a historically constituted world of cultural mean- Were such people semicultural, gearing up for ing or they do not: all human beings do, other an- history? How can one conceivably distinguish imals do not. There are no differences of degree. those actions and events that carried forward the Yet the very idea that humans inhabit separate, movement of human history from those that set cultural worlds implies that at some point, the it in motion in the first place? history of culture must have lifted off from a In recent years archaeologists have expended a baseline of full-blown, evolved human capacities. great deal of effort in revealing the origins of cul- Short of supposing some kind of unfathomable ture and history in what has come to be called the quantum leap, there is no alternative but to imag- “human revolution” (Mellars and Stringer 1989). ine a historical trajectory that rises inexorably This is now supposed to have taken place during from a point of emergence or origin. Figure 19.1 the Upper Palaeolithic, though archaeologists shows an early example of just such a view, taken remain perplexed by the apparent fact that so- from Kroeber’s classic paper of 1917 on the su- called modern humans—that is, beings equipped perorganic (Kroeber 1952: 50). Here, the history with the full suite of evolved capacities needed to of culture is seen taking off from organic evolu- set the cultural ball rolling—arrived on the scene tion at point B; by point C we have the rudimen- a good hundred thousand years before we find tary culture of “primitive man,” while by point D any evidence for the sorts of things with which (the present) the origins of culture have been left culture is usually associated: burials, art, complex far behind. Present-day social anthropologists and regionally diverse toolkits, language, and so may well frown at this picture, and scoff at its in- on. Indeed, the alleged revolution seems to have vocation of progressive development, but they taken about twice as long as the fifty-thousand- themselves have nothing better to offer. And one year history it is supposed to have inaugurated! of the reasons why they tend to steer clear of pre- Be that as it may, the argument I have set out historic archaeology, I suggest, is that it throws above suggests that the entire project of searching the spotlight on just those awkward questions for the genesis of some essential humanity is seri- From Complementarity to Obviation 263

ously misguided. We look in vain for the evolu- should be foremost in combating the pretensions tionary origins of human capacities for the simple of the origin-hunters. And they should help us to reason that these capacities continue to evolve in recognize that our humanity, far from having the very historical unfolding of our lives. been set for all time as an evolutionary legacy Of course, even an orthodox evolutionary the- from our hunter-gatherer past, is something that orist would have to admit that the evolution of we continually have to work at and for which we humankind did not exactly stop once history and ourselves must bear the responsibility. There is, culture were underway. However the conven- in short, no way of saying what a human being is tional view, exemplified in Kroeber’s diagram apart from the manifold ways in which human (figure 19.1) and reiterated by countless authors beings become. “Modern humans” have not orig- ever since, is that by comparison with the rate of inated yet, and they never will. historical change, this evolution has continued at snail’s pace, so that to all intents and purposes, Landscape and Environment contemporary human beings may be regarded as not significantly different from their predecessors This recognition that the forms of human being, of the Upper Palaeolithic.5 They are equipped and the capacities they entail, are continually with the same basic morphology, capacities and evolving as life goes on, helps to put paid to dispositions that, packaged in the genotype, have another dichotomy which has been particularly been passed on down the generations for tens of troublesome for archaeology. This is between the thousands of years. To be sure, the amount of natural and the artificial, and it lies at the source genetic change in human populations over this of the idea that archaeologists study artifacts. period may have been relatively small. My con- Now the very notion of artifact implies the work- tention, however, is that in their dispositions and ing up of some raw material to a finished form, capacities, and to a certain extent even in their corresponding to a preconceived design in the morphology, the humans of today are not at all mind of the artisan. Only once it has first been like their predecessors. This is because these char- made, in this sense, can it be brought into play in acteristics are not fixed genetically but emerge the ordinary business of life, in the course of within processes of development, and because the which it is used. This distinction between mak- circumstances of development today, cumula- ing and using is fundamental to what I have tively shaped through previous human activity, elsewhere (Ingold 1995b) called the “building are very different from those of the past. perspective”: the idea that life goes on within It is, I believe, a great mistake to populate the structures that have been constituted in advance, past with people like ourselves, equipped with the rather than these structures arising within the life underlying capacities or potentials to do every- process itself. Adopting such a perspective, it is thing we do today, such that history itself appears easy to imagine that the forms of objects recov- as nothing more than the teleological process of ered from archaeological sites correspond to de- their progressive realization. Indeed the very no- signs that were originally in the heads of their tion of an origin, defined as the point at which one-time makers. these capacities became established, awaiting However as Davidson and Noble (1993: 365) their historical fulfilment, is part of an elaborate have pointed out, it is a fallacy—and one that is ideological justification for the present order of found very frequently in archaeological writing— things and, as such, but one aspect of the intense to suppose that objects are ever finished in this presentism of modern thought. In so far as the sense. For one thing, their forms are not imposed task of archaeology is to illuminate the past by the mind, but arise within the movement of rather than legitimate the present, archaeologists the artisan’s engagement with the material; for 264 Tim Ingold

another, in the course of being used for one enfold, into their very formation and consti- purpose, objects may undergo further modifica- tution, the lives and works of their inhabitants tion that make them peculiarly apt for another. (Ingold 1993b: 156–157). To appreciate what is Whether, at any moment, we say the object is going on here, we need to adopt a different per- being used or made depends entirely on whether spective, one that recognizes that the forms peo- the reference is to a present or future project. Al- ple make or build, whether in their imagination though at a certain point, the artisan may claim or on the ground, arise within the current of their to have completed his work, that is certainly not involved activity, in the specific relational con- the end of the object he has produced. Indeed, texts of their practical engagement with their sur- an artifact can never really be said to be finished roundings. Building, in short, is encompassed by until it is of no further use to anyone and is finally dwelling, making by use. I call this the “dwelling discarded. The lesson to be learned from this is perspective” (Ingold 1995b). that the objects around us have histories which, Alongside the relatively recent anthropological in certain respects, are not unlike the life histories recognition, discussed earlier, that the human in- of persons. Just as persons continually come into tentional presence in the world is an embodied being through their involvement in relationships presence, there has been a remarkable upsurge with other persons and objects in their environ- of interest in the landscape. This is already prov- ments, so the forms and meanings of objects are ing to be a particularly fruitful area of collabora- generated within the contexts of their involve- tion between anthropologists and archaeologists ment in the diverse life-projects of the beings (for example, Bender 1993; Tilley 1994). Though (human and nonhuman) with which they are sur- the precise meaning of “landscape,” as that of rounded. In this respect they are never made but “body,” continues to be the subject of intense always in the making.6 controversy, there is a clear connection between We cannot, then, make a hard and fast distinc- the two concerns. For if persons inhabit intention between one class of things that are ready- tional worlds, and if bodies inhabit landscapes, made in nature, and another class of things that then to reunite persons with their bodies is also to have been made through the shaping of a nat- restore their intentional worlds to the landscape. urally given raw material into a finished arte- But this raises a parallel problem, too. How, if factual form. Nor can we adopt an analogue of at all, are we to distinguish landscape from envi- the complementarity thesis, and suppose that ob- ronment? Just as the body has come to be identi- jects, like persons, are in part naturally preconsti- fied with the cultural subject, and the organism tuted, and in part molded through the imposition with the residual biological object, so there is of cultural design. Just such an analogue is impli- a temptation to treat the landscape as an inter- cated in the unfortunate designation of artifacts subjectively constituted, existential space, while as objects of “material culture,” suggesting as it reducing the environment to a mere substrate does that to make an artifact you first take an ob- of formless materiality. In this vein, Weiner ject with certain intrinsic material properties and (1991: 32) speaks of how the bestowal of place then add some culture to it. I noted earlier that a names intentionally transforms “a sheer physical critical weakness of the complementarity thesis, terrain into a pattern of historically experienced applied to persons, is that it cannot offer a realis- and constituted space and time,” thereby creating tic account of ontogenetic development. In pre- “existential space out of a blank environment” cisely the same way, an approach that stresses the (my emphases). complementarity of natural and artificial—or Once again, the division between the sociocul- built—components of the environment cannot tural and the biophysical is reproduced rather begin to grasp the ways in which environments than dissolved. This, in my view, is a retrograde From Complementarity to Obviation 265

step. The environment of persons is no more re- our dwelling. In practice, of course, one cannot ducible than is their organic existence to pure mo- do one without the other, nor can either be done lecular substance. It is not merely physical, and it without regard to the inherent temporality of the is certainly not blank. For example, the ground I processes both of ontogenetic development and walk on is surely a part of my environment, but environmental formation. I have already shown in a physicalist description the ground, as such, how we can dispense with the distinction between does not exist; there are only packed molecules of social and biological anthropology. It is now pos- carbon, nitrogen, silicon, and so on. As Reed has sible to see how the anthropology/archaeology eloquently put it, “it is the earth on which we distinction might be thrown out as well. walk, and the soil in which we plant, that is relevant for us as perceiving and acting creatures; not the molecules discovered by scientists” (Reed Psychology and Anthropology 1988: 111). The environment, in short, is not the same as the physical world; that is to say, it is not One of the reasons why, up to now, it has proved describable in terms of substance. Rather, the en- so di≤cult to effect a reintegration of the subdivi- vironment is the world as it exists and takes on sions of anthropological inquiry, and particularly meaning in relation to the beings that inhabit it of its social and biological components, lies in the (Gibson 1979: 8). As such, its formation has to be fact that in the conventional complementarity ap- understood in the same way that we understand proach, the necessary link between the individual the growth of organisms and persons, in terms of organism and the cultural subject can only be the properties of dynamic self-organization of reestablished by way of a third term, namely what lational ﬁelds. But precisely because environment is called the “human mind.” The discipline that does not stand as material substance to the im- exists to study the human mind is, of course, psy- material forms of landscape—because it under- chology. Thus for advocates of the complemen- goes a continual process of formation with, and tarity approach, psychology would have to be around, its inhabitants—I see no basis on which along in any complete, synthetic account of the two terms, environment and landscape, may human existence. The synthesis would not be be distinguished. “biosocial” but “biopsychosocial.” As Mauss put Earlier, I suggested that the concept of the it, an exclusive focus on the relations between the “mindful body” should be replaced by that of biological and the sociological “leaves but little “living organism.” But the formulation remained room for the psychological mediator.” An ac- incomplete, since neither the body nor the organ- count of walking, for example, that rested solely ism, however conceived, can exist in isolation. on an anatomical or physiological base, or even We can now complete it with the proposition that on a psychological or sociological one, would be the mindful body in a landscape be replaced by inadequate. “It is the triple viewpoint, that of the the living organism in its environment. And this, ‘total man,’ that is needed” (Mauss 1979 [1934]: to conclude, offers the basis for a real synthesis of 101, my emphasis). In what follows I propose to archaeology and anthropology. Instead of being argue, to the contrary, that the human mind— separated by contrived divisions between past conceived as some kind of structured entity—is and present, or between artifacts and bodies, we as much an invention of modern science as is the might say that where anthropology studies the human genotype. Mind, as I have already sug- conditions of human living in the environment— gested with acknowledgment to Bateson (1973), or of what phenomenological philosophers call is not in the head rather than out there in the “being-in-the-world”—archaeology studies the world, but immanent in the active, perceptual en- formation of the environment of our living-in, gagement of organism-person and environment. 266 Tim Ingold

As the study of the conditions of such engage- mind of the individual, the latter is concerned ment, psychology should be no different from with the collective mind of society. Much recent anthropology. work in social anthropology, however, has To begin, we have to consider why a psycho- pointed to the inadequacy of the classical indi- logical mediator should be deemed necessary at vidual/society dichotomy (Strathern 1996). We all. The reason lies in the fact that the classical have begun to recognize (see, for example, Toren division between the biological and the social is 1993) that those capacities of conscious aware- not based on one opposition but is a compound ness and intentional response normally bracketed of two: between body and mind, and between under the rubric of mind are not given in advance the individual and the collectivity. Thus psychol- of the individual’s entry into the social world, ogy has traditionally shared with biological but are rather fashioned through a lifelong his- anthropology an exclusive focus on the individ- tory of involvement with both human and non- ual, and with social anthropology a focus on human constituents of the environment. We have mental rather than bodily states. I have already realized, too, that it is through the situated, inten- addressed the problem of the mind/body dichot- tional activities of persons, not through their sub- omy, but it remains to deal with that of individ- jugation to the higher authority of society, that ual versus collectivity. This latter dichotomy rests social relationships are formed and reformed. on a hierarchical conception of the relations With this, the hierarchical conception of part/ between parts and wholes that is very deeply whole relations simply collapses. Every particular embedded in the structure of our thought. An- person, in so far as it enfolds in its constitution thropologists have always professed their com- the history of its environmental relations, gathers mitment to a holistic approach, but they have the whole into itself.7 But that whole, so con- tended to take this to mean a focus on wholes— ceived, is not an entity but a movement or pro- conceived as total societies or cultures—as op- cess: the process of social life. Persons come into posed to their parts or members, individual being, with their specific identities, capacities, human beings. Following principles set out by and powers of agency, as differentially positioned Durkheim a century ago, it has generally been enfoldments of this process, and in their actions conceded that as the whole is more than the sum they carry it forward. Consciousness and social of its parts, so “society is not the mere sum of existence, though they appear at any particular individuals, but . . . a specific reality which has its moment to offer alternative perspectives on the own characteristics” (Durkheim 1982 [1895]: person, respectively inward-looking and out- 128–199). ward-looking (Ingold 1983: 9), turn out in their Now the very logic of summation invoked here temporal unfolding to be one and the same, like entails that every part is a self-contained, indivis- the single surface of a Möbius strip. Taking this ible, naturally bounded unit whose integrity and view, I can see no further intellectual justification constitution are already given, independently and for continuing to uphold the boundary that has in advance of any relations it may enter into with traditionally divided psychology from social an- others of its kind. These relations, in short, have thropology. The discipline that will be brought no bearing upon the constitution of the individ- into being through the dissolution of this bound- ual parts themselves, but are rather constitutive ary, whatever we choose to call it, will be the of a distinct entity, namely society, located at study of how people perceive, act, feel, remem- a higher level of abstraction. This Durkheimian ber, think, and learn within the settings of their view has long underwritten the academic division mutual, practical involvement in the lived-in of labor between psychology and social anthro- world. In the following paragraphs I should like pology: whereas the former is said to study the to review some of the consequences of this per- From Complementarity to Obviation 267

spective in three areas that have traditionally contexts of human engagement with the environ- been central to psychological inquiry: perception, ment while treating the latter as no more than memory, and learning. a kind of recording instrument, converting the stimuli that impinge upon it into data to be Perception processed. One of the most powerful critiques of this view has come from advocates of so-called Why do people perceive the world in the particu- ecological psychology, who have drawn inspi- lar ways that they do? Mainstream psychology ration above all from the pioneering work of has long regarded perception as a two-step oper- Gibson on visual perception (Gibson 1979). ation: in the first, sensory data are picked up Ecological psychologists reject the information- from the environment by means of the receptor processing view, with its implied separation of organs of the body; in the second these data the activity of the mind in the body from the re- are processed by a range of devices in the mind, activity of the body in the world, arguing instead to generate images or representations, internal that perception is an aspect of functioning of the models of an external reality. This processing is total system of relations constituted by the pres- known as cognition. By and large, psychologists ence of the organism-person in its environment. have been concerned to discover universals of Perceivers, they argue, get to know the world di- cognition, which are attributed to structures es- rectly, by moving about in the environment and tablished in the course of human evolution. discovering what it affords, rather than by repre- Anthropologists, by contrast, have wanted to ex- senting it in the mind. Thus meaning is not the plain why people from different cultural back- form that the mind contributes, by way of its ac- grounds perceive the world in different ways. quired schemata, to the flux of raw sensory data, They have done so by suggesting that human but is rather continually being generated within cognized models are constructed on the basis of the relational contexts of people’s practical en- programs or schemata that are acquired as part gagement with the world around them. of a tradition, and vary from one culture to an- It follows from this approach that if people other. What people see will therefore be relative raised in different environments perceive different to their particular framework for viewing the things, this is not because they are processing the world. At first glance, the universalistic claims same sensory data in terms of alternative repre- of psychology seem incompatible with the rela- sentational schemata, but because they have been tivistic stance adopted by social anthropology. trained, through previous experience of carrying But as several authors have pointed out (e.g., out various kinds of practical tasks, involving D’Andrade 1981; Sperber 1985; Bloch 1991), particular bodily movements and sensibilities, to the two perspectives are, in fact, perfectly com- orient themselves to the environment and to at- plementary. For unless innate processing mech- tend to its features in different ways. Modes of anisms are already in place, it would not be perception, in short, are a function of specific possible for human beings to acquire the pro- ways of moving around—of walking, of sitting grams for constructing their culturally specific or squatting, of tilting the head, of using imple- representations from the data of experience. ments, and so on, all of which contribute to what I will spell out the logic of this argument later Bourdieu (1977: 87) would call a certain “body on, because it bears on the issue of learning. My hexis.” And as we have already seen, these forms present concern is with the way in which the ap- of motility are not added to, or inscribed in, a proaches outlined above, both in psychology and preformed human body, but are rather intrinsic anthropology, reproduce the Cartesian duality of properties of the human organism itself, develop- mind and body, removing the former from the mentally incorporated into it modus operandi 268 Tim Ingold

through practice and training in a particular en- ous experience, within a given environmental vironment. Hence capacities of perception, as of context. Let me clarify the contrast by means of action, are neither innate nor acquired but un- a simple analogy. Suppose I play a record of one dergo continuous formation within processes of of Bach’s suites for unaccompanied cello. An ontogenetic development. This result is clearly in exceedingly complicated pattern is engraved on line with the conclusions to be drawn from an ob- the otherwise blank surface of the disc, but the viation approach to the relation between social mechanical processes—of rotation and amplifi- and biological phenomena. In their rejection, on cation—involved in the operation of the record the one hand, of the Cartesian view of action as player could hardly be more simple. Now sup- the bodily execution of innate or acquired propose I pick up my cello to perform the suite my- grams, and on the other hand, of the cognitivist self. In this case, the music issues directly from view of perception as the operation of the mind my own movement, a movement that involves the upon the deliverance of the senses, the obviation whole of my being indissolubly coupled with the approach in anthropology and the ecological instrument. The process of playing a musical in- approach in psychology find common cause. strument like a cello is enormously complex, and Both take the living-organism-in-its-environment calls for embodied skills that take years to ac- as their point of departure. This is why (contra quire. But whether the music exists at all as a Bloch 1991) I believe that an anthropology that structure in the head or mental score, indepen- sets out from this point has more to gain from an dently of the activities of practice and perfor- alliance with ecological psychology than from an mance, is a moot point. alliance with cognitive science. Now in introducing his distinction between complex structure and complex process meta- Memory phors, Rubin was actually concerned with the psychological study of memory. His point was Another way of expressing the difference between that in mainstream cognitive psychology, it is cognitivist and ecological approaches is in terms usual to regard memory as a kind of mental store, of a contrast suggested by Rubin (1988). One in which past experiences and received informa- may understand what is going on, he writes, in tion are engraved and filed, as on the grooves and terms of one or other of two alternative meta- bands of a record (or, to adopt a more contem- phors. The first is a complex structure metaphor, porary analogy, a computer disc). Remembering the second a complex process metaphor. The is then a rather simple process of searching or former, which is dominant in cognitive psychol- scanning, across a complexly structured cognitive ogy, works by converting what is observed in array. It is, moreover, a purely mental, inside-the- the world into a formal account, whether envis- head operation. Once a particular memory is re- aged as a script, schema, grammar, program, or trieved, it may or may not be expressed in overt, algorithm, and then has that account copied into bodily behavior. But every behavioral expression, the mind so that the observed behavior can be like every playing of a record, is no more than a simply explained as the expression of this mental replica run off from a preexisting template. With blueprint. The latter, the dominant metaphor a complex process model, by contrast, remem- in ecological psychology, imputes little or no bering is itself a skilled, environmentally situated structured content to the mind. Instead, behav- activity. It is in playing the Bach suite that I re- ior is explained as the outcome of a complex member it; the processes of remembering and process set in train by virtue of the immersion playing are one and the same. It follows that of the practitioner, whose powers of perception every performance, far from being a replica, is it- and action have been fine-tuned through previ- self an original movement in which the music is From Complementarity to Obviation 269

not so much reproduced as created anew. More preferred the approach of skill (one of his exam- generally, remembering is a matter not of discov- ples was of strokes in tennis or cricket). However ering structures in the attics of our minds, but as Connerton has pointed out (1989: 28), the cog- of generating them from our movements in the nitivist emphasis on looking for structures in the world. mind, and the concomitant reduction of action to Armed with this contrast, let me now turn to a simple process of mechanical execution, has left the role of remembering in social life. It is re- no conceptual space for the investigation of markable that the two pioneering figures in the bodily enskilment, or what he calls “habit mem- study of social memory, Halbwachs and Bartlett, ory.” It is true that most social anthropological took opposite sides on the issue. Halbwachs, a work on memory has actually been about com- committed Durkheimian, identified memory with memoration—the present reenactment of past the very framework of collective representations events in ritual practice, storytelling, writing, and that are supposed to give order and meaning to the like. And commemoration needs to be distin- the otherwise chaotic influx of raw sensation. If guished from memorization: the developmental memory is social rather than individual, it is be- incorporation of specific competencies (such as cause the complex structures that underwrite the playing a musical instrument) through repeated human capacity for recollection have their source trials. While the relation between memorization in a collective tradition. “Our recollections,” and commemoration has yet to be fully unraveled Halbwachs wrote, “depend on those of all our (Ingold 1996b: 203), the essential point to recog- fellows, and on the great frameworks of the mem- nize is that the one cannot occur without the ory of society” (1992: 42). For Bartlett, to the other. To commemorate the music of Bach, for contrary, what counted was not the structure of example, it must be possible to perform it, one memory, but the process of remembering. This cannot perform it without skill, and the develop- process, he argued, depends upon an organiza- ment of skill implies memorization (Connerton tion of what he called “schemata.” Ironically, 1989: 5).9 though it was Bartlett who introduced the concept of schema into psychology, he did not like it, Learning and warned explicitly against regarding schemata as static, maplike structures—which is precisely This is the point at which to return to the psy- how they are understood by most cognitive psychological version of the complementarity thesis, chologists and cognitive anthropologists today. namely that the acquisition of culture is possible According to Bartlett, the schema is an active or- thanks to innate mental processing devices. It is ganization of past reactions or experiences, which perfectly true that if culture consisted of a corpus is continually brought to bear, and at the same of transmissible knowledge, or in the words of time continually evolves, in the complex pro- Quinn and Holland (1987: 4), of “what [people] cess of our engagement with the environment must know in order to act as they do, make the (Bartlett 1932: 201).8 And it is because this is things they make, and interpret their experience largely an environment of other persons that rein the distinctive way they do,” then the mind membering is social. would have to be pre-equipped with cognitive de- Clearly, without the ability to remember, vices of some kind that would allow this knowl- human beings would be unable to learn anything edge to be reassembled inside every individual at all. But there is a world of difference between head through a processing of the raw input of learning as adding more to one’s internal, repre- sensory data. In other words, the programs or sentational structure, and learning as the devel- schemata that enable people to construct their opment of a skill (Rubin 1988: 379–380). Bartlett culturally specific representations of the world, 270 Tim Ingold

and to deliver appropriate plans of action, would gled story short, they boil down to two distinct themselves have to be constructed from the ele- claims. One is that the concrete mechanisms ments of experience, on the basis of certain rules making up the evolved architecture are reliably and principles. So how were these acquired? constructed, or wired up, under all possible cir- Perhaps in the same way, through the processing cumstances. The other is that these universal of experiential input according to yet another mechanisms proceed to work on variable inputs program. “You can learn to learn,” Johnson- from the environment to produce the diversity of Laird explains, “but then that learning would de- manifest capabilities that we actually observe. pend on another program, and so on. Ultimately, Consider the specific and much-vaunted example learning must depend on innate programs that of language acquisition. Here, the alleged univer- make programs” (Johnson-Laird 1988: 133). sal mechanism is the so-called language acquisi- Whence, then, comes the information that speci- tion device (LAD). It is assumed that all human fies the construction of the innate devices, with- infants, even those (hypothetically) reared in so- out which, it would seem, no learning could take cial isolation, come equipped with such a device. place at all? During a well-defined stage of development, this By and large, in the literature of cognitive psy- device is supposed to be activated, operating on chology, the postulation of innate structures is the input of speech sounds from the environment taken to require no more justification than vague so as to establish, in the infant’s mind, the gram- references to genetics and natural selection. Thus mar and lexicon of the particular language it is assumed that the design specifications for spoken in his or her community. It would thus what is often called the mind’s “evolved archi- appear that language acquisition is a two-stage tecture” (Cosmides, Tooby and Barkow 1992: 5) process: in the first, the LAD is constructed; in must form one component of the human geno- the second, it is furnished with specific syntactic type. I have already shown, however, that it is and semantic content. This model of cognitive impossible to derive a design specification for the development is summarized in figure 19.2. Notice organism from its genetic constitution alone, in- how the model depends on factoring out those dependently of the conditions of its development features of the environment that are constant, in an environment. For cognitive psychology this or reliably present, in every conceivable devel- problem is further compounded, for if the theory opmental context, from those that represent a of learning as the transmission of cultural infor- source of “variable input” from one context to mation is to work, the requisite cognitive devices another. Only the former are relevant in the first must already exist, not merely in the virtual guise stage (the construction of “innate” mechanisms); of a design, but in the concrete hardwiring of only the latter are relevant in the second (the ac- human brains. Somehow or other, in order to quisition of culturally specific capabilities). kick-start the process of cultural transmission, For comparative analytic purposes, it is some- strands of DNA have miraculously to transform times helpful to sift the general from the partic- themselves into data processing mechanisms. ular, or to establish a lowest common denomina- This is rather like supposing that merely by repli- tor of development. But real environments are cating the design of an aircraft, on the drawing not partitioned in this way. Let me continue for a board or computer screen, one is all prepared for moment with the example of language learning. takeoff. From well before birth, the infant is immersed Attempts in the literature to resolve this prob- in a world of sound in which the characteristic lem, insofar as it is even recognized, are confused patterns of speech mingle with all the other and contradictory. To cut a rather long and tan- sounds of everyday life, and right from birth it is From Complementarity to Obviation 271

Figure 19.2 The two stages of cognitive development according to the complementarity model. In the first stage the human genotype interacts with the constant component of the environment to produce the universal mechanisms of the mind’s evolved architecture. In the second, this architecture operates on variable environmental inputs to produce culturally specific capabilities. surrounded by already competent speakers who tablish, correspond of course to what appear to provide support in the form of contextually observers as the diverse languages of the world. grounded interpretations of its own vocal ges- In short, language—in the sense of the child’s tures. This environment, then, is not a source of capacity to speak in the manner of his or her variable input for a preconstructed device, but community—is not acquired. Rather, it is con- rather furnishes the variable conditions for the tinually being generated and regenerated in the growth or self-assembly, in the course of early de- developmental contexts of children’s involvement velopment, of the neurophysiological structures in worlds of speech. And if language is not ac- underwriting the child’s capacity to speak. As the quired, there can be no such thing as an innate conditions vary, so these structures will take language-learning device. manifold forms, each differentially tuned both to What applies specifically in the case of lan- specific sound patterns and to other features of guage and speech also applies, more generally, to local contexts of utterance. These variably at- other aspects of cultural competence. Learning to tuned structures, and the competencies they es- walk in a particular way, or to play a certain mu- 272 Tim Ingold

sical instrument, or to practice a sport like cric- sensitization of the entire perceptual system, ket or tennis, is a matter not of acquiring from comprising the brain and peripheral receptor or- an environment representations that satisfy the gans along with their neural and muscular link- input conditions of preconstituted cognitive de- ages, to particular features of our surroundings. vices, but of the formation, within an environ- Through this process, the human being emerges ment, of the necessary neurological connections, not as a creature whose evolved capacities are along with attendant features of musculature and filled up with structures that represent the world, anatomy, that underwrite the various skills in- but rather as a center of awareness and agency volved. This conclusion is once again concordant whose processes resonate with those of the envi- with the obviation approach developed earlier, ronment. Knowledge, then, far from lying in the and it undermines one of the key ideas of the relations between structures in the world and complementarity thesis—that cultural learning is structures in the mind, mediated by the person of like filling a universal, genetically specified con- the knower, is immanent in the life and con- tainer with culturally specific content. The notion sciousness of the knower as it unfolds within the that culture is transmissible from one generation field of practice set up through his or her presence to the next as a corpus of knowledge, indepen- as a being-in-the-world. dently of its application in the world, is untenable for the simple reason that it rests on the impossi- The three topics I have reviewed above—of per- ble precondition of a ready-made cognitive ar- ception, memory, and learning—are of course chitecture. In fact, I maintain, nothing is really closely connected. All of them could be addressed transmitted at all. The growth of knowledge in in terms of a complex structure metaphor, by the life history of a person is a result not of infor- imagining the world of our experience to be de- mation transmission but of guided rediscovery, composed into a myriad of ephemeral fragments, where what each generation contributes to the unit events, samplings of which the mind has then next are not rules and representations for the pro- to piece together into some coherent pattern by duction of appropriate behavior but the specific means of totalizing frameworks of social rather conditions of development under which succes- than individual provenance. I have argued, by sors, growing up in a social world, can build up contrast, for an approach that starts from rela- their own aptitudes and dispositions. tions and processes rather than structures and The process of learning by guided rediscovery events. Whether our concern be with perceiving, is most aptly conveyed by the notion of showing. remembering, or learning, the workings of mind To show something to someone is to cause it to are to be found in the unfolding relations be- be made present for that person, so that he or she tween organism-persons and their environments. can apprehend it directly, whether by looking, lis- There is no way of saying what the human mind tening, or feeling. Here the role of the tutor is to is, or of specifying its essential architecture, outset up situations in which the novice is afforded side of this unfolding. For the forms of human the possibility of such unmediated experience. knowledge do not stamp themselves upon the Placed in a situation of this kind, the novice is in- substance of human experience, but themselves structed to attend to this or that aspect of what arise within the complex processes of people’s en- can be seen, touched or heard, so as to get the feel gagement with their surroundings. In short, the of it for him- or herself. Learning in this sense is phenomena of mind are as much ecological and tantamount to what Gibson (1979: 254) called an social as they are psychological. To conclude this “education of attention.” Gibson’s point, in line section I should like to show why the approach with the principles of his ecological psychology, adopted here promises to shed an entirely fresh was that we learn to perceive by a fine-tuning or light on one of the most neglected areas of an- From Complementarity to Obviation 273

thropological inquiry, namely the knowledge and evolved potentials, to a final state of full-blown activities of children. cultural life, but nevertheless cannot countenance the possibility of a form of life that is semicul- Children tural, betwixt and between nature and history. Substitute “ancestral hominid” for “infant,” and In a paper presented some twenty years ago, the following characterization of childhood of- Theodore Schwartz spoke of his “sudden and be- fered by Goldschmidt (1993: 351)—“the process lated realization . . . that anthropology had ig- of transformation of the infant from a purely nored children in culture while developmental biological being into a culture-bearing one”— psychologists had ignored culture in children” would serve equally well to define the so-called (Schwartz 1981: 4). There are signs, today, of a human revolution. And just as it is di≤cult to see change of heart in both disciplines. The reasons how the events of this prehistoric revolution can for the anthropological neglect of children, how- possibly be distinguished from those of the his- ever, do not lie merely in a certain observational tory it is alleged to have inaugurated, so too, blindness—the failure of ethnographers in the there seems to be no obvious way of telling apart field to notice children, or to pay attention to the experiences, supposedly constitutive of child- their activities and what they have to say. Nor do hood, that make a human being ready for his- they stem from the real di≤culties, practical as tory, from those that belong to the historical well as ethical, of collaborating with children in process itself. An obviation approach, however, ethnographic research. To bring children back to enables us to dispense with such troublesome dis- where they belong—at the center of our inquiries, tinctions. The infant, who admittedly starts life as just as they are at the center of social life—will a “purely biological being,” remains so for the require more than just a different attitude on rest of his or her life. Yet right from the moment the part of ethnographers. For what is at stake of conception, this being is also immersed at the is the very framework of theory and concepts center of a world of other persons—a social that we bring to our scientific project. Once world—and participates in the historical process again, the source of the problem lies in the thesis of its unfolding (Toren 1993: 470). Surrounded of complementarity. Developmental psycholo- by its entourage of adults, the infant con- gists could afford to ignore culture, so long tributes—by way of its presence and activities— as they concerned themselves with supposedly to the latter’s growth and development, just as universal mechanisms of acquisition, whose they contribute to its own. structure and functioning were conceived to be To be sure, children are different. For one indifferent to the specificities of the acquired con- thing, they are physically smaller, so that the tent. But conversely, social anthropologists could environment affords them possibilities of action afford to ignore children, so long as they were re- that are not available to grownups, and of course garded as incomplete adults whose personhood constrains what they can do as well—especially if was not yet fully formed and who had still to take it is full of structures built to adult dimensions. It on the total complement of cultural knowledge is reasonable, too, to distinguish degrees of ma- from their predecessors. turity in the life histories of organism-persons. In a sense, anthropology would have rather Neurophysiologically, the brain of the adult hu- not had to deal with children for the same reason man is more complex than that of the small child. that it has shunned inquiry into human origins. It is not reasonable, however, to equate smallness In both cases, the received theoretical wisdom or immaturity with a state of incompletion. implies a transition from an initial state of bio- Persons, as I have shown, are never complete, logical existence, defined in terms of naturally never finished, but undergo continual develop- 274 Tim Ingold

ment within fields of relationships. The image of out of certain ways of doing things, and grows the child as an incomplete person has its source in into others. But no one has ever grown out of the complementarity thesis, with its assumption biology, nor has anyone grown into society or that humans come into the world with their ca- culture. pacities already in place, waiting to be filled up with cultural content. A classic statement to this effect comes from Geertz (1973: 50): “Between Conclusion what our body tells us and what we have to know in order to function, there is a vacuum we must Throughout this chapter, I have argued against fill ourselves, and we fill it with information (or the idea that human beings participate concur- misinformation) provided by our culture.” The rently in two distinct worlds, of nature and implication is that children’s ability to function in society, figuring as biological individuals in the the world is at best imperfect. Yet as Toren has former and as cultural subjects in the latter. In- rightly observed, “children have to live their lives stead, I propose that we consider humans as in terms of their understandings just as adults do; indistinguishably organisms and persons, partici- their ideas are grounded in their experience and pating not in two worlds but in one, consisting of thus equally valid” (1993: 463). Adults and chil- the entire field of their environmental relations. dren may, then, function differently from one Figure 19.3 illustrates schematically the contrast another, but no better or worse. between these two views. Needless to say, the The complementarity approach, in effect, hides environment of a person will include beings of children from view behind a category of child- many kinds, both human and nonhuman, to hood which marginalizes them, or even excludes which that person will relate in different ways de- them altogether, from full participation in social pending on their particular qualities and charac- life. The obviation approach, by contrast, brings teristics, and on the project in hand. As one children out into the open, but it does so by dis- passes from relations with humans to relations solving the categorical distinction between child- with nonhuman animals, plants, and inanimate hood and adulthood. Children and adults are no objects, there is no Rubicon beyond which we longer conceived to stand on either side of a can say of any relation that it is directed toward boundary between becoming a person and being things in nature rather than persons in society. one, between undergoing socialization and par- For as the edge of nature is an illusion, so too is ticipating in social life, between acquiring cul- the image of society as a sphere of life that exists tural knowledge and applying it in practice, or in beyond it (Ingold 1997: 250). But by the same short, between learning and doing. Children are token, in the project of scholarly research, there persons just as adults are, and their knowledge can be no absolute division of method and objec- and skills are likewise developed through partici- tive between studying the lives and works of hu- pation both with other children and with adults mans and of nonhumans. Why, then, should the in the joint practical activities of social life. This participatory and interpretative approaches of is not to say, however, that children and adults the arts and humanities be limited to the study of are the same. It is possible to speak of children human subjects? And why, conversely, should the without a special category of childhood, simply observational and explanatory approaches of sci- in recognition of the inherent temporality of hu- ence be limited to the domain of nonhuman “na- man life, of the fact that organism-persons grow ture”? Why, indeed, should these approaches be older, increasing in skill and maturity—and in separated at all? that sense also in knowledge—as they do so. Ever since its relatively recent inception, the In the course of this aging process, one grows credibility of “social science” has been compro- mised by the recognition that the observer of hu- From Complementarity to Obviation 275

Figure 19.3 A schematic comparison of the complementarity and obviation approaches (after Ingold 1996c: 127). In the complementarity approach (upper diagram), every human being is, in part, a person in society and, in another part, an organism in nature. In the obviation approach (lower diagram), the human being is a person-organism situated in an environment of human and nonhuman others. 276 Tim Ingold

man behavior is necessarily a participant in the dividual members for them to use in their every- field of observation. In this vein, numerous critics day lives, but are rather generated and sustained have pointed out that participant observation, within the contexts of people’s engagements with the methodological crux of social anthropologi- one another and with nonhuman components of cal inquiry, is a contradiction in terms. To partic- the environment. If this applies to knowledge in ipate, it is said, is to swim with the current, to general, in must apply to anthropological knowl- observe is to stand on the bank: how can one pos- edge in particular. In the field, anthropologists sibly do both at once? Now it is doubtless true learn; in the classroom they teach. This does not that scientific inquiry of any kind depends upon mean, however, that they are receiving knowl- observation. But there is more to observation edge in the first case and transmitting it in the than mere spectating. A disinterested bystander second. For in both, whether with local people or who did not, in some way, couple the movement with students, they collaborate in the dialogic of his or her attention to the surrounding cur- processes of its creation. It is through bringing rents of activity, who failed to watch what was the two dialogues, in the field and the classroom, going on, would see much, but observe nothing. into a productive interplay that anthropological Observation, in short, is itself an environmentally knowledge is generated. It follows that the dia- situated activity that requires the observer to logue in the classroom is as important, and as place himself or herself, in person, in a relation of integral to the anthropological project, as the active, perceptual engagement with the object of dialogue in the field. Belatedly, we have begun attention. It is from this kind of sensory parti- to recognize the contribution that local collabo- cipation, proceeding against the background of rators—erstwhile “informants”—have made to involved activity in the wider environment of the advance of our subject. It is high time we rec- human and nonhuman others, that all scientific ognized the contribution of students as well. knowledge grows. Thus, whether our concern be with humans or nonhumans, there can be no observation without Notes participation, no explanation without interpreta- 1. Because these others may be nonhuman as well as tion, no science without engagement. As one such human, there is nothing strange about the extension science, I believe that anthropology is destined of kinship relations across the species boundary that to take its place as part of a broader ecological is commonly taken for granted among non-Western study of the relations between organism-persons peoples. and their environments, premised on the inescap- 2. This position is beautifully epitomized, and paro- able fact of our involvement in the one world died, in the title of a recent article by Morton, “The in which we all live (Ingold 1992: 693–694). Any organic remains” (Morton 1995). divisions within this field of inquiry must be rela- 3. In this vein, for example, Goldschmidt (1993: 355) tive rather than absolute, depending on what is writes of the “dynamic relation between the genetic and selected as one’s focus rather than on the a priori the cultural, between biology and anthropology.” His separation of substantive, externally bounded equation of biology with genetic programming leads him to the bizarre thought that even the human em- domains. I hope, in this chapter, to have given bryo, to the extent that its development is affected by some idea of how we might proceed with recon- environmentally specific “intra-uterine experiences,” structing the discipline along these lines. could not be “purely biological” since it would already I would like to end, however, with a word have acquired a modicum of culture (1993: 357, n. 19). about the teaching of anthropology. I have 4. While recognizing that these processes, of individual shown that the forms of human knowledge are ontogeny and evolutionary phylogeny, are distinct, bi- not made by society, and handed down to its in- ologists do not deny that there are connections between From Complementarity to Obviation 277

them. Thus the circumstances of ontogenetic devel- Bender, B. (Ed.) (1993). Landscape: Politics and Per- opment, insofar as they affect genetic replication, may spectives. Oxford: Berg. exert an influence on evolution; conversely the evolved Bloch, M. (1991). Language, anthropology and cogni- genetic specification is supposed to establish a schedule tive science. Man 26: 183–198. for development (Hinde 1991: 585). Boas, F. (1940). Race, Language, and Culture. New 5. As Kroeber wrote, “All evidence directs us to the York: Free Press. conviction that in recent periods civilization has raced Bourdieu, P. (1977). Outline of a Theory of Practice. at a speed so far outstripping the pace of hereditary Cambridge: Cambridge University Press. evolution, that the latter has, if not actually standing Connerton, P. (1989). How Societies Remember. Cam- still, afforded all the seeming, relatively, of making no bridge: Cambridge University Press. progress” (1952: 51). Cosmides, L., J. Tooby, and J. H. Barkow. (1992). 6. The “finished artifact fallacy” has its precise coun- Introduction: Evolutionary psychology and conceptual terpart in standard notions of socialization or encultur- integration. In J. H. Barkow, L. Cosmides, and J. ation as the working up of human raw material into Tooby (Eds.), The Adapted Mind: Evolutionary Psy- finished forms, ready for entry into social life. chology and the Generation of Culture, pp. 3–15. New 7. I am compelled here to use the neuter pronoun “it,” York: Oxford University Press. with regard to persons, rather than “he/she,” in recog- Csordas, T. (1990). Embodiment as a paradigm for an- nition of the fact that gendering is itself an aspect of the thropology. Ethos 18: 5–47. way in which relations are enfolded in the consciousness and identity of the self. D’Andrade, R. G. (1981). The cultural part of cognition. Cognitive Science 5: 179–195. 8. “What is very essential to the whole notion,” Bartlett writes, is “that the organised mass results of Davidson, I., and W. Noble. (1993). Tools and lan- past changes of position and posture are actively doing guage in human evolution. In K. R. Gibson and T. something all the time; are, so to speak, carried along Ingold (Eds.), Tools, Language and Cognition in Human with us, complete, though developing from moment to Evolution, pp. 363–388. Cambridge: Cambridge Uni- moment” (Bartlett 1932: 201). There are, I think, strik- versity Press. ing similarities between Bartlett’s notion of schema and Durkheim, E. (1982). The Rules of Sociological Method. Bourdieu’s (1977) of habitus. Both terms suggest an ac- S. Lukes (Ed.), W. D. Halls (Trans.). London: Macmil- tive, dynamic organization of past experience, rather lan. (Original work published 1895.) than a passive, static framework for accommodating it. Geertz, C. (1973). The Interpretation of Cultures. New 9. Connerton prefers the notion of habit to that of York: Basic Books. skill, arguing that as a habit becomes established, Gibson, J. J. (1979). The Ecological Approach to Visual “awareness retreats,” leading ultimately to bodily au- Perception. Boston: Houghton Mi≥in. tomatisms (1989: 93–94). I do not think it is right to de- Godelier, M. (1989). Incest taboo and the evolution of scribe the movements of the musician or craftsman as society. In A. Grafen (Ed.), Evolution and its Influence, habitual in this sense. In such skilled activity, awareness pp. 63–92. Oxford: Clarendon Press. does not retreat but becomes one with the movement it- Goldschmidt, W. (1993). On the relationship between self. This movement, far from being automatic, carries biology and anthropology. Man 28: 341–359. its own immanent intentionality (Merleau-Ponty 1962: 110–111). Halbwachs, M. (1992). On Collective Memory. Chicago: University of Chicago Press. (Original work published in 1950). References Hinde, R. A. (1991). A biologist looks at anthropology. Man 26: 583–608. Bartlett, F. C. (1932). Remembering: A Study in Experi- Ingold, T. (1983). The architect and the bee: Reflections mental and Social Psychology. Cambridge: Cambridge on the work of animals and men. Man 18: 1–20. University Press. Ingold, T. (1990). An anthropologist looks at biology. Bateson, G. (1973). Steps to an Ecology of Mind. Man 25: 208–229. London: Fontana. 278 Tim Ingold

Ingold, T. (1992). Editorial. Man 27: 693–696. Mellars, P., and C. Stringer. (Eds.). (1989). The Human Ingold, T. (1993a). Technology, language, intelligence: Revolution: Behavioural and Biological Perspectives on A reconsideration of basic concepts. In K. R. Gibson the Origins of Modern Humans. Edinburgh: Edinburgh and T. Ingold (Eds.), Tools, Language and Cognition in University Press. Human Evolution, pp. 449–472. Cambridge: Cambridge Merleau-Ponty, M. (1962). Phenomenology of Percep- University Press. tion. (C. Smith, Trans.) London: Routledge & Kegan Ingold, T. (1993b). The temporality of the landscape. Paul. World Archaeology 25: 152–174. Morton, J. (1995). The organic remains: Remarks on Ingold, T. (1995a). “People like us”: The concept of the the constitution and development of people. In J. anatomically modern human. Cultural Dynamics 7: Morton and M. Macintyre (Eds.), Persons, Bodies, 187–214. Selves, Emotions, pp. 101–118. Special issue of Social Analysis Ingold, T. (1995b). Building, dwelling, living: How ani- , no. 37. Department of Anthropology, Univer- mals and people make themselves at home in the world. sity of Adelaide. In M. Strathern (Ed.), Shifting Contexts, pp. 57–80. Ots, T. (1994). The silenced body—the expressive London: Routledge. Leib: On the dialectic of mind and life in Chinese Embodiment Ingold, T. (1996a). The history and evolution of bodily cathartic healing. In T. J. Csordas (Ed.), and Experience: The Existential Grounds of Culture and skills. Ecological Psychology 8: 171–182. Self, pp. 116–136. Cambridge: Cambridge University Ingold, T. (1996b). Introduction to the 1992 debate, Press. “The past is a foreign country.” In T. Ingold (Ed.), Key The Ontogeny of Information: Devel- Debates in Anthropology, pp. 201–205. London: Oyama, S. (1985). opmental Systems and Evolution. Routledge. Cambridge: Cam- bridge University Press. (2d rev. ed., Durham, NC: Ingold, T. (1996c). Hunting and gathering as ways of Duke University Press, 2000.) perceiving the environment. In R. Ellen and K. Fukui (Eds.), Redeﬁning Nature: Ecology, Culture and Domes- Quinn, N., and D. Holland. (1987). Culture and cogni- Cultural tication, pp. 117–155. Oxford: Berg. tion. In D. Holland and N. Quinn (Eds.), Models in Language and Thought, pp. 3–40. Cambridge: Ingold, T. (1997). Life beyond the edge of nature? Or, Cambridge University Press. the mirage of society. In J. D. Greenwood (Ed.), The Mark of the Social, pp. 231–252. Lanham, MD: Reed, E. S. (1988). The affordances of the animate Rowman and Littleﬁeld. environment: Social science from the ecological point of view. In T. Ingold (Ed.), What Is an Animal? pp. IUAES (International Union of Anthropological and 110–126. London: Unwin Hyman. Ethnological Sciences). (1996). The IUAES and the new Unesco statement on race. Human Peace (IUAES Robertson, A. F. (1996). The development of meaning: Journal of the Royal Anthropo- Quarterly News Journal of the Commission on the Ontogeny and culture. logical Institute Study of Peace) 11(1): 18–20. 2: 591–610. Jackson, M. (1989). Paths toward a Clearing: Radical Rubin, D. C. (1988). Go for the skill. In U. Neisser and Remembering Reconsidered: Eco- Empiricism and Ethnographic Inquiry. Bloomington: E. Winograd (Eds.), logical and Traditional Approaches to the Study of Indiana University Press. Memory, pp. 374–382. Cambridge: Cambridge Univer- Johnson-Laird, P. N. (1988). The Computer and the sity Press. Mind: An Introduction to Cognitive Science. London: Ethos Fontana. Schwartz, T. (1981). The acquisition of culture. 9: 4–17. Kroeber, A. L. (1952). The Nature of Culture. Chicago: On Anthropological Knowledge. University of Chicago Press. Sperber, D. (1985). Cambridge: Cambridge University Press. Mauss, M. (1979). Body techniques. In Sociology and Psychology: Essays by Marcel Mauss (B. Brewster, Strathern, M. (1996). For the motion. Contribution to Trans.), part IV, pp. 97–123. London: Routledge & the 1989 debate, “The concept of society is theoretically Key Debates in Anthropol- Kegan Paul. (Original work published in 1934). obsolete.” In T. Ingold (Ed.), ogy, pp. 60–66. London: Routledge. From Complementarity to Obviation 279

Tilley, C. (1994). A Phenomenology of Landscape: Places, Paths and Monuments. Oxford: Berg. Toren, C. (1993). Making history: The signiﬁcance of childhood cognition for a comparative anthropology of mind. Man 28: 461–478. Weiner, J. (1991). The Empty Place: Poetry, Space and Being among the Foi of Papua New Guinea. Blooming- ton: Indiana University Press. Wikan, U. (1992). Beyond words: The power of resonance. American Ethnologist 19: 460–482. Wolf, E. (1994). Perilous ideas: Race, culture, people. Current Anthropology 35: 1–12.

V RESPONSES TO DEVELOPMENTAL SYSTEMS THEORY

On the Status and Explanatory Structure of Developmental Systems 20 Theory

Peter Godfrey-Smith

What kind of theory is DST? Is it a scientific the- Kitcher on the issue of holism (p. 382). Although ory or a philosophical theory? Is it an empirical Sterelny, Smith, and Dickison do not think the hypothesis, a suggested program of research, a problem is insoluble, they think DST does face philosophical gloss on our existing knowledge, or a problem from its apparent commitment to a what? What difference does it make whether or holistic view of causation in development and not the central ideas associated with DST are evolution.2 true?1 I will outline one way for DST to steer a path DST advocates and critics alike have won- through these problems. I will describe two intel- dered about this issue. Within the movement, lectual projects that need not be distinct, but that some, like Russell Gray, have tried to be very can be distinct, and that are certainly consistent explicit about the difference made to concrete sci- with each other. I will sort some of the main DST entific research by accepting DST (Gray 2000; ideas into different categories associated with this Gri≤ths and Gray 1994). Oyama, it seems to me, distinction, and in doing so I hope to lay to rest has been consistently more cagy about tying DST (among other things) the problem of causal to some specific direction of empirical research, holism for DST. More generally, I will outline a and has presented DST more as a very general way for DST advocates to conceive the place and abstract “way of seeing” the biological world their program occupies within science and philos- and our investigation of it. These two types of ophy. So the first part of the paper is entirely con- emphasis do not conflict, of course. DST can be structive, from the DST point of view. a general way of seeing, and that way of seeing The rest of this chapter will be slightly more can suggest specific projects of empirical work. critical. Although there is no question that Still, it can be hard to sort through the collection “DST” names a large and somewhat hetero- of ideas associated with DST, and work out what geneous collection of ideas, I suggest that one sort of role each idea plays. helpful way to think about DST is to think of it One critic of DST, Philip Kitcher, has com- as an assertion of very strong antipreformation- plained that DST does not “offer anything that ism about development. This idea is the origin of aspiring researchers can put to work” (Kitcher in much else in DST. In particular, it is helpful in press). Perhaps Kitcher’s overall view would be understanding the DST critique of the “informa- more accurately expressed by saying that he tional” perspective of the gene. For DST, the thinks that mostly DST does not offer anything informational gene is the preformationist’s last that aspiring researchers can put to work, and stand. No one can deny that many forms of pre- when it does offer something, what it offers is formationism turned out to be wrong. But I will an insistence on taking every causal factor in suggest in the second part of this paper that DST biology equally seriously in every investigation, might sometimes go too far in its rejection of the an insistence that would simply shut down scien- preformationist pattern of explanation. tific activity. Kitcher thinks that both DST and Lewontin’s “dialectical biology” tend toward an extreme holism that is inconsistent with the DST as Research Program and as Philosophy of practice of ordinary empirical research. The crit- Nature icisms of DST in Sterelny, Smith, and Dickison’s “The Extended Replicator” (1996) are much less DST is a collection of ideas about development, damning of DST than Kitcher’s, but Sterelny, causation, inheritance, and evolution. But is it a Smith, and Dickison converge somewhat with collection of scientific ideas, philosophical ideas, 284 Peter Godfrey-Smith

or both? I suggest that DST is a contribution to by a number of different sciences, some nonscien- two different kinds of intellectual project. Firstly, tific vocabulary will probably be needed in most DST can be regarded as a proposal for a scientific cases. Also, when scientific ideas are relevant to research program. DST contains a set of core other problems treated in philosophy—for exam- negative and positive ideas about biological sys- ple, ethical and political problems—the scientific tems. These ideas do have the ability to steer bio- ideas should be fed into such discussions in a logical research in particular directions, and they philosophically processed form, not in the raw have the ability to be confirmed and disconfirmed language of science. through empirical testing. The term philosophy of nature has a history The second project that some DST ideas con- that might be suspicious to many; it might sug- tribute to is what I will call philosophy of nature.3 gest a project in which the philosopher erro- When doing philosophy of nature in my sense, a neously tries to do science from the armchair, by writer comments on the overall picture of the nat- deploying special logical tools of dubious value ural world that science, and perhaps other types and arrogantly saying how the course of empir- of inquiry, seem to be giving us. But this com- ical science must necessarily go.4 I stress that mentary does not have to use language in the nothing like that is meant here. “Philosophy of same way that scientists find convenient for their nature” in my sense comes after empirical science own work. It can use its own categories and con- and tries to redescribe structures in the world that cepts, concepts developed for the task of describ- have already been described by the sciences. A ing the world as accurately as possible when a good philosophy of nature makes no empirical range of scientific descriptions are to be taken claims that are inconsistent with those made in into account, and when a philosophical concern the relevant sciences. Philosophers are at liberty with the underlying structure of theories is appro- to make novel empirical suggestions, of course. priate. The claims made by a good philosophy of They are also sometimes able to help science by nature do have to be consistent with the claims describing and exploring whole new kinds of the- made by science. But the concepts employed by a oretical structure, at a stage when these theoret- good philosophy of nature do not have to be the ical structures cannot yet be brought into much same as those used in the relevant science, and the contact with empirical methods. In recent work, organization and presentation of information in Michael Friedman has been making a very im- the two projects can be quite different. pressive case for the importance of this kind of I say the concepts “might” be different, and the “forward-looking” role of philosophy in relation presentation of information “can” be different— to science (1999). I agree with Friedman on the they need not be different. It might turn out that importance of this role, but this is a different role the concepts used every day by scientists are ex- from the one I am describing here. actly the right ones for doing philosophy of na- So “philosophy of nature” in the present sense ture as well. Any degree of overlap is possible. is not the same as the kind of “naturalistic phi- The key point is that a commentary on scientific losophy” that conceives all philosophical work as knowledge that serves the purposes of philosophy completely continuous with science. That sort of of nature can, if necessary, fashion its own way of philosophy, associated in particular with Quine describing the structure that has been uncovered (1969) but influential in various other forms, by science. The demands on a good scientific gives up the autonomy of philosophy with respect vocabulary can be quite different from the to its choice of questions, and hopes instead to demands on a vocabulary used in philosophy of make substantial contributions to the overall sci- nature. Indeed, because a philosophy of nature entific project. While holding that philosophy typically tries to accommodate the claims made requires constant input from science, I oppose Status and Explanatory Structure of DST 285

the kind of naturalism that requires philosophy philosophical descriptions of research programs to give up any aspiration to formulate and (though some of my language will recall those address its own distinctively philosophical set of discussions). In part this is because DST is some- questions. thing of a special case; it would be an error to say My claim that philosophy of nature comes that DST is a research program in just the same “after” science should not be taken to deny an sense as Newtonianism, Darwinism, or classical interaction between more philosophical and AI. But I do think there is some value in extract- more empirical commitments within science. ing a “research program” out of DST. Lewontin (personal correspondence) expressed I think of the DST research program as includ- the suspicion that the separation I am describing ing at least a set of core empirical claims, a set of between scientific description and philosophy of suggestions about which sorts of scientific work nature implies the view that science itself can are likely to pay off, and a set of concepts that are proceed in a purely empirical way unaffected by to be used in doing and describing the work. To philosophical doctrines and commitments. But I this basic set of elements, some might want to accept, with Lewontin, that empirical work with- add others, such as a set of habits of mind and a in science is guided—in different ways and in dif- set of standards for good explanations. Different ferent degrees—by philosophical ideas. As Kuhn philosophies of science differ about the “thick- (1970) stressed, abstract views about the proper ness” of anything that deserves to be called a relation between theory and evidence and the research program, tradition of normal science, or ultimate composition of the world are important consensus practice.There is no need to take sides parts of traditions of “normal science.” Still, I here on questions like that, as we can proceed hold that one valuable and distinct type of work with a simple and fairly “thin” concept of what a that philosophy can do is work that does aspire research program is like. to come “after” science. This is work that tries A research program has to contain sugges- to give a careful philosophical redescription of tions for empirical work. What sort of work is the picture of the world that science seems to be supposed to be done? Here DST is in an unusual delivering. This philosophical work might well situation, as part of what it does is pick out a come to have an effect on the science itself; it scattered body of existing scientific work, some of might change the hidden or overt philosophical it quite old, and claim: “there should be more of commitments of the scientists. But the absence of this.” In a footnote to his critical paper (2000, such an effect on science does not rob the philo- note 6), Kitcher says he would like to see what an sophical work of its value. example of DST-guided work in developmental Turning back to DST, my suggestion is that biology would look like, or an example of dialec- existing DST writings tend to combine contri- tical population genetics. The right DST reply is butions to two projects—describing a scientific to say that there is already a good deal of work in research program and outlining a philosophy of developmental biology and population genetics nature. As the two projects are associated with that does everything (or nearly everything) that different goals, different types of assessment are DST could ask for. Some of it is reviewed in part relevant in each case. I will sort through some 1 of this volume. This work exists, but in main- DST claims, and criticisms made of those claims, stream biological thinking it is often regarded in this light of this categorization. as describing oddities, details, and exceptions— In what sense is DST a scientific research pro- not as describing cases that provide models for gram? Is this a “research program” in the sense of thinking about development and evolution in Lakatos (1970) or Laudan (1977)? In this chapter general. In mainstream biological thinking, the I will not try to link DST with those existing developmental work that is supposed to provide 286 Peter Godfrey-Smith

a model is work on the expression of specific vey of these phenomena. In order to categorize genes, and on how genes and their products exert some of these cases as real extragenetic inheri- sensitive control on developmental processes.5 tance, DST has to engage in some conceptual According to DST, the mainstream view erro- battles over what counts as inheritance, and over neously holds that work focused on gene expres- possible mainstream redescriptions of these in sion provides a general model, while work on standard genetic terms.7 Those questions are very such things as the inheritance of cytosine methy- di≤cult. At least part of DST’s defence of its pre- lation patterns and endosymbionts merely deferred way of categorizing cases of inheritance scribes interesting oddities. must be the claim that if the DST categorization Lewontin can reply to Kitcher in similar terms; of types of inheritance is used, new patterns will there is lots of existing work in population genet- emerge and new insights will result. Perhaps there ics that does fit the framework of “dialectical are some risks of circularity here, but I do not see biology” reasonably well, and lots that does not. these risks as worse than they are in other cases For Lewontin, epistasis in fitness relations is to involving categorization in science. Here as else- be regarded as the rule rather than the excep- where, a categorization is to be judged, at least in tion. A high degree of sensitivity of evolutionary part, by its empirical fruits. And there are cer- predictions to the details of genetic systems and tainly some relevant phenomena that are agreed background parameters is to be expected. Lew- on all sides to constitute extragenetic inheritance. ontin also asserts the importance of “coupled” Either these cases are rare and make little differ- causal structures, in which variables affect each ence to the working of the biological world, or other symmetrically, rather than in an asymmet- they are not. rical way where one variable is always cause and So DST claims that extragenetic inheritance is the other is effect.6 common, and that once it is demarginalized we So part of the project in recent DST literature will improve our understanding of the way traits is an attempt to draw together a large number of appear and reappear across generations. This I existing pieces of scientific work and to suggest take to be perhaps the clearest case of a DST that when these are rightly interpreted and con- idea that constitutes part of a scientific research sidered together, they suggest a project for fur- program.8 I stress it here for illustrative purposes, ther empirical work. For DST advocates, there is but there are others as well. Other empirical already a scattered collection of successful scien- applications of DST, and explicit outlines of tific work that provides a set of models and con- DST-based research programs, are discussed by cepts to move ahead with. scientists in other chapters of this book. (See, for Earlier I said that DST as research program is example, the chapters by Nijhout and by Weber committed to a set of empirical claims. Which and Depew.) claims are these? Any suggestion will probably I turn now to the second project that DST can generate debate, but for purposes of illustration I be seen as undertaking, the project of “philoso- will focus on one key example. DST the research phy of nature.” As I said, a philosophy of nature program holds that extragenetic inheritance is is an attempt to describe the world in a way that more common and more causally important than is closely informed by scientific theories, but that mainstream views realize. Gray (in press) dis- is free to reject the vocabulary and perhaps some cusses a range of examples that he views as espe- of the classifications and interpretations of the cially suggestive for DST, and these range from world associated with the relevant sciences. I gut microorganisms through social behaviors to stress again that it is possible for someone to sug- habitats. Jablonka’s contribution to the present gest that standard scientific descriptions already volume also contains a very well organized sur- constitute an adequate philosophy of nature, or Status and Explanatory Structure of DST 287

that scientific descriptions need only a minimal theoretical language of biology. Terms like tran- additional commentary before they will func- scription, translation, proof-reading, and synony- tion as an adequate philosophy of nature. But the mous are all technical terms in biology, with quite other view can be argued as well—the view that precise conditions of use. It might be argued that scientific descriptions of the world typically need these concepts are now essential to ordinary, pro- a lot of massaging and interpretation before they ductive, empirical research in genetics. Though can be seen as our best-possible description. derived at least semimetaphorically from the case Why should this be? Why should there be a of language, they are just as deeply embedded as need to add to or modify the scientific descrip- any other specialized terms in the practice and tion? One’s answer to this will depend on general thinking of geneticists. views about how science works and about scien- DST has criticized the symbolic, informational tific language. I will illustrate with a simple case perspective on the gene extensively. I suggest that that is also highly relevant in the present context. there are two distinct criticisms DST makes here. Suppose one holds that metaphorical and ana- First, there is no doubt that if DST is a research logical language is essential to science, even in program, the program includes, as a negative highly developed fields. When I say “essential” I heuristic, a rejection of that conceptual frame- mean something strong; I mean that ordinary work. DST makes an empirical bet that biology scientific work requires, for communication be- would be better off without this way of talking tween scientists and for creative thinking, the about genes—or at least that it would be better constant deployment of a rich set of metaphors off with a drastically reduced role for such de- when thinking about some particular system. scriptions. Maybe DST is right about this. But Perhaps it is hard to imagine a case when the sci- suppose for a moment that DST is wrong, and entists will not be able to drop, perhaps with for a particular reason. Someone like Maynard some struggle, the metaphors. But it might be the Smith might argue that it is no accident that the case in some particular field that the metaphors introduction of the symbolic perspective on genes are never dropped, and hence that they become coincided with an avalanche of scientific progress tightly fused to the nonmetaphorical language in this area. And DST does not deny the ava- of that field. Then one task for philosophy of lanche. Maybe the symbolic perspective on genes nature will be to come along afterward and sort just happens to be a uniquely useful framework through the scientific descriptions, distinguishing for guiding empirical work in this area. If that is the parts that are properly taken literally from true, it is true not just because of what genes are the parts that must be regarded as merely like, but also because of what our minds are like. metaphorical.9 A conceptual framework like this might, for In the present context, the status of “genetic a mixture of reasons, turn out to be ideal for coding” is a relevant problem case. There has human rumination and communication about been a growing philosophical discussion of what genetics. That could be true even if genes have exactly we should make of the language of genet- many properties that are not captured by the ic coding in mainstream biology (Sarkar 1996; symbolic perspective. If genetics is complicated Godfrey-Smith 2000a), and this is associated with enough, and if everyday scientific work can only discussions of the role of other semantic pro- proceed with the aid of some simplifying frame- perties in biology, and the concept of a “genetic work or another, then the question becomes program.” (See also Moss 1992 and Keller’s con- which is the simplifying framework best suited tribution to this volume.) But as Maynard Smith for us. (in press) stresses, the symbolic perspective on I am not saying that the possibility in the pre- genes has become completely embedded in the ceding paragraph is true; it might be totally 288 Peter Godfrey-Smith

wrong, and certainly DST bets that it is wrong. framework, when we try to give the most accu- DST bets that genetics would benefit in many rate possible description of what we know about ways from giving up the special kinds of seman- genes, we should not use those terms. (The status tic and computational descriptions of genes that of genetic coding will arise again below.) are now so common. But suppose the argument During discussion of the first drafts of papers in the preceding paragraph is right. Then the sci- in this volume, two of the editors suggested that entist has no reason to drop the symbolic per- I compare DST with the “gene’s eye view” on spective on genes, when doing scientific work. evolution, with respect to the duality between But the fact that the scientist finds the framework research program and philosophy of nature. To indispensable does not stop the philosopher (or some extent the comparison is appropriate, as we the scientist when wearing the hat of philosopher do find a bifurcation of this kind in the literature of nature) from taking an entirely different atti- on genic selection. But DST should be wary of tude. The philosopher of nature should ask: even the analogy. if this way of talking has great practical value, The “gene’s eye view” is sometimes presented what is its real status? Is this a good literal de- as a specific program for empirical research. As scription of what genes are like, or is it a meta- such, it is intended to guide researchers away phor? Is it perhaps language that is being applied from looking for an organism-level benefit in all nonmetaphorically and being used to make liter- evolutionary explanations. Phenomena of meio- ally false claims, but that nonetheless has heuristic drive and junk DNA are taken to be instruc- tic value? In a situation when ease of scientific tive models that tell us how genes can proliferate communication and spurring of creativity count without helping organisms. This empirical orien- for nothing—that is, in a situation when the sci- tation is one aspect of the literature on the gene’s ence is done, for now—should we describe genes eye view, but this view coexists with Richard in these terms or not? Dawkins’s attempts to recast all of evolution in I said earlier that philosophy of nature also gene-selectionist terms, where this recasting is aims to “process” scientific descriptions prior justified by appeal to abstract and general argu- to their public assimilation and utilization in ments about the role of “replicators” (1976, other kinds of intellectual and practical decision- 1982). This latter type of argument for a gene’s making. The semantic perspective on genes eye view makes no particular claims about how appears, when used in nontechnical discussions often there will be a familiar organism-level in newspapers and the like, to give support to benefit involved in evolution; however common genetic determinist views widespread in the gen- these cases are, they can be redescribed as selec- eral population. Clarifying and criticizing such tion on genes. Even phenomena previously cate- connections is another role for philosophy of gorized as group selection can be redescribed in nature. this way. Dawkins’s The Selfish Gene can be seen So on a question like the status of “coding for” as outlining a philosophy of nature in which redescriptions, and other semantic talk in genetics, plicators are elevated to a position of primacy in I understand DST as saying two things. DST as evolution, and are regarded as the ultimate research program bets that science would be bet- beneficiaries of everything natural selection pro- ter off without such talk. But DST as philosophy duces. Replicators become the fundamental of nature should recognize that all sorts of differ- causal agents in the biological world.10 ent factors might influence the usefulness of a The analogy with the gene’s eye view illustrates conceptual framework in science. So DST as phi- some moves that DST should avoid. Dawkins losophy of nature should add that even if it turns and others have sometimes blurred the two con- out that genetics benefits from using the semantic struals of gene-selectionism in misleading ways. Status and Explanatory Structure of DST 289

A separate question is whether the gene’s eye ably inevitable that the researcher sort through view makes sense as a philosophy of nature. I causal factors and distinguish some as primary think it does not. The topic is too big for discus- and others as background conditions, and to sion here, but one reason is that, contra Dawkins, deny the researcher this strategy is to shut down replicators are not essential to evolution by natu- research. But holism per se is not a bad thing in ral selection (Godfrey-Smith in press b). philosophy of nature, because it is an error to I turn now to the specter of holism, which has demand that a philosophy of nature be a useful haunted DST in recent years and which has been tool in the laboratory, or a good heuristic for the subject of vigorous discussion. guiding research. In a philosophy of nature, If one is engaged in philosophy of nature, the holism is just one possible view about the causal unwieldiness of a description is no objection to structure of the world. that description. The heuristic emptiness of a As I stressed earlier, to make this point it is description is no objection to it either. Even if a not necessary to insist that the vocabularies and way of seeing the world would be positively stul- frameworks used in science and those used in tifying if adopted by the working scientist, that philosophies of nature are distinct. Whether they does not show anything wrong with that way of are distinct or not will depend on many facts, in- seeing things in the context of philosophy of cluding the flexibility of the language of scientific nature. research. There is nothing stopping a scientist So here we see one way of resolving the debates from first espousing a very holistic view of a sys- about holism. Critics of DST suggest that DST tem, and then going on to say he will relegate views about causation are holistic. And in some some causal factors to background status merely respects, they certainly are holistic. DST claims for immediate practical purposes. My separation that various causal factors are relegated to the of philosophy of nature from science is intended status of mere “support” or “background” by to show the possibility of a mismatch between the mainstream biological views, while other causal two, and the possibility of useful commentary on factors are regarded as primary or as the sources biology using a framework that is of little help to of form. But according to DST, this “privileging” an empirical scientist. of some causes over others makes no sense. Another way to put my point—a way more “What I am arguing for here is a view of causal- critical of DST—is this. One never knows in ity that gives formative weight to all necessary advance which heuristics and idealizations will be influences, since none alone is su≤cient for helpful in science. That is something only known the phenomenon or for any of its properties” after the fact, and it is an error to make strong (Oyama 1985: 15).11 For Oyama and others, it is predictions about such matters. Occasionally a crucial error to suppose that “some influences defenders of DST have done this, have said that are more equal than others” in the explanation of such-and-such is bound to be a useful heuristic in biological form. Rather than disregard some research.12 I suggest that we should never expect causal factors in order to “privilege” others, all to know such things in advance. But a redescrip- causal influences that bear on an event should be tion of development or evolution does not have accorded comparable status. These are the sorts to inspire scientists for it to be valuable. of claims to which the critics cry “holism!” and in Maybe defenders of DST will not want to some respects the label is entirely appropriate. accept the strategy offered in this section; they But holism per se is only a bad thing in situations might think that “philosophy of nature” in my in which it puts impossible demands on empirical sense is an empty exercise. Certainly some people investigation, situations in which it is paralyzing will think that it is empty or at least dubious. I do to scientific work. In empirical work, it is prob- suggest, however, that an appeal to the distinc- 290 Peter Godfrey-Smith

tion I have outlined is an important option to concerned to make a number of subtle and con- have on the table. troversial claims about what causation is like, Before leaving this topic, I will make a brief and how causal relations are structured within return to Kitcher’s criticism of DST. It is hard for biological systems. DST asserts strong views DST or any view like it to establish its superiority about causal connectedness, interpenetration, to a sparse and austere philosophy of nature. By “reciprocal selectivity,” the active and reactive “sparse” I mean a philosophy of nature that does properties of biological matter, and so on. These not recognize the reality of the distinctions that are all things that DST takes seriously which can DST makes at key places, and which treats some only be expressed, let alone defended, if one con- aspects of the DST description as mere color- ceives of causal facts in a rich and realist way. For ful talk. The opposition will be di≤cult for the someone like Kitcher, on the other hand, there richer view because the proponent of the sparser is little or nothing at issue when someone asks view will always claim to be able to reexpress the whether it is true that causation is “multiple, in- only significant parts of the richer view in his or terdependent and complex” (Oyama 1985: 32) her own sparse terms. Then a challenge will be and exhibits “reciprocal selectivity” (p. 15). To issued to the richer view to say why we should ask whether these things are true or not is to care about the remainder that cannot be sparsely choose between different colorful glosses that one expressed.13 In the context under discussion, might put on the world. The colors are different Kitcher has a sparse and austere view. Notably, but nothing substantial hangs on them. he is skeptical about strong claims about the DST is causally rich as a philosophy of nature. objective reality of causal relations. Kitcher is So is Lewontin’s dialectical biology, and so are prepared to engage in causal talk, but I think it various other views quite unlike these two. is fair to say he thinks that such talk should not Causally rich views must struggle hard, although be taken too metaphysically seriously. His own they might in the end prevail, when exposed to analysis of what is meant by causal talk is a so- the deflationary tendency of sparser philosophies, phisticated view, based on the concept of expla- like Kitcher’s. nation, which at bottom is an unusual version of the regularity theory of causation (1989). In science we collect and classify patterns that permit DST as Extreme Antipreformationism us to unify phenomena. The “causal” structure of the world is no more than this structure of pat- In this section I will discuss some DST views terns and regularities, viewed by us in a certain about the explanation of biological development. way. A consequence of this is that when writing I suggest that it is helpful to see DST as, among philosophically seriously about the nature of other things, an assertion of a very strong form of reality, for Kitcher a rather slim vocabulary is, in antipreformationism. principle, always su≤cient. All we need to do is Recall the eighteenth- and nineteenth-century to describe patterns in the flow of events—in debate between the preformationists and the 14 this respect Kitcher is an heir to the positivist defenders of epigenesis. Does the adult form tradition. preexist, in some way, in the fertilized egg, or If one has a view like this, one will be very is development a process in which order really resistant to views about biological systems (or does come from disorder? If the latter, it was anything else) that are driven by highly specific thought, this order must be brought about by and elaborate claims about causation and causal a special organizing force acting on the inert dependence. DST is a view of that kind. DST, matter. especially in Oyama’s and Gray’s versions, is Oyama (1985) discusses this debate, and sees her view as denying an assumption common to Status and Explanatory Structure of DST 291

both sides, the assumption that matter cannot within the fertilized egg. That view marks out acquire biological form without there being some one pole in the conceptual landscape. Consider external source of this form. So she does not ally next two ways to move away from this extreme herself with the traditional “epigenetic” form of view. opposition to preformationism. That is reasonable enough, but it is possible to recast this old Option 1: The zygote contains a blueprint or set dispute within a more modern and naturalistic set of instructions for the adult organism. of assumptions. Then when we ask the question Option 2: The zygote contains a complete set of of DST I think we get a different answer. Is devel- parts needed to build an adult organism, which opment a process in which, despite appearances, have the ability to self-assemble after assuming a good deal of the structure of the organism- the proper size and/or multiplying to the right to-come preexists in the fertilized egg, or is it a numbers. Here “self-assembly” is conceived as a process in which comparatively little structure is fairly trivial matter, once one has the parts. pre-given, and the active powers of biological Option 1 is familiar. Option 2 is a somewhat matter, working through local causal interac- modernized version of a historical version of pre- tions, give rise to complex eventual structures in formationism (Gould 1977: 20). Is there any truth reliable ways? Once expressed like this, it seems in either of these views? Certainly many people to me that much of the thrust of DST is to assert write as if option 1 is true. Some who write this the latter option, the antipreformationist option. way mean it literally and others think it is a use- The denial of preformationism is connected ful metaphor. I accept, with DST, that option 1 is to DST’s views about the attribution of infor- not literally true, or close to true, and not a very mational and semantic properties to genes. For helpful metaphor either. DST, the idea that genetic information is the Is there any truth in option 2? Here matters source of form in development is a lapse back are a bit more complicated. Certainly a simple into preformationist error. According to DST, version of 2 is not at all true, when given as an the informational gene is the preformationist’s explanation of development. The “parts” that last stand; it expresses the idea that although the historical defenders of this kind of view had in adult structure does not preexist in the egg, some- mind are certainly not present intact in the thing just as good does. Against this, Oyama zygote. But there is an element of truth in option asserts that information is not something that 2, and DST does not deny this. A zygote does preexists the biological interactions with which it contain structures that will also appear in the is associated. Rather, information (as her 1985 adult; the zygote contains intact some adult part title suggests) has its own ontogeny. Preforma- types—chromosomes, mitochondria, mem- tionism is denied even for information. branes, and so on. I suggest that DST might have a tendency to DST not only accepts this, but it also extends go too far on this point. The preformationist the list of inherited parts, via the extended DST pattern of explanation did not turn out to be view of inheritance (as Sterelny stressed to me). so completely misguided. To suggest this, I will The “persistent resources” in an organism’s envi- ﬁrst work through a series of different kinds of ronment are, for DST, genuine parts of DST’s preformationist view, and then show the con- analog of the adult organism—the late stages tinuity between a reasonable interpretation of in the life cycle. Resources such as habitats and modern theories, and preformationist styles of energy sources are parts of later stages, and they explanation. are also present at the initial stages. This is not One extreme kind of preformationism is true of all parts of course, or of global features of wholly false; this is the view in which a more or the organic structure; those do not preexist, but less complete adult organism exists in miniature 292 Peter Godfrey-Smith

are causal products of the interactions of various struggled with this elsewhere (2000a), and DST “resources” that exist at the earlier stages. So tends to oppose this type of description of genes. DST accepts that some key parts are present, I do not think the exact status of genetic coding is either as tokens or as members of a closely copied crucial to the present point. At least it is correct lineage, at the earliest stages. DST denies that to say that the DNA sequences contain structures the existence of these parts vindicates preforma- that act as templates for the proteins produced.15 tionism about the eventual biological forms—the That claim should su≤ce for the purposes of later stages of the life cycle. present discussion. In fact though, I hold that So option 2 is generally false, according to there are good reasons for claiming that genes DST and according to standard views, even code for proteins. though it has a few grains of truth. I now intro- I have argued so far that it is possible to pres- duce another descendant of the original naive ent modern genetics as having an explanatory preformationist view, one which combines as- structure that descends nontrivially from prefor- pects of option 1 and option 2 above. mationism. Consider now Oyama’s reaction to a related claim made by Stephen Jay Gould. Gould Option 3: A large and crucial set of components says “Modern genetics is about as midway as of the adult (and of all developmental stages) are it could be between the extreme formulations not materially present themselves in the zygote, [of preformationism and epigenesis] of the eigh- but are coded for in structures within the zygote. teenth century” (1977: 18; quoted in Oyama These components make up most of the machin- 1985: 24). Oyama rejects this “Golden Mean” ery by which the organism stays alive and con- resolution of the old debate, thinking the debate trols its passage through the developmental rested on false assumptions. But there is an im- sequence. portant element of truth in Gould’s claim. Our The relevant parts here are amino acid chains analysis of genetic causation should make sense —which when folded and processed become of the empirical fact that preformationism turned functional protein molecules. So this view is pre- out to contain an element of truth, of a perhaps formationist about parts, not the whole, and the unexpected kind. If a general analysis of causa- parts themselves are not physically present in the tion, development and information denies this, it zygote but instead are coded for. has probably overextended. We have arrived, of course, at one aspect of the Does DST overextend? It is hard to say. Many view found in mainstream modern genetics. This strands of the DST literature bear on this point, is not the only way to describe what modern and there is no simple resolution. At another genetics has taught us, but I claim that it is one point Oyama says: “What is transmitted [in re- reasonable way to describe the present picture. production] is macromolecular form, which, And this way of describing it shows the way in though it is necessary for the development of phe- which current mainstream views are continuous notypic form, neither contains it nor constitutes with older preformationist ideas. The more com- plans for it” (1985: 2). That is a good way to ex- mon formulations, in terms of a genetically en- press her position. I may also differ from Gould coded “program” or genetic “instructions” for on how an element of preformationism has been development, are not the only ways to express a vindicated. Gould says that the modern partial continuity between preformationism and modern vindication of preformationism is via our discov- views. ery of “coded instructions” in the genes. If Gould Of course, here we must confront the question: has in mind a set of instructions for the eventual Do the DNA sequences in the zygote really con- adult form, then I agree with Oyama that this is a stitute a coded representation of proteins? I have wrong turn. If all Gould has in mind here is the Status and Explanatory Structure of DST 293

specification of proteins, then his Golden Mean is Although I have tried here to raise a suspicion, rightly expressed. in this chapter I will not attempt to pin any de- Another interesting passage suggests that finite error on DST regarding preformationism. Gould does have in mind a strong version of the Matters are just not as clear as that. I will, how- “coded instructions” claim, which goes beyond ever, discuss another specific idea, associated the specification of proteins. He accepts, with the with DST, that appears to be linked to a strongly preformationists, that if the egg was truly un- antipreformationist message. organized, it could not develop into a complex Advocates of DST, and similar views, have adult without a directing entelechy. “[The egg] often been impressed by Gunther Stent’s com- does, and can only do so, because the informa- parison of development with ecological succes- tion—not merely the raw material—needed to sion (Stent 1981). Gri≤ths and Gray say this is build this complexity already resides in the egg” “perhaps the best metaphor” for biological devel- (1977: 21–22). Thus Gould squarely denies what opment (1994: 284; see also Oyama 1985: 177, Oyama calls the “ontogeny of information.” note 4; Francis in press.)16 What is the message of Both Oyama and the mainstream view (exem- the ecological analogy? The central point is that plified here by Gould) are very willing to express contingent causal processes without any guiding their views and conduct the dispute using the lan- plan or central controller can reliably produce guage of information: is developmental informa- specific outcomes. Each stage in the sequence tion preexisting or does it have an ontogeny? In gives rise to the next, but no part of the system my own statements of what is true in preforma- governs the whole process in a centralized way.17 tionism I avoid any formulation in terms of infor- DST sees the development of an organism as sim- mation. This is not because I think information is ilar, and the ecological analogy is a good one for a useless concept in biology. Rather, it is because making this point. I suggest that the clear and useful way to handle Another point is contained in the analogy, informational terms here is to restrict the concept however—an antipreformationist point. Eco- of information to the weak sense associated with logical succession is a process in which the late the mathematical theory of information—the stages owe their existence and structure to imme- sense developed by Shannon (1948) and others, diately prior stages, not to a special set of com- and sharpened up philosophically by Dretske ponents present at the first stage. Illustrations of (1981). Gri≤ths and Gray (1994) successfully the succession phenomenon sometimes use cases argued that this concept cannot be used to de- where weedy plants and then pines colonize a dis- scribe a special kind of specificity that genes have turbed area, to give way in turn to hardwoods. and nongenetic factors do not have. So I hold The hardwood forest (the analog of the adult that the partial modern vindication of preforma- organism) does not exist in miniature within the tionism lies in the fact that genes code for almost earlier stages, and neither does a representation all the proteins—and hence much of the cellular of, or recipe for, the eventual forest. machinery—that an organism will use during its The comparison is made complicated by the life. The same sort of claim can also be expressed fact that seeds of the hardwood trees do have to using the language of information; genes contain exist prior to the trees, and these seeds have to information specifying the proteins that the come from somewhere. Many components of the organism will use. But I hold that this formula- eventual “climax” ecology will have been present tion is more misleading, so I resist attempts to all along in the area. Some aspects of the soil and make the concept of information carry the main its chemistry are more or less stable and preexist- weight in philosophical discussions like the pres- ing, though other aspects are dependent on the ent one. process.18 And some of the kinds of organisms 294 Peter Godfrey-Smith

present in late stages will have been present dur- The issue is certainly di≤cult—what counts as a ing the initial stages, perhaps in different num- significant continuity with preformationism as bers. (Other components, like oak trees, will have opposed to an insignificant one? But a quotation had their seeds brought in from neighboring might help to illustrate my view. The quote is areas.) So there are certainly some structures in from Albert Matthews, in 1924, given by Robert the late stages that preexisted. Despite this, the Olby (1994). Matthews’s aim is to poke fun at ecological analogy is associated with a distinctive protein-based versions of preformationism, and pattern of explanation, and this is a pattern of he does so in a way reminiscent of caricatures of explanation stressing the way complex results can more traditional preformationism. arise reliably from a long process in which elaborate final structures need not be present, in any In the author’s opinion . . . the chromatin of spermatozoa is nothing else than the chromatin of spermato- sense, at the start. Structures present at time t : zoa; and that of an egg cell is the chromatin of an egg arise from conditions at t 1, which arise from cell. They are not nerve, muscle, epithelial chromatin, in : conditions at t 2, and so on. The transitions masquerade. . . . And it must be remembered that the from one stage to another are not explained in onus of proof is on those who assert that the chromo- terms of the continuing action of a key set of somes are . . . museums containing samples of all the components that were present at the start.19 In chromatin of all the cells of the body, not only all the the case of genetics and development, however, chromatins which develop during life, but all that the causal structure that we now recognize does infinite collection of old masters inherited from the give a special role, at every developmental stage, past, and all the infinite numbers of descendants yet to to protein molecules. These molecules are not appear in the eons before us. . . . They are concealed no doubt in the chromosomal attic, ready to be produced all materially present at the start of the process, when the occasion arises. (Quoted in Olby 1994: 80) but they have their structures coded for (or tem- plated for) by nucleic acid structures present at Matthews here takes chromosomes to contain the start. This is the feature of contemporary nucleic acid and protein, with most biological views that does justify something like Gould’s specificity residing probably in the protein. He “Golden Mean” interpretation of the situation. doubts that there is any sense in which the And this is a point of disanalogy between devel- structure of biologically important parts of, say, opment and ecological succession. muscle cells, are pre-given in egg or sperm. What does this tell us about DST? I have not Matthews changed his mind within a decade. given an argument against DST itself, or even But Matthews’s earlier critical position was co- against the overall DST enthusiasm for the eco- herently stated and false. The truth turned out to logical analogy. My suggestion is just that some be not so distant from the view he was making aspects of the analogy might be misleading, fun of here. Rather than actual samples of mus- because the analogy downplays a real continuity cle tissue, spermatozoa contain structures that between modern biological knowledge and pre- code (or template) for the proteins specifically formationist styles of explanation. needed eventually for muscles. Of course, subse- In correspondence about this issue, Richard quent discoveries did not in any sense vindicate Francis, one of the advocates of the ecological the picture of gametes containing “old masters” analogy, objected that the “continuity” that from past generations, and future generations as Gould and I allege between modern genetics and well. Still, the first part of the quote illustrates the older kinds of preformationism is so slight as to unobviousness, the nontriviality, of the prefor- be completely insignificant. If so, there has not mationist element in modern genetics. been any Golden Mean resolution, and DST cer- Before leaving this point, I will note another tainly cannot be faulted for failing to recognize it. possible objection to what I have said. Keller, Status and Explanatory Structure of DST 295

in her contribution to this volume, argues that interpreting and restating what we have learned the constantly growing explanatory role within via science. That project lies within philosophy of genetics for mechanisms of editing, regulation, nature; the empirical outlines of present-day biol- splicing (and so on) is undermining the idea that ogy are accepted at face value, but the signifi- “proteins are simply and directly encoded in the cance of the results and the fine details of the DNA” (p. 299; see also chapter 7; Sarkar 1996). explanatory structure are up for debate. Though Applied to the present point, this is an argument my own discussion here focused on the philos- that even fewer traces of what I have identified as ophy, the DST attitude to preformationist styles a preformationist pattern of explanation really of explanation also plays a role within DST as remain within genetics. The DNA exists in the research program. The DST resistance to prefor- zygote, but the eventual protein products of that mationist ideas, however subtle, is part of its DNA depend on a great array of other compo- attempt to steer work toward some scientific nents, many of which will only be built up in the options and causal possibilities, and away from course of development. So Keller might say that others. laying out a partial continuity between preformationism and modern genetics, as Gould did in 1977 and I do now, is wholly misleading. Cer- Acknowledgments tainly any view in this area must take account of the recently discovered complexity of gene regu- Thanks to all three editors for helpful com- lation, and of post-transcriptional and post- ments on this material. Thanks also to Richard translational modification of gene products. The Francis, Kim Sterelny, Rasmus Winther, and question is whether these facts completely under- Kritika Yegnashankaran for discussions and mine the value of familiar generalizations about correspondence. genetic specification of proteins—whether they completely undermine such claims as “almost all Notes the proteins that will be used by an organism during its life are coded for in the DNA of the 1. I use “DST” or “developmental systems theory” as zygote.” This objection, like that of Francis, is a convenient label here; nothing much hangs on the use important and hard to assess; if true it would of “theory” rather than “viewpoint” or some other refute my argument over the last few pages. For term. I take Oyama (1985) and Gri≤ths and Gray (1994) to be basic DST texts. I have given a simple out- now I will just leave this issue on the table. line of basic DST ideas in Godfrey-Smith (forthcoming). 2. See Gri≤ths and Gray (1997) for a rather forceful Conclusion reply. 3. I introduced this concept briefly in Godfrey-Smith This chapter has included a discussion of the sta- (2000), where it is applied to Lewontin’s “dialectical tus of DST and a rumination on one element of biology.” the DST perspective on biological development. 4. Naturphilosophie was an ambitious late eighteenth- How do the two discussions fit together? Is to early nineteenth-century approach to science and antipreformationism part of DST as research philosophy, influenced by German romanticism, which featured a good deal of unempirical argument about program or philosophy of nature? The issues how the course of science must go. Naturphilosophie involving coding, preformationism and expla- was based on strong doctrines of progressive develop- nation that I have tried to untangle are largely ment, and unity in nature’s principles. Schelling and philosophical. To try to locate present biological Oken were major figures. See Gould (1977) for a good knowledge in relation to a Golden Mean resolu- outline, which also happens to be relevant to the second tion of earlier extremes is to engage in a project of part of the present chapter. I use the term “philosophy 296 Peter Godfrey-Smith

of nature” despite, not because of, this historical con- ways in other DST discussions. Note also that Gri≤ths nection. Sometimes (as Oyama says elsewhere in this and Gray’s (1997) discussion was prompted by a rival volume) the good words are already taken, and have to set of claims about heuristic value in Sterelny, Smith, be reappropriated. and Dickison (1996). 5. Mainstream biology advertises, as a particularly 13. Quine, with his liking for “desert landscapes,” is important recent triumph in the general field of devel- often associated with this kind of challenge. opment, the discovery of the “homeobox” system of 14. A good outline is found in part 1 of Gould (1977). gene regulation. This is a rather general-purpose mech- 15. Note that we could have had a more literal “protein anism for gene regulation that has been found in genet- preformationism” as the truth—that is the thought ic systems affecting early development in a huge range experiment in Godfrey-Smith (2000a), based on actual of animals. Remarkably, homeobox-containing genes mid-century speculation. What if genes were little sam- have often been found in a similar layout on chromo- ples of all the proteins needed by the cell? somes in different animals, where the anterior-posterior axis of structures affected by these genes is mirrored by 16. Lewontin (1983: 100 and elsewhere) likes to use the linear layout of the genes on the chromosome. Such ecological succession to stress another theme, the pro- work provides a good deal of basis for great optimism found and systematic effects of organisms on their envi- at present about the power of mainstream developmen- ronments. Francis (in press) has a particularly detailed tal genetics. and picturesque discussion of ecological succession and its relation to development. 6. See Lewontin (1974), his two contributions to this volume, and his foreword to the second edition of 17. Stent’s aim in his interesting 1981 discussion was to Oyama (1985). oppose the overuse of the “program” concept in developmental biology. 7. On this point see Gri≤ths and Gray (1994); Sterelny, Smith, and Dickison (1996). 18. In some cases, such as recolonization of a recently deglaciated area, even the soil is largely a product of the 8. The idea stressed here comprises two parts of a list of early stages of the process. six empirical suggestions in Gray (in press). 19. Again, the point is made complicated by the fact 9. That philosophical work might in turn affect the sci- that a lower-level story could be told in terms of the ence, as scientists come to realize they have been using lives and deaths of individual organisms, and an ele- metaphors in certain ways—see above on the interac- ment of preformationism is true at that lower level. tion between philosophy of nature and science. 10. The version of the gene’s eye view outlined in Sterelny and Kitcher (1988) is best construed as show- References ing how a gene-level redescription is possible that (1) makes philosophical sense if combined with an austere Dawkins, R. (1976). The Selfish Gene. Oxford: Oxford and partially conventionalist view about causation (see University Press. below in this section), and that (2) might (but also Dawkins, R. (1982). The Extended Phenotype. Oxford: might not) have heuristic benefits for science via its Oxford University Press. unification of certain phenomena. Dretske, F. (1981). Knowledge and the Flow of Informa- G. C. Williams’s discussions of the gene’s eye view tion. Cambridge, MA: MIT Press. themselves contain some of the philosophical/scientific Francis, R. (in press). Genes, Brains and Sex in the In- duality I describe here (e.g., 1992), though he does not formation Age. Princeton, NJ: Princeton University try to establish the kind of very ambitious philosophi- Press. cal view found in Dawkins (1976). Friedman, M. (1999). Dynamics of reason: Kantian 11. In the 2000 edition of her 1985 book, Oyama has themes in the philosophy of science. Papers presented changed “necessary” to “operative.” This takes some, as the 1999 Kant Lectures at Stanford University. but not all, content out of the point. Certainly the ear- Godfrey-Smith, P. (in press). Explanatory symmetries, lier formulation was stronger. preformationism and developmental systems theory. 12. Gri≤ths and Gray (1997), section 6, is one exam- Philosophy of Science (supplement). ple. I think the suggestion is also made in more subtle Status and Explanatory Structure of DST 297

Godfrey-Smith, P. (2000). Organism, environment and Lewontin, R. C. (1974). The Genetic Basis of Evolu- dialectics. In R. Singh, C. Krimbas, J. Beatty, and tionary Change. New York: Columbia University Press. D. Paul (Eds.), Thinking about Evolution: Historical, Lewontin, R. C. (1983). The organism as the subject Philosophical, and Political Perspectives. Cambridge: and object of evolution. Cambridge University Press. Maynard Smith, J. (in press). The concept of informa- Godfrey-Smith, P. (2000a). On the theoretical role of tion in biology. In Proceedings of the 1999 International “genetic coding.” Philosophy of Science 67: 26–44. Congress for Logic, Methodology and Philosophy of Godfrey-Smith, P. (in press b). The replicator in retro- Science. Kluwer. spect. Biology and Philosophy. Moss, L. (1992). A kernel of truth? On the reality of Gould, S. J. (1977). Ontogeny and Phylogeny. Cam- the genetic program. Proceedings of the Philosophy of bridge, MA: Harvard University Press. Science Association 1992 vol. 1, pp. 335–348. East Gray, R. (1992). Death of the gene: Developmental Lansing, MI: Philosophy of Science Association. systems strike back. In P. E. Gri≤ths (Ed.), Trees of Olby, R. (1994). The Path to the Double Helix. (Rev. Life: Essays in Philosophy of Biology, pp. 165–209. edition.) New York: Dover. Dordrecht: Kluwer. Oyama, S. (1985). The Ontogeny of Information: Devel- Gray, R. (in press). Selfish genes or developmental sys- opmental Systems and Evolution. Cambridge: Cam- tems: Evolution without replicators and vehicles. In bridge University Press. (2d rev. ed., Durham, NC: R. Singh, C. Krimbas, J. Beatty, and D. Paul (Eds.), Duke University Press, 2000.) Thinking about Evolution: Historical, Philosophical, and Quine, W. V. O. (1969). Epistemology naturalized. In Political Perspectives. Cambridge: Cambridge Univer- Ontological Relativity and Other Essays, pp. 69–90. sity Press. New York: Columbia University Press. Gri≤ths, P., and R. Gray. (1994). Developmental sys- Sarkar, S. (1996). Decoding “Coding”—Information tems and evolutionary explanation. Journal of Philos- and DNA. BioScience 46: 857–864. ophy 91: 277–304. Shannon, C. E. (1948). A mathematical theory of com- Gri≤ths, P., and R. Gray. (1997). Replicator II: munication. Bell System Technical Journal 27: 379–423, Judgment day. Biology and Philosophy 12: 471–492. 623–656. Kitcher, P. S. (1989). Explanatory unification and the Sober, E., and R. C. Lewontin. (1982). Artifact, cause causal structure of the world. In P. S. Kitcher and W. and genic selection. Philosophy of Science 49: 157–180. Minnesota Studies in the Philosophy of Salmon (Eds.), Sober, E., and D. S. Wilson. (1994). A critical review of Science: Vol. XIII. Scientific Explanation. Minneapolis: philosophical work on the units of selection problem. University of Minnesota Press. Philosophy of Science 61: 534–555. Kitcher, P. S. (in press). Battling the undead: How (and Stent, G. (1981). Strength and weakness of the genetic how not) to resist genetic determinism. In R. Singh, C. approach to the development of the nervous system. In Thinking about Krimbas, J. Beatty, and D. Paul (Eds.), W. M. Cowan (Ed.), Studies in Developmental Neuro- Evolution: Historical, Philosophical, and Political Per- biology. New York: Oxford. spectives. Cambridge: Cambridge University Press. Sterelny, K., and P. S. Kitcher. (1988). The return of The Structure of Scientific Kuhn, T. S. (1970). the gene. Journal of Philosophy 85: 339–361. Revolutions. (2d ed.). Chicago: University of Chicago Press. Sterelny, K., K. Smith, and M. Dickison. (1996). The extended replicator. Biology and Philosophy 11: 377– Lakatos, I. (1970). Falsification and the methodol- 403. ogy of scientific research programs. In I. Lakatos Natural Selection: Domains, and A. Musgrave (Eds.), Criticism and the Growth Williams, G. C. (1992). Levels and Challenges. of Knowledge. Cambridge: Cambridge University Oxford: Oxford University Press. Press. Laudan, L. (1977). Progress and Its Problems. Berkeley: University of California Press.

21 Beyond the Gene but Beneath the Skin

Evelyn Fox Keller

The past decade has witnessed an e≥orescence of DST, “a”) locus of heredity constancy, but it can critical commentary deploring the excessively no longer be supposed to be its source: Particular genocentric focus of contemporary molecular genes (or sequences) persist as stable entities only and evolutionary biology. Among philosophers, as long as the machinery responsible for that sta- the best known work may well be that of de- bility persists. The dependence of gene function velopmental systems theorists (DST) Oyama, on complex epigenetic networks challenges (or Gri≤ths, and Gray (and sometimes including at least seriously complicates) the attribution of Lewontin and Moss), but related critiques have causal agency to individual genes. Finally, the also emerged from a variety of other quarters. third finding radically undermines the assump- Expressing diverse intellectual and philosophical tion that proteins are simply and directly encoded preoccupations, and motivated by a variety of in the DNA (indeed, it undermines the very no- scientific and political concerns, these analyses tion of the gene as a functional unit residing on have converged on a number of common themes the chromosome). and sometimes even on strikingly similar for- On this much, one finds a certain general mulations.1 Common themes include: conceptual agreement. But differences—deriving in part problems with the attribution of causal primacy from the different intellectual, scientific, and (or even causal e≤cacy) to genes;2 disarray in political perspectives of their authors—can also contemporary uses of the very term gene; con- be found. Sometimes these are matters of empha- fusions and misapprehensions generated by use sis, sometimes of focus, and sometimes of more of the particular locution of “genetic program.” substantive import. In this chapter, I want to Much of the impetus behind these critiques issues focus on what I believe to be a substantive issue from long standing concerns, and indeed, many distinguishing my own perspective from that of the critical observations could have been (and which tends to dominate the DST literature, and in some cases have been) made long ago. Why that issue can be put in the form of a question: Is then their particular visibility today? An obvious there a place on our biological map for the mate- answer lies close at hand: Critiques of geno- rial body of the organism, for that which lies centrism have found powerful support in many beyond the gene yet beneath the skin? And if so, of the recent findings of molecular biologists. where is that place? Indeed, I would argue that it is from these empirical findings that the major impetus for a reformulation of genetic phenomena now comes. The Body in Question Three developments (or findings) are of particular importance here: (1) the need for elaborate I share with proponents of DST the conviction mechanisms for editing and repair of DNA to that the oppositional terms in which the nature/ ensure sequence stability and fidelity of replica- nurture debate has historically been framed are tion; (2) the importance of complex (and non- both artificial and counterproductive. But the linear) networks of epigenetic interactions in the particular question I pose here reflects an addi- regulation of transcription; (3) the extent to tional source of unease, and that is over the tacit which the “sense” of the messenger transcript elision of the body implied not only by the fram- depends on highly regulated mechanisms of editing of the classical controversies, but at least par- ing and splicing. The implication of the first is tially continued in the solutions that have been that the structure of the gene (or sequence of thus far been put forth. DNA) may be the (or, from the perspective of Without question, the most conspicuous roots of my concern are to be found in the history of 300 Evelyn Fox Keller

genetics and neo-Darwinian evolutionary theory. telligence and other behavioral attributes; it With the emergence of genetics in the early part also helped pave the way for the mid-twentieth- of the century, debates over the relative force of century recasting of the nature-nurture debate nature and nurture (first framed as such by in one of its crudest forms, that is, as a bat- Francis Galton in 1874) were recast, initially, in tle between advocates of “Darwinian” and terms of heredity and environment (see, e.g., “Lamarckian” evolution (see, e.g., Keller 1991; Barrington and Pearson 1909; Morgan 1911; Jablonka and Lamb 1995). Conklin 1915), and soon after, in terms of genes To the extent that such debates imply a logical and environment (see, e.g., analyses of the re- disjunction (form or matter; nature or nurture; lative importance of heredity and environment genes or environment), they are clearly counter- in Fisher 1918 and Wright 1920). This second productive. But my particular argument here is reframing may have been an inevitable conse- that replacing an implied disjunction by an ex- quence of the terminological shift in the biologi- plicit conjunction (nature and nurture; genes and cal literature of the 1920s and later, in which the environment) does little to ameliorate the partic- term heredity came to be replaced by the newer ular problem of the role of the body that resides term genetics,3 but the consequence of this shift beyond the gene yet beneath the skin. To be sure, was more than terminological: it amounted to the disjunctive framing absolutely denies infor- a conceptual reduction of “nature” to “genes,” mational function to the material environment, and with that reduction, only one of two possible whereas the conjunctive framing advocated by statuses for the extragenetic body: either its com- DST clearly does permit such a function.5 But in plete elision or its relegation to the category of both, there is discernible tendency to figure the “nurture” or “environment.” Genetics further organismic body qua environment (and qua nur- contributed to this relegation with its recasting of ture), and accordingly, to leave any distinctively another and far older controversy, namely that informing role the organism’s “internal environ- concerning the relations between form (generally ment” might play in development and evolution construed as active) and matter (construed as concealed from view. passive). That ancient discussion could now be (and was) reconceptualized in terms of genes as the agents of “action” (later, as sources of “in- Which Body, Which Skin? And, Anyway, Why formation”), and of a cellular or extracellular en- Stop at the Skin? vironment that is simultaneously acted upon and informed, serving as passive material substrate Should the organismic body be singled out as for the development (or unfolding) of the organ- having particular biological significance? And ism.4 But where many of the discussions of hered- if so, which body, and which skin? Biology rec- ity and environment among geneticists focused ognizes many bodies, corresponding to many on their relative force in individual development, skins: in higher organisms, there is the multicellu- elsewhere, such debates more commonly focused lar body contained within an outer integument; on their relative force in shaping the course of in all organisms, cellular bodies are contained by evolution. Here, the neo-Darwinian synthesis cell membranes; and in eukaryotic organisms, was of particular importance. In identifying nuclear bodies are contained by nuclear mem- genetic continuity and change as the sole funda- branes. To avoid some of this ambiguity, I ment of evolution, it contributed powerfully to choose to focus on that moment in the life cycle the polarization of debates over the relative force of higher (metazoan) organisms in which the of genes and environment in such highly charged outer integument is the cell membrane, and the arenas as eugenics, or the “heritability” of in- organismic body is the cellular body—that is, in Beyond the Gene but Beneath the Skin 301

which the body in question is the fertilized egg or the face of all the vicissitudes it inevitably en- zygote. But there remains the question, why stop counters. This paradigmatic body may not be at the skin? Certainly, no biological integument autonomous, but as embryologists have always provides an absolute divide between interior and known, it is far more tolerant of changes in its exterior, and the cell membrane of a fertilized egg external environment than in its internal milieu. is, of necessity, more porous than most. Further- Indeed, were it not for their robustness, that more, because it regulates so much of the tra≤c is, for the tolerance of (at least many) early between inside and out, the cell membrane is embryos to being moved from one environment itself an active agent in shaping the body it to another, embryonic manipulation would not contains, indeed, in determining the very mean- be possible, and much of what we know of as ing of interiority. These facts constitute a warn- experimental embryology would never have ing against conceptualizing the organism as an come into being. autonomous individual, sealed off from an exte- All of this may seem too obvious to need say- rior world by a static or preexisting boundary. ing, but there are times when the obvious is what Yet even so, the cell membrane, dynamic and most needs saying. We have learned that no eli- permeable though it may be, defines a boundary sion is innocent. Nor, for that matter, is any which evolution has not only crafted into a cor- reminder of elision. Politics are everywhere. Just nerstone of biological organization but has en- as there are important political dimensions to the dowed with absolutely vital significance. And history of debates over genes and environment,6 given the dire effect the physical erasure of this so too, there are political dimensions to the eli- boundary would have on the survival of the sion of the body in genetical discourse,7 as there organism or cell, it scarcely seems necessary to also are, inevitably, to my insistence here on the elaborate upon the inappropriateness of its con- boundary of the skin. In fact, the title of this ceptual erasure. essay contains its own elision: it reminds us of In other words, the immediate and most obvi- another title (Barbara Duden’s The Woman ous reason for taking this boundary seriously is Beneath the Skin), while at the same time sup- grounded in its manifest indispensability for via- pressing the subject of that other title. There is, of bility. But this said, we are still not any closer to course, a reason. This is not an essay in feminist understanding why is it so important. By way of theory, nor is it about women. The “woman” in addressing this last question, I would like to sug- my title is signified only by its absence—intended, gest that the primary function of the cell mem- by that absence, to evoke nothing more than a brane (as of any other biological skin) is simply recognition of the trace of the woman beneath that it holds things together—more specifically, the skin that still lurks, if not in the body more that it keeps in proximity the many large mole- generally, surely, in the reproductive body of the cules and subcellular structures required for fertilized egg. Because, in sexual reproduction, growth and development. Proximity is crucial, the cytoplasm derives almost entirely from the for it enables a degree of interconnectivity and unfertilized egg, it is no mere figure of speech to interactive parallelism that would otherwise not refer to it as the maternal contribution. Further- be possible, but that is required for what I take to more, the representation of that body as “genetic be the fundamental feature of the kind of devel- environment,” as nothing more than a source of opmental system we find in a fertilized egg, name- nurture for the developing organism, is a bit too ly, its robustness. Prior to all its other remarkable reminiscent of conventional maternal discourse properties—in fact, a precondition of these—is for at least this author’s comfort. My title, in the capacity of a developmentally competent short, is deliberate in its allusivity: I want to indi- zygote to maintain its functional specificity in cate the possibility that gender politics has been 302 Evelyn Fox Keller

implicated in the historic elision of the body in speak at all of a developmental program, or of question, without, at the same time, reinscribing a set of instructions for development, in con- the woman in that or any other body. The pri- tradistinction to the data or resources for such a mary aim of this essay is, finally, a biological one, program, current research obliges us to acknowl- by which I mean that it is to reclaim the possibil- edge that these “instructions” are not written into ity of finding biological significance and agency the DNA itself (or at least, are not all written in in that no man’s land beyond the gene but be- the DNA), but rather are distributed throughout neath the skin. Contra Oyama (1992),8 I want to the fertilized egg. Indeed, if the distinction be- argue that taking the cell rather than the gene as tween program and data is to have any meaning a unit of development does make a difference: not in biology, it has become abundantly clear that it only does it yield a significant conceptual gain in does not align (as had earlier been assumed) the attempt to understand development, but also, either with a distinction between “genetic” and it permits better conformation to the facts of de- “epigenetic,” or with the precursor distinction be- velopment as we know them. tween nucleus and cytoplasm. To be sure, the informational content of the DNA is essential— without it development (life itself) cannot pro- Is There a Program for Development? And If So, ceed. But for many developmental processes, it Where Is It to Be Found? is far more appropriate to refer to this informational content as data than as program (Atlan Many authors have taken issue with the concept and Koppel 1990). Indeed, I want to suggest that of a program for development, noting its teleo- the notion of genetic program both depends upon logical implications, its metaphoric reliance on and sustains a fundamental category error in computer science, its implication of a unidirec- which two independent distinctions, one between 9 tional flow of information. But my concern here “genetic” and “epigenetic,” and the other be- is not with the concept of program per se: rather, tween program and data, are pulled into mis- it is with the more specific notion of a genetic pro- taken alignment. The net effect of such alignment gram, especially in contradistinction to its com- is to reinforce two outmoded associations: on the panion notion, that of a developmental program. one hand, between “genetic” and active, and, on In other words, for my particular purposes here, the other, between “epigenetic” and passive. I accept the metaphor of program, warts and all, Development results from the temporally and and focus my critical attention instead on the spatially specific activation of particular genes, implications of attaching to that metaphor the which in turn depends on a vastly complex net- modifier “genetic.” And I ask two questions: work of interacting components including not First, what is the meaning of a “genetic pro- only the “hereditary codescript” of the DNA, gram”? Second, how did this concept come to be but also a densely interconnected cellular ma- so widely accepted as an “explanation” of biolog- chinery made up of proteins and RNA molecules. 10 ical development? Necessarily, each of these systems functions in Taken as a composite, the meaning of the term relation to the others alternatively as data and as genetic program simultaneously depends upon program. If development cannot proceed without and underwrites the particular presumption that the “blueprint” of genetic memory, neither can a “plan of procedure” for development is itself it proceed without the “machinery” embodied written in the sequence of nucleotide bases. Is this in cellular structures. To be sure, the elements of presumption correct? Certainly, it is almost uni- these structures are fixed by genetic memory, but versally taken for granted, but I want to argue their assembly is dictated by cellular memory.11 that, at best, it must be said to be misleading, and Furthermore, one must remember that more than at worst, simply false: To the extent that we may Beyond the Gene but Beneath the Skin 303

genes are passed from parent to offspring. To for- genome, but, over the course of that decade, get this is to be guilty of what Richard Lewontin another notion of “program,” a “developmental calls an “error of vulgar biology.” As he reminds program,” also surfaced, and repeatedly so. This us, “an egg, before fertilization, contains a com- program was not located in the genome, but plete apparatus of production deposited there instead, distributed throughout the fertilized egg in the course of its cellular development. We in- (see, e.g., Apter 1966). By the 1970s, however, the herit not only genes made of DNA but an intri- “program” for development had effectively col- cate structure of cellular machinery made up of lapsed into a “genetic program,” with the alter- proteins” (Lewontin 1992: 33). native, distributed, sense of a “developmental Assuming one is not misled by Lewontin’s col- program” all but forgotten. loquial use of the term inherit to refer to trans- Francois Jacob, one of the earliest to use the mission over a single generation (as distinct from concept “genetic program” contributed crucially multigenerational transmission), none of this is to its popularization. In The Logic of Life, first either controversial or new, nor does it depend on published in French in 1970, Jacob describes the the extraordinary techniques now available for organism as “the realization of a programme pre- molecular analysis. Yet, however surprisingly, it scribed by its heredity” (Jacob 1976; 2), claiming is only within the last decade or two that the de- that “when heredity is described as a coded pro- velopmental and evolutionary implications of so gramme in a sequence of chemical radicals, the called “maternal effects” has begun to be appre- paradox [of development] disappears” (Jacob ciated.12 Current research now provides us with 1976: 4). For Jacob, the genetic program, written an understanding of the mechanisms involved in the alphabet of nucleotides, is what is respon- in the processing of genetic data that make the sible for the apparent purposiveness of biological errors of what Lewontin calls “vulgar biology” development; it and it alone gives rise to “the manifest. Yet, even when elaborated by the kind order of biological order.” (Jacob 1976: 8) He of detail we now have available, such facts are refers to the oft-quoted characterization of teleol- still not su≤cient to dislodge the confidence that ogy as a “mistress” whom biologists “could not many distinguished biologists continue to have in do without, but did not care to be seen with in both the meaning and explanatory force of the public,” and writes, “The concept of programme genetic program. The question I want therefore has made an honest woman of teleology” (Jacob to ask is, how come? What grants the “genetic 1976: 8–9). Although Jacob does not exactly de- program” its apparent explanatory force, even in fine the term, he notes that “[t]he programme is a the face of such obvious caveats as those above? model borrowed from electronic computers. It To look for answers, I will turn to history, more equates the genetic material of an egg with the specifically, to the history of the term itself. magnetic tape of a computer” (Jacob 1976: 9). However, equating the genetic material of an egg with the magnetic tape of a computer does “Programs” in the Biological Literature of the not imply that that material encodes a “pro- 1960s gram”; it might just as well be thought of as encoding “data” to be processed by a cellular The metaphor of a “program,” borrowed directly “program.” Or by a program residing in the from computer science, entered the biological lit- machinery of transcription and translation com- erature in the 1960s not once, but several times, plexes. Or by extranucleic chromatin structures and in at least two distinctly different registers. In in the nucleus. Computers have provided a rich its first introduction, simultaneously by Mayr source of metaphors for molecular biology, but (1961) and by Monod and Jacob (1961), the locus they cannot by themselves be held responsible for of the “program” was explicitly identified as the 304 Evelyn Fox Keller

the notion of “genetic program.” Indeed, as al- biology had revealed a stunningly simple me- ready indicated, other, quite different, uses of the chanism for the transmission and translation of program metaphor for biological development genetic information, but, at least until 1960, it were already in use. One such use was in the no- had been able to offer no account of develop- tion of a “developmental program”—a term that mental regulation. surfaced repeatedly through the 1960s, and that James Bonner, a professor of biology at stood in notable contrast to that of a “genetic CalTech, in an early attempt to bring molecular program.” biology to bear on development, put the problem Let me give an example of this alternative use. well. Granting that “the picture of life given to us In 1965, a young graduate student, Michael by molecular biology . . . applies to cells of all Apter, steeped in information theory and cy- creatures,” he goes on to observe that this picture bernetics, teamed up with the developmental biologist Lewis Wolpert to argue for a direct is a description of the manner in which all cells are similar. But higher creatures, such as people and pea analogy not between computer programs and plants, possess different kinds of cell. The time has the genome, but between computer programs come for us to find out what molecular biology can tell and the egg: us about why different cells in the same body are different from one another, and how such differences arise. if the genes are analogous with the subroutine, by spec- (Bonner 1965: v) ifying how particular proteins are to be made . . . then the cytoplasm might be analogous to the main pro- Bonner’s own work was on the biochemistry gramme specifying the nature and sequence of oper- and physiology of regulation in plants, in an in- ations, combined with the numbers specifying the stitution well known for its importance in the particular form in which these events are to manifest themselves. . . . In this kind of system, instructions do birth of molecular biology (see, e.g., Kay 1992). not exist at particular localized sites, but the system acts Here, in this work, published in 1965, like Apter as a dynamic whole. (Apter and Wolpert 1965: 257) and a number of others of that period, Bonner too employs the conceptual apparatus of auto- Indeed, throughout the 1960s, a number of de- mata theory to deal with the problem of develop- velopmental biologists attempted to employ ideas mental regulation. But unlike them, he does not from cybernetics to illuminate development, and locate the “program” in the cell as a whole, but almost all shared Apter’s starting assumptions rather, in the chromosomes, and more specifi- (see Keller 1995, chap. 3 for examples)—that is, cally in the genome. Indeed, he begins with the by they located the program (or “instructions”) for then standard credo of molecular biology, assert- development in the cell as a whole. ing that “[w]e know that . . . the directions for The difference in where the program is said all cell life [are] written in the DNA of their chro- to be located is crucial, for it bears precisely on mosomes” (Bonner 1965: v). Why? An obvious the controversy over the adequacy of genes to answer is suggested by his location. Unlike Apter account for development that had been raging and unlike other developmental biologists of the among biologists since the beginning of the centime, Bonner was situated at a major thorough- tury. By the beginning of the 1960s, this debate fare for molecular biologists, and it is hard to had subsided, largely as result of the eclipse of imagine that he was uninfluenced by the enthusi- embryology as a discipline during the 1940s and asm of his colleagues at CalTech. In any case, 1950s. Genetics had triumphed, and after the Bonner’s struggle to reconcile the conceptual identification of DNA as the genetic material, the demands posed by the problems of developmen- successes of molecular biology had vastly consol- tal regulation with the received wisdom among idated that triumph. Yet the problems of devel- molecular biologists is at the very least instruc- opment, still unresolved, lay dormant. Molecular tive, especially given its location in time, and I Beyond the Gene but Beneath the Skin 305

suggest it is worth examining in some detail for DNA is required to explain cell differentiation the insight it has to offer on our question of how (and this is his main point), but on the way to the presumption of a “genetic program” came— making this point, he has placed this “further in fact, over the course of that very decade— mechanism” in the nucleus, with nothing more by to seem self-evident. In short, I want to take way of argument or evidence than his “Clearly Bonner as representative of a generation of care- then.” Why does such an inference follow? And ful thinkers about an extremely di≤cult problem why does it follow “clearly”? Perhaps the next who opted for this (in retrospect, inadequate) paragraph will help: conceptual shortcut. The egg is activated by fertilization. . . . As division proceeds cells begin to differ from one another and to Explanatory Logic of the “Genetic Program” acquire the characteristics of specialized cells of the adult creature. There is then within the nucleus some From molecular biology, Bonner inherited a lan- kind of programme which determines the property [sic] sequenced repression and derepression of genes and guage encoding a number of critical if tacit pre- which brings about orderly development. (Bonner suppositions. That language shapes his efforts 1965: 6) in decisive ways. Summarizing the then current understanding of transcription and translation, Here, the required “further mechanism” is ex- he writes: plicitly called a “program,” and, once again, it is located in the nucleus. But this time around, a Enzyme synthesis is therefore an information-requiring clue to the reasoning behind the inference has task and . . . the essential information-containing com- been provided in the ﬁrst sentence, “The egg is ponent is the long punched tape which contains, in coded form, the instructions concerning which amino activated by fertilization.” This is how I believe acid molecule to put next to which in order to produce the (largely tacit) reasoning goes: If the egg is a particular enzyme. (Bonner 1965: 3) “activated by fertilization,” the implication is that it is entirely inactive prior to fertilization. At the same time, he clearly recognized that What does fertilization provide? The entrance of only the composition of the protein had thus the sperm, of course, and unlike the egg, the been accounted for, and not the regulation of its sperm has almost no cytoplasm: it can be thought production required for the formation of special- of as pure nucleus. Ergo, the active component ized cells, that is, cell differentiation remained must reside in the nucleus and not in the cyto- unexplained. As he wrote, “Each kind of spe- plasm. Today, the supposition of an inactive cialized cell of the higher organism contains its cytoplasm would be challenged, but in Bonner’s characteristic enzymes but each produces only a time, it would have been taken for granted as a portion of all the enzymes for which its genomal carryover from what I have called “the discourse DNA contains information” (Bonner 1965: 6). of gene action” of classical genetics (Keller 1995). But, he continues: “Clearly then, the nucleus con- And even then, it might have been challenged had tains some further mechanism which determines it been made explicit, but as an implicit assump- in which cells and at which times during develop- tion encoded in the language of “activation,” it ment each gene is to be active and produce its was likely to go unnoticed by Bonner’s readers as characteristic messenger RNA, and in which cells by Bonner himself. each gene is to be inactive, to be repressed” Bonner then goes on to ask the obvious ques- (Bonner 1965: 6). tions: “What is the mechanism of gene repression Two important moves have been made here. and derepression which makes possible develop- Bonner argues that something other than the in- ment? Of what does the programme consist and formation for protein synthesis encoded in the where does it live?” (Bonner 1965: 6) And he 306 Evelyn Fox Keller

answers them as best he can: “We can say that the Bonner takes it—namely, to elide the distinction programme which sequences gene activity must between genome and chromosome. The “genetic itself be a part of the genetic information since program” is saved (for this discussion) by just a the course of development and the final form are slight shift in reference: now it refers to a pro- heritable. Further than this we cannot go by clas- gram built into the chromosomal structure—that sical approaches to differentiation” (Bonner is, into the complex of genes and histones, where 1965: 6). that complex is itself here referred to as the In these few sentences, Bonner has completed “genome.” the line of argument leading him to the conclu- But the most conspicuous inadequacy of the sion that the program must be part of the genetic location of the developmental program in the information, that is, to the “genetic program.” genetic information becomes evident in the final And again, we can try to unpack his reasoning. chapter, in which Bonner attempts to sketch out Why does the heritability of the course of devel- an actual computer program for development. opment and the final form imply that the pro- Here, the author undertakes to reframe what is gram must be part of the genetic information? known about the induction of developmental Because—and only because—of the unspoken pathways in terms of a “master program,” pro- assumption that it is only the genetic material posing to “consider the concept of the life cycle as that is inherited. The obvious fact—that the made up of a master programme constituted in reproductive process passes on (or transmits) not turn of a set of subprogrammes or subroutines” only the genes but also the cytoplasm (the latter (Bonner 1965: 134). Each subroutine specifies through the egg for sexually reproducing organ- a specific task to be performed. For a plant, his isms)—is not mentioned. But even if it were, this list includes: cell life, embryonic development, fact would almost certainly be regarded as irrele- how to be a seed, bud development, leaf develop- vant, simply because of the prior assumption that ment, stem development, root development, the cytoplasm contains no active components. reproductive development. Within each of these The conviction that the cytoplasm could neither subroutines is a list of cellular instructions or carry nor transmit effective traces of intergenera- commands, such as, “divide tangentially with tional memory had been a mainstay of genetics growth”; “divide transversely with growth”; for so long that it had become part of the “mem- “grow without dividing”; and “test for size or cell ory” of that discipline, working silently but effec- number” (Bonner 1965: 137). He then asks the tively to shape the very logic of inference obvious next question: “[H]ow might these employed by geneticists. subroutines be related to one another? Exactly Yet another ellipsis becomes evident (now, how are they to be wired together to constitute a even to Bonner himself) as he attempts to in- whole programme?” (Bonner 1965: 135). Con- tegrate his own work on the role of histones in veniently, this question is never answered. If it genetic regulation. Not all copies of a gene (or a had been, the answer would have necessarily genome) are in fact the same: Because of the pres- undermined Bonner’s core assumption. To see ence of proteins in the nucleus, capable of bind- this, two points emerging from his discussion ing to the DNA, “in the higher creature, if it is to need to be underscored: First, the list of subrou- be a proper higher creature, one and the same tines, although laid out in a linear sequence (as if gene must possess different attributes, different following from an initial “master program”) attitudes, in different cells” (Bonner 1965: 102). actually constitute a circle, as indeed they must if The difference is a function of the histones. How they are to describe a life cycle. Bonner’s own can we reconcile this fact with the notion of a “master program” is in fact nothing but this “genetic program”? There is one simple way, and composite set of programs, wired together in a Beyond the Gene but Beneath the Skin 307

structure exhibiting the characteristic cybernetic live?” we would have to say, just as Apter saw logic of “circular causality.” long ago, that it consists not of particular gene The second point bears on Bonner’s earlier entities, and lives not in the genome itself, but of question, “Of what does the programme con- and in the cellular machinery integrated into a sist and where does it live?” The first physical dynamic whole. As Garcia-Bellido writes, “De- structures that were built to embody the logic of velopment results from local effects, and there computer programs were built out of electrical is no brain or mysterious entity governing the networks13 (hence the term “switching net- whole: there are local computations and they works”), and this is Bonner’s frame of reference. explain the specificity of something that is histor- As he writes, “[t]hat the logic of development is ically defined” (1998: 113). Thus, if we wish to based upon [a developmental switching] network, preserve the computer metaphor, it would seem there can be no doubt” (Bonner 1965: 148). But more reasonable to describe the fertilized egg as a what would serve as the biological analogue of an massively parallel processor in which “programs” electric (or electronic) switching network? How (or networks) are distributed throughout the are the instructions specified in the subroutines cell.14 The roles of “data” and “program” here that comprise the life cycle actually embodied? are relative, for what counts as “data” for one Given the dependence of development on the reg- “program” is often the output of a second “pro- ulating activation of particular genes, Bonner gram,” and the output of the first is “data” for yet reasonably enough calls the developmental another “program,” or even for the very “pro- switching network a “genetic switching net- gram” that provided its own initial “data.” Thus, work.” But this does not quite answer our ques- for some developmental stages, the DNA might tion; rather, it obfuscates it. The clear implication be seen as encoding “programs” or switches is that such a network is constituted of nothing which process the data provided by gradients of but genes, whereas in fact, many other kinds of transcription activators, or alternatively, one entities also figure in this network, all playing might say that DNA sequences provide data for critical roles in the control of genetic activity. the machinery of transcription activation (some Bonner himself writes of the roles played by his- of which is acquired directly from the cytoplasm tones, hormones, and RNA molecules; today, the of the egg). In later developmental stages, the list has expanded considerably to include enzy- products of transcription serve as data for splic- matic networks, metabolic networks, transcrip- ing machines, translation machines, and the like. tion complexes, signal transduction pathways, In turn, the output of these processes make up and so on, with many of these additional factors the very machinery or programs needed to embodying their own “switches.” We could, of process the data in the first place. Sometimes, this course, still refer to this extraordinarily com- exchange of data and programs can be repre- plex set of interacting controlling factors as a sented sequentially, sometimes as occurring in “genetic switching network”—insofar as the reg- simultaneity. ulation of gene activation remains central to development—but only if we avoid the implication (an implication tantamount to a category Into the Present error) that that network is embodied in and by the genes themselves. In the mid 1960s, when Bonner, Apter, and Indeed, it is this “category error” that con- others were attempting to represent development founds the very notion of a “genetic program.” If in the language of computer programs, automata we were now to ask Bonner’s question, “Of what theory was in its infancy, and cybernetics was at does the programme consist and where does it the height of its popularity. During the 1970s and 308 Evelyn Fox Keller

1980s, these efforts lay forgotten: cybernetics had of DNA.15 Artificial Life’s “genome” typically lost its appeal to computer scientists and biolo- consists of instructions such as “reproduce,” gists alike, and molecular biologists found they “edit,” “transport,” or “metabolize,” and the bio- had no need of such models. The mere notion of logical instantiation of these algorithms is found a “genetic program” su≤ced by itself to guide not in the nucleotide sequences of DNA, but in their research. Today, however, provoked in specific kinds of cellular machinery such as tran- large part by the construction of hard-wired par- scription complexes, spliceosomes, and metabolic allel processors, the project to simulate biological networks. Why then are they called “genetic,” development on the computer has returned in full and why is the full representation called a force, and in some places has become a flourish- “genome”? Is it not simply because it so readily ing industry. It goes by various names—Artificial follows from the usage the term “genetic pro- Life, adaptive complexity, or genetic algorithms. gram” had already acquired in genetics? But what is a genetic algorithm? Like Bonner’s Words have a history, and their usage depends subroutines, it is “a sequence of computational on this history, as does their meaning. History operations needed to solve a problem” (see, e.g., does not fix the meaning of words; rather, it Emmeche 1994). And once again, we need to ask, builds into them a kind of memory. In the field of why “genetic”? Furthermore, not only are the in- genetic programming, “genes” have come to refer dividual algorithms referred to as “genetic,” but not to particular sequences of DNA, but to the also “in the fields of genetic algorithms and arti- computer programs required to execute partic- ficial evolution, the [full] representation scheme is ular tasks (as Langton puts it, “one program per often called a ‘genome’ or ‘genotype’ ” (Fleischer gene”); yet, at the same time, the history of the 1995: 1). And, in an account of the sciences of term ensures that the word gene, even as adapted complexity written for the lay reader, Mitchell by computer scientists, continues to carry its Waldrop quotes Chris Langton, the founder of original meaning. And perhaps most impor- Artificial Life, as saying: tantly, that earlier meaning remains available for deployment whenever it is convenient to do so. [Y]ou can think of the genotype as a collection of little Much the same can be said for the use of the computer programs executing in parallel, one program terms gene and genetic programs by geneticists. per gene. When activated, each of these programs enters into the logical fray by competing and cooperating with all the other active programs. And collectively, A Recapitulation these interacting programs carry out an overall compu- tation that is the phenotype: the structure that unfolds during an organism’s development. (Waldrop 1992: I have taken some time in examining Bonner’s 194) argument for “genetic programs,” not because his book played a major role in establishing the Like their counterparts in molecular genetics, centrality of this notion in biological discourse, workers in Artificial Life are not confused. They but rather because of the critical moment in time well understand, and when pressed readily at which it was written and because of the relative acknowledge, that the biological analogs of these accessibility of the kind of slippage on which his computer programs are not in fact “genes” (at argument depends. The very first use of the term least as the term is used in biology), but complex program that I have been able to find in the biochemical structures or networks constituted of molecular biology literature had appeared only proteins, RNA molecules, and metabolites that four years earlier.16 In 1961, Jacob and Monod, often, although certainly not always, execute published a review of their immensely influential their tasks in interaction with particular stretches work on a genetic mechanism for enzymatic adaptation in E. coli, that is, the operon model. The Beyond the Gene but Beneath the Skin 309

introduction of the term program appears in their in our understanding of developmental logic. concluding sentence: “The discovery of regulator Arguably, it has also contributed to the endur- and operator genes, and of repressive regulation ance of such gaps. of the activity of structural genes, reveals that the So why did it prevail? If its popularity cannot genome contains not only a series of blue-prints, be accounted for in strictly scientific or cognitive but a coordinated program of protein synthesis terms, we must look elsewhere. I suggest we look and the means of controlling its execution” to the consonance of this formulation with the (Jacob and Monod 1961: 354). prior history of genetic discourse, particularly Three decades later, Sydney Brenner refers to with the discourse of “gene action” that earlier the belief “that all development could be reduced prevailed. Fortifying the “genetic program” in to [the operon] paradigm”—that “It was simply a the postwar era, with its easy and continuing matter of turning on the right genes in the right elision of the cytoplasmic body, were an entirely places at the right times”—in rather scathing new set of resources. Primary among these were terms. As he puts it, “[o]f course, while absolute- the new science of computers, the imprimateur of ly true this is also absolutely vacuous. The para- Schrödinger, and the phenomenal success of the digm does not tell us how to make a mouse but new molecular biology. Jacob’s Logic of Life was only how to make a switch” (Brenner et al. 1990: of key importance in the popularization of the 485).17 And even in the first flush of enthusiasm, concept of “genetic program.” Invoking the ap- not everyone was persuaded of the adequacy of proval of both Schrödinger and Wiener, Jacob this particular regulatory mechanism to explain endows the transition from past to future development.18 Lewis Wolpert was one of the metaphors with the stamp of authority.20 early skeptics. In the late 1960s, he seemed certain What’s in a word? As I have already suggested, that an understanding of development required quite a lot. Words shape the ways we think, and a focus not simply on genetic information, but how we think shapes the ways we act. In particu- also on cellular mechanisms.19 But by the mid lar, the use of the term genetic to describe devel- 1970s, even Wolpert had been converted to the opmental instructions (or programs) encourages notion of a “genetic program” (see, e.g., Wolpert the belief even in the most careful of readers (as and Lewis 1975). well as writers) that it is only the DNA that mat- What carried the day? Certainly not more in- ters; it helps all of us to lose sight of the fact that, formation about actual developmental processes. if that term is to have any applicability at all, Far more than most histories of scientific terms, it is primarily to refer to the entities upon which the history of genetic programs bears the conspic- instructions directly or indirectly act and not of uous marks of a history of discourse and power. which these instructions are constituted. The nec- Initially founded on a simple category error, in essary dependency of genes on their cellular con- which the role of genes as subjects (or agents) of text, not simply as nutrient but as embodying development was unwittingly conflated with their causal agency, is all too easily forgotten. It is role as objects of developmental dynamics, the forgotten in laboratory practice, in medical coun- remarkable popularity of this term in molecular seling, and perhaps above all, seduced by the genetics over the last three decades cries out for promise of utopian transformation, in popular an accounting. Certainly, it provided a conven- culture. ient gloss, an easy way to talk that rarely if ever trips scientists up in their daily laboratory work. But it does trip them up in their efforts to explain Notes development; indeed, the term has proven re- 1. See, e.g., Gri≤ths and Neumann-Held (1999); Keller markable effective in obscuring enduring gaps (1995, 1999). 310 Evelyn Fox Keller

2. See, e.g., Lewontin (1992); Moss (1992); Keller embryo exhibits characteristics intermediate between (1995); Strohman (1997). the two parental species. This dependency of genomic 3. Earlier, in the late nineteenth century, the term function on cytoplasmic structure follows as well from heredity had commonly been used far more inclusively, the asymmetric outcomes of reciprocal crosses de- encompassing both the study of both genetics and em- monstrated in earlier studies of interspecific hybrids bryology (see, e.g., Sapp 1987). Furthermore, in the (Markert and Ursprung 1971: 135–137). 1920s, the term genetics was largely understood to refer 12. “Maternal (or cytoplasmic) effects” refers only to solely to transmission genetics. the effective agency of maternal (or cytoplasmic) con- 4. For further discussion, see Griesemer (forthcoming); tributions (such as, e.g., gradients). Because such effects Keller (1995). need not be (and usually are not) associated with the 5. Indeed, the brunt of much of this literature is to existence of permanent structures that are transmitted argue for symmetry between the role of genes and other through the generations, they should not be confused developmental resources. Thus, for example, in arguing with “maternal inheritance.” against the conventional view that genes code for traits, 13. In modern computers, such networks are elec- Gri≤ths and Gray suggest that “we can talk with equal tronic. legitimacy of cytoplasmic or landscape features coding 14. Supplementing Lenny Moss’s observation that a for traits in standard genic backgrounds” (1994; 1998: genetic program is “an object nowhere to be found” 122). (Moss 1992: 335), I would propose the developmental 6. See, e.g., Kevles (1986) and Paul (1995, 1998) for dis- program as an entity that is everywhere to be found. cussions of the politics of eugenics debates, and Sapp 15. Executing a task means processing data provided (1987) for a discussion of the impact of Lysenko’s anti- both by the DNA and by the products of other pro- genetic crusade in the Soviet Union on American genet- grams—that is, by information given in nucleotide ics just before and during the cold war. sequences, chromosomal structure, gradients of pro- 7. See my discussion of the “discourse of gene action” teins and RNA molecules, the structure of protein com- in Keller (1995, chap. 1). plexes, and so on. 8. I refer in particular to Oyama’s discussion of 16. Simultaneously, and probably independently, Brenner’s abandonment of the concept of a “genetic Ernst Mayr introduced the notion of “program” in his program” and his emerging conviction that the proper 1961 article on “Cause and Effect in Biology” (adapted “unit of development is the cell.” She writes, “Having from a lecture given at MIT on Feb. 1, 1961). There given up genetic programs, [Brenner] now speaks of in- he wrote, “The complete individualistic and yet also ternal representations and descriptions. In doing so he species-specific DNA code of every zygote (fertilized is like many workers who have been faced with the con- egg cell), which controls the development of the central tradictions and inadequacies of traditional notions of and peripheral nervous system . . . is the program for the genetic forms and have tried to resolve them, not by behavior computer of this individual” (Mayr 1961: seriously altering their concepts, but by making the 1504). forms in the genome more abstract: not noses in the 17. As Soraya de Chadarevian points out (1994), genes, but instructions for noses, or potential for noses, Brenner had taken a critical stance toward the use of or symbolic descriptions of them. This solves nothing” the operon model for development as early as 1974 (see (Oyama 1992: 55). his comments in Brenner 1974). 9. See, e.g., Stent (1985), Newman (1988), Oyama 18. Or even of the appropriateness of the nomencla- (1989), Moss (1992), and de Chardarevian (1994). ture. Waddington, for example, noted not only that it 10. The remainder of this paper is adapted from Keller “seems too early to decide whether all systems control- (2000). ling gene-action systems have as their last link an 11. A vivid demonstration of this interdependency was influence which impinges on the gene itself,” but also provided in the 1950s and 1960s with the development redescribed this system as “genotropic” rather than of techniques for interspecific nuclear transplantation. “genetic” in order “to indicate the site of action of the Such hybrids almost always fail to develop past gastru- substances they are interested in” (Waddington 1962: lation, and in the rare cases when they do, the resultant 23). Beyond the Gene but Beneath the Skin 311

19. For example, Wolpert wrote in 1969: “Dealing as it Emmeche, C. (1994). The Garden in the Machine. does with intracellular regulatory phenomena, it is not Princeton, NJ: Princeton University Press. directly relevant to problems where the cellular bases of Fisher, R. A. (1918). The correlation between rela- the phenomena are far from clear” (Wolpert 1969: 2–3). tives on the supposition of Mendelian inheritance. 20. He writes: “According to Norbert Wiener, there is Transactions of the Royal Society of Endinburgh 52: no obstacle to using a metaphor ‘in which the organism 399–433. is seen as a message’” (Jacob 1976: 251–252). Two Fleischer, K. (1995). A Multiple-Mechanism Develop- pages later, he adds, “According to Schrödinger, the mental Model for Defining Self-Organizing Geometric chromosomes ‘contain in some kind of code-script the Structures. Doctoral dissertation, California Institute entire pattern of the individual’s future development of Technology. and of its functioning in the mature state. . . . The chro- Galton, F. (1874/1970). English Men of Science: Their mosome structures are at the same time instrumental in Nature and Nurture. London: Cass. bringing about the development they foreshadow. They The limits are law-code and executive power—or, to use another Garcia-Bellido, A. (1998). “Discussion.” In of reductionism. simile, they are architect’s plan and builder’s craft all in Chichester: John Wiley & Sons. one’ ” (Jacob 1976: 254). Griesemer, J. R. (forthcoming). The informational gene and the substantial body: On the generalization of evolutionary theory by abstraction. In N. Cartwright and References M. Jones (Eds.), Varieties of Idealization. Poznan Studies (Leszek Nowak, series ed.). Amsterdam: Apter, M. J. (1966). Cybernetics and Development. Rodopi Publishers. Oxford: Pergamon Press. Gri≤ths, P. E., and R. Gray. (1994). Developmental Apter, M. J., and L. Wolpert. (1965). Cybernetics and systems and evolutionary explanation. Journal of Phil- development. Journal of Theoretical Biology 8: 244– osophy 91: 277–304. 257. Gri≤ths, P. E., and E. M. Neumann-Held. (1999). The Atlan, H., and M. Koppel. (1990). The cellular com- many faces of the gene. BioScience 49: 656–662. puter DNA: Program or data. Bulletin of Mathematical Hull, D., and M. Ruse. (1998). The Philosophy of Biology 52(3): 335–348. Biology. Oxford: Oxford University Press. Barrington, A., and Pearson, K. (1909). A First Study Jablonka, E., and M. Lamb. (1995). Epigenetic Inheri- of the Inheritance of Vision and of the Relative Influence tance and Evolution. New York: Oxford University of Heredity and Environment on Sight. Cambridge: Press. Cambridge University Press. Jacob, F. (1976). The Logic of Life. New York: Bonner, J. (1965). The Molecular Biology of Develop- Vanguard. ment. Oxford: Oxford University Press. Jacob, F., and J. Monod. (1961). Genetic regulatory Brenner, S. (1974). New directions in molecular biol- mechanisms in the synthesis of proteins. Journal of ogy. Nature 248: 785–787. Molecular Biology 3: 318–356. Brenner, S., W. Dove, I. Herskowitz, and R. Thomas. Keller, E. F. (1995). Refiguring Life. New York: (1990). Genes and development: Molecular and logical Columbia University Press. themes. Genetics 126: 479–486. Keller, E. F. (2000). Decoding the genetic program. In Conklin, E. G. (1915). Heredity and Environment in the P. Beurton and R. Falk (Eds.), Genes, pp. 159–177. Development of Men. Princeton: Princeton University Cambridge: Cambridge University Press. Press. Kevles, Daniel J. (1986). In the Name of Eugenics: de Chardarevian, S. (1994). Development, programs Genetics and the Uses of Human Heredity. Berkeley: and computers: Work on the worm (1963–1988). Paper University of California Press. presented at the Summer Academy, Berlin, Germany. Lewontin, R. (1992). The dream of the human genome. Duden, B. (1991). The woman beneath the Skin. Cam- New York Review of Books, May 28, pp. 31–40. bridge, MA: Harvard University Press. 312 Evelyn Fox Keller

Markert, C. L., and H. Ursprung. (1971). Develop- Waldrop, J. M. (1992). Complexity. New York: Simon mental Genetics. Englewood Cliffs, NJ: Prentice-Hall. and Schuster. Mayr, E. (1961). Cause and effect in biology. Science Wolpert, L. (1969). Positional information and the spa- 134: 1501–1506. tial pattern of cellular differentiation. Journal of Theo- Monod, J., and F. Jacob. (1961). General conclusions: retical Biology 25(1): 1–48. Teleonomic mechanisms in cellular metabolism, Wolpert, L., and J. H. Lewis. (1975). Towards a theory growth, and differentiation. Cold Spring Harbor of development. Federation Proceedings 34(1): 14–20. Symposia Quantative Biology 26: 389–401. Wright, S. (1920). The relative importance of heredity Morgan, T. H. (1911). The inﬂuence of heredity and of and environment in determining the piebald pattern of environment in determining the coat colors in mice. guinea-pigs. Proceedings of the National Academy of Annals of the New York Academy of Science XXI: 87– Sciences 6: 320–332. 118. Moss, L. (1992). A kernel of truth? On the reality of the genetic program. Philosophy of Science Association 1: 335–348. Newman, S. A. (1988). Idealist biology. Perspectives in Biology and Medicine 31(3): 353–368. Oyama, S. (1989). Ontogeny and the central dogma: Do we need the concept of genetic programming in order to have an evolutionary perspective? In M. R. Gunnar and E. Thelen (Eds.), Systems and Development, pp. 1–34. Hillside, NJ: Lawrence Erlbaum Associates. Oyama, S. (1992). Transmission and construction: Levels and the problem of heredity. In G. Greenberg and E. Tobach (Eds.), Levels of Social Behavior: Evolu- tionary and Genetic Aspects, pp. 51–60. Wichita, KS: T. C. Schneirla Research Fund. Oyama, S. (2000). Evolution’s Eye: A Systems View of the Biology-Culture Divide. Durham, NC: Duke Uni- versity Press. Paul, D. (1995). Controlling Human Heredity. Atlantic Highlands, NJ: Humanities Press. Paul, D. (1998). The Politics of Heredity. New York: New York University Press. Sapp, J. (1987). Beyond the Gene. Oxford: Oxford Uni- versity Press. Stent, G. S. (1985). Hermeneutics and the analysis of complex biological systems. In D. J. Depew and B. Weber (Eds.), Evolution at a Crossroads, pp. 209–225. Cambridge, MA: MIT Press. Strohman, R. C. (1997). The coming Kuhnian revolution in biology. Nature Biotechnology 15: 194–200. Waddington, C. H. (1962). New Patterns in Genetics and Development. New York: Columbia University Press. Distributed Agency within Intersecting Ecological, Social, and Scientific 22 Processes

Peter Taylor

Societies emerge as changing alignments of social groups, selves—mental modules, optimizing or rational segments, and classes, without either ﬁxed boundaries or actors, or, most notably, genes. Similarly, to ex- stable internal constitutions. . . . Therefore, instead of as- plain the order of the world people have tradi- suming transgenerational continuity, institutional stabil- tionally posited a subject outside it, God, or, ity, and normative consensus, we must treat these as more recently, “the-forces-of-natural-selection.” problematic. We need to understand such characteristics In order to develop better explanations of de- historically, to note the conditions for their emergence, maintenance and abrogation. velopment, interaction, and perception, we need, —Wolf 1982: 387 Oyama implies, metaphors and concepts that do not rely on the dynamic unity and coherency of The anthropologist, Eric Wolf, proposes a con- agents, or on superintending agents within or ceptual inversion. Whenever theory has built on outside those agents. And, to the extent that such the dynamic unity and coherency of structures or patterns of thought persist because of their reso- units—in Wolf’s case, societies or cultures—con- nance with the experience agents have of their sider, instead, what would follow if those units relations and actions in the material and social were to be explained as contingent outcomes of world, we need different experience. Or, better, “intersecting processes.”1 This broad “Wolﬁan” we need to highlight submerged experience of heuristic informs this essay’s extensions of Devel- ourselves as “object-like” or “distributed,” that opmental Systems Theory (DST) to cases in the is, as agents dependent on other people and sociology of mental illness, social-environmental many, diverse resources beyond the boundaries studies, and social studies of science.2 I link the of our physical or mental selves. After all, the three cases in a project of reconceptualizing hu- primary experience of becoming an autono- man agents, in particular agents who are estab- mous subject is not “raw” experience, let alone lishing knowledge and engaging in change. I uniform and universal experience (Lebra 1984, show that viewing agents in terms of intersecting cited in Kondo 1990: 32), but experience medi- processes is also equivalent to teasing open their ated through particular social discourse. “heterogeneous construction,” that is, their con- There are circles here to be wrestled with. New tingent and ongoing mobilizing of webs of di- concepts and metaphors might emerge if we ex- verse materials, tools, people, and other resources. perienced ourselves differently, but what counts The importance for DST of reconceptualizing as our primary experience is mediated by prevail- agency is indicated by a the section of Susan ing conceptual schemes and shared metaphors. Oyama’s Ontogeny of Information on “Subjects And in current Western social discourse, these and Objects.” Oyama describes our primary ex- highlight our autonomy as subjects. Conversely, perience of ourselves as subjects maturing from when some of us seek to theorize developmental dependence and passivity to independence and systems or, in my case, to highlight distributed control—what I call “concentrated” agency. We agency, we foresake the facilitation afforded by come to experience temporal continuity and prevailing concepts and metaphors of concen- casual potency and are able to impart order trated agency. To so distance ourselves from the according to prior knowledge and plan. This ex- dominant discourse, however, requires a strong perience, however, “exaggerates our role as de- sense of “independence” and “causal potency” in tached subjects and denies our object-like status” attempting to impose an order—on one’s world (Oyama 1985: 76). Accordingly, when we try to and on one’s audiences.3 With these tensions ac- explain development, interaction, and percep- knowledged, but not resolved, let me move to the tion, we tend to posit another subject inside our- three cases. 314 Peter Taylor

and the subsequent attempt by the child to get by Case I. The Development of Severe Depression without the support of others. in a Sample of Working-Class Women The top three strands of figure 22.1 (class, family, psychology) combine the observations above A body of research initiated by the British soci- to explain the onset of serious depression.5 The ologists Brown and Harris in the 1960s has in- factors are not separate contributing causes, like terpreted the social origins of mental illnesses in spokes on a wheel, but take their place in the mul- a way that undercuts the persistent dichotomi- tistranded life course of the individual. Each line zation of genes versus environment. This “life should be interpreted as one contributing causal events and di≤culties” research, which is not well link in the construction of the behavior. The lines known in the United States, allows one to con- are dashed, however, to moderate any determin- clude that apportioning behavior to genes or en- ism implied in presenting a smoothed out or av- vironment is, at least for those seeking to reduce eraged schema; the links, while common, do not the incidence of mental illness, at best, not very apply to all women at all times, and are con- informative or helpful.4 To see how this follows, tingent on background conditions not shown in let me sketch their explanation of acute depres- the diagram. For example, in a society in which sion in working-class women in London (Brown women are expected to be the primary caregivers and Harris 1978, 1989). I will also work in the ex- for children (a background condition), the loss of tensions of their findings and generalized narra- a mother increases the chances of, or is linked to, tive contributed by Bowlby, a psychologist who the child’s lacking consistent, reliable support for focused on the long term effects of different pat- at least some period. Given the dominance of terns of attachment of infants and young children men over women and the social ideal of a hetero- to their mothers (Bowlby 1988). sexual nuclear family, an adolescent girl in a dis- Four factors are identified by Brown and Har- rupted family or custodial institution would be ris as statistically more common in women with likely to see a marriage or partnership with a man severe depression: a severe, adverse event in the as a positive alternative, even though early mar- year prior to the onset of depression; the lack of riages tend to break up more easily. In a society a supportive partner; persistently di≤cult living of restricted class mobility, working-class origins conditions; and the loss of, or prolonged sepa- tend to lead to working-class adulthood, in which ration from, the mother when the woman was living conditions are more di≤cult, especially if a child (under the age of eleven). Bowlby inter- a woman has children to look after and provide prets this last factor in terms of his and others’ for on her own. In many such ways these family, observations of secure versus anxious attachment class, and psychological strands of the woman’s of young children to caregivers. In a situation life build on each other. Let us also note that, as of secure attachment the caregiver, usually the an unavoidable side effect, the pathways to an in- mother, is, in the child’s early years, “readily dividual’s depression intersect with and influence available, sensitive to her child’s signals, and other phenomena, such as the state’s changing lovingly responsive when [the child] seeks protec- role in providing welfare and custodial institution and/or comfort and/or assistance” (Bowlby tions, and these other phenomena continue even 1988: 167). The child more boldly explores the after the end point, namely, depression, has been world, confident that support when needed will arrived at. be available from others. Anxious attachment, on Suppose now, quite hypothetically, that cer- the other hand, corresponds to inconsistency in, tain genes, expressed in the body’s chemistry, or lack of, supportive responses. The child is anx- increase a child’s susceptibility to anxiousness ious in its explorations of the world, which can, in in attachment compared to other children, even turn, evoke erratic responses from caregivers, Distributed Agency within Intersecting Processes 315

Figure 22.1 Pathways to severe depression in a study of working-class women. The dashed lines indicate that each strand tends to build on what has happened earlier in the different strands. See text for discussion and note 5 for sources. those within the same family. Suppose also that of onset of depression, for example, counseling this inborn biochemistry, or the subsequent bio- adolescent girls with low self-esteem, quickly act- chemical changes corresponding to the anxiety, ing to ensure a reliable caregiver when a mother rendered the child more susceptible to the bio- dies or is hospitalized, making custodial institu- chemical shifts that are associated with depres- tions or foster care arrangements more humane, sion. (This hypothetical situation is given by the increasing the availability of contraceptives for bottom strand of ﬁgure 22.1.) It is conceivable adolescents, increasing state support for single that early genetic or biochemical diagnosis fol- mothers, and so on. If the goal is reduction in lowed by lifelong treatment with prophylactic an- depression for working-class women, the un- tidepressants could reduce the chances of onset changeability of the hypothetical inherited genes of severe depression. This might be true without says nothing about the most effective, economi- any other action to ameliorate the effects of loss cal, or otherwise socially desirable engagement— of mother, working-class living conditions, and or combinations of engagements—to pursue. so on. There are, however, many other readily Notice also that many of these engagements have conceivable engagements to reduce the chances their downstream effect on depression via path- 316 Peter Taylor

ways that cross between the different strands. For features that I associate with the idea that example, if self-esteem counseling were some- something is “heterogeneously constructed,” or what effective, then fewer unwanted pregnancies an outcome of “intersecting processes.” (Most of and unsupportive partnerships might be initiated; these have equivalents in DST.) both effects could, in turn, reduce the incidence of single parenthood and di≤cult living conditions. a. Without any superintending constructor or These sequences of multiple causes, build- outcome-directed agent, ing on each other over the individual’s life his- b. many heterogeneous components are linked tory, permit a number of conclusions about the together, which implies that nature/nurture debate: c. the outcome has multiple contributing causes, and thus 1. Neither the unchangeability of genes nor the reliability of some gene- or biochemistry-based d. there are multiple points of intervention or intervention, such as the hypothetical prophylac- engagement that could modify the course of de- tic antidepressants, would prove that the genes velopment. In short, are the most signiﬁcant cause of the acute depres- e. causality and agency are distributed, not local- sion that has been occurring in the absence of ized. Moreover, such treatment. f. construction is a process, that is, the compo- 2. Critics of genetic explanations could dismiss nents are linked over time, the attribution of an individual’s behavior to g. building on what has already been con- genes (or 50 percent or 80 percent to genes) as a structed, so that technically meaningless partitioning of causes without placing themselves at the other pole from h. it is not the components, but the components genetic determination.6 That is, they would not in linkage that constitute the causes. Points (c) have to make the counterclaim that the environ- and (f–h) together ensure that ment determines behavior or that, if the right i. it is di≤cult to partition relative importance or environment were found, any desired behavior responsibility for an outcome among the different could be elicited. The Brown-Harris-Bowlby types of cause (e.g., 80% genetic vs. 20% environ- (BHB) account addresses malleability or immal- mental). Generally, leability of behavioral outcomes without ruling j. there are alternative routes to the same end, out genetic contributions. and 3. Similarly, critics would not need to rest their k. construction is “polypotent” (Sclove 1995), case on demonstrations that behavioral genet- that is, things involved in one construction pro- ics has been or still is methodologically ﬂawed cess are implicated in many others. Engaging in a (Lewontin, Rose, and Kamin 1984), on textual construction process, even in very focused inter- deconstructions of the categories and rhetoric ventions, will have side effects. Finally, points (f ) employed (Lewontin 1979), or on attributions of and (k) mean that political bias to the supporters of behavioral ge- l. construction never stops; completed outcomes neticists. These are all interesting, but, in light of are less end points than snapshots taken of ongo- the BHB account of the behavior, not necessary ing, intersecting processes. for a conceptual critique of genetic determinism. I am aware that there may be objections to the Over and above these conclusions, the BHB case I have chosen to make the preceding points. account of the origins of acute depression in In discussing depression among working-class working-class women also displays the following women, rather than in other groups, I could be Distributed Agency within Intersecting Processes 317

seen as perpetuating a male, professional-class lands, abandoning terraced lands, which then perspective. However, the politics of the case can eroded. The Indians adopted labor-saving prac- be viewed quite differently. Although depressed tices from the Spanish, such as cultivating wheat working-class women are the focus, the intersect- and using plows. As the population recovered ing processes account brings a range of other during the eighteenth and nineteenth centuries, agents into the picture. While the account does collective institutions evolved that reestablished not identify ways to cure the women studied, and maintained terraces and stabilized the soil other girls and women that follow them might dynamics. Erosion was reduced and soil accu- seek support from, or find themselves supported mulation was perhaps stimulated. This type of by—to pick up on the potential engagements landscape transformation also needed contin- mentioned earlier—counselors, hospital social uous and proper maintenance, since it introduced workers, people reforming custodial institutions, the potential for severe slope instability. The family-planning workers, social policy makers, collective institutions revolved around first the and so on. Moreover, these agents can view their church and then, after independence from Spain, engagement as linked with others, not as a solu- the rich Indians, caciques, mobilizing peasant tion on its own. For example, when women’s labor for key activities. These activities, in addi- movement activists create women’s refuges as a tion to maintaining terraces, included sowing step away from living in unsupportive house- corn in work teams, and maintaining a diversity holds, this makes it possible for therapists who of maize varieties and cultivation techniques. The specialize in the psychological dynamics of the caciques benefited from what was produced, but woman in her family to consider referring women were expected to look after the peasants in hard to refuges as a critical disruption to the family’s times—a form of moral economy (Scott 1976). dynamic. The politics of highlighting different Given that the peasants felt security in propor- kinds of causes and their interlinkages can be tion to the wealth and prestige of their cacique seen as promoting such exchange among the dis- and given the prestige attached directly to each tributed set of agents and contributing to the person’s role in the collective labor, the labor potential reformation of the social worlds inter- tended to be very e≤cient. In addition, peasants secting around the development of any given were kept indebted to caciques, and could not focal individual or outcome. readily break their unequal relationship. The caciques, moreover, insulated this relationship from change by resisting potential labor-saving Case II. The History of Soil Erosion in a technologies and ties to outside markets. Region of Oaxaca, Mexico The Mexican revolution, however, ruptured the moral economy and exploitative relationships In the mid-1980s resource economist Raúl by taking away the power of the caciques. Many García-Barrios and his ecologist brother Luis peasants migrated to industrial areas, returning studied severe soil erosion in a mountainous periodically with cash or sending it back, so that agricultural region near San Andrés in Oaxaca, rural transactions and prestige became moneta- Mexico, and traced it to the undermining of tra- rized. With the monetarization and loss of labor, ditional political authority after the Mexican the collective institutions collapsed and terraces revolution (García-Barrios and García-Barrios began to erode. National food-pricing policies 1990). The soil erosion of the twentieth century is favored urban consumers, which meant that in not the first time this has occurred in this region Oaxaca corn was grown only for subsistence of Oaxaca. After the Spanish conquest, when the needs. New labor-saving activities, such as goat indigenous population collapsed from disease, herding, which contributes in its own way to ero- the communities moved down from the high- 318 Peter Taylor

sion, were taken up without new local institutions teenth-century agro-ecology reveals the con- to regulate them. tingency that is characteristic of history. The Although this synopsis of the García- significance of such contingency rests not on the Barrioses’ account is brief and, like the first case, event of the revolution itself, but on the different smoothed out, it allows me to reiterate and elab- processes, each having a history, with which the orate on the intersecting processes viewpoint in revolution intersected. the context of social-environmental studies: 4. Structuredness. Although there is no reduc- 1. Differentiation among unequal agents. Sus- tion to macro- or structural determination in the tainable maize production depended on a moral above account, the focus is not on local, indi- economy of cacique and peasants, and the vidual/individual transactions. Regularities, for inequality among these agents resulted from a example, the terraces and the moral economy, long process of social and economic differentia- persist long enough for agents to recognize or tion. Similarly, the demise of this agro-ecology abide by them. That is, structuredness is discern- involved the unequal power of the State over ible in the intersecting processes. local caciques, of urban industrialists over rural 5. Distributed agency. The agency implied in interests, and of workers who remitted cash to the account of the García-Barrios brothers was their communities over those who continued distributed, not centered in one class or place. In agricultural labor. the nineteenth-century moral economy caciques 2. Heterogeneous components and inseparable exploited peasants, but in a relationship of recip- processes. As highlighted in figure 22.2, the rocal norms and obligations. Moreover, the local situation has involved intersecting processes op- moral economy was not autonomous; the na- erating at different spatial and temporal scales, tional political economy was implicated, by its involving elements as diverse as the local climate exclusion, in the actions of the caciques that and geomorphology, social norms, work rela- maintained labor-intensive and self-su≤cient tions, and national political economic policy. The production. Although the Mexican revolution processes are interlinked in the production of any initiated the breakdown in the moral economy, outcome and in their own ongoing transforma- the ensuing process involved not only political tion. Each is implicated in the others, even by ex- and economic change from above, but also from clusion (Smith 1984), such as when caciques kept below and between—semiproletarian peasants maize production during the nineteenth century brought their money back to the rural commu- insulated from external markets. No one kind of nity and reshaped its transactions, institutions, thing, no single strand on its own, could be su≤- and social psychology. cient to explain the currently eroded hillsides. In 6. Intermediate complexity. The García-Barrios this sense, an intersecting processes account con- brothers include heterogeneous elements in their trasts with competing explanations that center on account, but, as my synopsis and figure 22.2 in- a single dynamic or process, for example, climate dicate, different strands can be teased out. The change in erosive landscapes; population growth strands, however, are cross-linked; they are not or decline as the motor of social, technical, or torn apart. In this sense, the account has an in- environmental change; increasing capitalist ex- termediate complexity—neither highly reduced, ploitation of natural resources; modernization of nor overwhelmingly detailed. By acknowledg- production methods; or peasant marginalization ing complexity, the account steps away from de- in a dual economy (Peet and Watts 1996).7 bates centered around simple oppositions, e.g., 3. Historical contingency of processes. The role ecology/geomorphology versus economy/society. of the Mexican revolution in the collapse of nine- Similarly, by placing explanatory focus on the Distributed Agency within Intersecting Processes 319

Figure 22.2 Intersecting processes leading to soil erosion in San Andrés, Oaxaca. The dotted lines indicate connections across the different strands of the schema. See text for discussion. ongoing processes involved in the historically tion of traditional agricultural practices. They contingent intersections, the account discounts privilege multiple, smaller engagements, linked the grand discontinuities and transitions that are together within the intersecting processes.9 often invoked, for exmaple, peasant to capitalist agriculture, or feudalism to industrialism to This shift in how policy is conceived requires a Fordism to ﬂexible specialization.8 corresponding shift in scholarly practice. On the level of research organization, intersecting pro- 7. Multiple, smaller engagements. Distributed cesses accounts highlight the need, in brief, for agency, intermediate complexity, and the other transdisciplinary work grounded but not local- features of intersecting processes have implica- ized in particular sites. They do not underwrite tions, not only for how environmental degra- the customary, so-called interdisciplinary proj- dation is conceptualized, but also for how one ects directed by natural scientists, nor the eco- responds to it in practice. Intersecting processes nomic analyses based on the kinds of statistical accounts do not support government or social data available in published censuses. In all these movement policies based on simple themes, such different ways, representing intersecting pro- as economic modernization by market liberaliza- cesses is inseparably bound up with engaging or tion, or sustainable development through promo- 320 Peter Taylor

intervening10 in a way that further extends the an econometrician, continued his work on the idea of distributed agency. agricultural component of the institute’s forecast- ing model. The agricultural economist conducted extensive surveys of farm operations for forty Case III. The Simulated Future of a Salt- farms and acted as liaison with two senior agri- Affected Agricultural Region cultural extension o≤cers in the region who helped screen the production relationships and The “Institute” is an economic and social re- parameters used in the KFM. I was hired for search organization based in Melbourne, the fifteen months as a statistician and modeler to an- major city of the southern Australian state of alyze the farm surveys and to construct and oper- Victoria. The Kerang region, 240 kilometers ate the KFM. The ministry maintained oversight north of Melbourne, is an agricultural region of the project through its agricultural economist where farmers irrigate some pasture, which is and through regular meetings with the project grazed by beef or dairy cattle and sheep, and irri- team and an advisory committee. gate some crops. Soil salinization has been a The tangible products of the study included the chronic problem; during the middle 1970s, after survey and data analysis incorporated in one re- some very wet years, the problem was acute. The port to the ministry, the KFM and economic rise in salinity, following a decline in beef prices, analysis making up the second report, a technical threatened the economic viability of the region. monograph documenting the KFM, papers pre- The “Ministry” of the state government oversee- sented at two national conferences of agricultural ing water resource issues commissioned the insti- economists, and a public meeting in the Kerang tute in late 1977 to study the economic future of region to explain the results of the study (Fergu- the region. An agricultural economist from the son, Smith, and Taylor 1978, 1979; Taylor 1979). ministry and the principal investigator from the Although some refinements were omitted to Institute formulated a project to evaluate differ- meet the ministry’s deadline, the KFM was su≤- ent government policies, such as funding regional ciently flexible to allow evaluation of the required drainage systems, reallocating water rights, and range of factors, yet not so complex so as to be raising water charges. This evaluation would take unmanageable. into account possible changes in farming prac- At the public meeting to present the study’s tices, such as improvements in irrigation layout, findings some local agricultural extension o≤cers drainage, and water management, and changes in raised objections to the study’s having endorsed the mix of farm enterprises. The analysis was to irrigation of pasture over irrigation of crops. This be repeated for different macroeconomic scenar- ran contrary to the advice they had been giving ios as projected by the institute’s national fore- to farmers ever since the decline in beef prices. casting models. Subsequent reanalysis, incorporating generous The central part of the project was the con- increases in crop yields into the KFM’s parame- struction of what came to be known as the ters, was completed rapidly. This showed the re- Kerang Farm Model (KFM). Using an optimiza- sult favoring pasture irrigation was robust and tion technique called linear programming the could be attributed to beef prices having recov- KFM would determine for each of four compos- ered by this time in the late 1970s. The ministry, ite representative farms the mix of farming activ- meanwhile, focused its attention simply on re- ities that produced the most income. Different sults indicating that water charges were not a factors, such as water allocation, could be primary limiting factor on farm enterprises or vi- changed and the effect on the income and mix of ability. These results eclipsed others concerning activities ascertained. The division of labor in the the larger range of options that the institute had project was as follows. The principal investigator, Distributed Agency within Intersecting Processes 321

been commissioned to analyze, which suggests mize one objective: in the KFM, this objective that justifying an increase in water charges had was income. Furthermore, the use of a linear pro- been the ministry’s primary concern all along. gram for policy formation assumed that if the op- This last outcome could engender or reinforce timal mix of farming activities according to the cynicism or fatalism about social impact studies KFM were different from a farmers’ existing mix, commissioned by the authorities. However, if I the farmer would change accordingly and imme- were able to show the ways in which particular diately. Even though the economic future of the aspects influenced the results, I would be identi- region obviously entailed the farmers’ participa- fying how the research could have been done dif- tion, the study did not investigate why and how ferently. The possibility of identifying sites for farmers change, how directly and readily they re- possible modification of similar research informs spond to economic signs, or the extent to which the analysis to follow. any overriding economic rationality governed My entry point for analyzing the project will be their actions. around the modeling because that was the part The modeler questioned these assumptions. He that I, as a participant, observed more closely. I expressed interest in techniques that incorporated refer to myself in the third person as “the mod- more than one objective, but the principal inves- eler” to express some distance between my posi- tigator could not envisage modeling an alterna- tion and actions in 1978–79 and my interpretive tive objective to income. In any case, software for role today. I do not want to discount my obser- multiobjective analysis was not available at the vations and understandings as a participant, but computer center used by the Institute. The mod- it would be misleading to imply that during the eler designed the KFM to allow examination of Kerang study I had in mind a later analysis in the course over time of new investments needed, terms of the sociology of science. but when the project approached its deadline, this part of the model development was halted. Building and Probing the Kerang Farm Model The modeler learned of the existence of a sociological study on the factors influencing Kerang Diverse components went into the KFM: data on farmers to change their practices. This study had soil quality, expected crop yields, range of farm not, however, been released at that time and the sizes, technical assumptions used in the linear principal investigator lent no institutional sup- program, the status of the different agents in the port to obtaining advance access to it. These and project, the geographical distance between the other issues were, he maintained, outside the eco- institute and the Kerang region, the computer nomic specialization of the Institute and best left packages available, the terms of reference set by for others to deal with. the ministry, and so on. Moreover, many of these In a≤rming the technical assumptions in the components span the different realms of action KFM in response to the modeler’s questioning, of the various agents—from the modeler to the the principal investigator drew variously on his farmers—who are implicated in the building of senior and permanent position at the institute, the KFM. I need to put some order into this het- the Institute’s specialization in quantitative eco- erogeneity of components and assess their rela- nomic research, and the terms of reference and tive importance. Let me use my observations as deadlines that the ministry had set. These as- the modeler to unpack parts of the processes of sumptions, in turn, had several consequences. model-building here. They eliminated certain issues from investigation, Consider a central technical assumption in the for example, farmer’s objectives. They shaped KFM. The use of a linear program for economic the data that needed to be collected, for example, analysis assumed that farmers operate to maxi- obviating the need to investigate how farmers 322 Peter Taylor

change. And they colored the relationships put sentation of reality; neither were any farmers, into the model, for example, the time course of economists, or sociologists in a position to do so. investment became a secondary issue to locating the farming activities that optimized income. As Six Heuristics Drawn from the Reconstruction of an exercise in the authority of an experienced the Kerang Study principal investigator over a young researcher, this was not at all extraordinary. Nevertheless, The description of the building of the KFM, al- through such exchanges the principal investiga- though brief and clearly partial, is su≤cient to in- tor and the modeler were negotiating the different troduce six propositions concerning the processes components of what would count as a represen- of science in the making and interpreting those tation of reality and a guide to policy formation. processes. I begin with the observation that het- Of course, there were parties other than the erogeneous components from a range of realms principal investigator and the modeler potentially of social action are being drawn on by the differ- involved in accepting or disputing the KFM. The ent agents involved in the KFM (proposition 1). farmers might have objected to the way their Each of the other propositions follows more or behavior was modeled. The KFM could also less directly from the ones that have preceded it. have been disputed by economists interested in These propositions are advanced heuristically, multiobjective techniques, by sociologists inter- without expecting them to apply to all situations. ested in how people act, interact, and change, or 1. Science-in-the-making depends on heteroge- by agricultural policymakers interested in hav- neous webs, not unitary correspondence. From the ing the study’s results translated successfully into description above, it is clear that diverse com- changes in the state of farming in the region. ponents were involved in building the KFM. None of these potential disputes proved signiﬁ- Moreover, they were interconnected in practice, cant at the time. The farmers were separated forming heterogeneous webs. The assumption that from the formulation and operation of the KFM, farmers were subordinate to economic rationality and, conversely, the KFM was insulated from in the KFM facilitated the formulation of con- the farmers, by several considerations: by loca- clusions in the form of government policy option (the modeling was performed in the city); tions. The power of the government to enact its through a chain of personnel (modeler—agri- decisions rendered investigation of how farmers cultural economist—senior agricultural extension change less relevant, which shaped the data need- o≤cers—local agricultural extension o≤cers— ing to be collected. Generalized agronomic data, farmers); and by levels of abstraction and gener- rather than sociological insights, would su≤ce. alization. No one in the institute, the principal This, in turn, conditioned the relationships that investigator in particular, had training in mul- could appear in the model. Similarly, the mod- tiobjective economic analysis or ready access to eler’s mediated relationship with the modeled sit- suitable computer software. There were no soci- uation and his geographical separation from the ologists included in the project team or advisory region rendered it less relevant to model long- committee. The ministry, through the range of term options, such as selective reforestation and options established in the terms of reference for organic soil restoration. These possibilities, al- the study, indicated that change would be initi- though potentially of economic and ecological ated by government policy based on economic beneﬁt, would have required such things as ex- and engineering criteria. The farmers were, in ef- perimental plots, publicity, education, advocacy, fect, to be instruments, more than coparticipants, subsided loans for tree planting from the govern- in determining the future of the region. In short, ment, and other institutional changes before they the ministry did not dispute the KFM as a repre- could be adopted. With so many contingent fac- Distributed Agency within Intersecting Processes 323

In this sense I would say that science is constructed; science-in-the-making is an ongoing process of building from diverse components, as in building a house from the ground up using concrete, bricks, cement, wood, nails, and so on. This is social construction, but not “merely” social construction. Moreover, the associations that social construction has with reflection are not apt here. It might be possible to say that the model reflected all the different social components, but it would be stretching the metaphor of reflection. The heterogeneity and interlinkage of the components make it di≤cult and uninformative to collapse science-in-the-making to a unitary idea of reflection of society in theory, or, similarly, to an issue of correspondence of theory to natural reality. In short, science, I would say, is heteroge- Figure 22.3 11 Different components of socio-environmental model- neously constructed (Taylor 1995a). ing. See text for discussion of their interconnections. 2. Scientists represent-engage. In the process of building the model, the modeler, principal investigator, and other agents linked together tech- tors it was impossible even to estimate their costs. nical and social components in order to make Omission of such options from the modeling, in a model that worked for them. These scientific turn, helped ensure that such aspects of the future agents tended to make the different components reality would be less realizable, and the model’s reinforce, not undermine, each other, rendering account more real. Figure 22.3 presents a sche- both the model and the ongoing scientific activity matic picture of diverse components intercon- more di≤cult for others to oppose or modify in nected in the making of the KFM. practice (see proposition 1).12 This insight goes “Technical” considerations, such as the as- beyond the observation that representations of sumption of income optimization, and “social” natural reality support interventions in different considerations, such as the separation of the realms of social action, or the claim that repeat- modeler from the farmers, had implications in able interventions provide the basis for scienti- practice for each other. “Local” interactions were fic representations (Hacking 1983). Through the connected with activities at a distance. For exam- model’s heterogeneous construction, representa- ple, the modeler and the principal investigator tions and engagements were being formed simul- decided not to pursue sociological inquiry into taneously, and, moreover, jointly. “Interaction” how farmers change, which meant that the con- between “technical” and “social” considerations tent of and conduct of the survey of farms and fails to capture this relationship. Let me instead farmers could remain unchanged. No one com- speak of scientists representing-engaging. ponent in the web stood alone in supporting the 3. Scientists are practically imaginative agents. KFM as a representation of reality; in the actual The idea of representing-engaging implies that intersecting processes of building the model, tech- scientific agents are mindful both of nature and nical components could not be detached from so- of the social worlds in which they act, and that cial ones, nor local ones from those that spanned they project continuously between these realms. levels. This attention to their social situatedness is not 324 Peter Taylor

an accusation that scientists are corrupt, fallible, stead, we should inquire into the heterogeneity or lazily taking the path of least resistance. On of resources that facilitate agents acting as if the the contrary, it is an a≤rmation of the view that world were like their representations of it. During all human activity is imaginative, that is, the the Kerang study, the principal investigator may result of a labor process that has to exist in the well have believed deeply that economic decision- laborer’s imagination before the process com- making was of primary importance in people’s mences. Agents assess, not necessarily explicitly, lives. However, he was able to sustain this be- the practical constraints and facilitations of pos- lief against possible challenges by many prac- sible actions in advance of their acting (Robinson tical measures, such as not securing access to the 1984).13 sociological study on how farmers change, and Imagination in the sense I use it here is not at concentrating on his econometric investigations all like fantasy, in which worlds can be envisaged rather than developing skills in multiobjective and mentally inhabited so as to escape from the analysis. practical di≤culties of their realization. Achieving 5. Resources are causes. Up to this point in my some result in the material world, in contrast to description of the construction of the KFM, I in fantasy, requires human agents to be engaged have used the neutral term component to refer to with materials, tools, and, usually, other people. the diverse things that scientists link into webs The KFM modeler had to engage with pasture to support their theories and ongoing scientific growth, government sponsorship, an agricul- activity. But there are many components linked tural extension system, and so on. Moreover, together through the construction process that materials, tools, and other people confront scien- have little significance in explaining the develop- tists with their recalcitrance. So scientists project ment of theories and activity. The modeler, for themselves into possible engagements out in the example, used baking soda to clean his teeth world in order to imagine what will work easily at that time. But let me reserve the term resource for them and what will not. These constant pro- for components that make a claim or a course of jected confrontations with the components that action more di≤cult for others to modify. Re- personal and collective histories make available sources make a difference; that is, when resources lie behind all the actions people take, including are deployed they function as causes. In this light, scientists’ representing-engaging. Through them the term resource cannot be used descriptively people build up knowledge—not necessarily con- without also implying a claim about causes, and sciously articulated—about their changing capa- such claims invite analysis (see Taylor 1995a, ap- bilities for acting in relation to the conditions in pendix A). which they operate. 6. Counterfactuals are valuable for exposing 4. The agency of heterogeneously constructing causes. With the exception just now of the baking agents is distributed. If we focus on agents’ con- soda, the components of the construction process tingent and ongoing mobilizing of webs of mate- I have chosen to mention were significant re- rials, tools, people, and other components, we sources in the building of the KFM. Or so my ac- can think of their psychology or agency as count of the KFM would imply. But how can I distributed, not concentrated mentally inside support the causal claims that I have thus struc- socially autonomous agents. That is, although tured into my account of the KFM? For a start, agents work with mental representations of their let me note that, to support the causal claim that worlds, the malleability of those representations something made a difference logically requires an should not be understood merely in internal men- idea of what else could have been if the resource talistic terms related to belief or rationality. In- in question had been absent. There are many Distributed Agency within Intersecting Processes 325

sources for ideas about what else could have exploring the practical constraints on realizing been. Sociologists and historians of science listen counterfactual possibilities can, by a logic of in- to opposing parties in controversies (Collins version, expose the resources facilitating those 1981)—which include activists in movements who constructed what actually happened. for social change (Nelkin 1984)—undertake con- The emphasis on multiple, heterogeneous re- ceptual analysis or historical and cross-cultural sources means that the relevant counterfactuals comparisons (Harwood 2000), and give rein to are multiple and particular. We could formulate their sociological imagination (Hughes 1971). an all-encompassing counterfactual, in which, for Analyses of controversies have been popular; example, the Kerang study is replaced by a proj- they provide the clearest, most concrete evidence ect that could not be used for top-down govern- of alternatives, because the agents themselves ment policymaking. However, once we began identify the resources they consider important. to consider the practical implications of such a There is no logical reason, however, why the counterfactual, we would be challenged to iden- resources explicitly exposed during a controversy tify specific sites for possible modification of the constitute the full set used by a scientist. There research. This would be all the more the case if are resources taken for granted and shared by op- we focused on the practical implications for the posing parties and, moreover, resources that specific scientific agents involved. The modeler’s must be mobilized even when there is no appar- ability to produce results based on sociologically ent controversy. In short, ideas of what else could realistic processes of change was constrained, as I have been should not be limited by whether any- observed earlier, by his distance from the farm- one actually attempted to construct the alterna- ers’ realm of social action—distance given not tive situation. For all these reasons, explicit use of only by location, but also by the chain of mediat- counterfactuals may be needed in order to ana- ing personnel and degree of abstraction. The ge- lyze a more inclusive array of resources used in ographical and organizational distance was, in the construction of science. turn, related to the centralized character of gov- If we look back we can see that, although I ernment and intellectual activities in the one ma- began my account of the building of the KFM as jor city of each Australian state, something given a fairly neutral description, once I started to draw by the previous two hundred years of develop- connections among the heterogenous compo- ment. Toward the end of the project the modeler nents I began introducing counterfactuals. For considered a move counter to that centralization, example, in contrast to a single objective of namely, to live and work in the Kerang region as maximizing income in the modeled farms, I men- an agricultural consultant. He was aware that tioned the counterfactual possibility of multiob- this would raise practical issues such as purchase jective techniques. In explaining why this was not and maintenance of a car, long-distance access to incorporated in the KFM, I mentioned that the computer facilities and libraries, keeping abreast principal investigator’s training, his status rela- of discussions about the wider state of the rural tive to the modeler, software availability, and the economy, and other considerations of a more institute’s specialization were invoked during the personal nature. The modeler’s decision not to course of the study. These were constraints for move meant the representation of the Kerang re- anyone wanting to construct a multiobjective gion he was able to produce facilitated the mak- model. By identifying them I was implying that ing of policy based on simple economic grounds. the principal investigator’s training and so on This outcome did not flow from a political or in- were resources for constructing a model with a tellectual commitment to the economically based single objective function. In this general fashion, technocratic rationality; many practical, not only 326 Peter Taylor

intellectual or ideological, considerations would intersecting processes of different agents. At the have been entailed in producing a different result. same time, to the extent that agents need to explain their actions to others—and to themselves —when they attempt to mobilize different re- Conclusion: Persisting Tensions between sources or organize them in new directions, such Concentrated and Distributed Agency discursive themes may be valuable resources. The image that emerges is one in which agents For the three cases in this chapter I have pro- are always “vibrating” among their experiences vided overviews, in which the complexities have of concentrated and distributed agency. The chal- been smoothed or “disciplined.” The emphasis, lenge becomes to acknowledge the discursive however, on heterogeneous resources and on in- impact of simple themes, but to strengthen the vi- tersections of processes at different scales and brations in the direction of agents attending to types highlights the range of agents whose differ- their dependency on other people and many, dient engagements jointly might contribute to mod- verse resources beyond the boundaries of their ifying the focal outcomes and the social worlds physical or mental selves (Taylor 1999a).14 The intersecting around the development of those intersecting processes/heterogeneous construc- outcomes. As a result, the overviews do not tion framework introduced in this chapter clearly privilege interventions from a superintending or highlights distributed agency, but it would be bet- master position. ter for my case if I could move beyond text and I find the politics of distributed agency con- argument, to lead my audience into positive ex- genial, but recognize that a central question has periences of their distributed agency in establish- been left open—what would lead any agent to ing knowledge and engaging in change (Taylor engage so as to change the intersecting processes? 1990). To this end, the workshop processes devel- Actually, no one can simply continue to mobilize oped by the Institute of Cultural Affairs (ICA) the same resources as previously, because the have become my model.15 I could try to evoke the contingent intersection of different processes en- experience of ICA processes, but I am going to sures ongoing change and restructuring. So the leave it for interested readers to gain this experi- open question becomes what would lead any ence first hand. agent to try self-consciously to steer the restructuring in certain directions over others? One kind of answer would return us to concentrated Coda: Evolution in a Context of Unruly agency, in that we could point to the agent’s goal, Ecological Complexity such as preventing illness, soil erosion, and the production of models whose results can be ma- Given this open-ended conclusion, my argument nipulated by policy-makers, or convincing read- for pursuing and promoting the experience of our ers of the virtues of an intersecting processes distributed agency cannot on its own be expected framework. A different kind of answer, more to move readers to change their work and ideas. consistent with the spirit of this chapter, would Moreover, I have not explicitly bridged the gap stem from investigating the intersecting processes between the immediate focus of DST on develop- that have formed the particular agents in ques- ment and evolution and the areas addressed in tion (analogous to the first case, but without the this essay. I offer this coda, therefore, on theoriz- focal outcome of a mental illness). Indeed, the ing ecological complexity to nudge DST theorists Wolfian heuristic would have us subsume the first in the direction of exploring the kinds of inter- kind of answer into the second. That is, “goals” secting processes I have highlighted. Indeed, my become discursive shorthand for the particular own interest in intersecting processes grew from a Distributed Agency within Intersecting Processes 327

search for ways to theorize the complexity of eco- angle, models that distinguish among individual logical dynamics. Along the way I also observed organisms (in their characteristics and spatial that the structure and dynamics of this ecological location) have been shown to generate certain context have not been well integrated into de- observed ecological patterns, such as patterns of velopmental and evolutionary theory.16 The change in size distribution of individuals in a challenge of doing so needs eventually to be ad- population over time, where large scale, aggre- dressed—after all, all development and evolution gated models have not (Huston, DeAngelis and occurs within a dynamic ecological context Post 1988, Lomnicki 1988). And, the effects me- (Taylor 2001a). diated through the populations not immediately During the last decade or so a reassertion of in focus, or, more generally, through “hidden historical contingency, nonequilibrium formula- variables,” upset the methodology of observing tions, local context and individual detail has the direct interactions among populations and subdued the ambitions many ecological theorists confound many principles derived on that basis had in the 1960s and ’70s for identifying general (Strauss 1991; Taylor 2001b). principles (Taylor 1992; Kingsland 1995). Ecolo- To incorporate this new, or perhaps resurgent, gists have become increasingly aware that situa- emphasis, it has been suggested that ecology be tions may vary according to historical trajectories conceived as an “historical” science (Ricklefs and that have led to them; that particularities of place Schluter 1993). Like the fields of epidemiology, and connections among places matter; that time psychoanalysis, structural geology, paleontology, and place is a matter of scale that differs among and history proper, ecology faces the challenge of species; that variation among individuals can historical explanation: how to assemble a com- qualitatively alter the ecological process; that this posite of past conditions su≤cient for the subse- variation is a result of ongoing differentiation oc- quent outcomes to have followed, while, at the curring within populations (which are specifically same time, not obscuring the provisional quality located and inter-connected); and that interac- such accounts have, their being subject to compe- tions among the species under study can be arti- tition from other plausibly su≤cient accounts facts of the indirect effects of other “hidden” (Taylor 1987). The phrase “a composite of past species. conditions” could conjure up pure historical con- In patch dynamic studies, for example, the tingency, but I do not mean this. Like the ac- scale and frequency of disturbances that create counts of intersecting processes in this essay, his- open “patches” is now emphasized as much as toricity in ecological thought should preserve a species interactions in the periods between distur- place for regularities or structuredness of ecolog- bances (Pickett and White 1985). Studies of suc- ical patterns and processes. To say that ecological cession and of the immigration and extinction structure has a history is to say that it changes dynamics for habitat patches pay attention to the in structure and is subject to contingent events, particulars of species dispersal and the habitat while at the same time it constrains and facili- being colonized, and how these determine suc- tates the living activity that constitute any eco- cessful colonization for different species (Gray, logical phenomenon in its particular place. The Crawley, and Edwards 1987). On a larger scale, challenge facing ecology then is to theorize par- such a shift in focus is supported by biogeo- ticularity and contingency intersecting with struc- graphic comparisons which show that conti- ture, and of that structure changing in structure, nental floras and faunas are not necessarily in being internally differentiated and, because of equilibrium with the extant environmental condi- overlapping scales of different species’ activities, tions (Haila and Järvinen 1990). From a different having problematic boundaries—in short, to dis- 328 Peter Taylor

cipline, without suppressing this “unruly com- 2. Portions of this essay are adapted from other publi- plexity” of ecological processes (Taylor 1992, cations, with acknowledgment of the respective pub- 2001a, 2001b). lishers. Parts of Case I appeared in Taylor (1995a; It is within such unruly ecological complexity University of Chicago Press); Case II, in Taylor (1999; that organisms, for almost four billion years, Oxford University Press); Case III, in chapter 4A of Taylor (2001b; University of Chicago Press); and the have constructed their living and “evolved,” that coda, in Taylor (1997a; Taylor and Francis). Although is, given rise to descendants that differ from them. I have developed these cases without explicit reference It ought not be assumed that the ecological to DST, I have benefited from conversations with this context remains consistent, that is stable or re- volume’s editors since the middle 1980s. I welcome this peatable, with respect to evolution occurring in opportunity to let readers consider the convergences populations of individuals or, as in DST, of life between DST and ideas arrived at along some different cycles of organisms and resources (Gray 2001). paths. Although consistency of context may sometimes 3. Indeed, when we search for new concepts and meta- be the case, the relevant processes are not neces- phors, or more generally, use words and text to make sarily separable into “ecological” and “evolution- arguments and seek to convince others, we privilege ary” time scales (Taylor 2001a).17 The challenge three related and persistent “meta-metaphors”: “1) then for DST theorists is—as researchers con- metaphors are root, fundamental, underlying things that shape the surface layers; 2) mental things— ditioned by intersecting scientific and social thoughts, expectations, what we see—shape our ac- processes—to make sense of the appearance of tions; and 3) culture or society get into these thoughts organisms as intersections of ecological, develop- (and so we can be taught [or argued into] how to mental, and evolutionary processes. conceive/perceive the world” (Taylor 1997b: 222, note 37). These meta-metaphors discount our experience of thought being constructed in practical activity from Acknowledgments diverse resources. 4. Although associations between life events and di≤- I acknowledge valuable comments from Susan culties are also studied in the United States, conven- Oyama and Russell Gray, which helped me link tional quantitative epidemiology still dominates that my thinking to DST. research (Brown and Harris 1989: x–xi, 3–45). Associa- tions are thought of in statistical terms, that is, as if causality were a matter of adding up separate “effects” Notes (see note 6 below; chapter 15). In contrast, as the text and figure 22.1 indicate, Brown and Harris focus on the 1. Two comments on terminology: First, in other pub- development of life histories and the contingencies lications I use the term system or strong system to de- involved. note structures or units assumed to have dynamic unity 5. Figure 22.1 is adapted from Bowlby (1988: 177). His and structure (e.g., Taylor 1988, 1992, 1998, 2001b), schema is, in turn, adapted from Brown and Harris but, given the distinctions and arguments Oyama (1978: 265). The hypothetical genetics/biochemistry makes in this volume about the positive connotations strand is my addition. Its significance will become clear of system in DST, I have chosen not to cast the term in in due course. a critical light in this essay. Second, I use the term process in the sense of sequences of events that persist 6. The nonpartitionability of different kinds of biolog- or are repeated su≤ciently long for us to notice them ical and social causes, given the interdependence of and need to explain them. This contrasts with an essen- their effects, is demonstrated well by Lewontin (1974), tialist sense of process as a basic underlying causal when he argues that statistical partitioning of effects structure that allows people to explain events as in- (“analysis of variance”) does not constitute an analysis stances of the process or as noisy deviations from it. of causes. Of course, partitioning of biological and so- “Maturation,” “modernization,” and “population cial causes does have ideological meaning (Lewontin, growth” are examples of the latter sense of process. Rose, and Kamin 1984). Distributed Agency within Intersecting Processes 329

7. The combination of differentiation, historical con- eliminating problems in correspondence between the- tingency, and structuredness distinguishes this inter- ory and reality. secting processes view of socio-environmental change 2. Critics of scientific realism do not need to claim that from Vayda (1996). Although his approach shares construction is entirely a matter of social influences, many qualities with mine, he is more particularist and conventions or personal beliefs. Heterogeneous con- skeptical of theory based on social structures or struc- structionism addresses the malleability or immal- tured processes. leability of scientific knowledge without entailing such 8. Such discontinuities and transitions often rely on the relativism. sense of process that I want to avoid; see note 1. 3. Sociology of science’s analyses of methodology, in- 9. In the Oaxacan case, the changes of past centuries terests, and rhetoric are illuminating, but not strictly cannot be undone, but a more fine-grained intersecting necessary in the conceptual critique of scientific realism. processes analysis focusing on recent decades would More needs to be said to argue these propositions, expose a range of potential engagements—from Non- but not here (Taylor 1995a, 1995b); after all, analo- Governmental Organizations promoting conservation gies are meant to open discussions more than close of traditional cultivars to efforts to redirect interna- arguments. tional financial policies so as to support, rather than 12. The work of Latour (1994) and Pickering (1995) reduce, rural credit at the local level (de Janvry and shares with this essay an emphasis on process and sci- García-Barrios 1989). entific practice. Latour explores a Whiteheadean pro- 10. See Case III, heuristic 2. cess metaphysics for science studies. Heterogeneous 11. If we think of construction in terms of sequences construction might be read as Pickering’s “mangle” and of diverse, multiple causes, we can reject the terms of “impure dynamics,” and the imagination of scientists as the realism/relativism dichotomy persisting in explana- his “modeling.” In contrast, however, to my goals of tions of the course of science in similar ways to the ar- explanation and spanning levels of social practice, gument against nature vs. nurture from case I. Let us Pickering develops a position opposed to social con- read genes as underlying, “mind-independent” reality struction and causes and he theorizes practice mostly in and the environment as social influences on science. If terms of experimental practice. That is, it is mostly be- the outcome (mental illness, or, by analogy, some as- cause scientists tinker with tangible objects, whose re- pect of science, e.g., an established theory) is the result sistance requires accommodation, that their goals and of many heterogeneous components linked in a process, interests are subject to ongoing revision. Studies of in which each step builds on the outcomes of the pre- practice in that sense are reviewed by Golinski (1990); vious steps, then it is di≤cult to partition relative im- see also Pickering (1992). portance or responsibility for an outcome among the 13. Associating imagination and the labor-process is different contributing causes, or components in linkage. Marx’s idea. See Capital, volume 1, part 3, chapter 7, It becomes quite di≤cult to give meaning to determina- section 1, reprinted, e.g., in Tucker (1978: 344–345). tion by either nature or society, and not very helpful be- The convention in social studies of science has been to sides. Heterogeneous constructionism would, therefore, avoid reference to an agent’s psychology for fear of lead us to conclusions analogous to the three stated in shifting the terms of explanation from the social realm Case I for the nature/nurture debate: to an unobservable realm of the agent’s mind. I find du- 1. Suppose there are fundamental principles of nature bious both the equation of social with observable and that are di≤cult or impossible to modify (a tenet of sci- the empiricist rejection of unobservables, but, in any entific realism). This does not imply that this deep real- case, notice that imagination relies on a distributed, not ity predominantly or ultimately governs the actions of an internal, notion of mind and psychology. Further- scientific agents, in particular, their success in establish- more, psychological or cognitive models of the scientist ing some representation of this reality. Moreover, the as social agent are implicit in every explanation of the reliability of certain science-based interventions in the outcome of scientific activity. For example, Latour world (also important to most scientific realists) is in- (1987) depicts scientists building “networks” in re- teresting and worth explaining, but it does not justify sponse to the stimulus of others building competing the belief that sound scientific method is the most e≤ca- networks, and assumes that scientists seek to accu- cious route to exposing any unreliable knowledge or mulate resources, all of which results, if successful, in 330 Peter Taylor

“centers of calculation,” “obligatory passage points” Brown, G. W., and T. Harris. (1978). Social Origins of (Callon 1985), and their becoming macroactors (Callon Depression. New York: Free Press. and Latour 1981). Like the psychology of pigeons in Brown, G. W., and T. Harris. (1989). Life Events and the accounts of behaviorists, the psychology implied is Illness. New York: Guilford Press. both strong and minimal—the scientists are governed Callon, M. (1985). Some elements of a sociology of only by this egocentric metric of resource accumula- translation: Domestication of the scallops and the fish- tion; they are not assumed to have multiple projects in ermen of St. Brieuc Bay. In J. Law (Ed.), Power, Action, their lives and work. This, like most other models of Belief: A New Sociology of Knowledge? pp. 196–233. psychology and rationality implicit in social studies of London: Routledge & Kegan Paul. science, is quite restrictive, even when rationalized as a methodological tactic to highlight the flexibility of Callon, M., and B. Latour. (1981). Unscrewing the big agents’ actions and network building. Leviathan: How actors macro-structure reality and how sociologists help them to do so. In K. Knorr- 14. In this project I am inspired by DST-like work on Cetina and A. V. Cicourel (Eds.), Advances in Social the development of the self in relationship to others Theory and Methodology: Toward an Integration of (Fogel 1993) and to the “intentional scaffolding” others Micro- and Macro-Sociologies, pp. 277–303. Boston: provide (Hendriks-Jansen 1996). Routledge & Kegan Paul. 15. ICA workshops elicit insight from a large range de Janvry, A., and R. García-Barrios. (1989). Rural of participants in analyzing a situation and usually lead poverty and environmental degradation in Latin Amer- to plans that no one participant has envisaged before- ica: Causes, effects, and alternative solutions. Paper hand and that the participants are invested in carry- presented at the International consultation on environ- ing out. This is achieved by a neutral facilitator leading ment, sustainable development, and the role of small participants through four phases—objective, reflective, farmers, Institute for Food, Agriculture and Devel- interpretive, and decisional—a structure best repre- opment, Rome. sented in “focused conversations” (Spencer 1989; Eco- Stanfield 1997). For an elaboration of the basic pro- Ferguson, J., A. Smith, and P. Taylor. (1978). nomic Aspects of the Use of Water Resources in the positions of ICA facilitation and group process, see Kerang Region. http://omega.cc.umb.edu/~ptaylor/ICApropositions. (Vol. 1). Melbourne: Institute of Ap- html, which is adapted from workshop materials of plied Economic and Social Research. ICA Canada; see http://www.icacan.ca/. Ferguson, J., A. Smith, and P. Taylor. (1979). Eco- nomic Aspects of the Use of Water Resources in the 16. In a sense, ecological dynamics are implicit in any Kerang Region. evolutionary theory, but with “genetic (transmission),” (Vol. 2). Melbourne: Institute of Ap- “developmental,” “ecological” and “evolutionary” time plied Economic and Social Research. scales theoretically separated (Taylor 2001a). See note Fogel, A. (1993). Developing through Relationships: 17. Origins of Communication, Self, and Culture. New 17. Laland and Odling-Smee (see chapter 10), who ex- York: Harvester Wheatsheaf. tend the important emphasis of Lewontin (see chapter García-Barrios, R., and L. García-Barrios. (1990). En- 6) on organisms constructing the environments, recog- vironmental and technological degradation in peasant nize that the persistence of a constructed environment agriculture: A consequence of development in Mexico. (“ecological inheritance”) conditions subsequent evolu- World Development 18(11): 1569–1585. tion in the constructing species and others. Notice, Golinski, J. (1990). The theory of practice and the prac- however, that Laland and Odling-Smee do not other- tice of theory: Sociological approaches in the history of wise theorize the dynamics of environmental change or science. Isis 81: 492–505. explore the significance of those dynamics for the the- Gray, A. J., M. J. Crawley, and P. J. Edwards. (Eds.), ory of natural selection (Taylor 2001a). (1987). Colonization, Succession and Stability. Oxford: Blackwell. Gray, R. D. (2001). Selfish genes or developmental sys- References tems? Evolution without replicators and vehicles. In R. Bowlby, J. (1988). A Secure Base. New York: Basic Singh, C. Krimbas, J. Beatty, and D. Paul (Eds.), Books. Thinking about Evolution: Historical, Philosophical, and Distributed Agency within Intersecting Processes 331

Political Perspectives, pp. 184–207. Cambridge: Cam- Lewontin, R. C. (1983). The organism as the subject bridge University Press. and object of evolution. Scientia 118: 63–82. Gri≤ths, P. E., and R. D. Gray. (1994). Developmental Lewontin, R. C. (1985). Adaptation. In R. Levins and systems and evolutionary explanation. Journal of Phi- R. C. Lewontin (Eds.), The Dialectical Biologist, pp. losophy 91: 277–304. 65–84. Cambridge, MA: Harvard University Press. Hacking, I. (1983). Representing and Intervening. Cam- Lewontin, R. C., S. Rose, and L. J. Kamin. (1984). Not bridge: Cambridge University Press. in our Genes: Biology, Ideology and Human Nature. Haila, Y., and O. Järvinen. (1990). Northern conifer New York: Pantheon. forests and their bird species assemblages. In A. Keast Lomnicki, A. (1988). Population Ecology of Individuals. (Ed.), Biogeography and Ecology of Forest Bird Com- Princeton, NJ: Princeton University Press. munities, pp. 61–85. The Hague: SPB Academic Nelkin, D. (Ed.) (1984). Controversy: Politics of Techni- Publishing. cal Decisions. (2d ed.). Beverly Hills, CA: Sage. Harwood, J. (2000). National academic cultures: Oyama, S. (1985). The Ontogeny of Information: Devel- Science and scholarship in interwar Germany and opmental Systems and Evolution. Cambridge: Cam- the United States. In C. Charle, J. Schriewer, and P. bridge University Press. (2d rev. ed., Durham, NC: Wagner (Eds.), Transnational Intellectual Networks and Duke University Press, 2000.) the Cultural Logics of Nations, Oxford: Berghahn Oyama, S. (1988). Stasis, development and heredity. In Books. M-W. Ho and S. Fox (Eds.), Evolutionary Processes Hendriks-Jansen, H. (1996). Catching Ourselves in the and Metaphors, pp. 255–274. New York: John Wiley & Act. Cambridge, MA: MIT Press. Sons. Hughes, E. C. (1971). The Sociological Eye. Chicago: Peet, R., and M. Watts. (1996). Liberation ecology: Aldine Atherton. Development, sustainability, and environment in an Huston, M., D. DeAngelis, and W. Post. (1988). From age of market triumphalism. In R. Peet and M. Watts individuals to ecosystems: A new approach to ecologi- (Eds.), Liberation Ecologies: Environment, Develop- cal theory. Bioscience 38: 682–691. ment, Social Movements, pp. 1–45. London: Routledge. Kingsland, S. (1995). Afterword. Modeling Nature: Pickering, A. (Ed.) (1992). Science as Practice and Cul- Episodes in the History of Population Biology. (2d ed.). ture. Chicago: University of Chicago Press. Chicago: University of Chicago Press. Pickering, A. (1995). The Mangle of Practice: Time, Kondo, D. K. (1990). The eye/I. In Crafting Selves: Agency and Science. Chicago: University of Chicago Power, Gender, and Discourses of Identity in a Japanese Press. Workplace, pp. 3–48, 309–312, 329–339. Chicago: Pickett, S. T. A., and P. S. White. (Eds.) (1985). The University of Chicago Press. Ecology of Natural Disturbance and Patch Dynamics. Latour, B. (1987). Science in Action: How to Follow Sci- Orlando, FL: Academic Press. entists and Engineers through Society. Milton Keynes: Ricklefs, R., and D. Schluter. (Eds.) (1993). Species Di- Open University Press. versity in Ecological Communities: Historical and Geo- Latour, B. (1994). Les objets ont-ils une histoire? Re- graphical Perspectives. Chicago: University of Chicago contre de Pasteur et de Whitehead dans un bain d’acide Press. lactique. In I. Stengers (Ed.), L’effet Whitehead, pp. Robinson, S. (1984). The art of the possible. Radical 197–217. Paris: Vrin. Science Journal 15: 122–148. Lebra, T. (1984). Japanese Women: Constraint and Ful- Sclove, R. (1995). Democracy and Technology. New ﬁllment. Honolulu: University of Hawaii Press. York: Guilford. Lewontin, R. C. (1974). The analysis of variance and Scott, J. C. (1976). The Moral Economy of the Peasant: the analysis of causes. American Journal of Human Ge- Rebellion and Subsistence in Southeast Asia. New netics 26: 400–411. Haven: Yale University Press. Lewontin, R. C. (1979). Sociobiology as an adaptation- Smith, C. A. (1984). Local history in global context: ist program. Behavioral Science 24: 5–14. Social and economic transitions in Western Guatemala. Lewontin, R. C. (1982). Human Diversity. New York: Comparative Studies in Society and History 26(2): W. H. Freeman. 193–228. 332 Peter Taylor

Spencer, L. J. (1989). Winning through Participation. Africa. In F. Fischer and M. Hajer (Eds.), Living with Dubuque, Iowa: Kendall/Hunt. Nature: Environmental Discourse as Cultural Critique, Stanfield, B. (Ed.) (1997). The Art of Focused Conversa- pp. 121–134. Oxford: Oxford University Press. tion. Toronto: Canadian Institute of Cultural Affairs. Taylor, P. J. (2001a). From natural selection to natural Strauss, S. (1991). Indirect effects in community ecol- construction to disciplining unruly complexity: The ogy: Their definition, study and importance. Trends in challenge of integrating ecological dynamics into evolu- Ecology and Evolution 6(7): 206–210. tionary theory. In R. Singh, K. Krimbas, J. Beatty, and Thinking about Evolution: Historical, Taylor, P. J. (1979). The Kerang Farm Model. Mel- D. Paul (Eds.), Philosophical and Political Perspectives bourne: Institute of Applied Economic and Social . Cambridge: Research. Cambridge University Press. The Limits of Ecology and the Taylor, P. J. (1987). Historical versus selectionist expla- Taylor, P. J. (2001b). Re/Construction of Unruly Complexity. nations in evolutionary biology. Cladistics 3(1): 1–13. Chicago: Uni- versity of Chicago Press. Taylor, P. J. (1988). Technocratic optimism, H. T. Odum, and the partial transformation of ecological Taylor, P. J., and R. García-Barrios. (1995). The social metaphor after World War II. Journal of the History of analysis of ecological change: From systems to inter- Social Science Information Biology 21: 213–244. secting processes. 34(1): 5– 30. Taylor, P. J. (1990). Mapping ecologists’ ecologies of The Marx-Engels Reader. knowledge. Philosophy of Science Association 2: 95– Tucker, R. C. (Ed.) (1978). 109. (2d ed.). New York: W. W. Norton. Methods and Explanations in the Taylor, P. J. (1992). Community. In E. F. Keller and E. Vayda, A. P. (1996). Study of Human Actions and Their Environmental Lloyd (Eds.), Keywords in Evolutionary Biology, pp. Effects. 52–60. Cambridge, MA: Harvard University Press. Jakarta: Center for International Forestry Research. Taylor, P. J. (1995a). Building on construction: An ex- Europe and the People without History. ploration of heterogeneous constructionism, using an Wolf, E. (1982). analogy from psychology and a sketch from socio- Berkeley: University of California Press. economic modeling. Perspectives on Science 3(1): 66– 98. Taylor, P. J. (1995b). Co-construction and process: a response to Sismondo’s classification of constructiv- isms. Social Studies of Science 25: 348–359. Taylor, P. J. (1997a). Appearances nonwithstanding, we are all doing something like political ecology. Social Epistemology 11(1): 111–127. Taylor, P. J. (1997b). Afterword: Shifting positions for knowing and intervening in the cultural politics of the life sciences. In P. J. Taylor, S. E. Halfon, and P. N. Edwards (Eds.), Changing Life: Genomes, Ecologies, Bodies, Commodities, pp. 202–224. Minneapolis: Uni- versity of Minnesota Press. Taylor, P. J. (1998). How does the commons become tragic? Simple models as complex socio-political constructions. Science as Culture 7(4): 449–464. Taylor, P. J. (1999a). What can agents do? Engaging with complexities of the post-Hardin commons. In L. Freese (Ed.), Advances in Human Ecology vol. 8, pp. 125–156. Greenwich, CT: JAI Press. Taylor, P. J. (1999b). Mapping the complexity of social-natural processes: Cases from Mexico and 23 Niche Construction, Developmental Systems, and the Extended Replicator

Kim Sterelny

Lewontin’s World sometimes we should see the impact of organism on environment as the organism’s engineering its In the world of evolutionary biology, Richard own environment: The environment is altered Lewontin, perhaps more than anyone else, has in ways that are adaptive for the engineering or- emphasized the complexity, the subtlety, and the ganism. Thus some trees affect their local soil in variability of the relationship between genotype ways that improve their own prospects. Pine nee- and phenotype (Lewontin 1974, 1991). But he dles, for example, tend to suppress the growth of is also famous for rejecting a certain conception other plants that might compete for nutrients. of adaptation and adaptationism. Organisms, he But organisms do not just engineer their own maintains, do not just accommodate themselves environments, thus in part constructing their own to their environment. Organisms are active. They niches. They engineer their descendants’ envi- select their environment. For example, when an ronments. Downstream engineering establishes a insect selects a crack in bark in which to shelter, causal nexus between the generations additional it chooses to experience a range of temperature to that of genetic transmission. Thus we can see and humidity. These conditions might be quite Lewontin’s critique of externalist adaptationism different from the local norm. But they also as adding to his rejection of gene-centered biol- transform their environment, affecting both ogy. For environmental engineering creates what themselves and the biological communities of Laland, Odling-Smee, and Feldman call “ecolog- which they are a part (Lewontin 1983, 1982, ical inheritance” (Odling-Smee et al. 1996; see 1985). Just as environments alter organisms, or- also chapter 10). So, for example: ganisms alter their environments. Thus Jones et al. (1997) write: 1. Many animals make burrows, nests, shelters (like the beaver’s dams), and other structures. What does a tree do in a forest? Of course the living and These structures modify the impact of the envi- dead tissues are eaten . . . and the tree competes with ronment on both engineers and their offspring. other plants. . . . But the tree does much more than pro- Thus beaver lodges provide shelter against the vide food and directly compete for resources. The extremes of climate and protection against pred- branch, bark, root, and living and dead leaf surfaces ators. Beaver lodges, rabbit warrens, the tunnel make shelter, resting locations and living space. Small systems of naked mole rats, and some ant and ponds full of living organisms form where throughfall gets channelled into crotches . . . and the soil cavities termite nests are utilized and reengineered by that form as roots grow provide animals with places to those offspring as they become adults. live and cache food. . . . The leaves and branches cast 2. In selecting their microhabitat, organisms fil- shade, reduce the impact of rain and wind, moderate ter the way the world impinges on them. But they temperature extremes, and increase humidity for or- also often filter the way the world impinges on ganisms in the understory and soil. . . . Root growth aer- their offspring, for at least part of the life cycle of ates the soil, alters its texture and affects the infiltration those offspring. of water. . . . Dead leaves fall to the forest floor altering raindrop impact, drainage, and heat and gas exchange 3. Organisms do not just engineer the physical in the soil habitat, and make barriers or protection for world. They engineer other organisms, for seeds, seedlings, animals and microbes. . . . The roots themselves and their descendants. Parasites and can bind around rocks, stabilizing the substrate. (p. parasitoids often engineer the behavior or mor- 1946) phology of their host, often very gruesomely. These Some of these impacts are mere effects; they are effects on the host typically buffer or transform by-products of the organism’s way of life. But their offspring’s environment. Thus many wasps 334 Kim Sterelny

paralyze their prey without causing immediate nerable among them. Engineering the epistemic death, leaving the host incapable of defense and environment of offspring is probably much more preventing decay before the host is converted into common. There is very little evidence of explicit wasp. Less gruesomely, the same is true of mutu- teaching amongst animals (Caro and Hauser alisms. Leaf cutter ants are equipped with adap- 1992), though female cats of various species pro- tive specializations of many kinds in the care and vide opportunities for their young to learn to transmission of their fungal symbiont. There are hunt by providing them with particularly easy many other examples of obligate associations prey. But there are many ways animals can engi- of this kind. David Wilson discusses the mutual neer the epistemic environment of their young adaptation of the members of phoretic associa- that do not depend on explicit teaching. What tions with one another. A phoretic association is an animal does, where it goes, and what it eats a group of organisms with intrinsically poor pow- will provide information of itself—will provide ers of dispersal that rely on a winged host for learning opportunities—especially if young transport to the ephemeral and widely scattered animals are predisposed to pay special atten- resources they need. Because their own repro- tion to the behavior of its parents. And, of duction relies on the continued success of their course, many mammals monitor quite closely transportation system, Wilson notes, shared evo- the exploratory activities of their young, making lutionary interests can cause “some phoretic as- trial-and-error learning much safer that it would sociations to evolve into functionally organised otherwise be. teams that both feed and protect their insect car- This phenomenon is now, I think, widely rec- rier” (Wilson 1997: 2021). ognized. But its analysis and explanation remain 4. Sometimes plants change their environment highly contested. This chapter proceeds as fol- but do not engineer it. They make it less rather lows. In the next section I distinguish three takes than more suitable for themselves and their seeds, on environmental engineering. One downplays and hence promote a successional cycle. But we the cross-generation significance of environmen- should not suppose that this is the typical impact tal engineering, not treating it as an inheritance of plants on their immediate environment. For mechanism. Another focuses only on inheritance, trees and other plants do not just change their en- treating downstream ecological engineering as vironment, they adapt it. They change it in ways an aspect of cultural inheritance. Developmen- that promote their own survival, growth, or re- tal systems theory fuses the two, treating envi- production. Perhaps the most spectacular exam- ronmental engineering both as part of the ex- ples of such engineering is the way fire-resistant panded view of inheritance, but also treating an trees like eucalypts engineer their environment expanded view of inheritance as a reason for for fire: their distinctive menu of litter makes focusing on organism/environment interactions. their environment more fire-prone. A core feature of developmental system the- 5. Animals sometimes engineer their epistemic ory’s expanded view of inheritance is the “parity environment. They do this for themselves when thesis.” Genes play an important role in devel- they probe their environments, manipulating it opment and evolution, but they do not play a to access the information they need. For exam- unique role. The parity thesis is central to the ple, small fish like guppies sometimes deliber- third and fourth sections, where I develop a case ately inspect potential predators. Predators also for a substantially modified and qualified version inspect prey. Pursuit predators like African wild of the thesis. The usual way of rejecting parity is dogs are thought to probe the behavior of herds to argue that genes play an informational role of potential prey, attempting to induce in them in development, and other developmental re- behaviors that reveal the weakest and most vul- sources do not. That is not my strategy. Instead, Niche Construction and the Extended Replicator 335

I argue that an inheritance system must have a environmental engineering. It says nothing about particular suite of features if life is to be evolv- ecological inheritance. able. Genetic inheritance does have these features, but much of what is lumped into ecological Developmental Systems Theory and Niche- inheritance—though not all—does not. So al- Constructing Phenotypes though downstream environmental engineering is important, for the most part it should not be Developmental systems theorists wholeheartedly treated as an inheritance system. So parity is re- embrace Lewontin’s world. As I understand it, jected, or accepted in only a qualified way. developmental systems theory is characterized by three critical theses and a positive proposal. First, the critical theses: (1) We cannot simply assume Inheritance and Ecological Engineering that the organism/environment boundary is of theoretical significance for developmental and The Extended Phenotype evolutionary biology. Here there is a point of intellectual contact with the extended phenotype Dawkins (1982) reorganizes Lewontin’s insight. perspective. (2) It may be legitimate to fore- He argues that the effects of environmental engi- ground genetic structure and genetic change for neering should be seen as part of the phenotype specific explanatory or predictive purposes. But of the engineering organism, for these are the ef- in general, the genes an organism carries are just fects through which the responsible replicators one set of developmental resources among many. have been selected. I have argued elsewhere for Genes and gene changes are important both to the evolutionary significance of this conception. development and to evolution, but they are not of The single best reason for taking the gene/vehicle primary or privileged importance. (3) Develop- conception of evolution to be a genuine alterna- mental systems theorists are skeptical about the tive to the standard genotype/phenotype concep- project of explaining intergenerational similarity tion is the fact that the systematic, evolutionarily by appealing to the transmission of phenotype- significant phenotypic effects of genes are not making information across generations. In part, confined to the body of the organism carrying it is this scepticism about the idea of a preexisting them (Sterelny et al. 1996; Sterelny and Gri≤ths program that endogenously sequences develop- 1999, chap.3). For many extended phenotype ef- ment that underpins their deflationary view of the fects—the construction of shelters, the effects of evolutionary significance of genes. parasites and symbionts on their associates, the impact of plants on soil structure—are as de- The positive program of developmental systems velopmentally stable as standard features of an theory is that the fundamental unit of evolution is organism’s phenotype. Moreover, as Laland, the life cycle. In turn, the life cycle is the set of Odling-Smee, (and Feldman chapter 10) note, developmental resources that are packaged to- extended phenotypic traits can lead to evolution- gether and interact in such a way that the cycle ary cascades. The evolutionary invention of dam is reconstructed. The most obvious life cycle is building will trigger other changes in beaver phe- that of the organism plus its immediate environ- notypes, for life in the dam will modify the effects ment, but developmental systems theorists are of natural selection on many beaver traits. But open to the idea that cycles will exist at both finer this conception can be at most part of the analy- and coarser grains. For example, Gri≤ths and sis of the phenomena illustrated by 1–5 earlier. Gray (1997) argue that Keller and Ross’s work For although it captures the way biological enti- on pheromone-mediated transmission of fire ant ties transform their own environments, it does queen morphology is not just a good example of not capture the cross-generational significance of 336 Kim Sterelny

nongenetic inheritance; it is also a case where we dants, relative to their selected environments in should think of the life cycle as a nest life cycle, the standard way but also bequeath a legacy of not an organism life cycle (Keller and Ross 1993). modified natural selection pressures, as an eco- So in this respect, their view does depart from logical inheritance, relative to those same genes” that of Lewontin. For he does seem to have an (p. 125). Putting these ideas together, they argue: organism-centered view of evolution. Laland, Odling-Smee, and Feldman have de- When phenotypes niche-construct, they can no longer be thought of as simply “vehicles for genes,” since they veloped a view of the Lewontin phenomena that are now responsible for modifying some of the sources seems to me to be a cut-down or conservative of selection in their environments that may subse- version of developmental systems theory (see quently feed back to select their genes. Moreover, there Odling-Smee 1994; Odling-Smee et al. 1996; is no requirement for niche construction to result di- Laland, Odling-Smee et al. 2000). In these rectly from genetic variation before it can influence the papers, they develop two key ideas, one about selection of genetic variation. (Odling-Smee et al. 1996: evolutionary feedback loops, the second and 645–646) stronger about inheritance. The first idea is that They think that this is most obviously true of hu- ecological inheritance causes evolutionary cas- man culture, for we are ecological engineers par cades. So in constructing their own niches, excellence. But it is true of organisms in general. organisms do not just engineer their own envi- The result is a conception of evolution strikingly ronments and directly influence the environments similar to that of developmental systems theory. of their own offspring. Ecological inheritance— Thus they claim: the constructed niche—changes the way selection acts on genetic inheritance. There are feedback The entities that are selected, and between which there links between ecological and genetic inheritance. are fitness differences, are not well-described as “vehi- In chapter 10, Laland, Odling-Smee, and Feld- cles” or even “interactors” . . . but rather are “organism- man focus on this aspect of their model; that is, environment systems.” (Laland et al. 2000: p. 143) on the idea that ecological inheritance feeds back However, though their ultimate conception of onto genetic inheritance. So, for example, they evolution converges on developmental systems say, “We argue that through niche construction theory, in some respects their view is more con- organisms not only shape the nature of their servative. They do not reject the information- world but also in part determine the selection flow conception of cross-generation similarity, pressures to which they and their descendants are and hence genetic inheritance remains central exposed” (p. 117). The impact of organisms on to their conception. Moreover, they have some their environment is a product of evolution, but tendency to focus on just two streams of cross- it is also a cause of evolution, acting as a lens generation influence: genetic inheritance and cul- through which selection effects further changes in tural inheritance, with genetic inheritance being the lineage. in some way primary, but modified by a second- But they also view niche construction as a ary system. Other factors are important because bona fide inheritance system—as an independent they modify, and are modified by, the genetic in- channel through which intergenerational similar- heritance system. I think this picture is particu- ity is maintained. Thus, they write, “the effects of larly evident in Odling-Smee (1994). There he niche construction are frequently nontrivial, di- treats genetic inheritance as a flow of informa- rectional, accumulatory, pervasive and likely to tion, and other mechanisms as the provision of change the nature of the evolutionary process” resources (p. 178). But it emerges in many other (p. 118) and “the problem is that ancestral or- places in their argument; for example, in their dis- ganisms not only bequeath genes to their descen- Niche Construction and the Extended Replicator 337

cussion of Boyd and Richerson’s model of cul- These theories come in two versions. One version tural selection through individual imitation. defines cultural inheritance rather narrowly. In Thus the niche-construction view seems to our lineage, the flow of the skills of individuals come to something like this. Fundamentally, and of groups to the next generation through im- the history of evolution is the history of the evo- itation and explicit teaching constitute a cultural lution of organism/environment constructions. inheritance system. On this narrow view, there Nonetheless, we can usefully treat organisms as is rather little—perhaps no—cultural inheritance the units of selection (qua interactors), but they outside the human/great ape clade. Cultural pass similarity across the generations by both inheritance is a key part of the explanation of genetic and culturally transmitted information. human evolution. But it is not a phenomenon of So though their ultimate conception of evolution much general evolutionary significance. Genetic seems almost the same as that of developmental inheritance, on this view, is the core inheritance systems theory, they think that dual inheritance mechanism, supplemented (and perhaps even models are useful simplifications, whereas de- subverted) by meme-based inheritance in a few fenders of developmental systems theory do not clades. This conception seems to fit not just the (Gray 1992; Gri≤ths and Gray 1994). models of Boyd and Richerson (1985), but also Nonetheless, developmental systems theory the ideas of Dawkins (1989) and Dennett (1995). and the niche-construction conception are the But it will miss most of the examples discussed in only two approaches that identify an intrinsic 1–5 earlier. connection between environmental engineering The alternative is to try to shoehorn a very het- and a critical role for nongenetic inheritance. erogeneous range of examples into the category Thus, though the extended phenotype conception of cultural inheritance, treating, for example, the identifies an important role for ecological engi- fidelity of seabirds to the nesting sites, the fidelity neering, it was developed to sustain the theoreti- of phytophagous insects to the plant types on cal primacy of genetic inheritance. Conversely, which they hatched, or the pheromone-mediated the expanded inheritance models that I shall transmission of queen morphology in ants all as shortly consider are ex o≤cio silent on the signifi- instances of cultural transmission. There is noth- cance of environmental engineering, though they ing catastrophically wrong with treating cultural are certainly compatible with recognizing its transmission as a broad category. But it does significance. Only developmental systems theory have two disadvantages. First, 1–5 seem to form and its conservative cousin see these as two sides a very diverse set of examples, and it may be a of the same coin. mistake to choose a terminology that conflates their differences. Indeed, that will be the burden Expanded Inheritance of the fourth section of this chapter. Second, the choice of this paradigm can make an infor- The theory of the extended phenotype recognizes mational conception of inheritance seem less the pervasive ways in which genes and genotypes problematic than it is. When a young hominid is engineer their own environments. But though explicitly taught the technique of knapping a consistent with extragenetic inheritance, this view handaxe, there clearly is a flow of information is silent on it. Dual inheritance theories are the across generations. If teaching is not the trans- inverse of the extended phenotype model. Dual mission of information, nothing is. Other cases inheritance theories explain intergenerational are far more problematic. The symbiotic fungus similarity by an appeal to a flow of information on which leafcutters depend is transmitted from across the generations. But they recognize the ex- one nest to the next, and through exquisite adap- istence of a cultural as well as a genetic channel. tation of the winged queens. But it is much less 338 Kim Sterelny

obvious that this is transmission of information cance in itself, because (as has been pointed out across the generations. since Lehrman [1953]) selection-driven evolution- These points lead us to a less grudging recogni- ary change depends on the existence of such tion of the variety of inheritance mechanisms. On similarities but it is indifferent to their mecha- the “extended replicator” conception, the replica- nism. So ecological inheritance seems to be a tors are that set of developmental resources that channel across the generations that at least sup- are adapted from the transmission of similarity plements genetic inheritance. But downstream across the generations. So it is a multiple inheri- engineering has further evolutionary significance. tance model, but it takes the number of channels By modifying their offspring’s environment, or- determining similarity to be an open empirical ganisms modify selective pressures on their own question. Moreover, the centrality of genetic lineage. And hence we can expect to find evolu- transmission is not built into the model itself. It is tionary feedback loops of various kinds. Down- an empirical question whether genetic transmis- stream ecological engineering causes changes sion has deeper evolutionary roots—whether it in the stream of gene copying across genera- originated earlier in life’s history—than any other tions, and those changes in turn modify ecologi- kind of replicator. After all, it is quite certain that cal engineering. the mechanisms that construct, copy, and trans- I then sketched three different responses to this late DNA are one complex product of evolution. ecological, developmental, and engineering nex- DNA transmission and use is certainly not the us. One downplays the evolutionary significance original route of cross-generational similarity. In- of the organism-environment boundary but does deed, on Maynard Smith and Szathmary’s view, not rethink inheritance. Another rethinks in- the construction of the DNA/protein transcrip- heritance mechanisms but not the organism- tion system is the third of seven major transfor- environment boundary. Developmental systems mations in life’s history (Maynard Smith and theory and niche construction, in somewhat dif- Szathmary 1999: 17). Equally, it is an empirical ferent ways, make this boundary central both to question whether genetic transmission is more action in the environment (interaction) and the widespread across the tree of life than any other establishment of similarity across the generations set of replicators. It might be, for example, that (replication). On this perspective there is a very cytoplasmic factors transmitted in mitotic cell di- intimate linkage between the two phenomena. visions are as ancient and as widely used as DNA The issue is: is this a bug or a feature? In the replication and transcription itself. My own view fourth section I shall be suggesting that very is that DNA-based transmission of similarity is significant aspects of niche construction should of fundamental significance. But that is not built not be treated as inheritance systems. So, having into the structure of the theory. to date emphasized the similarities among 1–5— It is time to summarize the state of play. I the elements of ecological inheritance—I want began by describing a striking phenomenon: or- now to disaggregate, exposing some important ganisms engineer their own environments, often differences. in ways that are patterned across generations. In doing so, they typically also engineer the environment of their offspring. This downstream Hoyle’s World engineering has both developmental and evolutionary consequences. Developmentally, down- Suppose you are a biological engineer working stream environmental engineering is one causal for Hoyle & Co., and hence you are in the busi- channel through which parent/offspring simi- ness of supplying an empty planet with an indige- larity is generated. This has evolutionary signifi- nous biota. You plan to stock the planet with a Niche Construction and the Extended Replicator 339

seed of life and let evolution do its work. You minding us that evolutionary innovation depends want to produce a rich, complex, well-adapted, on cumulative selection. “No adaptation without and varied biota—biota in many ways like life cumulation” should be written on his tombstone. here. So you will design your seedling in ways He is right, of course. New complex structures that make selection and evolution effective. As we are very unlikely to arise through a single muta- all know, one necessary condition of selection- tional change. For if there are major changes to driven evolution is heritability. For no adaptive many aspects of an organism’s phenotype, it is change will take place unless the offspring of suc- very likely that at least one will lead to catastro- cessful organisms are likely to resemble their par- phe. Hence we need phenotype stability: biolog- ent(s) in the respects that make them successful. ical organisation must be reliably rebuilt over So you will need to design an inheritance mech- many generations, not a few. anism. To put it a little tendentiously, you will Third, selection depends on the generation of need some mechanism to insure that the biologi- variation. If evolution is to build a disparate cal characteristics of the parent are replicated in biota, or one characterized by adaptive complex- its descendants. One possible mechanism would ity, the replication system must have the potential be to ensure similarity by the transmission from to generate a large number of distinct pheno- parent to offspring of a set of replicators, that is, types. Given these desiderata, I suggest that a a set of factors each of which makes some pre- replication system should have the following dictable causal contribution to the organism’s characteristics. biological organization. Thus if two organisms received similar replicator sets, they would de- Anti-Outlaw Conditions velop a similar biological organization. What would the specifications for our replicator sets (C1) Replicators should be transmitted verti- look like? cally. Replicators are to be transmitted only to I shall argue that our inheritance system offspring and from parents. should meet three general specifications. First, it (C2) Replicators should be transmitted simulta- must somehow block outlaws. Complex living neously. systems depend on the cooperation of many com- (C3) The transmission of the replicator set ponents. The division of labor and adaptive spe- should not be biased. Either all an organism’s cialization seen in, for example, an ant nest replicators are transmitted to each descendant, depend on the linked reproductive fate of the ants or each replicator has an equal chance of being within the nest. If some ants in the nest could re- transmitted to each descendant. produce independently of others, very likely circumstances would arise in which the reproductive (C4) Replicators should have a “ballistic” com- interests of these individuals would not be identi- mitment to their biological role. Once some fac- cal to that of the nest, and defection would un- tor becomes adapted for its role as one of a set dermine cooperative integration. Similarly, if an of replicators, there should be no turning back. organism is to be built by a team of replicators Evolution-by-design should block any evolution- cooperating together, those replicators must have ary escape from the replicating role. a shared evolutionary fate. For otherwise the temptation to defect will undermine cooperative Stability Conditions organism building. Second, the replication system should insure (C5) The copy-fidelity of the generation of repli- the stable transmission of phenotypes over the cators from generation to generation should be generations. Richard Dawkins never tires of re- high. 340 Kim Sterelny

(C6) The replicator/organization map should be sion, has the same motivation, since one way a robust. Thus (1) the replicator/organization map replicator can go it alone is through horizontal or should have redundancies built into it, and (2) to oblique transmission. the extent that the causal channel from replicator One idea behind C2, the demand for simul- to organization depends on context, both internal taneous transmission, is that drip-feeding repli- and external, that context should be stable and cators from one organism to the next opens the predictable. door to transmission biases. For replicators that arrive first can filter those that cross later. Imag- Generation of Variation ine, for example, that (a) the replicator set for an organism—a cockroachoid, perhaps—includes a (C7) The array of possible replicator sets should set of bacterial symbionts that are critical to food be very large; possibly even unbounded. digestion and which are passed from parent to (C8) The effect of a replicator on the biologi- descendant; (b) the organism is fittest if it has sev- cal organization of its carrier should be well- eral different clones of symbiont; (c) symbionts behaved. That is, the replicator/organization compete to some extent within the roachoid, and map should be smooth. A small change in the hence each clone would be fitter if it were alone; replicator set should generate a small change in (d) different clones are passed at different times. biological organization. In that case, any mutation in an early passage clone that closed the door behind it would be fa- (C9) The generation of biological organization vored in it, even though it reduced roachoid fit- from the replicator set should be modular. The ness. As it happens, these are not biologically replicators as a whole should not generate the unrealistic assumptions. There are arthropod biological organization of the organism as a species which standardly have a suite of symbiont whole; rather, replicators, or small sets of repli- passengers (perhaps because different bacterial cators, should be designed so that they make a species have different peak e≤ciencies with differ- distinctive contribution to the generation of one ent foods) and competition between them within or a few traits, and relatively little distinctive the host is a possibility (Frank 1996). contribution to others. The stability criteria I take to be uncontrover- Some of these criteria are self-explanatory, but sial, though their application turns out to be less other may need some defence. Let me begin with transparent than the criteria themselves. C5 de- anti-outlaw criteria. As noted, the evolution of mands high fidelity copying of the replicator set, complex phenotypic structures is an evolutionary and C6 demands high fidelity use of that set. So achievement that depends for its success on the let’s turn to the generation of variation. Selection suppression of defection. This idea was originally depends on the existence of selectable variation. developed in the context of group selective expla- Hence we need a rich array of replicator packages nations of altruism (Williams 1966; Sober and (C7). Maynard Smith argues for the centrality Wilson 1998), and it has subsequently been ap- of this criterion: he argues that a crucial tran- plied to the evolution of multicelled organisms sition in evolution is the shift from limited sys- (Buss 1987; Michod 1999). So the most evolvable tems of replication (where the number of variants inheritance systems are those that suppress out- is smaller than the number of members of the laws. Outlaws, in turn, are replicators that go it evolving population) to unlimited systems where alone. Hence the ban on evolutionary escape (C4) the number of variants is as large or larger than and the ban on biased transmission—one in, all the population size. The argument links back in (C3). C1, the insistence on vertical transmis- to the importance of cumulative selection in Niche Construction and the Extended Replicator 341

adaptive change: new structures are built slowly, tinued evolutionary plasticity in a lineage and in small steps. Microeveolution requires a rich developmental modularity. For, as Wimsatt has supply of variation, and that requires a rich array shown, unless development is modular, phe- of replicator sets (Maynard Smith and Szathmary notypes will become generatively entrenched 1995; Maynard Smith and Szathmary 1999).1 (Wimsatt and Schank 1988). It is hard to change But that is not all it requires. We also need to developmental sequences if the development of consider the conditions under which replicator any characteristic is linked to the development of packages are mapped onto selectable variation. many characteristics. For a change is likely to I have been persuaded by Dawkins’s argument ramify, having many effects on the developed (1982, in the final chapter) for the evolutionary phenotype, and some of these are nearly certain importance of the life cycle. He argues that the to be deleterious. Thus, to the extent that devel- cycle from single-cell to organism to singlle-cell opment is holistic, the more complex the organ- is central to evolution. For an alteration to the ism, and the more it has been elaborated over replicator package that acts early, at the single- evolutionary time, the less significant further celled stage, can generate a genuinely novel struc- change there can be in that lineage. ture; it can have an overall effect on the organism. The point that adaptive change would be im- Dawkins suggests that without a developmental possible if development were holistic has been bottleneck mutations can have only a local effect. made before. Lewontin, for example, has pointed I think the importance of the single cell stage can out that such change requires traits to be “quasi- be overstated. Consider, for example, the cater- independent”; there are at least some develop- pillar/butterfly transition, or any other point in mental trajectories that allow one to be changed organism life cycles where major reorganization without affecting others (Lewontin 1978: 169). takes place. Even so, I think Dawkins is essen- Modularity adds something to this picture: mod- tially right. Hence the condition on simultaneous ules are (often with modification) reusable. Thus, and hence early transfer, rather than drip-feeding Muller and Wagner claim: replicators one by one. Because the whole replicator set is transferred early, alterations in the set The more we learn about molecular mechanisms of development in widely different organisms, the higher the have the potential to act early and have a global 2 number of conserved mechanisms that become known. effect. Some of them do indicate homology of morphologi- C8, the requirement for a smooth replicator/ cally divergent characters. . . . Still others illustrate that organization map, flows from the work of Kauff- highly conserved molecular mechanisms may be used in man (1993, 1995). Suppose a population of bald radically different development contexts, indicating pigs are trapped on an island in a period of global that the machinery of development consists of modu- cooling. Can selection save the population from lar units that become recombined during evolution. extinction by engineering furry descendants? (Mueller and Wagner 1996: 11) Only if slightly furrier pigs have a replicator set An important example is the invention of cell close in character to that of the bald pigs; only if types. For the most part, animals do not differ marginally still furrier pigs have a replicator set from one another because they are built from close to that of the slightly furry ones, and so on. different sorts of cells. Rather, they build differ- Hence the requirement for smooth mapping. ent structures from similar cell toolkits. Evolu- I am also impressed by the arguments of tion only had to discover the trick of making Muller, Wagner, Raff, and Dawkins on the sig- a certain type of cell once. This seems to be nificance of developmental modularity.3 First, a quite general phenomenon. Other examples they see an important connection between con- are the mechanisms for initiating eye and limb 342 Kim Sterelny

formation (Gilbert et al. 1996: 366). So evol- full Hoyle replication systems would be of special utionary/developmental complexes once dis- significance to evolution. Their existence and covered can then be co-opted and used for other character would be a central part of the explana- purposes. But that, of course, depends on devel- tion of the biological character of our world. opmental modularity. It is time now to forgo the pleasures of a priori Moreover, as Dawkins (1996) points out, there biology and return to actual mechanisms of are links between modularity, redundancy, and inheritance. I shall argue that: (1) no actual in- evolvability. If the replicator/organization map is heritance system is a full Hoyle; (2) genetic inher- modular, there is a fair chance that introducing itance has a high Hoyle score; (3) some of the redundancy into the replicator set through dupli- cross-generational channels lumped into ecologi- cation will result in the duplicated, hence redun- cal inheritance are close to Hoyle standard, but dant, structure in the organism’s organization. many are not; and (4) meeting the Hoyle stan- Most of the time, of course, such extra structure dard does not depend on taking the replicator set will be deleterious and selection will dispose of to carry information about the biological organi- the resultant replicator set. But not quite always, zation they will help build. and then the new structures can form the basis of genuine evolutionary novelty. Gould and Dawkins do not agree on much, but they do Gene-like Replication Systems agree on the importance of redundancy in scaffolding evolutionary change. And I am happy to The Hoyle conditions fit no actual inheritance join that consensus. So there are many reasons mechanisms perfectly. This is no surprise, inas- for thinking that the ideal inheritance system much as satisfying some conditions militates should satisfy C9. against satisfying others. It is, for example, well I suggest that inheritance mediated by a repli- known that fidelity trades off against variation. cation system satisfying C1–C9 would be evolu- Gri≤ths and Gray (1994) argue that the delivery tionarily potent; it would be highly evolvable. It of some developmental resources must be pre- would meet the Hoyle conditions; it would be (as cisely timed, and if they are right this militates I shall say) a full Hoyle. Such a replicator set against simultaneous transmission. But genetic would not, of course, be su≤cient to generate replication does fit the conditions quite well. a disparate and adaptively complex biota. The Gene transmission is vertical. It is simultaneous world must be cooperative too. Adaptive regimes and early, hence giving variant replicators a good must remain stable for evolutionarily significant chance to act early in an organism’s life cycle. It periods of time, for complex structure evolves is not quite outlaw-proof, of course. Meiotic only through cumulative selection. Moreover, drive and sex ratio-distorting genes are outlaws, “adaptive landscapes” must be continuous: small and hence transmission is not quite unbiased. changes in phenotype should make only a small The adaptation of genes to their role in replicator difference to fitness. Furry pigs will never evolve packages may not be quite an evolutionary sink if the fitness of somewhat furry pigs is not in- either. That depends on where viruses come termediate between bald and fully furry ones from. But genetic replication largely solves the (Lewontin 1978). It might also be the case that problem of preventing individual replicators some major evolutionary transitions—perhaps from going it alone. the invention of the eukaryote cell—depends on a Fidelity and richness are also uncontroversial. very lucky accident rather than being the pre- It is typically supposed that the fidelity of genetic dictable result of an appropriate replicator set replication is high, even in prokaryotic lineages and selection regime. Even so, the existence of that lack the error-correcting machinery of eukaryotes. But that depends on measuring fidelity Niche Construction and the Extended Replicator 343

by comparing the base sequence of a copy with of coarse-grained modularity in development re- the base sequence of its parent gene. Measuring mains an open empirical question. There are sim- fidelity that way, it is high. But if we take the ilarly open questions about robustness. There replicator to include not just the base sequence seems to be plenty of redundancy built into ge- but also the molecular machinery and relations to netic replication, at least with eukaryotes with other sequences through which it exerts its phe- paired genes. But the degree to which develop- notypic effects, then base sequence fidelity is not ment is canalized remain to be settled. fidelity simpliciter. Successful replication would In sum, it is clear that although genetic replica- involve the replication of this unit of phenotypic tion is not the full Hoyle, that system goes close action, not just the base sequence. Moreover, for to meeting the Hoyle conditions. Inheritance sexual organisms our measure of fidelity will de- based on gene-like replication yields a highly pend on the size of individual replicators. Long evolvable biota. The same cannot be said for all sequences are likely to be altered by recombi- that is packed into ecological inheritance. nation. So there are unanswered questions about It is somewhat ironic that when nongenetic the fidelity of genetic replication. Even so, on rea- inheritance is recognized at all, it is most com- sonable ways of settling these questions, genetic monly recognized in the guise of dual inheritance replication will probably count as a high-fidelity models in which cultural transmission supple- system. ments genetic inheritance. For cultural replica- There is a similar ambiguity with the respect to tion seems to me to have a low Hoyle score. It the richness of genetic replication. For though turns out that the conditions that allow cultural there is a truly vast number of different base replication restrict its richness and effect quite sequences—as Dennett puts it, the Library of sharply. Limited cultural inheritance is possible, Mendel is huge (Dennett 1995)—it is far from ob- but the conditions allowing the range of variation vious that that is the right way of counting repli- to increase also reduce fidelity. Perhaps imprint- cators. Presumably, we should count in terms of ing by phytophagous insects on their plant tar- phenotypic effect: Two sequences are different if gets, or the philopatry shown by sea birds to their they generate predictable differences in biological nesting sites, are valid instances of nongenetic organization.4 Even so, no one denies that genetic replication, but transmission of this type of be- replication is rich. In context, and in concert with havioral similarity from parent to offspring will much else, it generates a splendid range of bio- not generate cumulative change, for the system is logical organizations and could generate many not rich enough. Tomasello has pointed out that more. cumulative cultural evolution demands something Modularity and robustness are much more like the capacity for true imitation. He argues controversial. However, genetic replication is at that primate societies do not show cultural evolu- least minimally modular. For the DNA/RNA/ tion because apes do not ape. They learn from protein transcription and translation system is others by social priming—their attention, and ex- an instance of a modular replicator/organization ploratory trial-and-error learning, is directed in mapping, though it exemplifies modularity at part by what they see others do. So they can learn a fine-grained and local scale. The extent to what to be interested in: what constitutes a re- which we shall see a vindication of Wagner’s source or a danger. But they do not copy the hypothesis that there exists an array of gene- action patterns of others. Hence insight—the dis- development–adaptation modular complexes, covery of a new way to extract nuts, fish for ter- and that the existence of this array explains the mites, or reach inaccessible food—will not be continuing evolutionary plasticity of lineages, transmitted to other members of the group. In remains unknown. As far as I know, the extent effect, without imitation, copy fidelity is too low 344 Kim Sterelny

to sustain a ratchet effect, and hence for cultural cally. They are not replicated and transmitted evolution to generate cumulative change (Tom- en masse (like a computer being loaded with its asello forthcoming). There has been a recent doc- system disk) but are drip-fed. There is every op- umentation of the existence of a quite large set portunity for transmission biases. If this is a repli- of behavioral traditions distinctive of particular cation system at all, it is one made for banditry chimpanzee communities (Whiten et al. 1999), on a massive scale. So if ideas and the like are not but there is no evidence in this data either of cu- typically viruses of the mind (as Dawkins takes mulative cultural evolution or even of traditions religion to be) then that fact poses a serious prob- that have persisted over many generations. lem for theories of cultural evolution. Memes Yet once organisms have the cognitive sophis- should often be outlaws. If they are not usually tication to imitate, the robustness and modularity bad for their carrier’s fitness, this suggests to me of cultural transmission will be eroded. For once that cultural transmission should not be seen as cognitive sophistication su≤cient for imitation an inheritance mechanism at all. has evolved, the connection between experience So much for dual inheritance. The action of or- and behavior will be highly context-sensitive. ganisms in physically engineering their environ- Patterns of behavioral similarity will not be ment and that of their descendants seems not to transmitted deeply through the generations. The fit the Hoyle criteria well. John Maynard Smith causal influence of a replicator on biological has sharply distinguished between limited and organization is always context-dependent. It is unlimited systems of heredity, and at most the always scaffolded by features both internal and transmission of physically engineered surround- external. Hence if the cross-generational stability ings seems to be a limited system. But I think the condition is to be met, these contextual factors problem is more pressing. Transmission is not must be persistent, or reliably remade anew in vertical. Indeed, it is not even individual. It is each generation. These conditions may be met by diffuse. Groups of trees engineer their soil struc- a few elements in hominid cultures. Learning tures or a fire-prone understory; individual trees may be scaffolded, internally, by sensitive periods do not make their microenvironments for them- or externally by reliably reoccurrent and highly selves and their descendants. In most cases salient features of the environment. Language groups of animals make warrens, trackways, probably meets both these conditions. But many track-and-bowl systems, beaver lodges, termite of the elements of human and hominid cultures mounds, and other structures that are ultimately probably do not meet the criteria that underwrite taken over by the next generation. If this is trans- cross-generation stability. After all, the crux of mission at all, it is diffuse and development is the argument against behaviorism in psychology holistic. Contrast the case of the continuously oc- is that the effect of experience on behavior is cupied beaver lodge with a case where there really highly variable. So if an organism’s psychology does seem to be inheritance, transmission, and a is not captured by behaviorist models, cross- life cycle: cases in which mated queens leave ant generation similarities of behavior will not be and termite nests to found a new nest and build a well explained by dual inheritance models. Yet if new structure. In such cases, there obviously is a its psychology is captured by behaviorist models, new generation—an F2 nest—and a life cycle. its learning capacities are unlikely to be sophisti- But where the inherited system of resources is not cated enough to support imitative learning. tied to anything like a life cycle, there seems to be Set aside these problems. There is nothing in no moment of transmission at all. The next gen- the system of cultural transmission that filters eration just gradually comes to occupy, use, and outlaws. Ideas and other cultural constructs renovate the lodge as the previous generation dies spread horizontally and obliquely, not just verti- out. So especially when generations overlap, we Niche Construction and the Extended Replicator 345

have here a resource modified by many organ- Highly evolvable systems are rich: they have isms, for themselves and their descendants, and the capacity to produce many variants with used by many. Beavers are major ecological engi- which evolution can work. But though the princi- neers, and their engineering certainly has down- ple is clear, applying it is not. Maynard Smith has stream effects, but these effects do not constitute argued that highly evolvable replication systems an inheritance system. We might treat some of must be digital, for only such systems maintain these diffuse cases as individual vertical transmis- fidelity levels high enough for cumulative selec- sion by taking the units in question to be groups. tion (Maynard Smith 1996). However, the digi- But the conditions that allow group selection are tal/analog distinction, and hence the fidelity of quite onerous. So this is not a general solution to replication, can be assessed only in the context of the problem. replicator/reader systems. Our symbol reading On the other hand, some extragenetic inher- system determines the fact that itance systems do fit the Hoyle picture well. 8 8 Perhaps cytoplasmic factors in the egg that deter- 7 7 7 8 mine the basic positional layout of the early em- contains only two symbol types. The same point bryo do. The transmission of such factors may applies to counting replicator variation. It is the constitute a very important system, since it is system that uses genetic resources that determine both ancient and widespread. But my favorite ex- whether two base sequences are of the same type amples concern the transmission of obligate sym- or not. Does the leafcutter-fungus replication sys- bionts. Because symbiosis is a common biological tem count as a high fidelity/low variety system phenomenon, this category is important. It is not because it will only generate one kind of system: a minor quirk of a few clades. And here the fit a symbiotic association between an ant and a with evolvability criteria seems to be good. The fungus? Or will all the differences in ant and in transmission of these symbionts is often early. It fungus count as different variants? Suppose, for is typically unbiased. There is a considerable va- example, that the amount of fungus the queen riety of evolutionary specificity in these biological takes on her founding flight has some relation to relationships. Mycorrhizal associations between the probability of the fungus failing to grow, or fungi and their associated plants are not very on fungus growth rate. If so, this would add the species specific. But symbiosis is often highly evo- possibility of heritable variation to the system. lutionarily stable and mutually obligatory. The That variation might be digital if, say, these dif- symbiotic microorganisms cannot survive alone, ferences were governed by threshold effects. Or and nor can their partner. It is a mechanism with they might be analog if, say, growth rate varied high fidelity. It is reliable, with often delicate smoothly with sample size. adaptations to insure successful transmission. A Obviously, then, settling richness is both em- specific species associates with a specific species, pirically and conceptually complex. Depending sometimes so much so that the species branching on how these issues shake out, symbiont trans- pattern of the one models that of the associated mission may count as a limited replication system clade. However, because each member of the or, more likely, a large family of limited systems. partner retains a good deal of metabolic and de- Even if it does, they may be extraordinarily im- velopmental integrity, I assume that development portant. There is no doubt that symbiosis is eco- is both modular and robust. Many, perhaps logically important, inasmuch as nitrogen-fixing most, changes in leafcutter morphology, phys- bacteria in legumes and mycorrhizal associations iology, and behavior would have no impact on with trees are fundamental to terrestrial plant the specific association with their codependent ecology, as are coral reefs to the ecology of fungus. shallow tropical seas. But symbiosis may be of 346 Kim Sterelny

great evolutionary significance in the generation program that would otherwise be invisible. of novelty. It is now widely accepted that the Among the questions we should ask are: eukaryotic cell is an evolutionarily frozen symbiosis (Dyer and Obar 1994). 1. Can we find anti-outlaw mechanisms? These In general, symbiotic inheritance is less outlaw might include: simultaneous transfer; host adap- proof than genetic inheritance. Although some tations for blocking all but vertical transmission; symbiotic organisms have no future without their host adaptations for limiting the number of indi- hosts, there are many cases where the association vidual organisms transferred; and host adapta- is less rigid than this (Thompson 1994). Even tions for evolutionary capture of symbionts by when association is wholly obligatory, the trans- taking over the provision of critical metabolic re- mission of, say, symbiotic bacteria often involves sources to the associate. a significant number of organisms, and that 2. Is there evidence about the range of varia- opens the door to within-organism competition tion? Does the host phenotype differ, if different that reduces host fitness. Strikingly, it turns out combinations/quantities of symbionts are trans- that there are instances of host adaptation to ferred? Do genetic differences in host or associate minimize the outlaw problem, by segregating change the nature of their association? some symbionts into a germline group destined 3. The discovery of high Hoyle score replication for passage to the next generation, and a somatic systems on our world raises a profound evolu- group that will play a role in host metabolism. tionary problem. How and why could evolution For example, fulgoroid planthoppers segregate assemble such an inheritance mechanism? Repli- one of their bacterial symbiont species into a cators are adapted for their role of insuring that large, differentiated and reproductively disabled offspring are like their parents. That is no sur- form, and a small, undifferentiated form stored in prise, for most departures from similarity will be a separate part of their body and destined for bad news. But a system that insures accurate their eggs (Frank 1996). The upshot, though, is replication across a generation is one thing, an that while in some host/symbiont associations evolvable replication system quite another. It is anti-outlaw provisions are in force, genetic repli- possible that a highly evolvable replication sys- cation seems better screened against outlaws. tem might evolve in response to some local evo- In sum, then, some host/symbiont systems lutionary demand. But my working assumption meet anti-outlaw conditions quite well, with ver- is that a high Hoyle score replication system tical and early transmission. They meet stability is likely to be the result of some lineage level se- conditions well, for replication fidelity is high, lection for evolvability. That is an idea we can and the effect of symbiont on host phenotype is test. Can we find evidence of increased evolvabil- robust. They probably meet some of the con- ity by comparing symbiont rich clades with sym- ditions on the generation of variation, though biont poor sister clades? We might measure by I think many crucial details of the natural his- species richness and/or morphological and eco- tory of these associations remain unknown. So logical diversity. my best guess is that they have a high Hoyle score, though not as high as the genetic replica- Second, meeting the Hoyle conditions does not tion system. depend on the flow of information across a gen- I would like to extract two further points from eration. Indeed, the point of preformationist in- these examples. First, I think there is an impor- heritance mechanisms of this kind is that you can tant heuristic moral to be drawn. Treating sym- dispense with information. Because you have a biont transmission as an inheritance system with sample of the fungus that can be grown, you do a high Hoyle score suggests an empirical research not need information on how to make the fungus. Niche Construction and the Extended Replicator 347

That, of course, leaves open the status of gene- Hoyle score. Third, I claim that downstream based inheritance. We might reasonably conjec- environmental engineering is not in general an ture that gene-based inheritance systems are inheritance system or cluster of inheritance information-based systems precisely because they systems. Fourth, I argued that some elements are not based on sample-to-product inheritance. lumped into environmental engineering are in- This is a thought that must be held for another heritance mechanisms with a high Hoyle score. time. However, I will suggest one way it might be And finally, I decoupled Hoyle scores from the developed. idea that replicator sets carry information speci- We rightly think that perception involves a fying phenotypes. For sample-based systems of flow of information from distal events to the inheritance are not information flow systems, but mind. For perceptual representation tracks the on certain empirical bets, genetic systems might world despite great flux in the proximal chan- be.5 nel between, say, perceived tiger and the tiger percept. Color constancy mechanisms, for example, seamlessly compensate for great variation in Notes illumination conditions. So there is a robust rela- 1. Thus, for example, they argue: “One could argue tionship between distal source and internal regis- that for replicators that are not modularly replicated, tration, and only between them. Along similar variants can arise only through ‘macromutations’. . . . lines, if there were a stable relationship between Hence what members of autocatalytic cycles as limited genetic structure and phenotypic structure de- hereditary replicators lack is the ability to undergo spite variation in developmental route, we would microevolution: hereditary is almost always exact” rightly see the genetic structure as carrying in- (Szathmary and Maynard Smith 1997: 559). formation about the phenotype. On that way of 2. Peter Godfrey-Smith has pointed out to me that reading the situation, the idea of genes as infor- early transfer does not entail the existence of a develop- mation carriers would depend on the existence of mental bottleneck. Vegetatively reproducing plants presumably meet C2 through asexual cell division. So if a certain type of canalization, not of genetic vari- a bottleneck were a necessary condition on a highly ation but of developmental variation. It would evolvable biota, even setting aside necessary features of depend on the right type of robustness (C6). For the abiotic environment, evolvability may not be fully now, though, I will leave the issue open. It is, captured through a specification of the replication sys- however, important to see that if the argument of tem. There remains, though, the possibility that it is this section is right, gene-based inheritance is of captured indirectly. Maynard Smith has argued that the special importance in explaining life’s disparity bottleneck is a consequence of the anti-outlaw condi- whether the idea of the informational genome tions; it is an adaptation to ensure common fate (May- can be vindicated or not. nard Smith 1988). 3. See Mueller and Wagner (1996), Wagner (1995), Wagner and Altenberg (1996), Raff (1996), and the dis- Recapitulation cussion of “kaleidoscope embryology” in Dawkins (1996). I discuss modularity in development further in Let me briefly summarize the state of play as it Sterelny (forthcoming), and the issue is also taken fur- now is, if the argument of this chapter is right. I ther in Brandon (1999). have (first) identified a cluster of characteristics 4. Thus treating genes as difference-makers, in line with of inheritance systems that support evolvability. Sterelny and Kitcher (1988) and Sterelny and Gri≤ths Inheritance systems with most of these character- (1999, chap. 4.3). They make a difference, of course, only relative to an appropriate genetic, cellular, and en- istics have a high Hoyle score. Second, I have ar- vironmental context. But even given this qualification, gued that genetic inheritance does have a high the identification of gene types is far from uncontrover- 348 Kim Sterelny

sial and straightforward; see Sterelny and Gri≤ths Jones, C., J. Lawton, and M. Shaclak. (1997). Positive (1999, chap. 4.1–4.3). and negative effects of organisms as physical ecosys- 5. Thanks to the participants in the ANU 1999 genetic tems engineers. Ecology 78: 1946–1957. information workshop, and especially to Russell Gray, Kauffman, S. A. (1993). The Origins of Order: Self- Paul Gri≤ths, Susan Oyama, and James Maclaurin for organization and Selection in Evolution. New York: Ox- their comments on an earlier version of this paper. ford University Press. Kauffman, S. A. (1995). At Home in the Universe. New York: Oxford University Press. References Keller, L., and K. G. Ross. (1993). Phenotypic plastic- Boyd, R., and P. Richerson. (1985). Culture and the ity and “cultural transmission” in the fire ant, Solenop- Evolutionary Process. Chicago: University of Chicago sis invicta. Behavioural Ecology and Sociobiology 33: Press. 121–129. Brandon, R. N. (1999). The units of selection revisited: Laland, K. N., Odling-Smee, F. J., and Feldman, M. The modules of selection. Biology and Philosophy 14: W. (2000). Niche construction, biological evolution and 167–180. cultural change. Behavioral and Brain Sciences 23: 131– 175. Buss, L. (1987). The Evolution of Individuality. Prince- ton, NJ: Princeton University Press. Lehrman, D. S. (1953). Critique of Konrad Lorenz’s theory of instinctive behaviour. Quarterly Review of Caro, T., and M. Hauser. (1992). Teaching in non- Biology 28(4): 337–363. human animals. Quarterly Review of Biology 67: 151– 174. Lewontin, R. (1974). The analysis of variance and the analysis of causes. American Journal of Human Genetics Dawkins, R. (1982). The Extended Phenotype. Oxford: 26: 400–411. Oxford University Press. Lewontin, R. C. (1978). Adaptation. Scientific Amer- Dawkins, R. (1989). The Selfish Gene. (1st ed. 1976). ican 239: 156–169. Oxford: Oxford University Press. Lewontin, R. C. (1982). Organism and environment. In Dawkins, R. (1996). Climbing Mount Improbable. New H. C. Plotkin (Ed.), Learning, Development and Cul- York: W. W. Norton. ture, pp. 151–170. New York: Wiley. Dennett, D. C. (1995). Darwin’s Dangerous Idea. New Lewontin, R. C. (1983). The organism as the subject York: Simon and Schuster. and object of evolution. Scientia 118: 65–82. Dyer, B. D., and R. A. Obar. (1994). Tracing the His- Lewontin, R. C. (1985). Adaptation. In R. Levins and tory of the Eukaryotic Cell: The Enigmatic Smile. New R. Lewontin, The Dialectical Biologist, pp. 65–84. York: Columbia University Press. Cambridge: Harvard University Press. Frank, S. A. (1996). Host control of symbiont trans- Lewontin, R. C. (1991). Biology as Ideology: The Doc- mission: The separation of symbionts into germ and trine of DNA. New York: HarperCollins. soma. American Naturalist 148: 1113–1124. Maynard Smith, J. (1988). Evolutionary progress and Gilbert, S. F., J. M. Opitz, and R. Raff. (1996). Resyn- the levels of selection. In M. Nitecki (Ed.), Evolutionary thesising evolutionary and developmental biology. Progress, pp. 219–230. Chicago: University of Chicago Developmental Biology 173: 357–372. Press. Gray, R. (1992). Death of the gene: Developmental Maynard Smith, J. (1996). Evolution—natural and systems strike back. In P. E. Gri≤ths (Ed.), Trees of artificial. In M. Boden (Ed.), The Philosophy of Artificial life: Essays in philosophy of biology, pp. 165–209. Life, pp. 173–178. Oxford: Oxford University Press. Dordrecht: Kluwer Academic. Maynard Smith, J., and E. Szathmary. (1995). The Gri≤ths, P. E., and R. Gray. (1994). Developmental Major Transitions in Evolution. New York: W. H. systems and evolutionary explanation. Journal of Phi- Freeman. losophy 91: 277–304. Maynard Smith, J., and E. Szathmary. (1999). The Gri≤ths, P. E., and R. D. Gray. (1997). Replicator II: Origins of Life: From the Birth of Life to the Origins of Judgement day. Biology and Philosophy 12: 471–492. Language. Oxford: Oxford University Press. Niche Construction and the Extended Replicator 349

Michod, R. E. (1999). Darwinian Dynamics: Evolution- Williams, G. C. (1966). Adaptation and Natural Selec- ary Transitions in Fitness and Individuality. Princeton, tion. Princeton, NJ: Princeton University Press. NJ: Princeton University Press. Wilson, D. S. (1997). Biological communities as func- Mueller, G. B., and G. P. Wagner. (1996). Homology, tionally organized units. Ecology 78: 2018–2024. hox genes and developmental integration. American Wimsatt, W. C., and J. C. Schank. (1988). Two con- Zoologist 36: 4–13. straints on the evolution of complex adaptations and Odling-Smee, F. J. (1994). Niche construction, evo- the means of their avoidance. In M. H. Nitecki (Ed.), lution and culture. In T. Ingold (Ed.), Companion Evolutionary Progress, pp. 231–275. Chicago: Univer- Encyclopedia of Anthropology, pp. 162–196. London: sity of Chicago Press. Routledge. Odling-Smee, F. J., K. N. Laland, and M. W. Feldman. (1996). Niche construction. American Naturalist 147: 641–648. Raff, R. (1996). The Shape of Life: Genes, Development and the Evolution of Animal Form. Chicago: University of Chicago Press. Sober, E., and D. S. Wilson. (1998). Unto Others: The Evolution of Altruism. Cambridge, MA: Harvard Uni- versity Press. Sterelny, K. (forthcoming). Development, evolution and adaptation. Philosophy of Science, (supplementary volume). Sterelny, K., and P. Gri≤ths. (1999). Sex and Death: An Introduction to Philosophy of Biology. Chicago: Univer- sity of Chicago Press. Sterelny, K., and P. Kitcher. (1988). The return of the gene. Journal of Philosophy 85: 339–360. Sterelny, K., K. Smith, and M. Dickison. (1996). The extended replicator. Biology and Philosophy 11: 377– 403. Szathmary, E., and J. Maynard Smith. (1997). From replicators to reproducers: The ﬁrst major transitions leading to life. Journal of Theoretical Biology 187: 555– 571. Thompson, J. N. (1994). The Coevolutionary Process. Chicago: University of Chicago Press. Tomasello, M. (forthcoming). Two hypotheses about primate cognition. In C. Heyes and L. Huber (Eds.), Evolution of Cognition. Cambridge, MA: MIT Press. Wagner, G. P. (1995). The biological role of homo- logues: A building block hypothesis. Neues Jahrbuch für Geologie und Palaontologie 195: 279–288. Wagner, G. P., and L. Altenberg. (1996). Complex adaptations and the evolution of evolvability. Evolution 50: 967–976. Whiten, A., J. Goodall, and W. C. McGrew. (1999). Culture in chimpanzees. Nature 399: 682–685.

Developmental Systems Theory and Ethics: Different Ways to Be 24 Normative with Regard to Science

Cor van der Weele

Biological knowledge and biological technology The concern that such theoretical fights are not have become immensely consequential. Genetics necessarily fruitful for ethical practice has led is already a central part of the development of Beauchamp and Childress (1979 and later edi- biotechnology, and, with the Human Genome tions) to design a medical ethics that could be Project approaching its conclusion, will become helpful in practice. They left the goal that ethics ever more powerful. Because genetics and its should be founded by (a single) ethical theory be- social implications such as prenatal and presymp- hind; instead, they focused on common morality, tomatic genetic diagnosis touch upon life and the great advantage of which they felt was that death, health and illness, development and iden- agreement can be found there that is forever lack- tity, it need not surprise that they generate great ing on the level of theory. From common moral- hopes and fears, and many normative questions. ity they abstracted common principles, which are DST is one approach to raise normative con- general and universal rules and which form the siderations. It is critical of an overemphasis on core of their approach, that has been extremely genetics in biology, with an important eye to the influential. Justification for these principles is not social consequences of such one-sidedness. But it a matter of giving them a theoretical foundation; does not involve ethical norms, against which to justification is found through an approach called test, for example, applications of biotechnology. reflective equilibrium (Rawls 1971), in which co- What does DST have to do with ethics? In herence is sought between concrete judgements order to answer that question, this chapter will and more abstract ethical notions, such as princi- distinguish three different ways to be normative ples. This looking for coherence is essentially an with regard to science. The three normative en- ongoing social affair that has no real endpoint. terprises to be considered are ethics, science criti- There is no absolute justification. cism, and STS (Science and Technology Studies). This approach certainly takes ethics away from abstract realms, bringing it in closer contact with the world. It remains a philosophical under- Ethics and Science taking, however, in that the approach has mainly been developed, discussed and justified Ethics belongs to philosophy. It is that part of by philosophers. philosophy that systematically reflects on moral- Beauchamp and Childress focus on medical ity, be it in a normative, a descriptive or a meta- ethics, and the common morality they refer to is sense, the latter involving analysis of moral thus domain-specific; it is the morality associated language and reasoning, the former two involv- with medical practice. For this domain, they pre- ing the question how we ought to live. Ethical sented four principles: autonomy, nonmalifi- theory is a philosophical framework within which cence, benificence, and justice. With the help of to reflect on moral actions, moral judgements, the principles many practical dilemmas in bio- or moral character. Classical notions on which medical ethics can be tackled, as Beauchamp ethics is built are obligations (Kant), utility (util- and Childress show in their book. Automatic so- itarianism), or virtues (Aristotle). Major efforts in lutions are not the result, because the different ethics are directed to rendering these respective principles can point in the direction of different approaches coherent and convincing, discussing solutions. But the approach does enable decision- their differences or finding more convincing alter- making on the basis of argument concerning natives, all directed toward finding conceptual the application and balancing of relevant ethical frameworks within which the question how we principles. In other words, practical ethics is a ought to live can be answered. 352 Cor van der Weele

matter of the ongoing application and elucida- prohibit research will only produce suspicion and tion of moral principles. will almost always backfire. How does such an ethics relate to science? The practical implications of scientific knowl- It is not controversial that, like medical prac- edge are a completely different matter. Many tice, scientific inquiry is subject to ethical evalua- widely held general moral principles, all the prin- tion; various principles are clearly relevant and ciples of biomedical ethics among them, are rele- applicable. In the case of scientific knowledge, vant when it comes to social implementation, and domain specific principles include general rules ethics is productive and fruitful here. For exam- specifying that science should extend the domain ple, the ethical studies surrounding the Human of knowledge, that the gathering of knowledge Genome Project are almost totally directed to a should be done in fair ways, and that (human) study of its implications, as is reflected in the subjects of scientific research should be dealt with name of the undertaking, ELSI, which stands for respectfully. Research on humans is subject to “Ethical, Legal and Social Implications.” Apart ethical evaluation in that fundamental human from questions about research ethics (e.g., con- rights have to be respected. Increasingly, the in- cerning privacy and safety of research subjects) terests of nonhuman animals are also subject to the great majority of questions call for studies ethical consideration. Thus, ethical principles are of the implications of the Genome Project for relevant to scientific research and lead to restric- public health activities, genetic testing, conceptions on doing research. tions of humanity, concepts of race and ethnicity, When it comes to the content of knowledge, it and so on. In all these areas, policies should be is more problematic what ethics could have to worked out in which society accommodates the say. Ethicists, by their training as moral philoso- challenges posed by genetic knowledge and tech- phers, are not in a position to judge the specific nology in an ethically responsible way. contents of science, but they often do reflect on Thus, in dealing with science, ethics is fruitful scientific knowledge in more general ways. For with regard to implications and with regard to re- example, in controversial areas such as IQ research ethics, while it has not been successful in search, the Human Genome Project, cloning, em- dealing with the content of scientific knowledge. bryo research, or xenotransplantation, it can and has been argued that there should be restrictions on knowledge gathering because of potential Science Criticism worrisome social consequences. Scientists tend to resist such proposals. There is a strong internal Precisely because it does not deal with the con- normativity in science that says that knowledge is tents of scientific knowledge, Ruth Hubbard does a worthwhile goal, irrespective of its content or of not have faith in ELSI. The program, she writes its consequences. Ethicists in their turn often feel in Exploding the Gene Myth (Hubbard and Wald that scientists claim an amount of freedom that is 1993: 159), will not affect the decisions made in not appropriate given the enormous social im- science. It will not question the current emphasis pact of their work. However, there is no generally on genes as determining our development, health, shared common morality from which to derive and behavior, and it will not get in the way of sci- widely recognized principles to override the in- ence. On the contrary, the human genome man- ternal normativity of science. In this situation, agers will decide which ethical and legal questions many ethicists prefer not to push. Peter Singer will be asked. (1996: 227–228), for example, writes that there Hubbard’s approach to science is normative, are strong pragmatic reasons not to put ethical but her domain is not ethics. She is one of those restrictions on the content of science: attempts to critical scientists who do address the contents of science. Developmental Systems Theory and Ethics 353

Science criticism, as Proctor (1991: 232) uses Cranor’s book (1994: 95–97) that though some the term in his book Value-Free Science? and as I of the experimental studies in the context of the shall use it here, rests on a rejection of the neu- Human Genome Project are inevitably bound trality of science, by making a connection be- to undermine the simplistic notions on which tween the implications of science and content. A they rely, the project certainly encourages the major theme in science criticism is that because trend that genes are now often seen not only as scientific knowledge heavily influences social at- causing disease, but as defining disease. tention, it should be responsible knowledge. Ex- Against the overemphasis on genetics and its actly what this means may of course vary, since tendency for causal reductionism, most critical criticism in science is many-voiced, but a per- biologists stress an interactive, systemic nature of sistent theme within biology is criticism on ge- causality. Genes are players in a system. This is netic determinism. It is the main subject of Ruth where DST fits in, in ways that are elaborated Hubbard’s normative approach. By focusing too elsewhere in this book. heavily on genes, scientists draw our attention The content of science thus matters because away from societal causes regarding health and of its context. Alan Garfinkel has argued that disease. There must be a balance, she writes because of this context, science cannot be value (Hubbard and Wald 1993: 61), between individ- free. I think his argumentation and analysis are ual health care and public health measures. An congenial to many critical scientists. In chapter 5 excessive preoccupation with individual concerns of his book Forms of Explanation, called “The and responsibilities is detrimental to health when Ethics of Explanation,” Garfinkel (1981) argues it encourages us to neglect the social conditions that science cannot give value-free explanations. that affect us all. It does not lead to necessary His point of departure is Weber’s view that sci- cleaning up of factories; instead, it leads to iden- ence is like a map that can tell us how to get tification of suspectibilities of individual workers somewhere but that cannot tell us where we to various dangerous substances. should go. This view is shared by Hempel and According to this line of reasoning, the present others who hold that a division of labor exists overemphasis on genes is represented in the Hu- between value-free causal accounts, given by man Genome Project. Therefore, the American scientists, and value judgments, made by policy- Council for Responsible Genetics (1990) has been makers and others. Garfinkel disagrees. If scien- warning for many years against the Human tists are concerned with statements of the form Genome Project, criticizing the reductionist view “A causes B”, he says, this implies by no means of causality that detracts from other biological that science is value-free. The reason is that processes and from social factors. The council causal statements are always made in a social has commented on the comparatively large context. It can be wrong to make true causal amounts of money spent on mapping and study- statements. For example, it is morally wrong to ing genes, saying that “the genome project vastly say, “If you look in the attic, you’ll find Anne exaggerates the importance of genes, especially at Frank” in the presence of Nazi search parties, this time, when a deteriorating environment and even if it is “merely” a causal statement. It is not economy make it increasingly di≤cult for most the presence of value-words but the context that people to live healthful lives” (1990: 4). makes the statement non-neutral (Garfinkel The theme of the increasing geneticizing of 1981: 137). health and disease is prominent in science criti- Simple as it is, says Garfinkel, the point is often cism in biology, such as in the volume Are Genes missed by scientists who want to forget about the Us? (1994), edited by Carl Cranor. For exam- context of their research, as in war industries. ple, Evelyn Fox Keller argues in her paper in When the context of relevance is not so clear, and 354 Cor van der Weele

when applications of research are not in sight, it with Rose and Kamin (Rose, Lewontin, and is more di≤cult to evaluate the situation; after all, Kamin 1984), as well as with Levins (Levins and “any fact may end up aiding some evil cause. So Lewontin 1985), he has argued that genetic de- what are we to do?” (p. 138). But even in seem- terminism flourishes as part of a bourgeois, re- ingly neutral situations, Garfinkel maintains, “A ductionist, Cartesian world view that does not causes B” is not value-free. Why not? Because ex- encourage questions about the interdependence planations are incomplete and this incomplete- of organisms and their environment. A dialectical ness is always potentially relevant. Ever since worldview is needed, of which a dialectical biol- Mill, it has been clear that explanations typically ogy is an element. mention only one or two causal factors as the Not all critical scientists present such analyses. cause of a phenomenon, giving all the other fac- Their normative context is often relatively un- tors the status of background. Explanations unspecified, nor do they analyze in detail exactly avoidably involve choices. The incompleteness how better knowledge will lead to a better soci- of explanations becomes morally relevant when ety. And this cannot be surprising, since social or science becomes a guide for social choices. Any political normativity is a motivating element but explanatory framework recognizes only certain not the primary domain of analysis for most crit- alternatives and therefore guides you to specific ical scientists. With an eye on relations between solutions. The choices involved need certainly science and society, science criticism focuses pri- not be made for moral or political reasons: “The marily on the content of science. As Susan Oyama value ladenness is a fact about the explanation, writes in one of her papers: “Some of my reasons not its proponents. It is value laden insofar as it for working on the nature-nurture problem stem insists . . . that change come from this sector from concerns about publicly contested issues rather than that” (Garfinkel 1981: 141). All kinds of, say, intelligence, race or sex, but most have to of nonmoral mechanisms exist to explain the do with the kinds of distinctions that are made in choices involved, such as tradition, available the scientific work that draws on and feeds these equipment, expertise, and so on. larger disputes” (Oyama 2000, chap. 11). This The same need for choices arises with regard to balance applies clearly not only to her reasons methodological criteria; Levins (1966, 1968) as but also to her critical efforts. Social concerns well as Van der Steen (1993a, 1993b) have argued are there, but they are mostly in the background. that methodological trade-offs cannot be avoided This background is also present when she says, because theories cannot satisfy all methodologi- for example, that the analysis of the concept of cal criteria at the same time. nature is important in part because the concept is The unavoidability and consequentiality of “ethically resonant” (Oyama 1999). This suggests choices is a core insistence in science criticism: that (some) concepts help shape social problems scientific choices matter in social contexts. In and solutions, but leaves it open exactly how this many cases this is clear; many historical exam- happens, and it does not involve explicit moral ples have been analyzed, by critical scientists and normativity. others, that show how socially relevant biolog- In order to be in a position to address the pre- ical knowledge can be. The history of eugenics cise and technical content of science, a high level and of IQ-testing are among those examples. of expertise is required. Thus, critical scientists But describing context involves again choices. are typically scientists themselves. Lewontin’s ex- Garfinkel specifies his social normativity in po- pertise in population genetics enabled him to litical terms, calling his view of reality Marxist. present a detailed critique of the neglect of gene/ Lewontin too has been very explicit, seeing sci- environment interaction in causal analyses in ence as part of wider social ideology. Writing biology, and to insist that the proper object of Developmental Systems Theory and Ethics 355

study is the relation of genotype, phenotype, and developmental biology as a whole. If it were not, environment, expressed in norms of reaction Rifkin would not have had to notice, as he does, (Levins and Lewontin 1985: 114). Norms of reac- that “still little research has been done, to date, tion describe biological outcomes of given geno- on how genetic predispositions interact with toxic types as functions of environmental variables. materials in the environment, the metabolizing of The argument is a piece of theoretical biology for different foods, and lifestyle to effect genetic mu- which Lewontin’s expertise as a population ge- tations and genotypic expressions” (Rifkin 1998: neticist and theoretical biologist is the relevant 229). But he is certainly right that developmental expertise. Likewise, Oyama’s criticism on deter- biology is a field in which systemic complexity is minist reasoning in The Ontogeny of Information increasingly drawing attention, with DST as an (Oyama 1985/2000) is based on relevant expertise important attention drawer. and gains its influence from being a clear and well In systemic approaches, the findings of the Hu- informed expert analysis, and the same applies to man Genome Project are not the Holy Grail of Ruth Hubbard’s argument in Exploding the Gene biology. They do not have much meaning in Myth (Hubbard and Wald 1993) that there is themselves, and they have to be interpreted in more to health and disease than meets the genetic a larger network of heterogeneous causal inter- eye. actions. Such a view implies that the Human That the content of science matters for society Genome Project is not, or at least should not be, is an assumption that is shared by many, and just an everlasting gene-centered affair by defini- much hope is often placed on a non-(genetically) tion. Much depends on what is done with the se- reductionist biology. For example, in The Biotech quence information. If the next step is an enor- Century, Jeremy Rifkin (1998) points out that a mous effort into functional genomics in which biology that talks in reductionist terms leads to gene/environment interaction are the center of at- very different kinds of biotechnological practices tention (through the study of norms of reaction than a biology that takes a more systemic, inte- and the molecular mechanisms underlying them), grative approach to nature. The former encour- the Human Genome Project could be a step to- ages a biotechnology that, for example, looks ward a systemic biology, though it could still be at genetically engineered plants not as parts of questioned whether it is a good or necessary step. larger environments but as self-contained devices Rifkin hopes that a new approach in biology in isolation. Rifkin sees especially promising de- will make a difference for biotechnological prac- velopments in the direction of a more integrative tices, and this is suggestive of the ways in which biology in developmental biology. In this field, he science matters. First, its reigning paradigms says, with reference to the work of Stuart New- matter, because they act as searchlights and be- man, a new middle ground (not a DST-style for- stowers of meaning. Second, its precise results mulation) between nature and nurture is being matter, for example on how development is influ- found in which understanding of the subtle rela- enced by toxins and other environmental causes. tionships between genotype and phenotype, and Study of reaction norms is a valuable instrument between environmental triggers and genetic ex- for the latter. The information generated by the pression, is growing. In developmental biology, study of reaction norms has its restrictions from the idea of the almighty master molecules is giv- a DST point of view, as Russell Gray (1992) ing way to “a more sophisticated understanding has pointed out: reaction norms do not look at of genes as integral components of more com- processes but only at outcomes of development; plex networks that make up both an organism therefore they are not sophisticated enough to and its environment” (Rifkin 1998: 156–157). This generate a picture of development that “includes view is no doubt on the overoptimistic side for a range of codeveloping life-history trajectories 356 Cor van der Weele

over a range of codeveloping environments” genics. Rifkin (1998) is one example among (Gray 1992: 174). Nevertheless, in my view (Van many. He refers to Beckwith when he says that “a der Weele 1999, chap. 5), norms of reaction do more balanced presentation of the relationship have an important place in the study of develop- between genetics and environment needs to be mental systems. They can be seen as a first step, made in the public arena, lest we risk the new sci- which can be followed by more detailed studies ence becoming the handmaiden for a eugenics- on life trajectories, or on the mechanisms of gene based politics” (Rifkin 1998: 158). But here, too, expression. (For another defense of norms of re- the point can only be that a more complex biol- action, see Kitcher forthcoming.) ogy “does not so readily lend itself to such uses.” What can a more systemic biology accomplish? It would be much too strong to say that a more Is it going to change the world for the better? complex biology would or should prevent eu- This assumption, apart from being too simplistic, genic policies, for several reasons. is not always shared by those who are expected to First, eugenic policies are feasible on the basis benefit from a richer biology. For example, many of good and rich biology, too—although presum- homosexuals have not shown themselves too un- ably with less oversimplification. The fact that happy with deterministic genetic accounts of ho- gene/environment interactions are important mosexuality, even if the accounts turned out to be does not imply that the outcome of every genetic very problematic from a scientific point of view, defect becomes completely unpredictable. Al- because such accounts make homosexuality seem though it is important to acknowledge that the ef- less fraught with complex and highly moralized fects of genes depends on all kinds of other genes choices. Instead, it would be a matter of sim- and environmental conditions, there are many ple, unavoidable nature. Susan Oyama, when ex- cases in which these effects will be fairly pre- pressing her anxiety about the biological turn dictable in a normal range of genetic and envi- (Oyama 2000, chap. 10) is at the same time sensi- ronmental conditions. Within DST it is stressed tive to the complexities of social processes, saying strongly that the contingency of causal influences that while the disempowered may be unable to in development does not imply unpredictability. prevent a swing back to biology anyway, the dis- On the contrary, in normal situations in which all cussion might encourage scrutiny of the whole the normal causal factors are there, outcomes issue, which could be a good thing. But given the of development are highly predictable. In other dichotomy between insides and outsides on words, given a normal environment, the influence which this discussion rests, her preference is, here of chromosomal abnormalities and genetic muta- as elsewhere, to question this dichotomous con- tions will to a certain degree be predictable in ceptual foundation, rather than “trying to turn many cases. Thus, an extra chromosome 21 leads such a dangerously loaded dichotomy to one’s to Down syndrome, even though its precise man- advantage.” In her view, the entanglement of bi- ifestations are quite unpredictable. Whether a ological arguments in beliefs about the fixity of person will eventually develop Huntington’s dis- nature has too often served to prejudge complex ease can also be predicted on the basis of genetics matters, while a reconstructed biology “does not alone, in the environments we are familiar with. so readily lend itself to these uses” (Oyama 2000, In short, predictions based on information about chap. 10). genes will not vanish completely on the basis of a Oyama’s careful discussion rightly suggests sound biology. that a more complex biology cannot accomplish Second, there is more to eugenics than biology. all kinds of social work by itself. Let me take eu- The old eugenics was formed by not only biolog- genics as a further example. A nonreductionist ical forces but by social and political forces as biology is often seen as a weapon against eu- well. While eugenics once took the form of co- Developmental Systems Theory and Ethics 357

ercive measures, later proponents of eugenics does not look at scientific disciplines from the emphasized free individual choice, arguing for point of view of the discipline itself, but from the a “democratic eugenics.” Whether new forms of point of view of social science: science is de- genetic testing, predicting, and prevention are to scribed as a social activity in a social context. For be called eugenic is a matter of definition. If the a long time, normative reflection was almost ab- term eugenics is reserved for the old ways, in sent from large parts of STS, as the discipline em- which race and coercion were central, they are phasized that empirical understanding was its not. If everything that is directed toward genetic goal. Normative reflection was associated with health in future generations is to be seen as eu- taking sides and was explicitly avoided. Many genics, they are. In his book The Lives to Come STS-ers, like many critical scientists, might be in Philip Kitcher talks about “laissez-faire” eugen- part motivated by normative social considera- ics with respect to genetic counseling and testing tions, but they were careful not to let this show up (Kitcher 1996: 196). Laissez-faire eugenics, in in their work. The difference is that for STS-ers, which everyone is to be her or his own eugenicist, this work is not, say, biology, but social studies of attempts to honor individual freedom and choice. biology. This is an attractive feature of the new eugenics. But the antinormative atmosphere of STS is On the other hand, individual choices are not lessening, partly through explicit reconsideration made within a vacuum, and disturbing or even of normative issues. Thus, Hans Radder (1992, disastrous scenarios are still feasible. Yet in this 1998a, 1998b) has argued that questionable nor- era of genetic knowledge, some form of eugenics mative consequences go unnoticed as the result of is inescapable, says Kitcher, and it is important to the absence of normative reflection. For example, find the safest options. I agree, and it seems clear actor-network theory only emphasizes the side that in order to find safe options it is necessary to of winners by being interested exclusively in how consider the science that is made use of as well as chains of power in networks are strengthened. the social and political conditions, the goals that The voice of the losers is not represented in actor- the technologies are used for, and how they affect network theory; actors vanish from sight, even people’s lives. cease to be actors, as soon as they are unsuccess- ful in building chains. A more symmetrical treatment of all the relevant actor perspectives would Appropriate Technology be normatively desirable, according to Radder (1992: 162). How exactly do science and society interrelate? It Radder himself has been practicing construc- is not surprising that the analyses of critical sci- tive normative reflection in his book In and About entists regarding this question are often general the World (Radder 1996). In this book, he pres- or tentative, given that their primary target of ents a strategy for the development of “appropri- criticism is science itself. A separate academic dis- ate technology.” Recommendations for an ap- cipline exists, science and technology studies, propriate agricultural biotechnology, he thinks, STS (e.g., Bijker, Hughes, and Pinch 1987; require that we do not stay within the boundaries Latour 1987), which is dedicated to detailed of common ethical approaches, since that would study of how science shapes society and is shaped yield recommendations that are too general and by it, with a special emphasis on relations of unpractical. Instead, we have to make a detour power and knowledge. This is the third domain through the complexities of modern technolo- of normativity with regard to science that I will gical systems and their realization in the living consider. world. He proposes to ask questions concerning Like science criticism, STS is concerned with the desirability of the realization of technological the content of science. Unlike science criticism, is 358 Cor van der Weele

systems, concerning the desirability of the choices the input of public norms and values with the resulting from this realization, and concerning help of which normative decisions about techno- the desirability of the conditions needed for the logical proposals are to be made. Such a division realization. In addition, we should ask whether of tasks reflects the division of work between sci- we know enough to answer these questions in ence generating knowledge, and ethics (derived a sensible way. In order to answer those ques- from common morality) generating norms. A tions, an integration of knowledge and expertise different possibility is that laypeople also have in the fields of biotechnology, technology assess- things to say with regard to knowledge, for ex- ment, science and technology studies, and ethics ample because they possess relevant knowledge is needed. In the process of implementation, ac- of their own. Brian Wynne (e.g., Wynne 1996), tive participation of the people involved is a pro- writing about risk assessment, argues for this cedural necessity. view. The role of knowledge with regard to risks Participation is a key element in the growing is to control them. It is often thought that while amount of normative proposals concerning the science aims for control over life circumstances social implementation of science and technology. and develops the knowledge needed for it, lay- Participatory methods are increasingly developed people resist such control and have no relevant and implemented, such as “Constructive Tech- knowledge either. Wynne attacks his view. Lay- nology Assessment,” to which Radder’s proposal people do aim for control, but since in daily life, is close in atmosphere. It is a field in which the unlike in scientific experiments, there is no possi- world of STS meets the world of policy studies bility to keep conditions under control, the kind and Technology Assessment (TA). Newer brands of knowledge that is useful for control in daily life of TA stress the importance of participation by is much more local and contextual than scientific all the social groups somehow involved with or knowledge typically is (Wynne 1996: 70). Per- affected by the development of new technologies. sonal agency and responsibility play a large role Because participation is an area full of pitfalls in it, and exactly these are in danger when people and practical di≤culties, much explorative and become dependent on scientific knowledge. The experimental work is being done in this area, in fundamental risk that people run in our “risk order to enable participatory strategies to be suc- society,” according to Wynne, is a risk to their cessful. Various instruments to encourage public identity, through dependency upon expert sys- participation are developed, such as models and tems. What is needed is a search for more legit- instruments for participatory debate and strategy imate, more contextual forms of knowledge in development. The underlying normativity is a public domains, in which laypeople are involved democratic one, and the direction in which these (Wynne 1996: 78). Wynne’s normative project, efforts point are aptly called the democratization clearly, is the protection of human agency and of technology development. control in a world full of expert systems. One important issue in finding fruitful methods With regard to genetics, there are senses in of participation concerns the fundamental ques- which the risks he mentions are far from far- tion which roles the various participants have, fetched. For example, information about genetic and can have, in such procedures. For example, defects typically has the character of risk figures with regard to scientific knowledge, what are lay- which are not only very hard to interpret but are people assumed to contribute? One possibility is also in themselves devoid of personal meaning. that they contribute values, with which they try to The possibilities for genetic analysis are growing judge technological projects. On this view, the rapidly. In the near future, DNA-chips (devices relevant knowledge involved in public debates that contain many pieces of single-stranded comes from science, while the public is there for DNA) will become available that enable rapid Developmental Systems Theory and Ethics 359

analysis of many parts of the genome at once. If can fruitfully deal with the problems of how to such analysis is going to inform you that you relate scientific and lay knowledges is a major have an increased (but not exactly known) risk to challenge for the implementation of appropriate get diseases X, Y, and Z, this information may in- technology. herently suggest that you should take preventive measures. Whether you want to have this information and to become a more or less full-time Integration? risk evader is a question that is intimately related to the question of what constitutes personal con- I have considered three different normative un- trol. The widespread assumption that more dertakings with regard to science. The aim has knowledge enables better choices may diminish not been to be complete as to possible ap- the social space for such questions and causes proaches, but, by showing distinct possibilities, people to become fearful and dependent in deci- placing in perspective DST’s position in the nor- sion making. As yet this is an underrecognized mative landscape. problem. Let me recapitulate and sketch some potential Participatory normative projects, especially if integrative developments. First, ethics and sci- critical questions about the sources of legitimate ence criticism. The approach taken by ethics pri- knowledge are involved, recall a theme of the pre- marily comes from philosophy. It is characterized vious section: good science does not automati- by close attention to a conceptually coherent nor- cally lead to a good society. Along Wynne’s lines mative framework. Ethics does not itself judge of thought, the authority of science is even a scientific knowledge; with regard to scientific threat to human agency and control. This threat knowledge, it is in an outsider’s position. Science comes from scientific expertise in general and is criticism, on the other hand, focuses on science. not confined to only some kinds of such exper- Its normativity is aimed at sound and responsible tise. Yet at the same time, it may be possible to knowledge. It concentrates on methodology, distinguish among more and less threatening metaphors, blind spots, missing facts, interpre- forms of science. Scientific knowledge that prom- tation of data, problem definition, statistical eval- ises to be uniquely authoritative in solving prac- uation, or other dimensions in which different tical questions is more threatening to human choices can make social differences. An impor- agency and control than scientific knowledge as it tant motivation is that knowledge is an important is encouraged by DST, that actively acknowl- factor in shaping society, and that sound knowledges the complexity and multiple layers of the edge is important to help shape a sound society. world. Such knowledge, again, “does not so eas- The two approaches thus appear to be very dis- ily lend itself to such uses.” The latter kind of similar, the one stressing ethical norms, while science will tend to be more hospitable to the im- taking scientific knowledge for granted, the other portance of local knowledge as well as the impor- stressing sound science. As far as critical scien- tance of human agency in the determination of tists talk about ethics it is either in a tentative normatively desirable social outcomes. An im- form, by saying, for example, that science is “eth- portant question for scientists as well as policy- ically resonant,” or by referring to general politi- makers is whether or not they acknowledge the cal analysis and worldviews. The direct relation possibility that in the implementation of science between those worldviews and biological knowl- and technology important knowledge does not edge is just as problematic as that between ethical only come from science but from other domains norms and biological knowledge. of human life as well. In other words, devel- For some purposes, the differences between opment of participatory social instruments that ethics and science criticism do not really seem 360 Cor van der Weele

to be a problem, because they go with a nice divi- scientific attention that is their object. Further, sion of tasks: science criticism deals with science, they do not deal with the problem that we cannot while ethics deals with its moral implications. attend to everything at once; their writings at However, through this division of labor each times suggest that we can. But what does make undertaking has only very limited legitimacy, their work interesting with regard to knowledge and a thorough analysis of the ethical resonance is that for them, the core of morality is not in of knowledge cannot be undertaken. It would moral theory, principles, or rules, but in percep- be interesting to see whether some rapproche- tion. Moving to a less personal sphere, it is not so ment between scientific and ethical normativity is big a step to substitute research, or research ques- possible. tions, for perception. It is interesting to consider A potential area of rapprochement of these the thought that for science, a fair and rich dis- two approaches may be framed by focusing on tribution of attention, however that could be the concept of attention. Science criticism cen- judged, could be an ideal that comes close to trally invokes the issue of selective attention, say- Hubbard’s implied ideal of a balance of atten- ing for example that genetic determinism draws tion. It is also in line with many of the emphases undue attention to genetic causes and solutions in DST. For example, in her chapter “The Con- for social problems. I have suggested elsewhere ceptualization of Nature,” Susan Oyama (2000) (Van der Weele 1999: 135) that being normative discusses choices that biologists face in their about strategies of directing scientific attention search for knowledge. Such choices include what could be called an ethics of attention. Interest- to pay attention to and what to problematize, ingly, an approach which by some has been called says Oyama. Referring to Donna Haraway, she an ethics of attention does exist within ethics. Its argues for a biology of multiple and embodied source is in the work of Iris Murdoch (1991, first perspectives. Though cast in different terms, this published in 1970), who refers to origins of her seems to me close in spirit to the ideal of a biol- view in the work of Simone Weil. The context is ogy in which fair and rich distribution of atten- not science, but the directing of personal attention is a central challenge. tion. Starting from the view that our energy is Likewise, but more briefly: There are also clear naturally directed in selfish ways, Murdoch sees points of contact between science criticism and the essence of moral excellence as outward atten- STS. Though the first undertaking focuses pri- tion. Clarity of vision is the normative ideal: marily on scientific normativity and the second unsentimental, detached, unselfish, objective on social analysis in the construction of science attention. The direction of such attention is out- and technology, both are concerned with con- ward, away from the self, toward the world, and tents of knowledge. The development of socially the ability so to direct attention she calls love responsible technology requires a thorough anal- (Murdoch 1991: 66). Martha Nussbaum (1990) ysis in the intermediate area of relations between takes a much similar approach. In her many dis- various kinds of knowledge, be it lay knowledge cussions of the novels of Henry James, it is rich- in relation to scientific knowledge or knowledge ness of perception that is the primary condition from various scientific disciplines. The aim to use of moral excellence. Following James, she writes more inclusive and integrative knowledge in the that our moral task is to be “people on whom construction of responsible technology could be nothing is lost,” who are “finely aware and richly fused with the aim to find more democratic ways responsible” (Nussbaum 1990: 148). of developing technology, as in participatory The work of Murdoch and Nussbaum is not forms of technology assessment. tailor-made for thinking about selective attention Further analysis and development of such inte- in science. In the first place, it is personal, not grative work is beyond this chapter. The main Developmental Systems Theory and Ethics 361

purpose here has been to distinguish and com- Levins, R. (1966). The strategy of model building in pare different normative approaches to science, population biology. American Scientist 54: 421–431. in order to shed light on relations between DST Levins, R. (1968). Evolution in Changing Environments. and ethics. After what has gone before, the con- Princeton, NJ: Princeton University Press. clusion can only be that while DST is clearly Levins, R., and R. Lewontin. (1985). The Dialectical Bi- ethically resonant, and though integration can be ologist. Cambridge, MA: Harvard University Press. imagined, DST and ethics in their present forms Murdoch, I. (1991). The Sovereignty of Good. London: are clearly distinguished as normative enter- Routledge. prises, or, if you prefer, are worlds apart. Nussbaum, M. C. (1990). Love’s Knowledge: Essays on Philosophy and Literature. New York: Oxford Univer- sity Press. References Oyama, S. (1985). The Ontogeny of Information: Devel- opmental Systems and Evolution. Cambridge: Cam- Beauchamp, T. L., and J. F. Childress. (1994). Princi- bridge University Press. (2d rev. ed., Durham, NC: ples of Biomedical Ethics. (4th ed.). New York: Oxford Duke University Press, 2000.) University Press. Oyama, S. (1999). The nurturing of natures. Unpub- Bijker, W. E., T. P. Hughes, and T. J. Pinch. (Eds.) lished manuscript. (1987). The Social Construction of Technological Sys- tems. Cambridge, MA: MIT Press. Oyama, S. (2000). Evolution’s Eye: A Systems View of the Biology-Culture Divide. Durham, NC: Duke Uni- Council for Responsible Genetics (1990). Position versity Press. Paper on Genetic Discrimination. Boston: Council for Responsible Genetics. Proctor, R. N. (1991). Value-Free Science? Purity and Power in Modern Knowledge. Cambridge, MA: Har- Cranor, C. F. (Ed.) (1994). Are Genes Us? The Social vard University Press. Consequences of the New Genetics. New Brunswick, NJ: Rutgers University Press. Radder, H. (1992). Normative reflections on constructivist approaches to science and technology. Social Garfinkel, A. (1981). Forms of Explanation. New Studies of Science 22: 141–173. Haven: Yale University Press. Radder, H. (1996). In and About the World: Philosophi- Hubbard, R., and E. Wald. (1993). Exploding the Gene cal Studies of Science and Technology. Albany: State Myth: How Genetic Information Is Produced and Ma- University of New York Press. nipulated by Scientists, Physicians, Employers, Insur- ance Companies, Educators and Law Enforcers. Boston: Radder, H. (1998a). The politics of STS. Social Studies Beacon Press. of Science 28: 325–333. Keller, E. F. (1994). Master molecules. In C. F. Cranor Radder, H. (1998b). Second thoughts on the politics of (Ed.), Are Genes Us? The Social Consequences of the STS. Social Studies of Science 28: 344–348. New Genetics, pp. 89–98. New Brunswick, NJ: Rutgers Rawls, J. (1971). A Theory of Justice. Cambridge, MA: University Press. Harvard University Press. Kitcher, P. (1996). The Lives to Come: The Genetic Rev- Rifkin, J. (1998). The Biotech Century: Harnessing the olution and Human Possibilities. New York: Simon and Gene and Remaking the World. London: Victor Schuster. Gollancz. Kitcher, P. (forthcoming). Battling the undead: How Rose, S., R. C. Lewontin, and L. J. Kamin. (1984). Not (and how not) to resist genetic determinism. In R. in Our Genes: Biology, Ideology and Human Nature. Singh, C. Krimbas, J. Beatty, and D. Paul (Eds.), Harmondsworth: Penguin. Thinking about Evolution: Historical, Philosophical, and Singer, P. (1996). Ethics and the limits of scientific free- Political Perspectives. Cambridge: Cambridge Univer- dom. The Monist 79: 218–229. sity Press. Van der Steen, W. J. (1993a). A Practical Philosophy for Latour, B. (1987). Science in Action. Milton Keynes: the Life Sciences. Albany: State University of New Open University Press. York Press. 362 Cor van der Weele

Van der Steen, W. J. (1993b). Towards disciplinary dis- integration in biology. Biology and Philosophy 8: 385– 397. Van der Weele, C. N. (1999). Images of Development: Environmental Causes in Ontogeny. Albany: State Uni- versity of New York Press. Wynne, B. (1996). May the sheep safely graze? A re- ﬂexive view of the expert-lay knowledge divide. In S. Lash, B. Szerszynski, and B. Wynne (Eds.), Risk, Envi- ronment and Modernity: Towards a New Ecology. London: SAGE Publications. 1 Index

Adaptation, 7, 20, 63–65, 117, 122, 199, 202, 207–209, Arthur, W., 225 211, 223, 333 Artiﬁcial life, 308 externalist, 333 Ashby, W., 168 lock-and-key model of, 203 Assemblies, repeated, 209–210, 220, 228 result of natural selection, 209, 214 Atlan, H., 302 Adaptationism, 155, 168, 241, 333 Aufderheide, K. J., 105 Adaptive landscapes, 65, 204, 227, 342 Autocatalysis, 224, 242–244, 246 Adaptive thinking, 204 Autopoietic process, 240 Adriaanse, M. S. C., 32 Avital, E., 100, 110 Agar, N., 245 Agency Babies, 153–155 concentrated, 326 Baboon, gelada, 152–153 distributed, 313–330 Baerends, G. P., 25, 32 Akeson, R., 96 Baerends-van Roon, J. M., 32 Alcock, J., 122 Balaban, M., 42 Algorithm, genetic, 308 Baldwin effect, 250 Alleles, 86–87, 131–132 Baldwin, J. M., 250 additivity of, 133 Balf, D. F., 110 dominance of, 133–134, 139–140 Barker, D. J. P., 81, 153–155 frequency of, 136 Barkow, J. H., 270 Alligators, 152 Barrington, A., 300 Altenberg, L., 203, 347 Bartlett, F. C., 269, 277 Altruism, reciprocal, 211, 340 Bateson, G., 258, 265 Amino acids, 292 Bateson, P. [P. G.], 4, 8, 9, 156, 157, 190, 191 Anastasi, A., 32 Beach, F. A., 17, 18, 25, 31 Anokhin, P. K., 43 Beards, green, 211 ANOVA, 156, 180–181, 190, 328 Beatty, J., 209 Anthropology, 255–276 Beauchamp, T. L., 351 /archaeology boundary, 265 Beavers, 119, 211–213, 333, 335, 344–345 /psychology boundary, 266 Beer, C. G., 20 Antipreformationism, 283, 290–294 Behavior Antireductionism, 76. See also Reductionism acquired, 15, 18 (see also learned) Antonovics, J., 206 adaptiveness of, 19, 20, 159 Ants, 8, 35–36, 152, 339 of ant colonies, 8, 142–148 /acacia symbiosis, 210 appetitive, 25, 33–34 colony behavior in, 142–148 categories of, 35–36 ﬁre, 198–199, 335 consummatory, 25 leafcutter, 334, 345 development of (see ontogeny of) role of queen, 143, 337 and environment, 29, 149 Apes, 337, 343 evolution of, 34–35 Aphids, endosymbionts of, 198, 201 functional approach to, 159–160 Apolipoprotein B, 72–73 genetically controlled, 31, 149, 169 Appel, S., 45–46 inherited (see innate) Apter, M. J., 303, 304, 307 innate, 15–21, 25–27, 30, 33–34, 109, 123 Archaeology learned, 15, 18–20, 29–31, 33–34, 109–110, 112, building vs. dwelling perspective in, 263 122–124, 143, 201, 249, 343–344 natural/artifactual distinction in, 264 learned/innate distinction in, 19, 26 Aristotle, 241, 351 ontogeny of, 16, 19–20, 27, 28–33, 42–43, 45, Armstrong, E. A., 25 149–151, 156, 163 364 Index

Behavior (cont.) Burns, J. A., 131–133 “programmed,” 142 Burrows, 120 stereotyped, 26–27 Buss, D. M., 179 unity of, 30, 32 Buss, L., 213, 340 Behavioral embryology, 42 Behavioral endocrinology, 15, 20 Callebaut, W., 4, 225, 230 Behaviorism. See Psychology, behaviorist Callerts, P., 203 Bender, B., 264 Callon, M., 330 Benzer, S., 169 Campbell, D. T., 225 Berg, P., 71–72, 76 Canalization, 62, 347 Berger, P., 75 Caporael, L. R., 179, 209, 220 Bernard, L. L., 16 Carlson, E. A., 69, 85 Bertalanffy, L. von, 51, 186–187 Carmichael, L., 17, 30, 31 Beurton, P. J., 82 Caro, T. M., 157, 334 Bidirectionality of structure-function relationship, Cascades, evolutionary, 336 43–44, 46 Cartesian duality, 260, 267–268, 354 Bijker, W. E., 357 Catalano, S. M., 44 Birch, H. G., 28 Cats, inﬂuences on the gene activity of, 49 Bird, C., 27 Causal parity/symmetry, 182–184, 188, 195, 286, 289, Black, J. E., 44 334 Blanchard, B. D., 167 Causality, 6, 82, 178, 180, 181–184, 190, 197, 290, 299 Blanden, R. V., 248 Cavalli-Sforza, L. L., 122 Bleske, A. L., 179 Cells, 103–106, 301–302, 341 Bloch, M., 267, 268 machinery of, 303 Blowﬂies, 198 Central dogma (of molecular biology), 46–47, 48, 50, Blueprint, genetic. See Genetic blueprint 248 Blum, A. S., 185 Chalmers, N. R., 157 Boakes, R., 16 Channels, 5, 337 Boas, F., 259–260 Chaos theory, 226 Bock, W. J., 118 de Chardarevian, S., 310 Body, organismic/extragenetic, 300–302 Chein, I., 32 Bogen, J., 251 Chen, Y., 89 Bolhuis, J. J., 159 Chick Bonner, J., 304–308 embryonic development in, 42 Bonner, J. T., 229, 243 imprinting in, 157–159 Bourdieu, P., 267, 277 pecking behavior in, 27, 28–30, 32, 33–34 Bowerbirds, 113 vocal recognition in, 44 Bowlby, J., 314, 316 Children, 273–274 Boyd, R., 201, 226, 337 Childress, J. F., 351 Brandon, R., 203, 206, 208, 210, 233, 249, 347 Childs, J. L., 167 Breast cancer, 88–89 Chimpanzees, 201, 344 Breed, F. S., 27 Chisholm, J. S., 169, 170–171 Brenner, S., 309, 310 Chromatin, 4, 106, 108–109, 195–197, 199, 200, 247, Brown, G. W., 314, 316 294 Brown, R. Z., 36 Chromosomes, 86, 106, 247, 294, 304, 306 Brown Jays, 231–232 Cilia, 106–107 Browman, L. G., 28 Clark, M. M., 109 Brunner, H. G., 81 Clausen, J., 62 Buchnera bacteria, 198, 201 Cleaner wrasse, 211 Burian, R. M., 69, 243 Clements, Frederic, 185 Index 365

Coevolution, 6, 125, 206 Delacour, J., 32 Coghill, G. E., 27, 42 Denenberg, V. H., 20 Constructionism, 57, 75–76 Dennett, D. C., 221, 244, 246, 250, 337, 343 Competitive exclusion, 157 Depew, D. J., 8, 85, 195, 286 Complementarity, 255–258, 264, 274 Depression, 314–317 Complexity, 157, 163, 186, 221, 225, 326 Desai, M., 154 Computers, 157, 303 Descartes, R., 57, 59 Conklin, E. G., 300 Design space, 226 Connerton, P., 258, 269, 277 Development, 20, 55, 103, 150, 229. See also Ontogeny Construction, 7 additive, 159 adaptation as, 1, 204–205 of ant colonies, 141–148 Context-dependence, 87, 131, 197 anti-instinct movement and, 16 Context-sensitivity, 200 constraints on (see Developmental constraints) Contingency, 9, 94, 139–140, 186, 226–229, 234 cooking metaphor of, 157 cycles of, 118 Darwinian approaches to, 159–162 developmental, 3, 356 environment in, 153, 172 Convergence, 65 epigenetic, 47 Cooper, R. M., 151 facultative, 207 Cooperation, 339 genes in, 50, 151 Cornish, E. C., 130 human, 153–155, 168, 170–172 Cosmides, L., 270 internalist models in, 188 Cradleboards. See Navajo child rearing jazz metaphor of, 157 Cranor, C. F., 353 as Markov process, 63 Crawley, M. J., 327 modularity of, 203, 215 Crews, D., 152 neural, 44, 46 Crick, F., 47–48 prenatal, 41, 43, 44 Crop breeding, 55–56 sensitive periods in, 159 Cruze, W. W., 27, 30 systems view of, 5, 18 (see also Developmental Csordas, T., 258 Systems Theory) Culture, 122, 124, 153, 169, 184, 196, 199 Developmental evolution of, 201, 229–232 age, 41 human, 256, 262, 336 bottleneck, 341, 347 influence of genes on, 259–260 constraints, 65, 219 transmission of, 270–272 genetics, 220 Cybernetics, 185 manifold concept, 45 Cystic fibrosis, 88–89 modularity, 341 Cytoplasm, 4, 202, 305–306, 310, 338, 345 “noise,” 62 physiological system, 48 Dams, 119 resources, 5, 89, 91, 195, 206, 207, 239, 242, 245, 292, D’Andrade, R. G., 267 335 Darwin, C., 15, 59–61, 66, 99, 120, 162, 186, 198, 204, Developmental systems, 100, 118, 125, 134, 140, 214, 261 208–209, 222–223 evolution of, 207 Darwinism, 199, 208, 214, 239, 240–242. See also Developmental Systems Approach (DSA) or perspec- Neo-Darwinism tive, 74–76, 129, 160. See also Developmental Dawkins, R., 70, 96, 99, 119, 122, 125, 157–158, 160– Systems Theory 162, 196, 199, 211–212, 239, 246, 288, 296, 335, 337, Developmental Systems Theory (DST), 1, 15, 76, 85, 339, 341–342, 344, 347 90, 149, 162, 177–189, 195–216, 239–251, 299–300, Davidson, I., 263–264 313, 334–338, 353, 355, 356, 359, 361 Deacon, T. W., 111, 250 critique of “coding” perspective on genes, 287–288 DeAngelis, D., 327 definition, 1–6 366 Index

Developmental Systems Theory (DST) (cont.) as steady-state device, 185–186 as empirical research program, 219, 284–286, 295 Edwards, P. J., 327 and generative entrenchment, 219–220, 224, 226, Egg, fertilized, 301, 304 229–230, 233 Egyházi, E., 46 as philosophy of nature, 283–286, 289–290, 295 Ehrlich, P. R., 118 Dewsbury, D. A., 17 Eibl-Eibesfeldt, I., 19 Dialectical biology, 178–182, 184, 189, 190, 283, 286, Embryology, 94, 222, 301 290, 354 eclipse of, 304 Dialects, 113 Emergence, 139, 189, 191, 244, 246 Diamond, J., 113 Encoding, 100, 102, 113 Dichotomies, 117, 163, 179, 188, 195, 206, 356. See Endosymbionts, 4, 286. See also Gut microorganisms, also Nature/nurture; Genes/environment; Symbiosis Genotype/phenotype; Replicator/vehicle Energy, 124, 185 Dickison, M., 91, 183, 224, 233, 283, 296 Entelechy, 31, 293 Diffusion gradients, 133–138 Entrenchment Disease, genetic, 129 differential, 221 DNA, 4, 45, 47, 48, 69–82, 86, 88, 90, 100, 106, 161, generative (see Generative entrenchment) 195, 212, 302, 304, 310, 338 Entropy, 221 causal role of, 247 Entwicklungsmechanik, 59 editing and repair of, 299 Environment, 17, 32, 63, 105, 162, 205 environmental interaction with, 72, 74 adaptation to, 63 information encoded in, 3, 102, 164, 261, 270, 292, constructed/modified by organism, 4, 64–65, 117, 295, 299, 302 (see also Genes) 119, 141–142, 147, 203, 215, 333–338, 344 “junk,” 288 ecological, 206, 208 mitochondrial, 199, 201, 202, 250–251 epistemic, 334 privileging of in ontogenesis, 177–178 evolution of, 57, 125 regulation of, 72 external, 206, 208 Dobzhansky, T., 62 influences on genes of, 47, 49, 50–51, 55–57 Doherty, P., 88, 93 organism and, 18, 31, 56, 59, 66, 118, 207–209 Dretske, F., 293 phenotype and, 61, 188–189 Drieschner, M., 82 as “problem” for organism, 65–66, 117, 122, 203– Drosophila, 2–3, 61–63, 86, 94, 139, 169 204, 208 influences on the gene activity of, 49 selective, 121, 206 mutations in, 56, 203, 225 social, 172 Ducklings universal physical, 206, 215 imprinting in, 41–42 (see also Development, prenatal) variation in, 136–138 vocal recognition in, 44 Environmental determinism, 179 Duden, B., 300 “Environmentalism,” 50 Dunbar, R. I. M., 153 Enzymes, 130–140 Dunlap, K., 16, 17 sensitivity coefficient of, 131–132 Dupré, J., 248 summation property of, 131–132 Durkheim, E., 266 temperature sensitivity of, 136–138 Dyer, B. D., 346 Epialleles, 247 Epigenesis, 4 E. coli, 102, 309 new, 241, 291 Earthworms, 120, 204 predetermined, 42–43, 46 Ecology, 185–187, 204, 294, 326–330 probabilistic, 42, 43–45, 46, 48, 50–51 Ecosystem, 124, 186–187 Epigenetic, 4 engineering of, 119, 124 development, 198, 299 as organism, 185, 187 systems, 85 (see also Inheritance systems, epigenetic) Index 367

Epp, C. D., 70, 82 Frank, S. A., 340, 346 Ethics, 9, 351–352, 359 Freeman, G., 130 medical, 351 Friedman, M., 284 Ethology, 17–19 Fruit flies. See Drosophila Eucalyptus, 195–197, 204–205, 207, 334 Eugenics, 356–357 Gage, F. H., 44 Eukaryotes, 210, 346 Galápagos finch, 122 Eveleth, P. B., 155 Galef, B. G., 109, 110, 123 “Evo-devo,” 220 Galton, F., 300 Evolution, 65, 94, 99, 103, 113–114, 124, 135, 207, 214, García-Barrios, L., 317 328, 337 García-Barrios, R., 185, 317 as change in gene frequency, 195, 215 García-Bellido, A., 307 as construction, 6 Garfinkel, A., 353–354 criteria for, 220–221, 223 Gautier, T., 168 cultural, 225–226, 230–232 Geertz, C., 274 developmental criteria for, 221 Gehring, W. J., 203 Darwinian, 149, 300 “Gene-centrism,” 80, 87. See also Genocentrism externalist models in, 188 “Gene for” locution, 3, 8, 87, 149, 183 human, 337 Gene pool, 215 population structured, 211 Generative entrenchment, 219–235, 341 quasi-independence in, 202 Generativity, 221 as series of solutions, 61 Generators, 221, 223 Evolutionary biology, 60 Genes, 47, 59, 62, 69, 79–80, 92, 99, 120, 125, 196, 211, Evolutionary psychology. See Psychology, evolutionary 224, 260, 287, 299, 334–335 Evolvability, 203, 233, 342, 347 correlation with phenotypes, 134–136 Exaptation, 226, 232, 249 definitions of, 7, 69–71, 80, 85 Exons, 992–93 as difference-makers, 347 Extracellular resources, 195 DNA-centered concept of, 70–71, 79 Eyes, 94 /environment dichotomy, 6, 43, 85, 91, 117, 149–150, 155–156, 163, 184, 206, 212, 300, 314 Falconer, D. S., 62 environmental influences on, 46–47, 49, 50, 56 Falk, R., 7, 69–70, 96 epistatic interactions among, 133, 286 Fausto-Sterling, A., 9 evolutionary concept of, 70, 76, 96, 239 Feedback, 121, 139, 146, 157, 162, 336, 338 expression of, 47–48, 286 Feldman, M. W., 7, 113, 333, 335–336 “fuzzy” concepts of, 70 Ferguson, J., 320 Gene-D (developmental) concept of, 88–93, 95–96 Fertilization, 305 Gene-P (preformationist) concept of, 87–92, 96 Fidelity, 339–340, 342 information coded in, 157, 195, 240, 283, 287–288, Fischer, E. P., 70 292, 300, 334 Fisher, J., 110 Mendelian concept of, 70, 86–87 Fisher, R. A., 300 molecular concept of, 69–70, 73, 76, 87, 90, 96, 239 Fitness, 57, 61, 198, 204, 209, 229 mutation in, 31 of trait groups, 212 privileging of in ontogenesis, 178, 180, 182, 188, Flavell, R. B., 247 240–241, 299, 335, 338 Fleischer, K., 308 process molecular (PMG) concept of, 76–82, 90–91 Fogel, A., 18, 330 regulation of activity of, 47, 103, 295 Fogle, T., 70, 82 selfish, 160–162, 211, 230, 233, 239, 241, 244–245, Foley, J. P., Jr, 32 288 Food preferences, 110 sex ratio-distorting, 342 Francis, R., 294 Genetic blueprint, 5, 21, 47, 90, 156, 302, 309 368 Index

Genetic determinism, 61, 85, 179–181, 184, 288, 354, Griffiths, P. E., 2, 5, 7, 8, 9, 69–70, 76, 77, 82, 85, 96, 356 100, 114, 117, 160, 164, 178, 183, 190, 219, 225, 228, Genetic drift, 228 239–242, 244, 245, 249, 283, 293, 295, 296, 299, 309, Genetic essentialism, 181 310, 335, 337, 342, 347–348 Genetic imprinting, 82 Grimes, G. W., 105 Genetic research programs, 81 Grohmann, J., 26, 34 Genetics, 300, 304 Gubernick, D. J., 167 developmental, 56 Guerrero, L., 168 evolutionary, 57 Gunther, M., 168 Mendelian, 117 Gupta, A. P., 62 molecular, 59 Gut microorganisms, 109, 195, 198–199, 213, 345 population, 62, 65, 122, 161, 219, 229 Guthrie, E. R., 32 Genocentrism, 240–241, 246. See also Gene-centrism Gutmann, M., 82 Genome, 79, 106, 304 Genotype, 8, 56–57, 61–63, 76, 78, 179–181 Hacking, I., 323 definition of, 129 Hagen, J. B., 185–189, 191 in genetic algorithm literature, 308 Haila, Y., 187, 327 myth of, 260–261 Haines, J. L., 129 /phenotype distinction, 85–86, 99, 333, 335 Halbwachs, M., 269 reading off from phenotype, 61 Halder, G. P., 203 Gerbils, 109 Hales, C. N., 154 Gerhart, J., 85, 90, 94–95 Hamburger, V., 42 Germline, segregation of, 213, 346 Hamilton, W. R., 160 Gibson, J. J., 189, 265, 267, 272 Hamsters, influences on the gene activity of, 49 Gilbert, W., 92–94, 195, 239, 251, 342 Hansell, M. H., 119 Gillespie, Charles, 59 Haraway, D. J., 360 Glynias, M., 92–93 Harris, T., 314, 316 Goal-seeking behavior, 168 Hartley, P. H. T., 25 Goats Harwood, J., 325 dominance in, 169–170 Haselton, M. G., 179 parental care in, 169–170 Hauser, M., 334 Godelier, M., 261 Hebb, D. O., 17, 18 Godfrey-Smith, P., 8–9, 90, 182, 190, 215, 347 Heinroth, O., 32 “Golden Mean” (between preformationism and Height, 150, 155 epigenesis), 292–295 effect of nutrition on, 155 Goldschmidt, W., 273, 276 Heisey, W. W., 62 Golinski, J., 329 Hendricks-Jansen, H., 189, 330 Goodwin, B. C., 195, 239 Heppel, L. A., 28 Gordon, D., 4, 8 Hepper, P. G., 110 Gos, E., 34 Heredity, 7, 16–17, 19, 31–32, 61, 99–100, 103, 215, Gottlieb, G., 4, 7, 8, 9, 18, 167, 168, 169, 188, 190, 251 223, 310. See also Inheritance Gould, S. J., 209, 222, 226, 291–295, 296, 342 limited, 110–111, 200–201, 340, 344 Grasshoppers, 152 unlimited, 102, 106, 108, 112, 200, 340, 344 Gray, A. J., 327 vs. environment (see Nature/nurture) Gray, R. D., 2, 7, 8, 9, 75, 85, 100, 114, 117, 161, 164, Heritability, 3, 8, 62, 150–151, 155–156, 180–181, 198, 178, 180, 183, 219, 228, 239–242, 244, 283, 286, 290, 227, 241, 244, 300, 306, 339 293, 295, 296, 299, 310, 328, 335, 337, 342, 355–356 ratio, 150 Greenough, W. T., 44 Hermit crabs, 208 Griesemer, J., 100, 114, 220, 224–225, 234, 310 Hess, E. H., 19, 20, 169 Index 369

Heterocatalysis, 224 in ecosystems, 185 Heuristics, 289, 322 epigenetic, 108–109 Heutink, P., 82 intentional, 197 Heyes, C. M., 110, 111 mathematical theory of, 5, 215, 293 Hierarchy, 221 nonmodular, 105 Hinde, R. A., 9, 19, 20, 110, 159, 277 ontogeny of, 293 Hinton, G. E., 250 teleosemantic theory of, 5 Holism, 78, 91, 105, 108, 110, 112, 185–187, 202, 229, transmission of, 105–106, 109–110, 196, 338 266, 283, 289, 341, 344 Ingold, T., 8, 149, 164, 257, 261, 263, 264, 266, 269, Holland, D., 269 274 Holt, E. B., 43, 44 Inheritance, 9, 15, 102, 105, 163, 177, 195–198, 214, Holton, T. A., 130 233, 286, 303, 338–339, 343 Homeobox, 94, 296 cultural, 201, 334, 337 Homeosis, 94 definition of, 196 Homology, 93–94 dual (genes and culture), 199, 337 Honey, R. C., 159 ecological, 118–122, 125, 333, 335–336, 338 Hormones, 168 epigenetic, 4, 200–201, 206, 286 Horn, G., 158 expanded/extended, 3–4, 6, 199–203, 214, 291, 334, Howells, T. R., 29, 31 336–337 (see also Inheritance, epigenetic) Hoyle conditions (for evolvability), 342–347 “extended replicator” model of, 196, 200 Hubbard, R., 352, 353, 360 of extracellular resources, 196 Hughes, E. C., 325 genetic, 336, 347 Hughes, T. P., 357 informational concept of, 337 Huh, D. L., 155 multiple channel model of, 196–197, 200, 206, 338 Hull, D. L., 99, 211, 223, 234 modes of, 101 Human Genome Project, 21, 351–353, 355 structural, 105–106 Humans vertical, 202 as evolved entities, 257, 263 Inheritance systems, 100, 112, 335, 345 influences on the gene activity of, 49 behavioral (BIS), 109–111 Hunt, E. L., 34 epigenetic (EIS), 103–109, 246–247 Hunter, T., 95 genetic (GIS), 100–103 Huston, M., 327 symbolic (SIS), 111–112 Hutchinson, G. E., 118, 167, 187 Innate releasing pattern, 26, 35 Huxley, T. H., 242 Innateness, 7, 19, 25, 27–29, 33–34, 36 Hydén, H., 46 levels of, 33 Instinct, 7, 15–20, 25–26, 31, 36. See also Behavior, Imitation, 110–112, 343. See also Behavior, learned innate Immunoglobulin, 93 “physiological,” 17 Imperato-McGinley, J. L., 168 Interaction, 164, 171, 177–179, 190 Imprinting, 41–42, 157–159, 160, 167–169 dynamic, 187 Individual, 209–210, 214 gene/environment, 151–155 origin of the, 213 Interactionism, 1, 3, 9, 178–191 Inducing-substance transfer, 109–110 bloodless/liberal, 179, 184 Inertia, evolutionary, 228 constructivist, 4, 178–179, 182, 184, 188–189 Information, 6, 19, 87, 102, 111, 112, 187 conventional, 2, 178, 188 adaptive, 21 democratic, 189, 191 in central dogma, 46, 48 Interactors, 99, 211, 337. See also Vehicles definitions of, 100 Interdependence, 182, 188 developmental, 5, 197 Internal/external distinction, 60, 62 370 Index

Internalism, 187–188 King, M., 89 IQ, 155–156, 181, 354 King, R. C., 129 Irminger-Finger, I., 89 Kingsland, S., 327 Isolation experiments, 18, 20, 28–29 Kinship, 257 Kinzler, K. W., 129 Jablonka, E., 7, 85, 100, 197–198, 220, 241, 246–248, Kirk, K., 89 286, 300 Kirkpatrick, M., 122 Jackson, M., 259 Kirschner, M., 85, 90, 94–95 Jacob, F., 303, 308–309, 311 Kitcher, P., 6, 7, 69–70, 91, 177–184, 189, 190, 283, Janzen, F. J., 152 285–286, 290, 296, 356, 347, 357 James, H., 360 Klingenberg, C. P., 134 James, W., 15–16 Klinghammer, E., 19 Janich, P., 75, 80 Klopfer, M. S., 169–170 Järvinen, O., 327 Klopfer, P. H., 4, 8, 9, 41, 167, 168, 169–170, 172, 191 Jennings, H. S., 32 Knight, R. D., 2, 191, 195, 240, 245 Jensen, D. D., 20 Knippers, R., 71, 73 Jervis, R., 186 Koehler, O., 25 Jilling, T., 89 Kondo, D. K., 313 Johannsen, W., 85–88, 99 Koppel, M., 302 Johnson, 158 Kramer, S., 19 Johnson-Laird, P. N., 270 Krieg, P., 88, 96 Johnston, C. A., 119 Kroeber, A. L., 262–263, 277 Johnston, T. D., 7, 9, 20, 179, 187, 190, 195 Kruger, A. C., 201 Jones, C. G., 119, 124, 333 Kuenen, D. J., 26 Jones, F. G. W., 32–33 Kuhn, H. G., 44 Jukebox, development as, 153, 164 Kuhn, T. S., 2, 285 Kulig, A. W., 143 Kacser, H., 131–133 Kuo, Z-Y., 7, 9, 16–18, 21, 27, 42, 43, 44, 50 Kagan, J., 168–169, 171 Källander, H., 113 Lakatos, I., 285 Kamin, L. J., 178, 185, 187, 316, 328, 354 Laland, K. N., 7, 113, 330, 333, 335–336 Kant, I., 351 Lamarckism, 59, 122, 247–249, 300 Karpiuk, P., 109 Lamb, E. J., 7, 85, 100, 103, 109, 113, 241, 246–248, Kauffman, S. A., 149, 157, 163, 221–222, 233, 243, 341 300 Keck, D. D., 62 Lande, R., 122 Keeler, C., 31 Landscape, adaptive, 204 Keightley, P. D., 131 Langton, C., 308 Keller, E. F., 3, 9, 69–70, 73–74, 85, 220, 240–241, Language, 111–112, 124, 153, 201, 202, 232, 250, 294–295, 300, 305, 309, 310, 353 270–272, 344 Keller, L., 335–336 Lashley, K. A., 17 Kelley, J. C., 119 Latour, B., 329–330, 357 Kelso, J. A. S., 18 Laudan, L., 285 Kempermann, G., 44 Laughlin, R. M., 28 Kerem, B., 89 Laws of nature, 228 Kerem, E., 89 Lawton, J. H., 119, 124 Kerang Farm Model, 320–326 Lawton, M. F., 232 Kermicle, J. L., 247 Learning. See Behavior, learned Kevles, D. J., 310 Lebra, T., 313 Kinder, E. F., 28 Lee, T. M., 152 King, A., 45 Lees, A. D., 35–36 Index 371

Lehrman, D. S., 7, 9, 15, 17–21, 42, 50, 85, 158, 188, Maynard Smith, J., 5, 99, 102, 121, 196, 199, 200–201, 198, 202, 338 210, 287, 338, 340–341, 344–345, 347 Lerner, R. M., 149 Mayr, E., 32, 239, 240, 248–249, 303, 310 Lerner, T. M., 62 McCabe, B. J., 158 Leung, W., 89 McCabe, T. T., 167 Levin, B., 71 McDougall, W., 15, 16, 17 Levins, R., 6, 178, 182, 219, 223, 354–355 Meadows, D., 186 Lewis, J. H., 309 Mehdiabadi, N., 146 Lewontin, R. C., 3, 6, 9, 75, 85, 113, 117, 119, 122, Meiosis, 102 142, 177–178, 180, 182, 185, 187–188, 202–205, Meiotic drive, 212, 288, 342 208–210, 219, 220, 227, 244, 249, 283, 285–286, 290, Mellars, P., 262 296, 299, 303, 310, 316, 328, 333, 335–336, 341–342, Membranes, 4 354–355 Memes, 199, 224, 230–231, 245, 337, 344 Lickliter, R., 51 Memory, human, 268–269 Life cycle, 196, 207, 209–210, 222, 261, 328, 333, Mendel, G., 59–61 335–336, 341 Mendel, Library of, 343 Life, origins of, 244 Mendelism, 85–87 Lindee, M. S., 21 Merleau-Ponty, M., 258–259, 277 Lindeman, R., 187 Metabolic pathways, 131–140 Lindley, R. A., 248 flux in, 131–133, 138 Lindquist, S. L., 105 Meta-engineering, 221 Lineage, 214 Metaphors, 3, 9, 55, 57, 59–61, 156–157, 164, 287, 291, Lomnicki, A., 327 303, 307, 313 Lorenz, K., 15, 17–18, 19–21, 25–26, 29, 30–36, 167 Methylation, 106, 108–109, 197–198, 247, 286 Lorenzen, P., 75 Mice Luckmann, T., 75 dwarfism in, 31–32 Luhmann, N., 75 influences on the gene activity of, 49 Lumsden, C. J., 172 parental care in, 167 Lundelius, J. W., 130 Michel, G. F., 250 Lysenko, T. D., 310 Michod, R. E., 340 Microevolution, 341 MacDowell, E. C., 31 Microtubules, 94–95, 195 Mackintosh, N. J., 155 Miki, Y., 89 Maeinschein, J., 15 Miklos, G., 90 Maier, N. R. F., 27, 34 Miklos, I., 200 Manning, A., 169 Miller, G., 226 Margulis, L., 210 Miller, N. E., 25 Markert, C. L., 310 Millikan, R., 5 Marler, P., 157 Mills, S., 209 Martin, P., 150, 156 Mind/body dichotomy, 266 Marx, K., 329 Mind, “evolved architecture” of, 270 “Master molecule,” 1 Minsky, M., 92 Mataric, M., 191 Mitchell, H. K., 138 Maternal assembly calls, 44–45 Mitchison, T., 95 Maternal effects, 303, 310 Mittelstrass, J., 75, 80 Matthews, A., 294 Models, 57, 159 Maturana, H. R., 75, 240 Modules, 85, 91–93, 96, 105, 111, 340, 342 Maturation (of behavior), 17–18, 21, 29, 30, 34, 36, 43, in development, 215 46, 179, 185, 187, 188 Monod, J., 303, 308–309 Mauss, M., 257, 265 Montgomery, H. E., 81 372 Index

Moore, C. L., 250 regeneration of, 113 Morgan, C. L., 15, 41 vacant, 204 Morgan, C. T., 28 Nijhout, H. F., 4, 8, 9, 85, 286 Morgan, T. H., 300 Nissen, H. W., 25 Morning glories, 130 Noble, W., 263–264 Morphogenesis, 56 Nowlan, S. J., 250 Morphogenetic ﬁelds, 94–96 Non-DNA entities, 72, 73, 80, 162, 196, 199–200, 241 Morton, J., 276 Non-equilibrium systems, 85 Moss, L., 7, 9, 85, 220, 287, 299, 310 Nonlinearity, 131, 133, 139–140, 155, 186 Moxon, E. R., 102 Normative approach to science, 351–361 Mrosovsky, N., 152 Normative reﬂection, 357–358 Mueller, G. B., 341, 347 Norms of reaction, 61–62, 179–182, 184, 190, 355–356 Murdoch, I., 360 Nucleotides, 106, 303 Mutation, 56, 78, 102–103, 138, 180, 199, 200, 208, Nursing 225, 340 in goats, 170 Mutualism, 334 in humans, 168 Myopia, 149–151 Nussbaum, M. C., 360

N-CAM. See Neural cell adhesion molecule Œ (developmental system symbol), 207 Naiman, R. J., 119 Obar, R. A., 346 Natural selection. See Selection Observation, 274–276 Nature/nurture, 1, 4, 6, 31, 43, 45, 123, 156, 163, 179, Obviation, 255–258, 273 183, 186, 190, 240, 299–300, 354 O’Dell, M., 247 Navajo child rearing, 169, 170–171 Odling-Smee, F. J., 7, 113, 204–208, 330, 333, 335–336 Necker cube, 161 Olby, R., 294 Nelkin, D., 21, 325 Ontogeny, 59, 129, 181 Nelson, P., 228 recapitulating phylogeny, 228 Nematodes as unfolding, 60–61 developmental complexities of, 220 Oppenheim, R., 42 influences on the gene activity of, 49 Organisms, 64, 187, 202, 205, 212 Neo-Darwinism, 9, 99, 114, 195, 204–205, 215, 219, as cause of their own development, 62–63 239, 300 /environment boundary, 337–338 Nesse, R. M., 151 subjects and objects of evolution, 59, 62 Nests, 120, 161, 337 Opitz, J. M., 93, 195, 239, 251 ant, 141–148, 212–213, 333, 336, 344 Orwell, G., 178 bee, 213 Ots, T., 259 variation and selection of, 161, 202 Outlaw replicators, 340 Networks, 307 Oyama, S., 2–4, 6, 8, 9, 20, 74–75, 85, 100, 114, 138, Neumann-Held, E. M., 7, 90–91, 96, 164, 220, 224, 149, 164, 177, 184, 187, 197, 205, 215, 239–242, 259, 309 261, 283, 289–293, 296, 299, 302, 310, 313, 354–356, Neural cell adhesion molecule (N-CAM), 88, 90, 360 92–93 Ozanne, S. E., 154 Neural nets, 158–159 Neutrality, contingent, 136, 139–140 Pal, C., 200 Newman, S. A., 310, 355 Paley, W., 209 Newton, Isaac, 60, 230 Pangloss, Doctor, 154 Niche, ecological, 6, 63, 204 Paramecium, 107 construction of, 4, 6, 113, 118–125, 141–142, 203– Parasites, 333–334 204, 206, 214, 333, 336–338 Parental care, 167–172 definition, 118 early fixation model of, 167–168 Index 373

gradual shaping model of, 168 Plasticity, evolutionary, 341 in humans, 167–168 Pleiotropy, 203 metamorphosis model of, 168 Plomin, R., 190 regulation by interaction model of, 168 Polygenic control, 61 “Parity thesis,” 3. See also Causal parity Polymorphism, 132, 207 Patrick, J. R., 28 Polypeptides, 69–80, 90–91. See also Protein synthesis Paukstis, G. L., 152 “under-determination” of by DNA, 72 Paul, D., 310 Populations, 6, 205 Paulsen, S. M., 134, 137 Porteous, J. W., 131 Paulson, G., 45–46 Portin, P., 69–70 Pearson, K., 300 Post, W., 327 Peet, R., 318 Poverty, effect on development of, 154–155 Peirce, C., 111 Power, C., 155 Pener, M. P., 152 Predators, 334 Penetrance, variable, 139 Preformationism, 4, 31, 36, 85, 87–91, 239–242, Penrose, L., 81 291–295 Perception, human, 267–268 Prigogine, I., 242 Pericak-Vance, M. A., 129 Prions, 105–106, 199, 241 Peterson, N. S., 138 Probabilistic-epigenetic framework. See Epigenesis, Peterson, R. E., 168 probabilistic Petrunkevitsch, A., 32 openness of, 78–80 Phenocopying, 2 problems with, 90–91 Phenotype, 8, 61–63, 76, 78, 88, 109, 122, 179, 201, Process Molecular Gene (PMG) concept, 76–82 (see 203, 207, 339 also Gene, process molecular concept of) deﬁnition, 130 Proctor, R. N., 353 environmental contingency of, 62 Program extended, 119, 125, 335, 337 computer, 304 genetic causes of/correlation with, 81, 86–87, 89, developmental, 296, 301, 304 130–140 genetic, 6, 55, 299, 301–310, 335 (see also DNA, insensitivity of, 139 Genes) niche-constructing, 335 Promoters, 79 ontogeny of, 129–140 Protein synthesis, 45, 292, 299 reconstruction of, 100 Provenza, F. D., 110 “thrifty,” 154–155 Provine, W., 239 Phenylketonuria (PKU), 78, 81 Prusiner, S., 241 Philosophy, 55 Pseudoneutrality. See Neutrality, contingent of biology, 70, 160, 239 Psychology, 15–16, 20, 25, 32, 258, 265, 267 of nature, 283–285 behaviorist, 16, 344 of science, 80, 248 developmental, 16, 273 Phylogeny, 60, 66 ecological, 267–268 Physics, 60 evolutionary, 92, 215, 248–249 Phosphorylation, 50, 95 functionalism in, 15 Piaget, J., 190 Pickering, A., 329 Quine, W. V. O., 284, 296 Pickett, S. T. A., 327 Quinn, N., 269 Pigliucci, M., 3 Pinch, T. J., 357 Radder, H., 357–358 Pinker, S., 153 Raff, R., 93, 195, 220, 228, 239, 251, 341, 347 Pinnipedia, 65–66 Rasmussen, N., 225, 228 Pinto-Correia, C., 4 Ratner, H. H., 201 374 Index

Rats Roughgarden J., 118 genes vs. environment in development of, 151 Ross, K. G., 335–336 influences on the gene activity of, 49 Ross, S., 20 maternal behavior in, 27–28, 32, 33–34 Roux, W., 191 Rawls, J., 351 Rowe, D. C., 190 Ray, W. S., 34 Rowell, C. H. F., 152 Reaction norms, 138 Rubin, D. C., 268–269 Read, A. P., 72, 82 Rubin, G., 90 Reconstruction, 1 Rushdie, S., 163 Red Queen effect, 63, 227 Reductionism, 5, 60, 185, 220, 232, 239, 248–249, 353, Salthe, S. N., 239 355 Sano, H., 247 Redundancy, 342 Santamaria, P., 138 Reed, E. S., 265 Sapp, J., 15, 310 Rehmann-Sutter, C., 81, 82 Sarkar, S., 7, 9, 82, 287 Reichman, O. J., 113 Schaffner, K. F., 2, 70, 191, 220, 233, 251 Releaser. See Innate releasing pattern Schank, J. C., 219, 221–222, 225, 341 Rendel, J. M., 62 Schlicht, E., 103, 111 Replication, 233, 343 Schlichting, C. D., 3 cultural, 343 Schluter, D., 327 fidelity of, 201 Schmalhausen, I. I., 61 modular, 102, 343 Schmidt, C. L. A., 28 Replicators, 99–100, 114, 162, 164, 196, 212, 224, 234, Schmidt, S. J., 75 240, 288, 339–340 Schneirla, T. C., 7, 9, 18, 19, 20, 27, 29, 33, 34, 35–36, extended, 196, 338 42, 43–44, 50 multiple, 200 (see also Replicators, extended) Schoolland, J. B., 29 /vehicle dichotomy, 99, 114, 212, 223, 335 Schrödinger, E., 172, 309, 311 Reproducers, 220, 224 Schubert, E., 89 Retroviruses, 47 Schwartz, T., 273 Reyes, A., 96 Science criticism, 353, 357, 359–361 Reznick, J. S., 169 Science and Technology Studies (STS), 351, 357–359, Rheinberger, H-J., 70 360 Richardson, R., 243 Sclove, R., 316 Richerson, P. J., 201, 226, 337 Scott, J. C., 317 Ricklefs, R., 327 Scriver, C. R., 78, 81 Riess, B. F., 25, 28 Scully, J. L., 81 Rifkin, J., 355 Seed caching, 113 RNA, 46, 47, 48, 50, 72, 74, 92, 310 Segmentation, 94 splicing and editing of, 72–73 Selection, 59–60, 62, 63, 65–66, 99, 102–103, 113, Robertson, A. F., 261 117–118, 120, 122, 124, 204, 214 Robertson, D. F., 122 artificial, 135–137, 203 Robertson, F. W., 62 cause of, 195 Robinson, S., 324 chemical, 244 Rodgers, B., 155 criteria for, 198, 220, 244 Romanes, G., 15 cumulative, 199–201, 203, 339, 342, 345 Romer, A. S., 32 of developmental systems, 198 Rose, S. [P. R.], 46, 178, 185, 187, 316, 328, 354 environmental, 249 Rosenberg, A., 209 frequency-dependent, 57, 65 Rosenblatt, J. S., 18 gene, 199, 212, 288 Index 375

group, 202, 210–212, 250, 251, 345 Steele, E. J., 248 kin, 211 Stent, G. S., 9, 293, 296, 310 levels of, 114, 210, 241 Sterelny, K., 5, 9, 69–70, 76, 82, 91, 160, 183, 190, sexual, 209 199–202, 204, 210, 212–213, 224, 233, 241, 244, 249, trait-group, 210–213 283, 291, 296 units of, 162, 203, 240, 246, 337 Sticklebacks, 29, 32 Self-assembly, 291 Stimuli, 158–159 Self-organization, 228 Stone, C. P., 17, 31 Sensitivity analysis, 132, 134 Stotz, K., 4 Sex determination (in reptiles), 152, 160–161 Strachan, T., 72, 82 Shachak, M., 119, 124 Strathern, M., 266 Shackleford, T. K., 179 Strauss, S., 327 Shannon, C. E., 215, 293 Stringer, C., 262 Shapiro, J. A., 103 Strohman, R. C., 47, 310 Shatz, C. J., 44 Structure-function relationship, 43–44, 46 Sheard, N. M., 28 Structures Shepard, J. F., 27 adaptive, 224–225 Sherry, D. F., 110, 123 dissipative, 242 Sickle-cell anemia, 124, 130–131 Sturtevant, M. H., 130 Siegel, B., 89 Succession, ecological, 64, 185–187, 293–294, 334 Silver, R., 17 metaphor for development, 293–294 Simon, H. A., 221 Super-organisms, 210, 212–213, 241 Simulations, computer, 135 Swimming, adaptations to, 66 Singer, M., 71–72, 76 Symbiosis, 198, 200, 202, 203, 210, 337, 340, 345–346 Singer, P., 352 Symbols, 111–112 Small, S., 96 Systems, 177, 184–189 Smith, A., 320 autocatalytic dissipative, 242, 245 Smith, B. H., 178 complex, 186, 233 Smith, C. A., 318 general/open, 187 Smith, C. C., 113 interactive construction in, 188–189 Smith, H. G., 113 unpredictable, 186 Smith, K. C., 80, 91, 183, 224, 233, 283, 296 Systems thinking, 17, 156. See also Developmental Smith, P. E., 31 Systems Theory Smith, S., 32 Szathmáry, E., 102, 105, 196, 197–198, 199, 200–201, Smotheran, W. P., 110 210, 338, 341 Snidman, N., 169 Sober, E., 190, 210–212, 340 Taming (in birds), 158 Sociobiology, 179, 184, 248, 250, 257 Tanner, J. M., 155 Software, 157 Tansley, A., 185, 187 Soil erosion, 317–318 Taxis, 25, 34 Song, bird, 157, 201 Taxonomy, 32–33 Songbirds, inﬂuences on the gene activity of, 49 Taylor, P. J., 9, 185–186 Spassky, B., 62 Teaching, 334 Speciation, 200 Technology, appropriate, 357 Spelt, D. K., 34 Technology Assessment (TA), 358 Sperber, D., 267 Teleology, 15–17, 103, 160, 162, 303 Sperry, R. W., 34 Thelen, E., 18 Stansﬁeld, W. D., 129 Thompson, D. W., 228 Steady-state systems, 104 Thompson, J. N., 346 376 Index

Thorpe, W. H., 20, 32–33 Vitalism, 16 Tilley, C., 264 Vogelstein, B., 129 Tinbergen, N., 15, 17, 19, 20, 25–26, 30–32, 35 Voltaire, 154 Tits, milk bottle opening by, 110–111, 123–124 von Baer, 222, 225 Tobach, E., 42 von Weizsaecker, C. F., 82 Todd, P., 226 Vrba, E., 226 Tomasello, M., 201, 343–344 Tooby, J., 270 Waddington, C. H., 2, 9, 62, 113, 247, 310 Tools, 337 Wagner, D., 144 Toren, C., 273–274 Wagner, G. P., 203, 341, 343, 347 Trait groups, 210–213 Wahlsten, D., 51 Traits, 203 Wakeﬁeld, J. C., 179 complex, 129–130, 138 Wald, E. 352, 353 Transcriptional complexes, 90 Waldrop, M. M., 308 Transitions, evolutionary, 200, 342 Walking, 257–258, 260–261 Transmission Walsh, F., 88, 93 of culture, 270–272, 343 Washburn, S. L., 32 diffuse, 202, 344 Waterlow, J. C., 154 stable, 339 Waters, C. K., 70, 82, 90 vertical, 339–340, 344 Waters, P. J., 78, 81 Trends, evolutionary, 203 Watson, J., 16 Trivers, R., 211 Watts, M., 318 Tucker, R. C., 329 Weaver, W., 215 Tuite, M. F., 105 Weber, B. H., 8, 85, 195, 286 Turner, M., 225 Webs, 120 Turtles, 152, 160–161 Webster, G., 239 Twins, monozygotic, 3 Wecker, S. C., 167 “Two worlds,” 255 Weil, S., 360 Weiner, J., 169, 264 Ulanowicz, R., 239 Weingarten, M., 80 Ursprung, H., 310 Weismannism, 234 Uyldert, I. E., 34 Weiss, P., 51 Welsch, P., 89 Vacuum activity, 35–36 West, M., 45 Value-free causal accounts, 353–354 Whales, 199, 201, 202, 250–251 van der Steen, W. J., 354 Wheeler, W. M., 35 van der Weele, C., 9, 47, 81, 180, 183, 354, 356, 360 White, P. S., 327 Van Valen, L., 63, 117, 227 Whitehead, H., 199, 202, 250–251 Vandenbergh, J., 44–45 Whiten, A., 201, 344 Varela, F. [J.], 75, 240 Whiting, J. M., 167 Variance, analysis of. See ANOVA Whitman, C. O., 32 Variation, 56, 59, 62, 100, 102–103, 114, 129, 131, 156, Whorf, B. L., 169 162, 181, 198, 203, 208, 339, 341 Wiener, N., 309, 311 environmental, 136–138 Wiesner, B. P., 28 patterned, 103, 106 Wikan, U., 258 Vayda, A. P., 185, 329 Wilkinson, R. G., 155 Vehicles, 99, 122, 162, 164, 212, 336. See also Williams, G. C., 151, 215, 227, 296, 340 Interactors Wills, C., 48 Vicarious selectors, 225–226 Wilson, D. S., 209–213, 251, 334, 340 Vine, D. O., 29 Wilson, E. O., 152, 172, 240 Index 377

Wimsatt, W. C., 5, 8, 179, 341 Wolbachia bacteria, 200, 202 Wolf, E., 259, 313 Wolf, Ulrich, 76–78 Wolff, C. F., 59 Wolpert, L., 304, 309, 311 Woodward, J., 251 Wright, B. E., 102 Wright, S., 51, 103, 223, 300 Wynne, B., 358–359

Yntema, C. L., 152 Yerushalmi, Y., 152

Zadnik, K., 150 Zamenhof, S., 49 Zentall, T. R., 110 Zorn, A., 88, 96 Zubek, J. P., 151 Zucker, I., 152