Above the Gene, Beyond Biology

Home , Christopher Zeeman, Epigenetics, Eva Jablonka, Frank Rattray Lillie, Operon, Richard Lewontin

ABOVE THE GENE, BEYOND BIOLOGY

TOWARD A PHILOSOPHY OF EPIGENETICS

JAN BAEDKE

University of Pittsburgh Press Published by the University of Pittsburgh Press, Pittsburgh, Pa., 15260 Copyright © 2018, University of Pittsburgh Press All rights reserved Manufactured in the United States of America Printed on acid-free paper 10 9 8 7 6 5 4 3 2 1

Cataloging-in-Publication data is available from the Library of Congress

ISBN 13: 978-0-8229-4521-5

Cover art: (foreground) Line drawing adaptation of the epigenetic landscape into an ontogenetic landscape for locomotion, from E. Thelen and L. B. Smith, “Dynamic Systems Theories,” in Handbook of Child Psychology, 5th ed., vol. 1, edited by W. Da- mon and R. M. Lerner (New York: Wiley, 1998); (background) original art courtesy of DNA Art Gallery LLC. Cover design by Joel W. Coggins For A, U, and J, my epigenotype

”Epigenetics.” The science concerned with the causal analysis of development. CONRAD HAL WADDINGTON, 1952

I suppose the process of acceptance will pass through the usual four stages: 1. This is worthless nonsense, 2. This is an interesting, but perverse, point of view, 3. This is true, but quite unimportant, 4. I always said so. JOHN BURDON SANDERSON HALDANE, 1963

CONTENTS

Acknowledgments xi

INTRODUCTION. WHAT IS EPIGENETICS? 1

1. HOW EPIGENETICS DEALS WITH BIOLOGICAL COMPLEXITY 13

2. CHALLENGES OF EPIGENETICS IN LIGHT OF THE EXTENDED EVOLUTIONARY SYNTHESIS 41

3. CAUSAL EXPLANATION 90

4. MECHANISTIC EXPLANATION 129

5. ASSESSING THE EXPLANATORY POWER OF EPIGENETICS 168

CONCLUSION. A PHILOSOPHY OF EPIGENETICS 201

Notes 217 References 245 Index 301

ACKNOWLEDGMENTS

. . . try with a little help from my friends THE BEATLES, 1967

I AM GRATEFUL TO a number of people who have helped in writing this book. Their intellectual, material, and emotional support is deeply appreciated. I benefited from exciting exchanges on causality, mechanisms, and scientific explanation, as well as on past and present epigenetics, with Mas- simo Pigliucci, Eva Jablonka, Siobhan F. Guerrero Mc Manus, Jani Raerinne, James DiFrisco, and Maurizio Meloni. In addition, I am very thankful to Daniel Brooks, Frank Paris, Kirsten Schmidt, Melinda B. Fagan, Fridolin Gross, Maria Kronfeldner, and Marcel Weber for making constructive comments on ideas presented in this book. I am especially grateful to Helmut Pulte and Christina Brandt for their comprehensive and valuable comments, as well as their kind advice and support. This book was written in different places. I would like to thank the Phi- losophy of Science Group (Helsinki University), including Petri Ylikoski and Jaakko Kuorikoski, the Biolosophy Group (Bielefeld University), the ACKNOWLEDGMENTS

PhiBio Group (Metropolitan Autonomous University, Mexico City), and the members of the Centre for the Study of Life Sciences (Egenis, University of Exeter), including Daniel Nicholson, John Dupré, and Thijs van Stigt, who commented on several ideas initially presented in the form of conference papers. The research behind this book was supported by the German Research Foundation (DFG), the Foundation Mercator, the Ruhr Univer- sity Research School, the Heinrich Hertz Foundation, and the German Aca- demic Exchange Service (DAAD). Many thanks to Dusa McDuff for sharing with me translations of Rus- sian papers originally prepared for her father, C. H. Waddington. Also, as a non-native speaker of English, I am especially grateful to Daniel Brooks and Maureen Creamer Bemko, who made this book readable and possibly enjoyable by editing my “German-style” English. At the University of Pitts- burgh Press, I thank Abby Collier for her support and enthusiasm for the project and for making this such an enjoyable process. Also, I am grateful for the comprehensive comments of the reviewers, which were insightful and very helpful in correcting omissions and clarifying arguments. Finally, I owe an immeasurable debt to Ulla Quadbeck-Baedke and Jochen Baedke for supporting this project a long time before it actually began, as well as to Abigail Nieves Delgado for making me think and smile every single day since I met her.

xii ◂ ABOVE THE GENE, BEYOND BIOLOGY

INTRODUCTION WHAT IS EPIGENETICS?

The potential is staggering. . . . The age of epigenetics has arrived. TIME, JANUARY 2010

THIS BOOK IS ABOUT how biologists in the booming field of epigenetics explain living systems. It directly responds to an idea seemingly omnipres- ent in the academic and non-academic world: the view that epigenetics imposes a major theoretical shift on modern biology by invoking previously neglected phenomena and new levels of biological complexity. More particularly, it addresses the question of whether epigeneticists explain differently—both from how other biologists explain their phenomena and from how philosophers of science usually conceptualize biological explanations. Today, epigenetics is usually described as the investigation of regulatory non-DNA factors that are taken to be causally responsible for realizing genetic information. These factors are addressed not only to explain developmental phenomena, like phenotypic plasticity or, more specifically, cancer, schizophrenia, obesity, alcoholism, and aging, but also to aid in the search for successful associated medical applications, like stem cell therapy INTRODUCTION

and cloning. In addition, epigenetic factors are highlighted in investigations of heredity phenomena, like disease etiology and sex-linked inheritance patterns, and in studies of the role of development in evolution. In short, epigenetics is currently one of the hottest topics in biology. Th e number of paper titles containing the word “epigenetic(s)” has increased more than tenfold since 2000, thus gradually chipping away at the predominance of genetics (fi g. I. 1). Moreover, both the highly ambitious Human Epigenome Project and the fi eld’s own journalEpigenetics have been launched since 2000. However, despite its current topicality, the term “epigenetics” is anything but new. It was introduced by the prominent British embryologist Conrad Hal Waddington back in the 1940s. According to Waddington, epigenetics should, on the one hand, refer to the Aristotelian theory of epigenesis, which understands development as consisting of both gradual and qualitative changes. On the other hand, it should also highlight the need to investigate processes “above” the gene, as implied by the prefi x epi, which means “over” or “upon.” More specifi cally, Waddington (1952b, vi) understood epi-

FIG. I.1. Relative frequency of articles with the word “epigenetic(s)” in their title (using ISI Web of Knowledge, 1950 to 2015). A frequency index of 1 means there is one title including the word “epigenetic(s)” for every one hundred titles including “genetic(s).” In total numbers, until the year 2000 there are fewer than one hundred articles for each year, and in 2015 there are more than twenty-four hundred.

2 ◂ INTRODUCTION genetics as the “science concerned with the causal analysis of development,” especially the causal role of networks of interacting genes and how these networks bring phenotypes into being. If we compare Waddington’s classical epigenetics and contemporary epigenetics, we find a few general views that seem to have survived over the decades. First, both Waddington’s epigenetics as well as substantial parts of its modern counterpart investigate development in a systemic, networklike manner. Waddington called this environmentally sensitive network of interactions the “epigenotype”—a web of processes that jointly gives rise to the phenotype. This network view currently reappears in epigenetic studies, such as the Human Epigenome Project, in which researchers seek not to “genotype” humans but to “epigenotype” them (i.e., to screen their whole epigenome). Second, epigenetics remains closely linked to study of causal analysis. For example, in the mission statement of the journal Epigenetics, the editors define contemporary epigenetics as the field that “studies heritable changes in gene expression caused by mechanisms others [sic] than the modification of the DNA sequence” (Epigenetics 2017). In other words, while classical epigenetics was focused on the causal role of genes, many modern epigeneticists investigate how nongenetic changes are caused (e.g., through environmental influences) and how they lead to developmental and hereditary (transgenerational) effects. Thus, despite the fact that the causal factors of interest might have changed over the decades, from genes to everything but or, as one might now rightfully say, everything above—genes, the cornerstones of epigenetics’ causality-based research program seem to have survived. These general similarities should not convey the view thatclassical epigenetics was a success story. Figure I.1 clearly shows that it was not. In fact, until today almost no concepts and central ideas of the original field were picked up by mainstream biology. This is unsurprising, since Waddington’s systemic view was not considered in line with the reductionist one-cause- one-effect thinking popular during the rise of molecular biology. Moreover, as Patrick Murray, a colleague of Waddington in the early 1930s, once noted, reading Waddington’s books is like “wading through mud up to the armpits” (quoted in Hall 1992, 116). As a consequence, not only Waddington’s view of the complex dynamics between gene interaction and development but also his attempt to unite genetics, embryology, and evolutionary theory were largely forgotten. That is, until recently.

▸ 3 INTRODUCTION

Given this mostly unsuccessful and somewhat discontinuous history of the field, one might wonder not only what contemporary epigenetics exactly is but, more specifically, what some authors mean when they say that epigenetics currently introduces something new to modern biology—a novelty they describe as “epi-geneticization” (Van Speybroeck et al. 2007) or the “epigenetic turn” (Jablonka and Lamb 2010; Nicolosi and Ruivenkamp 2012). This recent epigenetic turn, which some may hold to be quantitatively expressed by Figure I.1, is usually considered as taking place in various biological fields, from molecular biology to evolutionary biology. One could characterize this development as one in which the nongenetic factors that epigeneticists have identified in regulating and editing genetic information on the molecular level seem to be taking over genes’ causal and explanatory supremacy. The supremacy of the gene was established in a step-by-step process over the course of the twentieth century: through recognition of the usefulness of the gene concept as an organizing instrument in early evolutionary studies on population changes in the 1930s; through awareness of the possibility of identifying genes, both with increasing precision and as material entities (namely as DNA sequences), since the late 1940s; and through the adoption of a gene-centered view of evolution that made it possible to unify natural selection and heredity under a sole causal unit, namely the gene, since the 1960s. The triumphant march of this genetic framework came to a (possibly interim) halt in the late 1990s, when biologists increasingly realized that humans have not only far fewer genes than expected but that on the level of the genome humans are nearly identical.1 If nothing else, these results led to a new theoretical framework becoming established, one that is commonly described as postgenomics (see, e.g., Griffiths and Stotz 2006; and Stotz 2008). This framework states that when and how genes are expressed is not determined intrinsically but rather by the genes’ cellular and organismic environment. This new perspective gives epigenetic factors more causal and explanatory relevance in the biosciences. Importantly, these factors are thought to explain not only organisms’ plastic and environmentally responsive development but also how changes in ontogenetic pathways can affect population dynamics and thus evolutionary trajectories. The latter refers to the idea that the nonrandom responsiveness of regulatory epigenetic factors to environmental cues, as well as their heritability, may drive and bias evolutionary change. Eva Jablonka and Marion Lamb (1995, 2005, 2008, 2010) have famously

4 ◂ INTRODUCTION argued in a number of books and papers that the evolutionary relevance of contemporary epigenetics has a somewhat unforeseen historical punch- line—the return of Lamarckian ideas to modern biology. This argument has two components. First, epigenetic variation is produced and inherited with a degree of autonomy from the DNA level. In other words, the emergence and transmission of genetic and epigenetic variation are causally decoupled to a certain degree. Accordingly, epigenetic variation might offer a distinct substrate for evolutionary change, which, so the argument goes, may guide genetic change. Thus, epigenetics challenges gene-centrism and, in stark contrast to the gene-centered view of evolution (Williams 1966; Dawkins 1976, 1982), invokes a broader notion of heredity that should be taken into consideration in accounts of evolution. Second, because of the dual nature of epigenetic regulatory systems— being involved in development and inheritance—inducible and heritable epigenetic variation resembles Lamarckian “soft inheritance,” or inheritance of acquired characteristics. This claim of the “Lamarckian dimension” of epigenetics has provoked a wide discussion, both in academia—among historians of science and biologists interested in the historical aspects of their field (see, e.g., E. Richards 2006; Gissis and Jablonka 2011; and Y. Wang et al. 2017)—as well as in the wider public (see, e.g., Young 2008; and Burkeman 2010), on whether the neo-Darwinian framework of evolutionary theory should be expanded to incorporate this “Lamarckian dimension.” This historical debate is in itself an important dimension of the discussion of epigenetics. In the 1990s not many biologists would have put their money on Lamarck’s resurrection. Moreover, for some authors Lamarck still is a red flag of sorts. For example, Richard Dawkins (1982, 164) once noted that “a return to the theory of evolution that is traditionally attributed to Lamarck . . . is one of the few contingencies for which I might offer to eat my hat.” Given this delicate promise, some authors might be even more inclined to consider giving Lamarckism a (last) chance. However, as this book attempts to show, assessing epigenetics’ special theoretical structure, as well as the character of the current “epigenetic turn,” should not be left completely to the participants in this Lamarckism debate. This is particularly important, because in recent years a less historically oriented, more general debate about possible expansions of evolutionary theory and their associated challenges has developed. It shows increasing focus on the the-

▸ 5 INTRODUCTION

oretical structure, assumptions, and predictions of novel developmental approaches such as epigenetics. Against this background, let us redirect our focus of attention away from the second claim (the Lamarckian turn) and toward the first claim (what may be described as the epigenetization of biological theory). This (at first glance) historically less radical argument for overcoming theoretical orthodoxy in biology carries with it a variety of still widely neglected philosophical issues concerning the explanatory characteristics of epigenetics, its methodological hurdles, and its conceptual challenges for theoretical integration. These issues are foreshadowed, for example, by the debate currently being revisited on how to conceptually grasp evolutionary “ultimate” causes and developmental “proximate” causes with respect to novel, developmentally oriented evolutionary explanations. While many epigeneticists criticize this traditional dichotomy, those who adhere to traditional conceptual frameworks often do not attack epigenetics’ Lamarckian dimension but rather epigenetics’ explanatory framework more generally. For example, the population geneticist Michael Lynch claims, “There are more things to explain, but I think a lot of us are happy with the fundamental framework to do that explaining in” (quoted in Grant 2010). By taking such scattered discussions and comments as first stirrings of a growing philosophical debate on the conceptual foundations and explanatory standards of modern biology, this book argues that the so-called “epigenetic turn” is, above all, a shift in scientific explanation and in conceptualizing living systems. This idea does not presuppose that one can sharply demarcate epigenetics from other fields, such as genetics, or that one can identify a unique kind of explanation or conceptualization in this field, distinct from all other biological explanations and concepts. As is usual in young research fields, epigenetics does not exhibit clear-cut and sharp boundaries. Nevertheless, it will be shown that it is possible to identify a bundle of crucial features of epigenetics. These features make three things possible:

1. They allow us to broadly characterize epigenetics and epigenetic explanation, respectively. 2. They allow us to better understand current trends in biology (especially in molecular and evolutionary biology). 3. They allow us to fill gaps in our current philosophical theories of scientific explanation.

6 ◂ INTRODUCTION

Let me briefly describe these points. As for the first, we may understand the concept of epigenetics not as sharply circumscribed, or as showing necessary and sufficient conditions that demarcate this domain of study, but as a cluster concept, in the way described by Ludwig Wittgenstein (1958). This means that although epigenetics shows common features, certain particular features may be more relevant than others to single instantiations of the concept. In addition, some features may sometimes be entirely absent from single instantiations of the concept. Take, for instance, the concept of games. While a common feature of games is that there is usually a win- ner, some games do not necessarily possess this feature, soccer and soli- taire being examples. In a similar manner, we may come across a particular epigenetic investigation that exhibits a number of features similar to other approaches we usually consider to be “epigenetic,” but it also shows features distinct from (maybe all) other epigenetic investigations. Thus, in Wittgen- stein’s words, epigenetics only exhibits family resemblance. However, this result does not render meaningless any use of the concept of epigenetic(s). Nor does it suggest that one cannot get an idea about the set of features that is characteristic of epigenetics and how this set differs from those of other fields. What is more, it is wrong to suggest that “epigenetics” is only “a useful word if you don’t know what’s going on—if you do, you use something else,” as Adrian Bird once commented (quoted in Nature Biotechnology 2010, 1031). “Epigenetics” can be characterized in further detail. In order to trace the characteristics of epigenetics, a number of features widely shared in the field will be discussed. These features are, first and foremost, related to how epigeneticists explain. I will show that, as a consequence of their interest in grasping the complexity of developmental phenomena, many epigeneticists explain in a different manner than do other biologists, especially those in molecular and evolutionary biology. More specifically, they trace particular dependency relations and highlight as explanatorily relevant certain kinds of dependencies that are usually not central to explanations in other fields.2 Regarding the second point, investigating this set of features allows better understanding of the current so-called epigenetic turn. Here, it is important to mention that our Wittgensteinian perspective does not presuppose that the current trends in biology summarized by this term are driven only by epigenetics. In fact, certain elements characteristic of the epigenetic turn may be found in other fields as well, such as systems biology. Nonetheless,

▸ 7 INTRODUCTION

it will be argued that this shift is driven in a crucial and important manner by epigenetics. This makes epigenetics a privileged playground for philosophers of science interested in the dynamics of biology. On the third point, there are many more virtues of philosophical investigations of epigenetics. It will be shown that a number of epigeneticists use particular concepts of causation and biological mechanism, as well as concepts of causal explanation and constitutive mechanistic explanation, that cannot be easily integrated with the way philosophers of science commonly conceptualize the explanatory relations traced in biology. Thus, we should also turn to epigenetics in order to improve our theories of scientific explanation. Against this background, this book will consider various models and modeling strategies in epigenetics and investigate how they function in conceptualizing and explaining the properties and dynamics of complex epigenetic systems in development and heredity. In the chapters that follow, I use the notions of explanation and modeling interchangeably in those cases in which we are concerned with models identifying a causal or constitutive (usually mechanistic) dependency relation as an explanatory relation between explananda and explanantia. This does not include descriptive models or model organisms, which by themselves do not explain. The latter two may, however, play significant, yet more indirect heuristic roles in explanatory practices.3 For example, I consider models that serve as theoretical or material scientific representations (Hesse 1966; Giere 1988, 1999; Griesemer 1990; Baetu 2014), are used as instruments with which a biologist can intervene on the models’ associated phenomena and theories (M. Morgan 2003; Leonelli 2007), and mediate the transfer of knowledge from a general class of objects (or a theory) to a more particular phenomenon (Morrison and Morgan 1999). More specifically, I investigate experimentation-based causal or mechanistic models, mathematical models, models inferring causality from statistical dependencies in observational data, and heuristic (visual) models, as well as their associated methodologies and modeling practices in a wide array of fields, ranging from molecular epigenetics and stem cell biology to ecological or evolutionary epigenetics. This book contains five main chapters and a concluding chapter. The first two chapters (1 and 2) address mainly biological and historical topics, while the remaining ones (3–5 and the conclusion) deal with philosophi-

8 ◂ INTRODUCTION cal issues. Subsequent to the introductory notes presented here, chapter 1, “How Epigenetics Deals with Biological Complexity,” reviews crucial challenges for biologists in general, and epigeneticists in particular, in modeling the complexity of living systems. These problems are accompanied by issues posed by philosophers of biology regarding how to conceptualize explanatory relations traced at higher levels of organization that physics alone cannot describe. Epigenetics intensifies these problems in a molecular context, since it investigates biological complexity in an expanded manner, both with respect to the number of components and their relations investigated in a particular system (i.e., structural complexity), as well as with respect to the nonlinearity of these relations and thus the seemingly unpredictable behavior of the system (i.e., dynamic complexity). This new layer of biological complexity is labeled “epigenetic complexity.” In contrast to Wadding- ton’s foundation of classical epigenetics as a science concerned with genetic (or genomic) complexity, modern epigenetics is identified as the complexity science primarily concerned with nonsimple relations (i.e., not one-to-one relations between two factors but one-to-many or many-to-many relations) between nongenetic factors and their role in plastic development and heredity. In addition, the Lamarckian dimension of the latter relations across cell divisions and especially between organisms is discussed by reviewing various cases of epigenetic inheritance. In chapter 2, “Challenges of Epigenetics in Light of the Extended Evo- lutionary Synthesis,” we step out from the shadow of Lamarck and the prevailing Lamarckism debate by discussing a number of inter- and intradisciplinary conflicts currently hindering theoretical integration. They arise due to both the explananda and explanantia chosen by epigeneticists and the modeling strategies applied to establish explanatory dependencies. Inter- disciplinary conflicts mainly concern the theoretical integration of epigenetics into the neo-Darwinian framework of the modern synthesis and the theoretical expansion of the latter, respectively. This includes the issues of whether (and how) molecular epigenetic explanations can address evolutionary explananda although they are based on highly artificial experimental setups and focus on proximate causes rather than on ultimate causes. The latter problem in particular is central to the current so-called “extended evolutionary synthesis” debate. In addition, intradisciplinary conflicts are discussed. These concern issues on how to understand the field of epigenetics methodologically and conceptually, as well as on how epigeneticists should

▸ 9 INTRODUCTION

explain. This internal tension of epigenetics is revealed by comparing approaches of prominent advocates of stem cell epigenetics and molecular epigenetics. This analysis shows that Waddington’s traditional methodological program—to develop mathematical models of developmental dynamic stability in his so-called “epigenetic landscape” framework—is still alive in epigenetics. Chapters 3 and 4 discuss genuine philosophical issues related to the structure of epigenetic explanations, including both causal and mechanistic explanations. In particular, these chapters describe how epigenetics currently partakes in establishing a new explanatory agenda in molecular and cell biology. In chapter 3, entitled “Causal Explanation,” I present an account of scientific explanation in line with philosophical orthodoxy, whereby generalizations are explanatory if they are invariant under intervention. Then this interventionist account is applied to elucidate how epigeneticists in molecular and cell biology conceptualize causation and causal explanation. Contrary to a mechanistic theory of causation, it is argued that an invariant generalization does not necessitate supplementary information taken from more fundamental (i.e., lower) levels of organization that explains why the relation under study holds. In other words, the appropriateness of higher- level generalizations in molecular epigenetics, which no longer consider genes the primum movens in development and heredity, will be justified. Lower-level genes neither possess a unique ontic or epistemic status, which would necessitate listing them in every explanans of causal explanations, nor do they share the explanatory realm in which epigenetic causes actually make a difference. In short, epigenetic complexity is assessed in a particular way by tracing causes without mechanisms. At the same time, this does not mean that epigeneticists seek to eliminate genetic dependencies from the “causality landscape” of complex living systems. Rather, genes and their causal roles are reassessed in epigenetic causal explanation in a specific and novel manner. In chapter 4, “Mechanistic Explanation,” I discuss the concept of mechanism in epigenetics. It is argued that this concept, which is central for so-called constitutive explanations of molecular and cell biologists, has been decisively “demachinized” by epigeneticists. In this new field, biological mechanisms are often conceptualized in a nonmachinelike manner, in contrast to how traditional molecular biologists and many of today’s philosophers of science conceive of mechanisms. Independently, some epigenetic

10 ◂ INTRODUCTION mechanisms cannot be described adequately in terms of the interventionist account of causation, in contrast to what several philosophers of science think about mechanisms in biological explanation more generally. Partic- ularly in models representing dynamic stability of complex epigenetic systems, there are descriptions of mechanisms that entail explanatorily relevant relations and properties that cannot be considered as “causal” according to the standard accounts of biological mechanisms. These cases are labeled mechanisms without causes. While the earlier chapters primarily deal with the explanatoriness of epigenetics (i.e., whether and how epigenetics explains), in chapter 5 we turn to the concept of explanatory power. In this chapter entitled “Assessing the Explanatory Power of Epigenetics,” I seek to offer a solution to the problem of how to integrate novel and orthodox explanations into the pluralistic framework of an extended evolutionary synthesis. In particular, I present a contrastive framework, which is able to evaluate the explanatory value of distinct biological explanations. This account is able to give precise guidance to the advocates of an extended evolutionary synthesis by means of which criteria (why) and in which explanatory context (when) epigenetic explanations are legitimately chosen over orthodox molecular and evolutionary explanations. By reviewing the results of the above investigation and discussing some loose ends, in the conclusion I argue for the necessity to establish a philosophy of epigenetics, hence the subtitle for the concluding chapter. This historically informed field should address philosophical issues on the interrelationship between methodology, modeling, biological concepts, and scientific explanation in epigenetics. Moreover, it should not only be understood as a philosophical appendage of the recent epigenetic turn but also establish itself as a field offering valuable insights both for epigeneticists, on the philosophical foundations of their still immature field, and for philosophers of science interested in biological concepts and scientific explanation in biology. It should do so by making use, first and foremost, of the philosopher’s toolbox, not the empirical scientist’s. This means, in short, a philosophy of epigenetics should be established as a genuine philosophical domain that goes beyond biology, so to speak. As just initiated, however, this book is intended to be of interest not just for philosophers of biology or even philosophers of science more generally. It also seeks to address philosophically minded biologists in molec-

▸ 11 INTRODUCTION

ular, developmental, and evolutionary biology interested in clarifying the explanatory structure and power of epigenetic explanations and how they are shaping and possibly changing modern biology. Moreover, although the main purpose of this book is to approach the previously neglected philosophical dimensions of epigenetics, the historical dimension is not left out. A number of chapters (e.g., 1 and 2) explicitly focus on the methodological and conceptual history of epigenetics. In addition, throughout these pages the reader will find various historical case studies. For example, epigenetic explanations are discussed in the context of positions ranging from nineteenth-century experimental physiology (Claude Bernard), early twentieth-century genetics (Thomas Hunt Morgan), and developmental genetics (Conrad Hal Waddington) to the advent of molecular biology (Francis Crick and James Watson, François Jacob and Jacques Monod) and topological approaches of embryogenesis in the 1960s and 1970s (René Thom, Israel Gel’fand). However, especially in chapter 2 these historical cases are presented in a general manner in order to suit the systematic issues discussed. Nevertheless, these historical “hints” may motivate the historically interested reader to elaborate on still underrepresented topics in the history of epigenetics.4 From a biographical perspective, this book ties in with the interdisciplinary interest of epigenetics’ founding father. Conrad Hal Waddington was not only a renowned embryologist and geneticist but also one of the most central figures in establishing the field of theoretical biology—a field that is understood to be genuinely concerned with interdisciplinary issues on the boundary between biology and philosophy. By taking his influential multi- volume work Towards a Theoretical Biology (Waddington 1968, 1969b, 1970b, 1972) as a guiding light, this book was written under the slogan “Toward a Philosophy of Epigenetics.”

12 ◂ ONE HOW EPIGENETICS DEALS WITH BIOLOGICAL COMPLEXITY

One can best feel in dealing with living things how primitive physics still is. ALBERT EINSTEIN, 1947

The word explanation occurs so continuously, and has so important a place in philosophy, that a little time spent in fixing the meaning of it will be profitably employed. JOHN STUART MILL, 1843

EXPLAINING COMPLEX LIVING SYSTEMS

UNTIL RELATIVELY RECENTLY, JOHN Stuart Mill’s call for dealing with scientific explanation was addressed by philosophers of science as well as scientists in a rather monistic way. Many were convinced that a phenomenon is explained only if some law covers it as a special case. Accordingly, philosophers usually focused on phenomena in the exact sciences or held HOW EPIGENETICS DEALS WITH BIOLOGICAL COMPLEXITY

the view that all scientific explanations can be reduced to a small number of laws or theories at a physical level. This view of scientific explanation has imposed an overstated criterion of how certain explanations do, in fact, explain, which has led some to describe biologists’ scientific program as a technological application of basic laws of nature, similar to that of, for example, engineering (Smart 1959). However, due in no small part to the increasing success of molecular biologists in “decoding” the genome and the importance of their findings for humans, as well as the emergence of the field of philosophy of biology in the 1960s to 1970s, this view has gradually changed.1 Two consequences have to be considered in this context. First, both the applicability of the so-called covering-law account of explanation in biology and the existence of biological laws have been called into question (see chap. 3). Second, new accounts of scientific explanations that fit the unique explanatory challenges in biology and the special sciences more generally have been (re)developed. In philosophy of biology, these new explanatory accounts draw on a long tradition of conceptualizing relationships in and between living systems (between structures of various size, from DNA to organisms, and of different strength), as well as analyzing the role with which they figure into explanation. For example, by revising the Aristotelian concept of teleology (Aristotle 1984, 195a23–24, 198b8–9), Ernst Mayr (1961) developed an idea of final cause that he considered to be explanatorily relevant exclusively for evolutionary biologists. In addition, it has been debated how the seemingly goal-directedness of living beings can be addressed properly by functional explanations (Lehman 1965; Wimsatt 1972). More recently, philosophers of science have focused on clarifying the nature of causal dependency relations at higher levels of organization. These dependencies are usually cited as explanantia in the special sciences in general and in biology in particular. Causal relations have been grounded in complex regularities, as argued by John Leslie Mackie (1974), or in counterfactual dependencies described in the form “if event c had not occurred, event e would not have occurred” (Lewis 1973, 2000). In addition, they have been also understood as relations of probabilistic relevance (Pearl 2000; Williamson 2009) and as relations involving manipulability (Woodward 2003). The latter of these has been highly influential in recent philosophy of science. This approach is exemplified by the so-called interventionist account of James Woodward, which integrates counterfactual and probabi-

14 ◂ HOW EPIGENETICS DEALS WITH BIOLOGICAL COMPLEXITY listic theories with the view that causes are (probabilistically) dependent on an intervention and are thus exploitable for purposes of control. In particular, interventionism has been invoked in philosophy of biology to specify causal relations and how they figure in causal explanation in a variety of research fields, ranging from molecular genetics to ecology (see chap. 3). Moreover, this theory has been used extensively to make sense of how biological mechanisms constitute a phenomenon and how mechanistic explanation functions in biology (see chap. 4). These more recent accounts of causal and mechanistic explanation have been in large part developed to address problems that arise due to the complexity of living systems investigated in biology. For example, models of these systems often include events with low probabilities. This means that there are dependencies in the living world—between genes, cells, organisms, and their environment—that only hold occasionally or even rarely. Are these dependencies causal? In order to answer such a question, we first have to understand what constitutes a living system’s complexity. Complexity has been defined in various ways (see S. Lloyd 2001; Emmeche 1997; Holland 1998, 225–31; Mitchell 2003, 4–7; and Northrop 2010, 2–4). With respect to the issue at stake, the following webs of dependencies in cells, organisms, or ecosystems will be understood as complex:

Structural complexity. Dependencies may be located in living systems that can be extended along any spatial dimension. This means the system as a whole as well as the dependencies between its parts are determined by n elements or objects. For example, organisms are constituted by n number of cells. Additionally, these cells stand in nonsimple relations (i.e., one-to- many or usually many-to-many relations) and are organized in a nonrandom way. Dynamic complexity. Dependencies may relate events (e.g., transcription, mitosis, reproduction) or causal properties of objects (e.g., the structure of the DNA, the ability of cells to form tissues, the maturation time of an organism) in a nonlinear way. Dynamic systems entailing such dependencies are no longer conceived in a Newtonian way as simple systems with linear or proportional relations between causes and effects (which are, in principle, completely computable in a Laplacian manner) but as complex ones whose stability characteristics are affected nonlinearly. For example, small changes in the trajectory of a cell during differentiation may cause

▸ 15 HOW EPIGENETICS DEALS WITH BIOLOGICAL COMPLEXITY

major effects (and severe damages) to an organism. Conversely, major changes may cause minor effects or even no effect at all in the complex system.

In short, complex biological systems are multicomponent (usually multilevel) systems that are constituted through various nonlinear dependencies. Additionally, such systems cannot be separated from their environment. They are open. In other words, they constantly exchange both energy and matter with their environments and consequently exist in a continuum with their environments. For them, this exchange is necessary in order to maintain themselves displaced from thermodynamic equilibrium. In the moment this open exchange is stopped and equilibrium is reached, death results.2 As a consequence of extensive structural and nonlinear relationships with the environment and thus the irregularity and unpredictability of dynamic patterns, it is not possible to offer a fixed, complete, and formal description of a complex system. Robert Rosen (1999, 292; original emphasis) describes this idea: “A system is simple if all its models are simulable. A system that is not simple, and that accordingly must have a nonsimulable model, is complex.” Rosen argues that a system is complex when it acts in an unexpected or unanticipated way so that its behavior does not match the predictions of our models. Thus, according to him, complexity is a relational property of a system resulting from the comparison of the system with its model. In a similar manner, Bruce Edmonds (1999a, 1999b) states that complexity hinders representing a system’s overall behavior, even when its components and their interrelations are completely known. We may call this problem the representation problem of complexity. Closely related, models that simplify or frame complexity are unavoid- ably limited, since the information that these models omit has nonlinear and barely predictable effects that reverberate throughout the system. This problem is deeply interlinked with, for example, the experimental methodologies applied to conduct modeling, as noted by Joseph Needham (1929, 82) with respect to organisms: “An organism is something which the scientific method cannot deal with; it is a hard, round, smooth nut, which experimental analysis can neither crack nor lever open any point. As soon as a hole is made in it, it explodes like a Prince Rupert drop and vanishes away.” The representation problem of complexity is usually accompanied by the

16 ◂ HOW EPIGENETICS DEALS WITH BIOLOGICAL COMPLEXITY epistemological problem that taking every bit of data into account when modeling complex phenomena—that is, the idea to model wholes without, as far as possible, omitting their parts—does not necessarily increase our knowledge of these phenomena. This idea has been defended by a number of philosophers of science who emphasize the importance of abstraction (i.e., omitting details), especially to explain complex systems (Batterman 2002; Strevens 2008; M. Weber 2008; Levy and Bechtel 2013).3 The need to abstract from complexity can be metaphorically visualized by a short story, titled “On Exactitude in Science,” from the Argentine writer Jorge Luis Borges, in which an empire is imagined where the science of cartography becomes so exact that cartographers developed a map on the same scale as the empire itself:

In that Empire, the craft of Cartography attained such Perfection that the Map of a Single province covered the space of an entire City, and the Map of the Empire itself an entire Province. In the course of Time, these Exten- sive maps were found somehow wanting, and so the College of Cartogra- phers evolved a Map of the Empire that was of the same Scale as the Empire and that coincided with it point for point. Less attentive to the Study of Cartography, succeeding Generations came to judge a map of such Mag- nitude cumbersome, and, not without Irreverence, they abandoned it to the Rigours of sun and Rain. In the western Deserts, tattered fragments of the Map are still to be found, Sheltering an occasional Beast or beggar; in the whole Nation, no other relic is left of the Discipline of Geography. (Borges 1975, 131)

Metaphorically speaking, we may refer to this problem as the “Borges problem.” It is crucial for those explanatory attempts, like analyses of big data sets, which seek to understand complex wholes as wholes. Since the mid-twentieth century, biologists modeling living systems have had to deal increasingly with the representation problem of complexity and the Borges problem, especially at the molecular level as the number of elements of organismic life, and the knowledge of their nonsimple dependencies and nonlinear dynamics, steadily increased. Early attempts to face these challenges include, for example, François Jacob and Jacques Monod’s (1961) operon model of gene regulation; Brian Goodwin’s (1965) models of a genetic oscillator, showing how interactions in regulatory gene

▸ 17 HOW EPIGENETICS DEALS WITH BIOLOGICAL COMPLEXITY

circuits allow periodic fluctuations; Stuart Kauffman’s (1969, 1971b) model of the pathways of cell differentiation (and distinct cell types) as changes in “nets” of genes; René Thom’s (1969, 1975) topological approach to describing discontinuities like neurulation arising in embryonic development; and Robert MacArthur’s (1955) first ecological model of the complex dynamics underlying community stability. A new layer of complexity was added to biologists’ picture of nature as a result of the Human Genome Project. This project was originally viewed as an effort to open the “book of life”—as Richard Dawkins (1976) famously called the genome in his book The Selfish Gene—and thus to rid humans of genetic diseases. As it was revealed in 2001, however, the human genome contains only thirty-five thousand genes (and estimates of the number have even dropped to around twenty thousand since then), while original estimates were for around one hundred thousand genes. Even more surprising, geneticists realized that at the DNA level humans are 99.9 percent identical. Obviously, biologists were not able to fully understand the diversity of nature by considering genes alone. They could not yet read the “book of life” from DNA sequence alone: something was missing. As a consequence, research in molecular biology shifted toward investigating how genes function within complex wholes. In other words, if the effect of a gene depends first and foremost on its context, understanding how this gene influences a trait means at the same time understanding how other factors in the gene’s genomic, cellular, organismic, and extraorganismic environment cause the trait.4 Thus, the idea of the genome shifted, from an entity with discrete and stable units—the primum movens in organismic causality—into a complex dynamic system with a worryingly high number of components and interactions. This means that not only did the number of causal agents increase significantly (numerous non–protein-coding RNA transcripts have been added to the picture) but the interactions among them became more complex. The interactions no longer occur in a linear (and often unidirectional) way between DNA sequences and the phenotype but in a nonlinear way between various regulatory factors, the DNA, and especially the environment. This increase in complexity has motivated some authors, like Michel Morange (2000b), to claim that we can no longer precisely define what a gene is. Others, like Eva Jablonka and Marion Lamb (2005, 6), have embraced the new challenges the Human Genome Project presents: “Just as in an earlier century, when the telescope opened up new

18 ◂ HOW EPIGENETICS DEALS WITH BIOLOGICAL COMPLEXITY horizons for astronomers and the microscope revealed new worlds to biologists, the revelations of molecular biology [through the Human Genome Project] cannot be neatly slotted into the existing framework of thought. They do not make the old genetics more complete; rather, they highlight the simplifying assumptions that have been made and reveal vast areas of unanticipated complexity.” This new molecular complexity isepigenetic complexity, and the research field that addresses its causal and mechanistic content is called epigenetics.

FROM GENETIC TO NONGENETIC COMPLEXITY

In order to understand today’s studies of epigenetic complexity, we have to understand how epigenetics came to be the field it is today. Right from the beginning, two concepts have been at the center of epigenetics: one of them is the complexity of development; the other is causality. As described in the introduction, Conrad Hal Waddington (1952b, vi), the leading British geneticist and embryologist from the late 1930s to 1950s, defined the term “epigenetics” as the “science concerned with the causal analysis of development.”5 This definition raises the question, Which factors, exactly, are involved in development and how do they causally work? In the early twentieth century, the causal role of genes in development was anything but certain. In fact, many researchers were not convinced that genes control development. Those who were convinced were described by Ernest Everett Just (1939) as “geneticists,” in contrast to the doubters, whom he called “biologists.” The biologists’ skepticism was based on the idea that in general the problems of genetics and the problems of embryology were fundamentally different. As argued famously by Thomas Hunt Morgan (1926), geneticists are concerned with the inheritance of traits while embry- ologists have concerned themselves with the expression of traits. In addition, following Morgan’s theory of inheritance, according to which all cells of higher organisms have the same genes, Frank Rattray Lillie (1927, 367) wondered how genes could account for the various differences in development, as in the fate of cells or the development of muscles and nerves: “Those who desire to make genetics the basis of physiology of development will have to explain how an unchanging complex can direct the course of an ordered developmental stream.”

▸ 19 HOW EPIGENETICS DEALS WITH BIOLOGICAL COMPLEXITY

This is the very issue from which Waddington’s epigenetic research program starts. His answer to the problem posed by Lillie—known as “Lillie’s paradox” (Burian 2005)—was the idea of genetic complexity. Like others focusing their study on developmental genetics, he was interested in how genetic differences lead to phenotypic variation in development. However, he also wanted to understand why, as Jablonka and Lamb (2002, 85) put it, “very often genetic and phenotypic variations are not coupled.” Therefore, he adopted a special systemic perspective on how genes work in development. He argued that genes do not act separately, as isolated and distinct entities, but together, as clusters of genes, in causing a particular trait. Only through this idea of gene clusters, in which are found differential gene expression and regulation (in today’s parlance), can one understand the various changeability and stability characteristics of developmental pathways that puzzled Lillie. By taking this more complex view of what is usually described as the genotype-phenotype map as a background, Waddington (1942a, 1957) investigated how regularly occurring developmental pathways of a dynamic system, like a cell or the whole embryo, are changed or maintained in the presence of genetic or environmental perturbations. He described such a pathway as “canalized” or “buffered” if some phenotype was maintained, although the genotype or the environment might have changed to some extent (see also Nijhout 2002; and S. Gilbert and Epel 2009). According to Waddington, this kind of developmental robustness allows the buildup of hidden variability in the genome that is not expressed phenotypically until a certain threshold or level of disturbance is reached. For example, if embryos of the fruit fly Drosophila melanogaster are raised at different temperatures, the insect’s body segmentation patterns do not change, although the so-called bicoid protein, which usually in large part determines the development of body segmentation, is expressed at different rates and thus distributed differently throughout the organism (see, e.g., Houchmandzadeh et al. 2002; and Lepzelter and Wang 2008; reviewed in S. Gilbert and Epel 2009). This concept ofcanalization matches John Hol- land’s (1995) definition of a complex system. Holland describes complexity as the property of an adaptive system by means of which it can maintain its coherence and structure given a variety of changes. Waddington’s skepticism against simplified causal models of the relationship between genes and trait development not only promoted the con-

20 ◂ HOW EPIGENETICS DEALS WITH BIOLOGICAL COMPLEXITY cept of canalization—that is, many-to-one-dependencies between genes and their output—but also the opposite concept of phenotypic plasticity (see Nilsson-Ehle 1914; Waddington 1942a, 1952a; Pigliucci 2001; and West- Eberhard 2003). While canalization supports the basic idea that a web of causes controls one output and thus renders changes in a single cause less relevant, phenotypic plasticity holds that a trait of an organism can react to an environmental input in various ways. Th is means that the genome codes for a wide range of potential phenotypes. For example, in a number of dung beetle species the horn length and body size of males are determined by the quality and quantity of food they eat during larval development (Moczek and Emlen 1999; Moczek 2010) (fi g. 1.1A). Somewhat related, so-called polyphenic development oft en leads to

A ▸

◂ B

FIG. 1.1. Phenotypic plasticity in the Asian rhinoceros beetle, Trypoxylus dichotomus (A), and in prairie bird locust, Schistocerca emarginata (B). For description, see the text (A: Warren et al. 2014, e3; B: Sword 2002, 1642; A: reproduced with permission from Warren et al.; B: reproduced by permission of the Royal Society).

▸ 21 HOW EPIGENETICS DEALS WITH BIOLOGICAL COMPLEXITY

environmentally determined discontinuous phenotypes. Examples of this plastic “either/or development” are seasonal polyphenism in butterflies (Brakefield and Frankino 2009), environmental sex determination in reptiles (Janzen and Phillips 2006), predator-induced polyphenism in water fleas (Laforsch and Tollrian 2004; Engel et al. 2014), and morphological and behavioral changes in locust species (short-winged, uniformly colored, and solitary versus long-winged, brightly colored, and gregarious), induced by changes in population density (Simpson and Pener 2009) (fig. 1.1B). As these examples showed, genotype and phenotype seem to be decoupled to a certain degree. This result contradicts those views—still prominent today—of the genotype-phenotype map according to which the causal effects of genes on traits show separability and additivity. These ideals draw on old atomistic perspectives according to which the organism is nothing but (or at least similar to) a machine (Nicholson 2013, 2014a). Famously, this perspective has been defended by a number of advocates of the modern synthesis, as it allowed population geneticists not only to ascribe to singular genes particular fitness values but also to legitimate the idea that development is of little causal relevance for evolution, as it is only a direct readout of the genotype (Pigliucci 2010). In other words, the genotype can be understood as a “genetic program” or a “genetic blueprint” of the phenotype. This view has been adopted by gene(centric) selectionists, such as George C. Williams (1966, 1992) and Richard Dawkins (1976, xxi), the latter being one who understands natural selection as an engineer and organisms as “robot vehicles blindly programmed to preserve the selfish molecules known as genes.” Against these simplistic views of genetic causality the concepts of canalization and phenotypic plasticity support a more complex causal picture of the developing organism, which more seriously includes both multifactorial causation—many-to-one dependencies—and common causes—one-to- many dependencies (fig. 1.2) or a combination of the two: many-to-many dependencies.6 In addition, the complexity of the genotype-phenotype map is increased by considering inner- and extraorganismic environmental factors as causally and explanatorily relevant. Thus, Waddington’s epigenetics rests on a view of developing organisms as complex systems, whereas their complexity results from both the high number of their parts and their nonsimple relations (structural complexity), as well as the nonlinearity of these relations and openness of the systems investigated (dynamic complexity).

22 ◂ HOW EPIGENETICS DEALS WITH BIOLOGICAL COMPLEXITY

A ▸ ◂ B

FIG. 1.2. Two simple causal (directed) graphs showing the causal relationships described by the concepts of canalization (A) and phenotypic plasticity (B). C1–3 = causes; E1–3 = effects. In (A) a change in, for example, C1 does not necessarily change

E, and in (B) a change in C, for example, changes not only E1 but E2 and E3 as well. For examples of these causal relations in developing organisms, see the text.

Besides his interactionist perspective on genes and his systemic view of the causal relationships that build up organisms, Waddington’s (1957) classical epigenetics opposed traditional understandings of the relationship between genes, development, and evolution. In line with other authors, such as Ivan I. Schmalhausen (1949), Waddington criticized Morgan’s cleavage of heredity and development and the negligence of the latter in the framework of the modern synthesis. His classical epigenetics sought to shed a light on the origins of the regulatory switches between alternative phenotypes and on the phylogenetic routes that make organisms’ traits canalized or plastic. In other words, epigenetics should not only be embryological or developmental genetics but should also inform evolutionary biology. Both central features of classical epigenetics—a more complex understanding of genetic causality and the genotype-phenotype map, respectively, as well as the idea of interlinking this view with evolutionary issues—can also be found in modern epigenetic studies. At the same time, however, a number of things have changed. Most important, the development from Wadding- ton’s epigenetics to modern epigenetics is a shift from investigating complex genetic systems to complex nongenetic systems, as well as a move toward increasingly considering a variety of nongenetic inheritance phenomena.

▸ 23 HOW EPIGENETICS DEALS WITH BIOLOGICAL COMPLEXITY

Since the 1990s, the concept of epigenetics has more and more been used to address phenomena of nongenetic inheritance across cell divisions and, more recently, between organisms.7 Simultaneously, the definition of epigenetics as “nuclear inheritance which is not based on differences in DNA sequence” by Robin Holliday (1994, 454; emphasis added) started to spread in molecular biology (see Russo et al. 1996; Wu and Morris 2001; and Egger et al. 2004).8 In fact, Holliday’s own research on the molecular mechanisms regulating gene activity, especially DNA methylation, and on the inheritance of cell phenotypes through so-called “cell memory” had a major influence on a change in the meaning of the term “epigenetic(s)” during this time. A recurring motive became the idea that epigenetics is the study of those regulating factors of gene activity that can lead to the inheritance of this very gene activity. For example, in a special issue on epigenetics in Nature in 2007, Alex Eccleston and colleagues (2007, 395; emphasis added) define epigenetics as “the study of heritable changes in gene expression that are not due to changes in DNA sequence.” And more recently, Trygve Tollefsbol (2011, 1; emphasis added) notes in his Handbook of Epigenetics that “a consensus definition is that epigenetics is the collective heritable changes in phenotype due to processes that arise independent of primary DNA sequence. This heritability of epigenetic information was for many years thought to be limited to cellular divisions. However, it is now apparent that epigenetic processes can be transferred in organisms from one generation to another.” This idea of epigenetic inheritance includes developmentally induced, highly stable, nongenetic cell heredity (Jablonka and Lamb 1995, 2005; Probst et al. 2009) and less stable transmission between generations of organisms (Jablonka and Raz 2009; Wang et al. 2017). The former process is taken to be responsible for cells with the same genetic makeup to develop reliably into functionally distinct cell types and tissues in multicellular organisms. The latter includes a number of channels of nongenetic inheritance, for example, through the germ line or as a result of developmental interactions between mother and offspring (see next section). Through these channels, developmental non-DNA factors are transmitted, which then lead to non- Mendelian inheritance patterns in traits. Despite the common conceptual distinction between “intragenerational epigenetics” and “intergenerational epigenetics,” both fields are concerned with the very same molecular processes causing phenotypic plasticity and canalization by altering the expression of genes without changes in the base-

24 ◂ HOW EPIGENETICS DEALS WITH BIOLOGICAL COMPLEXITY pair nucleotide sequence. In past decades, a number of such epigenetic processes have been identified that integrate and mediate genomic input, such as stochastic variation in gene expression, and environmental input, such as variation in nutrition, temperature, and stress, in the development of phenotypes.9 These include:

1. Regulation of gene activity through chromatin marking, like DNA methylation (Wigler 1981; Suzuki and Bird 2008) or histone modification (Barth and Imhof 2010; Cheedipudi et al. 2014). This refers to some chemical groups, also called marks, that are attached directly to the DNA or to proteins like histones. Histones act as spools around which DNA winds, and histone modification regulates gene expression (fig. 1.3). Subsequently, these marks are reconstructed in daughter cells. The patterns of a specific category or marks, those of imprinted genes, are constructed in a parent- of-origin-specific manner. For example, for some genes in humans only the copy of the father ever gets expressed, while for other genes the copy of the mother is expressed. 2. Three-dimensional structures that act as a template for the production of similar structures, such as in templating of membrane structures and prions (Cavalier-Smith 2004; Shorter and Lindquist 2005). 3. Self-sustaining metabolic loops, in which components of the cycle, like metabolites and proteins, are transmitted to daughter cells (Smits et al. 2006; Zordan et al. 2006). 4. Small heritable RNAs that mediate and transmit patterns of gene expression. The best-known case is the posttranscriptional silencing mechanism called RNA silencing in which various regulatory RNA molecules are involved (Bernstein and Allis 2005; Siomi and Siomi 2009; Rassoulzade- gan 2011).

Collectively, the four regulatory processes described above yield a second layer of biological information, mainly responsible for editing and realizing genetic information, by switching genes on or off.These processes should not be understood as distinct from each other but as constantly interacting. Together these regulatory processes constitute the complex web called the epigenome. To visualize the size of the epigenome and thus its structural complexity, consider the following characteristics: in humans each of the approximately two hundred cell types has its own epigenome, which

▸ 25 HOW EPIGENETICS DEALS WITH BIOLOGICAL COMPLEXITY

FIG. 1.3. DNA methylation and histone modifications. Gene expression is repressed by enzymes placing methyl groups on cytosines of the DNA (left). Enzymes change the way the DNA is wrapped around histones by attaching or removing various chemical groups to or from histone tails (center). This affects whether certain genes can be expressed (M. Baker 2010, 181; reprinted with permission from Macmillan Publishers Ltd).

changes a number of times during development, especially in response to environmental cues. The so-called DNA methylom (i.e., all DNA methyla- tions) alone has approximately 27 million methylation locations (see Lister et al. 2009). Understanding this new dimension of biological complexity—epigenetic complexity—necessitates revealing the molecular basis of the ramified causal network of developing organisms, both structurally and dynamically. This includes investigating epigenetic regulatory systems both from a developmental perspective, in fields like cell differentiation, reprogramming, and cancer research, and from a heredity perspective, for example, in epidemiology, ecology, or evolutionary biology. The latter research fields are primarily focused on whether and how chromatin modifications and epigenetically regulated patterns of gene expression are maintained transgenerationally in the absence of an inducing factor and changes in DNA sequences and whether this heritable variation has an impact on population dynamics (see, e.g., Jablonka and Lamb 2005, 2010; Bossdorf et al. 2008; E. Richards 2008, 2009; C. Richards et al. 2010a; Bossdorf and Zhang 2011; Bonduriansky et al. 2012; Jablonka 2013; Zhang et al. 2013; and Cortijo et al. 2014). Thus, epigenetics combines two dimension types: molecular-developmental and ecological-evolutionary.10 Both dimensions of modern epigenetics are addressed in this book, although it seems that approaches focusing on epigenetic inheritance rather than epigenetic regulation of development have reached supremacy in shaping the academic and public understanding of epigenetics in the more

26 ◂ HOW EPIGENETICS DEALS WITH BIOLOGICAL COMPLEXITY recent past. This is due to the fact that the spread of Holliday’s heredity- centered definition of epigenetics has coincided with the radical boom of epigenetics since 2000 (see the introduction). This rapid development of modern epigenetics was likely driven by the fact that around this time not only the mechanisms of gene regulation and their heredity potential were unraveled but the disappointing results of the Human Genome Project made many researchers realize that nongenetic complexity has to be considered more seriously in order to understand the development of traits and their transmission. With respect to epigenetic inheritance, Jablonka and Lamb in particular have emphasized that epigenetic inheritance resembles Lamarckian inheritance, as, for example, the title of their pathbreaking work Epigenetic Inheritance and Evolution: The Lamarckian Dimension (1995) indicates. In the years since, they have argued repeatedly that because of the dual nature of complex epigenetic systems as developmental and inheritance systems, inducible and heritable epigenetic variation resembles the concept of inheritance of acquired characteristics or, as Mayr (1980) calls it, “soft inheritance” (see also Jablonka et al. 1998; Jablonka and Lamb 2005, 2007, 2008; and Gissis and Jablonka 2011). This claim has led to a lively historical debate on how one should understand the heritability of developmental complexity, a concept that was black-boxed for so long by evolutionary theory. To address this issue, let us have a closer look at how epigenetic inheritance between cells and organisms works and, more generally, how we can interpret the role played by the suggested Lamarckian dimension of epigenetic complexity in evolution.

IN THE SHADOW OF LAMARCK

As described above, epigenetics has always been a developmentally oriented research program that is genuinely concerned with evolutionary issues. Waddington (1942a, 1953, 1956) claimed that the genetic capability to respond, for example, to environmental influences during development is crucial for what he called “genetic assimilation.” According to this pseudo- Lamarckian process, an environmentally induced phenotypic character is, through selection and recombination, taken over by the genotype, so that finally the initial stimulus is no longer necessary to develop the new phe-

▸ 27 HOW EPIGENETICS DEALS WITH BIOLOGICAL COMPLEXITY

notype (see also Schmalhausen 1949). In other words, the novel induced phenotype is progressively more directly controlled by genes. As Brian K. Hall (1992, 118) notes, “There is nothing Lamarckian about genetic assimilation.” In fact, according to Waddington, genetic assimilation fully rests upon neo-Darwinian or Mendelian genetics: (1) upon the genetic capability of organisms to respond plastically to environmental changes, (2) upon the two conditions—the induced phenotype has greater fitness and natural selection favors fitter individuals—that produce the new phenotype via the genome (by preexisting, yet hidden or cryptic genotypic variation) rather than the idea that individuals acquire the new phenotype from the environment, and finally, (3) upon the idea that a gradual change in the frequencies of alleles influences the likelihood of producing the new phenotype.11 However, the underlying idea that phenotypic variation and the flexibility of organisms’ responses to environmental cues, respectively, are not random and thus may direct morphological evolution to a certain degree has been invoked by Waddington (1957) and subsequently by a number of biologists and philosophers of biology in order to criticize the narrow framework of the modern synthesis (Ho and Saunders 1979; Schlichting and Pigliucci 1998; West-Eberhard 2003; Jablonka and Lamb 2005; Pigliucci et al. 2006; Moczek 2007; Minelli and Fusco 2010; Bateson and Gluckman 2011; Piersma and van Gils 2011; Jablonka 2013). Rather than studying evolutionary change primarily as a change in genotype frequencies in population genetics (Wright 1922; R. Fisher 1930; Dobzhansky 1951), as done in the framework of the modern synthesis, these authors argue for reconsidering the capability of nongenetic, developmental variation to bias and guide evolutionary change. This includes not only environmentally induced changes in regulatory processes but also the physical constraints and self-organizing properties of the developing embryo. For instance, Mary Jane West-Eberhard (2003, 2005) claims that genes are probably more often followers in evolutionary change than leaders.12 Instead, it is the complex, environmentally responsive system of the developing organism that takes the lead. It introduces in a nonrandom and biased manner how new phenotypes arise in populations and are subsequently stabilized by genes. I will call this view the “development first view” or, as it is sometimes called, the “phenotype first view” of evolution. It rests on the assumption that the dynamic regulatory flexibility and plastic responsiveness of com-

28 ◂ HOW EPIGENETICS DEALS WITH BIOLOGICAL COMPLEXITY plex developing systems carry explanatory weight in evolutionary theory. This view has been accompanied by a common call for a so-called “extended (evolutionary) synthesis” (Jablonka and Lamb 2005; Pigliucci and Müller 2010b) or, as some have even termed it, a “post-Darwinian synthesis” (Huang 2012a).13 Such a call for theoretical expansion or even revolution is not new in evolutionary biology. As Ulrich Kutschera and Karl Niklas (2004, 273) emphasize, “The ‘evolution’ of evolutionary theory remains as vibrant and robust today as it ever was.” In fact, in the twentieth century the modern synthesis was targeted by a number of critics. For example, the exclusion of developmental biology from the theoretical framework of the modern synthesis (Harrison 1937; De Beer 1954; Hamburger 1980), the mode and speed at which evolutionary change takes place (Eldredge and Gould 1972; Gould and Eldredge 1977), the explanatory autocracy of the adaptationist program in evolutionary biology (Gould and Lewontin 1979), and the omission of a satisfying theory of evolutionary novelty (Goldschmidt 1940) have been criticized. Accordingly, the modern synthesis has been blamed for constraining the direction of research in evolutionary biology (Provine 1989). As we can see, attacking the modern synthesis is anything but novel. This raises a crucial question: What distinguishes the recent critical movement from older criticism, and particularly, what makes epigenetics a distinctive critical and novel approach to opposing the modern synthesis? Eva Jablonka and Marion Lamb would answer that epigenetics reintroduces long- neglected Lamarckian concepts into evolutionary theorizing. This claim needs clarification, since, as described above, the Waddingtonian concept of genetic assimilation—once a cornerstone of epigenetics—can easily be reconciled with neo-Darwinian views of heredity and change of allelic frequencies in populations. In modern epigenetics, however, the underlying idea of developmental directionality in evolution is significantly expanded by Lamarckian views.

LAMARCKISM AND EPIGENETIC INHERITANCE

In his Philosophie Zoologique, Jean-Baptiste de Lamarck (1809) famously defended an account of evolutionary change through inheritance of acquired characteristics.14 According to this model (which is only one of many elements of Lamarck’s comprehensive account of evolution), organisms, influenced by their environment, change their plastic physiology

▸ 29 HOW EPIGENETICS DEALS WITH BIOLOGICAL COMPLEXITY

through use and disuse of respective traits and pass on these changes to subsequent generations. Darwin (1859) refers to this model of use and disuse in his Origin of Species in order to explain the source of variation. Moreover, he speculated that if body tissues increase in size or change in novel environmental conditions, they produce more or modified “gemmules,” which circulate through the body and penetrate germ cells (Darwin 1868). Thus, the acquired features of a certain body part were inherited. By the end of the nineteenth century, this idea of soft inheritance had come under attack, particularly through the efforts of August Weismann (1892). He argued that the germ line was spatially separated from the somatic lineages of cells and that only information in the germ line figures into heredity.15 By showing that cutting off the tails of mice for several generations does not lead to a tailless new race, he proposed that anything acquired by an individual during development in only somatic cells could not influence heredity. As James Griesemer (2002, 98) puts it, “The body or phenotype is a causal dead end.” Later this Weismannian heredity model and the separation between germ line and soma became a cornerstone of the modern synthesis. Subsequently, a molecularized version of this separation model, best known as Francis Crick’s (1958, 1970) “central dogma,” was developed. According to the central dogma, there is only a one-way infor-

FIG. 1.4. Causal pathways of inheritance and development according to molecular Weismannism. One genome (G) causally influences another genome transgenerationally, and G also causally influences the development of the organism, that is, its somatic cell lineages (S). Lamarckian influences across generations through causal pathways from organism to genome (S → G) or from organism to organism directly (S → S) are rendered impossible (after Maynard Smith 1966, 67).

30 ◂ HOW EPIGENETICS DEALS WITH BIOLOGICAL COMPLEXITY mation flow from DNA to protein. In Dawkins’s (1976, 1978, 1982) gene’s-eye view of evolution, the movement is from replicator to vehicle. A simplified representation of this so-called “molecular Weismannism” is shown in figure 1.4. The separation of germ line from soma not only rejects the idea of soft inheritance but restricts the explanatorily relevant causal pathways of heredity in the theory of evolution. Methodologically speaking, the Weis- mannian model of heredity may be understood as an attempt to deal with the complexity of inheritance phenomena by excluding a set of developmental causal factors. For example, the model is confined to germ cell formation and separation in taxa such as mammals and insects, while many other animal and all plant taxa lack a segregated germ line or their germ line separates late in development. Moreover, it ignores various examples of soft inheritance, which were reported well into the twentieth century (Kam- merer 1906, 1924; Plate 1931).16 Today, epigenetics once again opens this box of hidden complexity by reassessing Lamarckian pathways of inheritance without adopting Lamarck’s principle of “use and disuse.”

FIG. 1.5. Causal pathways of epigenetic inheritance in multicellular organisms. Germ-line inheritance (a) involves a single-cell “bottleneck,” that is, a gamete or spore in sexually or asexually reproducing organisms. Somatic inheritance (b) includes transmission through the effects of developmental interactions between offspring and/or between parent and offspring, like parental (often maternal) behavior and social learning. Epigenetic variation can be induced directly in the F0 generation in both soma and germ cells (arrows not depicted). In so-called “somatic induction” an induced change in the soma affects the germ line in F0 (see arrow soma → germ) and is then inherited through the germ line. Thus, in the next generations the new phenotype of F0 is developed (after Jablonka and Raz 2009, 133).

▸ 31 HOW EPIGENETICS DEALS WITH BIOLOGICAL COMPLEXITY

As outlined above, various phenotypic traits can, in fact, be plastically modified through environmental induction and subsequently passed on over generations through a number of channels of epigenetic inheritance, both in asexual clonal lineages and sexually reproducing organisms (S. Gil- bert and Epel 2009; Jablonka and Raz 2009; Kappeler and Meaney 2010; Shea et al. 2011; Ledón-Rettig et al. 2012; Aiken and Ozanne 2014; Wang et al. 2017). In contrast to the concept of genetic assimilation, transgenerational epigenetic inheritance does not necessarily require environmental induction over several generations. Rather, novel DNA methylation patterns and other chromatin modifications can be induced by the early life (intrauterine or postnatal) environment. This may lead to new phenotypes that can be passed on to future generations. These epigenetic modifications are transmitted through both germ-line and somatic inheritance (fig. 1.5). Usually the faithfulness of transmissions of epigenetic marks and thus the stability of induced phenotypic traits across generations is lower in soma-to-soma inheritance, since chromatin modifications of both male and female germ lines are (more or less completely) reset during germ cell development and gonadal sex determination, respectively (Lange and Schneider

2010; Jablonka 2013). Accordingly, novel “acquired” phenotypes of F0 that are

transmitted through somatic inheritance are usually reset in generation F1, given that they are not environmentally induced in every generation anew. In contrast, transmission of epigenetic variations that are directly induced in germ-line cells can avoid this problem, because these variants are maintained by the organisms’ mitotic machinery.17 Thus, they show more stable patterns of inheritance. Moreover, induced epigenetic variation in somatic cells can be transmitted, for example, through small RNAs to the germ line, where epigenetic marks and thus gene activity are changed (see fig. 1.5). As a consequence, the new phenotype is inherited faithfully across several generations. Germ-line transmission following “somatic induction” resembles best the Lamarckian idea that acquired characteristics can be inherited. Below is a list of examples of epigenetic inheritance through soma-to-soma transmission and germ-line transmission.18

SOMATIC INHERITANCE • Rats receiving extensive maternal care (i.e., licking and grooming; fig. 1.6A) develop a more stress-resistant phenotype. Changes in stress resistance are due to induced variation in methylation patterns in gluco-

32 ◂ HOW EPIGENETICS DEALS WITH BIOLOGICAL COMPLEXITY

corticoid receptor genes in the hippocampus. Females receiving more maternal care develop into mothers who give their offspring more care, thus behaviorally stabilizing the variation in gene expression transgenerationally (Meaney 2001; Weaver et al. 2006; Kappeler and Meaney 2010). In addition, the level of maternal care modifies expression of the sex hormone estrogen of the female offspring and thus their sexual behavior (Champagne et al. 2003). • Exposing mice to an “enriched” environment (i.e., novel objects, a high number of social interactions, and voluntary exercise) for two weeks during development enhances synaptic plasticity and memory formation not only in these mice but in their offspring, too. This is most likely a case of in utero transmission. In addition, in offspring of mutant parents with a knock-out causing a reduced learning and memory capacity, this inherited genetic “bias” is epigenetically masked or buffered in early development, if the parents are exposed to the “enriched” environment (Arai et al. 2009; Arai and Feig 2011). • In litter-bearing mammals, adult behavior is influenced by variations in prenatal hormonal exposure due to the position of the fetuses in utero. For example, female fetuses located between two males tend to show masculinized traits, including persistently higher levels of testosterone and aggressiveness, in contrast to those females that develop without adjacent males. This intrauterine effect is due to a transmission of sex steroids between male and female fetuses. In addition, this effect seems to be self-perpetuating in future generations, because “masculinized” adult females more likely produce more male pups than female pups, which increases the likelihood of a similar intrauterine environment (Ryan and Vandenbergh 2002).

GERM-LINE INHERITANCE • If pregnant rats are exposed to endocrine disruptors (i.e., environmental chemicals, widely distributed in humans’ environment, that disrupt hormone production and function), such as the fungicide vinclozolin, their male offspring exhibit abnormalities in their testes and immune systems for at least four generations. In addition, differences in the

methylation patterns in several genes are found in generation F3, although only the parental generation was exposed to vinclozolin. This

▸ 33 HOW EPIGENETICS DEALS WITH BIOLOGICAL COMPLEXITY

suggests that these epigenetic variations are transmitted through the germ line (Anway et al. 2005; Guerrero-Bosagna et al. 2010; Skinner et al. 2013). In addition, maternal exposure to vinclozolin can also induce changes in methylation patterns in rats, leading to differences in anxi- etylike behaviors in offspring (Skinner et al. 2008). Moreover, females show aversion to males whose great-grandmothers were exposed to vinclozolin (Anway et al. 2005; Crews et al. 2007). Studies on the effects of the endocrine disruptor bisphenol A in rats show similar transgenerational impairment of spermatogenesis and aberrant DNA methylation in testes (Salian et al. 2009; Doshi et al. 2011). • Early-life stress (i.e., repeated and unpredictable maternal separation) induces depressivelike behavior and altered response to novel environments in rats and in their offspring, although the offspring are reared normally. This behavioral modification is most likely due to alterations in DNA methylation patterns of candidate genes for depressive behavior in the germ line (Franklin et al. 2010). • Maternal high-fat diet exposure in mice induces an increase in body

size and reduces insulin sensitivity. This new phenotype persists in F3 females. It is transmitted only via the paternal lineage, likely through inherited alterations in gene expression patterns (Dunn and Bale 2011). • Epigenetic variation in the germ line is also induced by genetic mutations and is inherited independent of these mutations. If male mice have mutations in genes coding for genetic programming (i.e., for DNA methyltransferase and a chromatin remodeler protein), this results in epigenetically mediated phenotypic changes in the offspring (e.g., in coat color) (fig. 1.6B). This variation is passed on through the paternal germ line, although the mutant allele is not inherited (Chong et al. 2007). • Injecting double-stranded RNA into the nematode Caenorhabditis elegans, which contains the homolog gene ceh13, results in this gene being silenced and thus a small and dumpy phenotype. This novel phenotype of the worm can be inherited for more than forty generations. Silencing and inheritance are most likely mediated by small RNAs that change chromatin organization (Vastenhouw et al. 2006; see also Wang et al. 2017). Small RNAs are also involved in the transmission of antiviral response in C. elegans. If the soma of these worms is infected with a virus, then small, interfering RNAs silence the viral genome. This resistance

34 ◂ HOW EPIGENETICS DEALS WITH BIOLOGICAL COMPLEXITY

persists for fifty generations, in the absence of exposure, by small RNAs being transmitted through the germ line (Rechavi et al. 2011). It seems that a “transgenerational timer” based on a feedback loop (in which small RNAs are involved) controls the duration of this transgenerational effect in C. elegans (Houri-Ze’evi et al. 2016). • In the fruit fly Drosophila melanogaster, eye transformations (i.e., ecto- pic outgrowth in eyes) can be induced by geldanamycin treatment. If, for example, flies of a strain that is susceptible to eye transformation by epigenetic changes in chromatin are exposed to geldanamycin, this new trait is transmitted to at least thirteen generations in the absence of the drug (Sollars et al. 2003; Ruden et al. 2005).

Recently, these and other findings have allowed various biologists, historians and philosophers of biology, and popular science authors to juxtapose epigenetic inheritance with Lamarckian inheritance of acquired characteristics or soft inheritance (see, e.g., Jablonka and Lamb 1989, 1995, 2005, 2007, 2008; Jablonka et al. 1992, 1998; Por 2006; E. Richards 2006; Koonin and Wolf 2009; Martienssen 2010; Sano 2010; Gissis and Jablonka 2011; Hall 2012; Wei et al. 2014; and Wang et al. 2017).19 A prominent study widely

A ▸ ◂ B

FIG. 1.6. Examples of induced phenotypes that are transmitted through somatic or germ-line inheritance. (A) Lactating female rat licking and grooming one pup while nursing litter. (B) Isogenic mice after germ-line transmission of epigenetic variations show different coat-color phenotypes, yellow left( ), mottled (middle), and agouti (right) (A: Meaney and Szyf 2005, 105; B: Chong et al. 2007, 616; A: reproduced with permission from Les Laboratories Servier [Suresnes, France] from Dialogues in Clin- ical Neuroscience; B: reproduced with permission from Macmillan Publishers Ltd.).

▸ 35 HOW EPIGENETICS DEALS WITH BIOLOGICAL COMPLEXITY

discussed, especially by those flirting with the ghost of Lamarck, shows that

mice in F1 and F2, whose F0 fathers have gone through a painful experience associated with a certain odor, exhibit fearful behavior to this odor, even though they had never encountered it before (Dias and Ressler 2014; see also Szyf 2014; and Welberg 2014). Other authors have cast doubt on the adequacy of describing epigenetics as a Lamarckian enterprise (Hall 1998; Dawkins 2004; West-Eberhard 2007; Haig 2007, 2011; Coyne 2011a, 2011b; Dickins and Rahman 2012; Deichmann 2016b). For example, Pigliucci and Finkelman (2014) have argued that modern epigenetics does not offer a Lamarckian theory of evolution, despite presenting Lamarckian processes of inheritance. They emphasize that for epigeneticists the unit of selection still remains the population and not, as Lamarckism claims, the individual. In this book, the “Lamarckian dimension” of modern epigenetics is understood as a loose reference to a cluster of complex explanandum phenomena that cannot be easily grasped by the Mendelian and Weismannian tenets of the modern synthesis.20 These phenomena are based on phenotypic plasticity, and they bias the direction of evolutionary change through nongenetic regulatory agents and complex nongenetic systems of inheritance. Accordingly, they specifically differ from the neo-Darwinian concept of genetic assimilation, which has been a conceptual cornerstone of classical epigenetics. This means that by considering epigenetic complexity, phenotypic plasticity can now be stretched temporally, across mitoses and especially mei- oses. Genes no longer have to take plasticity by the hand. That is, for phenotypic variation to occur, it no longer has to be environmentally induced in every generation anew, thanks to epigenetic germ-line inheritance. What is more, developmental directionality can occur in the complete absence of genetic assimilation—for example, in the cases of self-perpetuating parental effects and somatic inheritance, respectively. In other words, epigenetics uncovers a special, previously neglected kind of biological complexity underlying the general idea of evolution being guided by development.

OTHER “DEVELOPMENT FIRST VIEWS”

The Lamarckian touch of epigenetics introduces an interesting historical element into the theoretical framework of a so-called extended evolutionary synthesis and its “development first view” on evolution. From a super-

36 ◂ HOW EPIGENETICS DEALS WITH BIOLOGICAL COMPLEXITY ficial point of view, expanding the modern synthesis by epigenetics means expanding evolutionary theory by including Lamarckian ideas. However, if we adopt the more nuanced understanding of “epigenetic Lamarckism” described above—a view of directing epigenetic complexity in evolution— one can demarcate epigenetics not only from neo-Darwinian orthodoxy and classical epigenetics but also from other related biological fields, like evolutionary developmental biology (known as evo-devo) and developmental niche construction, which are also challenging the theoretical framework of the modern synthesis.21 Like epigenetics, evo-devo emphasizes the neglected role of developmental biology and embryology in the theory of evolution (Love 2003; Amund- son 2005; Laubichler 2007, 2010; Sansom and Brandon 2007; Müller 2007; Minelli 2009, 2010; Pigliucci and Müller 2010b). Although this field also has a focus on the complexity of developmental systems, especially their evolvability (Kirschner and Gerhart 1998; Hendrikse et al. 2007), it has not yet developed a theory of heredity distinct from that of the modern synthesis.22 As a first consequence, evo-devoists favor a far more “gene-centric” view of developmental complexity—one that resembles Waddington’s interest in gene networks, like the Hox gene cluster. This includes issues like how genes jointly regulate modifications of developmental processes (usually associated with major shifts in morphological “body plans”) and produce evolutionary novelty. A second consequence is that epigenetic complexity is first and foremost investigated as a case of intragenerational phenotypic plasticity, not as a heredity phenomenon (see, e.g., Müller 2010). In contrast, in niche construction theory the intergenerational heredity perspective of epigenetic complexity is recognized, although with a different emphasis than in epigenetics. These theorists seek to understand the self-perpetuating and reciprocal effects of organisms that construct their own niche (and/or that of other species) during development (Lewontin 1982; Sterelny 2001; R. Day et al. 2003; Odling-Smee et al. 2003; Laland and Sterelny 2006; Laland et al. 2009, 2011; Odling-Smee 2010).23 A common example is beavers: they shape their ecosystem and thus alter the selection pressure acting on them and their offspring. Such cases include processes like parental effects, social learning, and even symbolic communication— processes that often include epigenetic soma-to-soma transmission. How- ever, the emphasis in niche construction theory is on the explanatory relevance and centrality of ontogenetic agency.24 This means, with respect to

▸ 37 HOW EPIGENETICS DEALS WITH BIOLOGICAL COMPLEXITY

epigenetics, that it focuses on the acting organism as an inducer and mediator of complex epigenetic systems rather than on the molecular system itself. In short, there are related accounts on the market that adopt a “development first view” of evolution and/or its accompanying concepts, such as phenotypic plasticity, canalization, or genetic assimilation. Additionally, these accounts (at least in part) draw on what I have called epigenetic complexity. However, they focus on different aspects of regulatory complexity, and thus its conceptualization differs from that of epigenetics. Some (i.e., evo-devoists) do not consider epigenetic complexity to be a heredity phenomenon and highlight its intragenerational effects on phenotypic plasticity; others (niche construction theorists) understand these systems as a minor component of a broader concept of nongenetic inheritance, which widens the scope of organisms’ actions in shaping their environment. This distinction of contemporary developmentally oriented theories commonly calling for an expansion of the prevailing framework of the modern synthesis shows that epigenetics has developed a novel research program, focusing on particular explananda. These explananda often carry with them a “Lamarckian dimension.” However, in order to fully understand what constitutes the special theoretical structure of epigenetics, as well as the possible theoretical shift in modern biology imposed, in large part, by epigenetics, we cannot focus only on historically neglected “Lamarck- ian” phenomena-to-be-explained. To put it differently, we cannot simply ask questions like, To what extent does epigenetic germ-line transmission resemble the traditional concept of inheritance of acquired characteristics? or, Why has “Lamarckian” epigenetic inheritance flown under the radar for so long? I suggest that to fully understand the “nature” and potential novelty of epigenetics we have to draw attention to the way in which the relationship between explananda and explanantia is established in this field. In other words, we have to turn the spotlight on the structure of epigenetic explanations. This includes asking other questions: Do contemporary epigeneticists explain differently than other biologists? How do they model what one might call “epigenetic dependencies”—that is, dependencies usually between intra- or extraorganismic environmental cues and heritable regulatory factors—in light of the complexity and changeability of the genotype-

38 ◂ HOW EPIGENETICS DEALS WITH BIOLOGICAL COMPLEXITY phenotype map? According to which explanatory standards is such an epigenetic dependency relation across generations, for example, considered to be explanatorily relevant? Does the introduction of developmental epigenetic complexity as explanans in evolutionary explanations reshape (or even distort) evolutionary theory? Have epigeneticists developed special methodologies to solve general problems of complexity sciences, like the representation problem of complexity or the Borges problem? And finally, can the conceptual toolbox of philosophers of science make sense of the structure of epigenetic explanation and of the causal roles epigenetic factors play in development and heredity? Answering these questions is less a historical than a philosophical project—one that steps out from the shadow of Lamarck and the prevailing Lamarckism debate. This project will be under- taken in the chapters that follow. ▶ ▶ ▶

Biologists explain complex living systems at higher levels of organization. Accompanied by the advent of philosophy of biology, these explanations have become subject to non-nomological accounts of explanation. How- ever, both biologists and philosophers of biology who seek to conceptualize and explain structurally and dynamically complex living systems face at least two major challenges. First, complexity hinders representing a living system’s overall behavior, even when its parts and their interrelations are fully and completely known (the “representation problem”). Second, our knowledge of a complex system is not necessarily increased if we seek to grasp it in its entirety (the “Borges problem”). These two problems have been intensified by epigenetics, since this field seeks to investigate molecular complexity in an expanded manner, both with respect to the number of components and their relations studied (structural complexity), as well as with respect to the nonlinearity of these relations and thus the spontaneous behavior of the system (dynamic complexity). This new layer of biological complexity is labeled “epigenetic complexity.” This trend can be described historically as one in which Waddington’s classical systemic perspective on living systems and his interest in developmental plasticity and robustness, as well as in their roles in evolution, have been further developed by contemporary epigenetics. Most important, in

▸ 39 HOW EPIGENETICS DEALS WITH BIOLOGICAL COMPLEXITY

modern epigenetics Waddington’s focus on genetic (or genomic) complexity has shifted toward a research interest on the complex interplay between nongenetic factors and their role in plastic development and heredity. This heredity dimension of epigenetic complexity has been discussed intensively, because complex epigenetic relations across cell divisions, and especially between organisms, carry with them a certain “Lamarckian dimension.”

40 ◂ TWO CHALLENGES OF EPIGENETICS IN LIGHT OF THE EXTENDED EVOLUTIONARY SYNTHESIS

In the early days of molecular biology, it was an evangelical movement. Most people were against us. Most of the bio- chemists didn’t understand the nature of the problems that we thought were interesting and important. They had a completely different set of attitudes. SYDNEY BRENNER, 1988

There is not the slightest evidence that the findings of epigenetics will dispel the main ideas of neo-Darwinism. JERRY COYNE, 2011

I now groan audibly when a journalist (usually from conti- nental Europe, where they spend too much time learning philosophy rather than science) asks me the now inevitable “what about epigenetics?” question. . . . I am heartily sick of the “epigenetics” bandwagon and almost look forward to the next one, whatever it turns out to be. RICHARD DAWKINS, 2011 CHALLENGES OF EPIGENETICS

NOT EVERYONE HAS WELCOMED the recent rise of epigenetics. In fact, some evolutionary biologists doubt whether epigeneticists should have a say in evolutionary biology. This reluctance may be due in part to sen- sational reporting of new epigenetic research results and simplistic public debates about “Lamarck’s return” and “Darwin’s death.” Besides these issues, however, there are more serious challenges currently hindering theoretical integration of epigenetics. In this chapter, I discuss two kinds of tensions that arise due to the explananda and explanantia chosen by epigeneticists and due to the methodological strategies chosen to investigate the relationships between these explanantia and explananda. First, I give an overview on some challenges the developmentally oriented view of epigenetics poses for the field of evolutionary biology or, more specifically, for orthodox neo-Darwinian explanation. This includes the issues of whether (and how) molecular epigenetic explanations that are based on highly artificial experimental setups and that focus on developmental proximate causes rather than on ultimate causes can address evolutionary explananda. The latter problem in particular becomes apparent in the recent debate about a so-called “extended evolutionary synthesis.” Second, in addition to the tensions between epigenetics and other disciplines—above all evolutionary biology—we also find intradisciplinary tensions concerning how to understand the field of epigenetics conceptually or methodologically, as well as how (and what) epigeneticists should explain. These latter issues display epigenetics’ status as a young and developing research field. Its inner tension and heterogeneity are reconstructed by means of a historical analysis of approaches using Waddington’s epigenetic landscape images as a methodological tool to mathematically model the role of epigenetic determinants in cell differentiation and reprogramming.

TOWARD AN EXTENDED EVOLUTIONARY SYNTHESIS

Since the 1990s, an increasing number of epigenetic factors and processes have been revealed. Due to their importance for research in both development and heredity they have often been juxtaposed with the somewhat antiquated concept of inheritance of acquired characteristics. As described in the previous chapter, this Lamarckian claim has recently been accom-

42 ◂ CHALLENGES OF EPIGENETICS panied by a call for an “extended evolutionary synthesis” or even a “post- Darwinian synthesis” that includes not only epigenetics (West-Eberhard 2003; Kutschera and Niklas 2004; Jablonka and Lamb 2005, 2007, 2010; Jablonka 2006; Pigliucci 2007, 2009; Cabej 2008; Gissis and Jablonka 2011; Mesoudi et al. 2013; Laland et al. 2014, 2015) but also related fields such as evo-devo (Johnson and Porter 2001; Müller 2007; S. Gilbert and Epel 2009; Kirschner and Gerhart 2010; Laubichler 2010; Müller 2010; Newman 2010; Laland et al. 2014, 2015) and niche construction theory (Odling-Smee et al. 2003; Laland et al 2009, 2014, 2015; Danchin et al. 2011; Piersma and van Gils 2011; Mesoudi et al. 2013).1 This novel extended theory is thought to be based on a “development first view” on evolution, which emphasizes the evolutionary significance of developmental responsiveness. In addition, it should help to overcome gene-centrism and relegate genes to the role of “followers” in evolutionary processes (see, e.g., Pigliucci and Müller 2010a, 14). The leading role is now allocated to developmental interactions, their inheritance, and the mobilization of phenotypic traits through the plastic responses of environmentally sensitive and adaptive developmental systems. Accordingly, genes are mainly responsible for routinization and progressive fixation of traits in evolution. The various developmentally oriented approaches calling for an expansion of evolutionary theory have been challenged in at least three ways. First, particularly orthodox-minded population geneticists have doubted whether novel concepts, like evolvability and robustness in evo-devo, actually offer anything substantially new compared to the conceptual framework of the modern synthesis (Lynch 2007; Coyne quoted in both Pennisi 2008 and Whitfield 2008; Welch 2017).2 Second, it has been claimed that the current “developmental state” of epigenetics and evo-devo does not yet necessitate a theoretical shift in population genetics–driven evolutionary biology, because, for example, evidence has to be strengthened (Pál and Hurst 2004; Coyne 2011b; Futuyma 2011; Wray et al. 2014; Deichmann 2016a). Finally, Lindsay Craig (2010) has argued that since the new concepts of an extended evolutionary synthesis (especially those of evo-devo) pose insurmountable conceptual challenges for theoretical population genetics, this new framework is not an extension of the theory of evolution, strictly speaking.3 In order to face such arguments against the theoretical integration of the “development first view” into the prevailing framework of evolutionary

▸ 43 CHALLENGES OF EPIGENETICS

biology, a number of developmentally minded (evolutionary) biologists and philosophers of biology have called for this novel view to be supported by a new methodological and explanatory framework that revises traditional disciplinary boundaries—especially those between molecular and developmental biology on the one hand and evolutionary biology on the other. This framework, however, has yet to be developed. With respect to epigenetics, it faces at least two major challenges. The first problem concerns how to interrelate explanations about dependency relations on the molecular level, which are usually brought about in highly artificial experimental situations, with statistical models about natural population dynamics that make inferences from observational data. The second problem is whether to accept explanations in evolutionary biology that are based on developmental proximate causes rather than ultimate causes as explanans variables. Both problems refer to the structure and appropriateness of epigenetic explanations in evolutionary contexts. I now address each of them in detail.

EPIGENETIC ARTIFICIALITY AND EVOLUTIONARY RELEVANCE

In 2010, the evolutionary psychologist Michael Badcock wrote an article in Psychology Today entitled “Epigenetics: It’s Nonsense without Evolu- tion!” This title nicely summarizes a common concern about contemporary research in epigenetics. Despite the vast amount of new data and the number of epigenetic heredity models coming from manipulationist experiments in molecular biology, some critics ask for an evolutionary story to be told that addresses why (in the adaptationist sense of “what for”) epigenetic regulatory systems regulate development the way they do.4 Others doubt that transgenerational transfer of epigenetic information does have an impact on natural population dynamics and is therefore relevant for evolutionary biology. The argument for the latter concern relies on the fact that, contrary to genetic variation, body-to-body transfer of epigenetic variation shows relatively limited heritability due to unstable states and high mutation rates (J. Walsh 1996; Hall 1998; Dawkins 2004; Pál and Hurst 2004; Coyne 2011b; Deichmann 2016a).5 To address these critics, additional research efforts stretching out to ecologically relevant circumstances have to be made. Such observational studies are crucial

44 ◂ CHALLENGES OF EPIGENETICS in order to estimate how stable or invariant transgenerational epigenetic effects are in natural populations. It is at this point that epigenetics and evolutionary biology face a major explanatory and methodological challenge. If epigenetics wishes to become a discipline involved in evolutionary biology, it has to be able to interrelate molecular biologists’ causal explanations with the statistical explanations of ecologists and evolutionary biologists (see C. Richards et al. 2010a; Griesemer 2011a, 2011c; and Latzel et al. 2013). This problem has to be solved in order to link observational studies in the emerging field of ecological or evolutionary epigenetics (Bossdorf et al. 2008; C. Richards et al. 2010b; Pennisi 2013) to causal reasoning in experimental molecular epigenetics. In the philosophical literature on causation, the molecular biologists’ view on causal explanation is commonly treated as a mechanistic view of causation (see chap. 3). According to such an account, explanation is achieved when one shows how a phenomenon is brought about by an underlying mechanism. Statistically relevant correlations alone are considered to be insufficient for a causal explanation. In contrast, ecologists (and epidemiologists) develop statistical models in order to make inferences from observational data (see Shipley 2000; and Royle and Dorazio 2008). In addition, many ecological mechanisms are still not well known (see Raerinne 2011). This asymmetry between the two fields is reinforced by the nature or genesis of the epigenetic dependencies that are investigated. Biologists are usually interested in preserving the natural properties of living systems as effectively as possible. In other words, biological explanation requires small “surgical” interventions and not, as the philosopher James Woodward (2003, 284) calls them, “extreme circumstances.” Does molecular epigenetics sat- isfy this general explanatory rule? The complexity and openness of epigenetic inheritance systems necessitate an “extreme” level of control, and thus the system investigated has to be artificially manipulated in order to address issues like developmental plasticity and environmental responsiveness. For example, all genetic variation has to be fixed while tweaking epigenetic processes so as to preclude the causal influence of any genes on the observed (transgenerational) phenotypic effects. In addition, environmental variables as well as other developmental variables have to be controlled for. Experimental studies on RNA interference (RNAi) in C. elegans and mice are illustrative examples of this high level of artificiality.6 For instance, Minoo Rassoulzadegan and colleagues (2006) demonstrate that

▸ 45 CHALLENGES OF EPIGENETICS

the micro-injection of small RNA molecules into fertilized eggs of genetically modified mice causes a heritable epigenetic modification in male mice: white tail-tips and paws. Nadine Vastenhouw and colleagues (2006) showed that in a mutant strain of C. elegans with a gene ( gfp) that expresses green fluorescent protein, a new transgenerational phenotype (that does not glow green) is established if these nematodes are fed genetically modified bacteria that express double-stranded RNA homologous to thegfp - DNA sequence.7 Thus, it seems that epigeneticists’ reliance on the power of multifactorial manipulation makes them more likely to commit an explanatory oddity described by Kenneth Waters (2007, 576): “Biologists[’] interests in manipulating causal processes in organisms sometimes motivates them to seek causal generalizations that hold under conditions that could be artificially brought about in organisms and their environment.” Against this backdrop, explanations of experimental epigeneticists are of no relevance for evolutionary biology if they are not interested in whether their generalizations hold under less artificial circumstances. If they simply presuppose that their generalizations are relevant for evolutionary phenomena studied outside the lab, their explanations are likely to distort the theory of evolution. Somewhat related is the critique that, due to small sample sizes, the results of artificially induced epigenetic dependencies are often too good to be true. Gregory Francis (2014) holds this view with respect to a widely discussed study by Brian Dias and Kerry Ressler (2014), which shows that when male mice are conditioned to fear a particular odor, this fearful behavior can be transmitted for two generations.8 According to the above, we have to ask how to relate molecular epigenetic and evolutionary models—the translation problem—and thus how to test whether the dependency relations observed in artificial epigenetic systems occur under natural conditions as well—the artificiality problem.9 Let us first focus on the latter problem before turning to the former. The loss of similarity between natural and artificial epigenetic systems in epigenetics plays into the hands of those critics who claim that epigenetic variation probably has no impact on natural populations (see, e.g., Pál and Hurst 2004). According to this argument, molecular epigeneticists use invalid counterfactuals, or, more precisely, they use generalizations holding solely in counterfactual situations. It is not assumed that such counterfactuals are too vague to lead to an appropriate experimental design (since they

46 ◂ CHALLENGES OF EPIGENETICS are in fact able to do so; see below), nor is it usually claimed that generalizations on causal relationships in molecular epigenetics are not stable. Rather, it is argued that the artificial situations going along with epigenetic causal reasoning are not factual in nature, as there is no real-world counterpart for them. In fact, with regard to the dependency between micro-injection of RNAs and the heritable epigenetic modification, a paramutation, traced in mice by Rassoulzadegan et al. (2006, 473; emphasis added), the paper’s authors themselves had to admit that “the initial event inducing paramutation is not known.” In other words, there is no “real environment” value known that the cause variable could take in a model. However, some phenotypic traits caused by natural heritable epigenetic variation are already known (see, e.g., Bossdorf et al. 2008; C. Richards et al. 2010b; Cortijo et al. 2014). To gain more knowledge of “real-world” causal processes underlying natural epigenetic variation and in order to estimate the evolutionary relevance of epigenetic inheritance, it is important that the population or ecological level of epigenetic phenomena is made accessible to causal reasoning, too.

OUT OF THE LAB— MODELING IN EVOLUTIONARY EPIGENETICS

How frequently are epigenetic variants environmentally induced and how often are they transmitted between generations? What impact does this transmission have on natural population dynamics? To answer these essential evolutionary questions, the subfield of evolutionary epigenetics—also called population epigenetics or ecological epigenetics—has emerged. It includes different lab and field study techniques, like studies conducted under controlled conditions—including greenhouse studies (Bossdorf et al. 2008; Pennisi 2013) and so-called epiRILs studies (Reinders et al. 2009; Johannes et al. 2009; Cortijo et al. 2014)—and purely observational studies on natural populations (Herrera and Bazaga 2011; see also C. Richards et al. 2010b; and Bossdorf and Zhang 2011).10 However, only the latter approach, also known as natural experiment, addresses these questions adequately, since it is the only method that avoids manipulation of the population under study. In contrast to epigeneticists working in the lab, ecologists performing natural experiments use nonexperimental methods to infer causation from statistical relevance. These methodologies seem to follow Leonardo da Vin-

▸ 47 CHALLENGES OF EPIGENETICS

ci’s (McCurdy 1906, 54) advice: “There is no result in nature without a cause; understand the cause and you will have no need for the experiment.” The above methodological distinction leads us to this question: Do ecological and evolutionary epigenetics provide methodological strategies that allow translating their statistical models into the models of molecular epigeneticists and vice versa? I argue here that the answer to this question is yes. To illustrate this, let us have a look at what may be the first truly ecological epigenetics study, a natural experiment done by the community ecologists Carlos Herrera and Pilar Bazaga (2011). They use tools essential for building and modifying models in observational studies. First they observed a population over a long period of time and then statistically controlled for a certain variable on paper without any physical changes in the field population. The modeling technique they use is called structural equation modeling (SEM).11 It is a multivariate statistical analysis method often used in quantitative population research in ecology or evolutionary biology (see Shipley 2000; and Pugesek et al. 2003). SEM models allow translation of a set of hypothesized causal relationships between variables into a model concerning patterns of statistical dependencies. It offers methods for modeling complex systems with many assumed causal relationships between latent (not directly observed) variables. In structural equations, regression-dependent variables regress on independent variables. In other words, the dependent variables are predicted by the independent ones. Herrera and Bazaga observed that epigenetic variation in a wild population of the southern Spanish plant Viola cazorlensis, a long-lived nonmodel organism, is significantly correlated with long-term differences (over two decades) in herbivory (magnitude of browsing damage) but only weakly with herbivory-related DNA sequence variation. The correlation pattern is depicted in figure 2.1A. This study includes, first, a natural trajectory experiment that includes collecting herbivory data of a twenty-year series of annual herbivory for each of the plants in the same population and, second, a natural snapshot experiment, which means observing a final steady state of the community under study. For the latter, data on genetic and epigenetic characteristics of the plants after twenty years is compared with differences in substrate type and long-term herbivory. From the correlation pattern (fig. 2.1A) the investigators finally derived the causal thesis that natural epigenetic variation (besides variation in

48 ◂ CHALLENGES OF EPIGENETICS

A ▸

◂ B

FIG. 2.1. A simplified path diagram representing the correlation pattern observed by Herrera and Bazaga (2011) in a wild population of V. cazorlensis before and after statistical manipulation on an environmental variable. (A) The causal system before manipulation; (B) the causal system after manipulation (i.e., statistical populationi). Epigenetic variation refers to multilocus differences in DNA methylation patterns among individuals measured by using methylation-sensitive molecular markers; DNA sequence variation refers to multilocus genetic differences among individuals exclusively in those loci known to be significantly related to herbivory. The question mark indicates the assumed causal relationship in which epigeneticists are primarily interested. In diagram (B) an intervention (I) changes the value of the substrate type variable, holding it fixed at value {i} (depicted as three arrow-breaking events). The substrate type variable does not change in value in populationi (A, B: Baedke 2012, 166).

▸ 49 CHALLENGES OF EPIGENETICS

genetic and environmental variables) is likely to be an (at least partly) independent cause, which influences interactions between plants and herbivores (i.e., herbivory damage) in the field. Because the amount of empirical data and reliable theoretical models usable as a starting point for this population study was very low, Herrera and Bazaga had to build and then modify a number of models that increasingly fit the observed data. Therefore, they intervened on a part of the assumed causal structure by changing the environmental substrate type variable to a certain non-actual value {i}. The substrate type variable correlates with the measured association between the variable of special interest (i.e., epigenetic variation and herbivory damage). It takes a number of corresponding

FIG. 2.2. A simplified path diagram representing the two best-fitting causal models of Herrera and Bazaga (2011). The question mark indicates a nonrecursive (bidirectional) relation that clarifies the difference between the two causal models (each contains one of the recursive [unidirectional] relationships “epigenetic variation → herbivory damage” and “herbivory damage → epigenetic variation”). Both causal models (i.e., the bidirectional relation) could in fact be correct. Path coefficient 0.09 indicates a relatively weak causal influence (only 9 percent of relevant epigenetic variation explained by herbivory-related genetic variation) while the path coefficient 0.44 indicates that 44 percent of herbivory damage is explained by herbivory-related genetic variation (Baedke 2012, 166).

50 ◂ CHALLENGES OF EPIGENETICS values {rock, cliffs, sandy soil} among individuals in the population under study. Statistically holding this variable fixed at value {i} helps to stabilize the total effect on the potential output variable of herbivory damage. Figure 2.1B shows a path diagram of the causal system after statistical manipulation. This method of statistical manipulation can be understood best as being governed by counterfactual questions, such as “What would happen if plants did not live on their natural substrate type, like rocks?” This means creating a counterfactual, artificial population (populationi) of plants living in an environment that does not interact with them through substrate types. In a second step, Herrera and Bazaga designed a competing model situation that includes four alternative, most-probable causal models (not depicted) varying in their fit to the data. These models correspond to different theoretical positions on the potential causal role of epigenetic variation. This model specification procedure makes concessions to the fact that the theory of natural epigenetic inheritance is relatively vague and is backed up by very few empirical data. As a result, the authors changed the causal relationships between variables in the model in various ways, linking the variables through different causal routes. Each time, the model was reevaluated using standard goodness-of-fit statistics. By this modification procedure they were able to rule out a couple of nonfitting, hypothesized causal models. Unfortunately, in the end, there were two different models left with equally good fit to the data. These two models are represented together in one path diagram (fig. 2.2). As the bidirectional relationship in figure 2.2 correctly indicates, the manipulative model building and modification procedures used in this study carry with them a certain coarseness or fuzziness, since the correct causal relationship between epigenetic variation and ecologically relevant traits must remain to some degree unclear. This is likely to be a general problem that ecological epigenetics has to deal with in future natural snapshot experiments. In snapshot data sets, phenotypes measured in the field at a particular time reflect both heritable variation as well as environmentally induced phenotypic plasticity (i.e., less stable and reversible variation).12 For this reason it is difficult to decide between alternative models trying to explain the process that leads to adaptive heritable epigenetic variation. Does this mean that the association “epigenetic variation/herbivory damage” does not causally explain anything, since it offers no more information than the measured correlation of these variables? Before we address this question in more philosophical terms (in chapter 3), let me at this juncture

▸ 51 CHALLENGES OF EPIGENETICS

emphasize that the model in fact offers valuable knowledge about whether certain causal factors, like habitat parameters, are relevant for the association under study. As long as the dependency relation does not break down during changes in the values of these variables, they can be excluded from a causal explanation. In other words, the dependency relation between two variables holds in some specific causal situations or settings brought about by a statistical manipulation, while in others it does not.13 In this sense, it offers crucial information for subsequent field and lab studies.

OUT OF THE LAB—HEURISTICS

As we have seen, proving relationships among variables in observational snapshot data sets usually necessitates additional experimental approaches to tease apart different causal models and to disentangle intragenerational plasticity from heritable epigenetic variation: “A comprehensive research programme in ecological epigenetics must include molecular studies and controlled experiments, but also field studies that test whether epigenetic patterns in natural populations are consistent with theoretical predictions and the results of more controlled, but less realistic, experiments” (Bossdorf and Zhang 2011, 1573). Such a research program must be able to interlink different methods in experimental and observational studies and to develop reciprocal transparent models in ecological epigenetics and molecular biology. In addition, causal modeling in the young discipline of ecological epigenetics has major advantages in being connectable to reliable causal explanations coming from experimental studies to secure understanding of complex correlational patterns in the field. Even when the derived causal model exhibits a certain fuzziness, as in the case of the SEM mentioned above, following manipulative experiments can smoothly compensate for this handicap by continuing to uncover the correct causal relationship. Here, statistical manipulations and counterfactual scenarios in observational studies can easily be connected to a particular experimental design in additional controlled experiments. In short, they allow specifying test criteria or strategies for reciprocal transparent studies. Below are some of the heuristic research strategies relevant for epigenetics.14

1. Synchronizing lab experiments, field studies, and natural experiments. Questions like “What would happen if the plants did not live on rocks?”—

52 ◂ CHALLENGES OF EPIGENETICS

when asked in observational studies to create counterfactual statistical populations—can act as starting points to design connectable hybrid lab-field studies or common environment studies, like greenhouse studies with fixed environmental variables. These investigations can perform similar tests, though with different manipulation tools. In addition, methodological constraints in manipulative common environment approaches can have reflexive influence on natural environment studies, especially on considerations regarding which natural population or ecosystem to choose. Is it, for example, possible to preserve relevant real-world values of inducing environmental variables of a given ecosystem in a certain common environment approach? 2. Generating transferable information. Statistical manipulations in natural snapshot experiments likely lead to fuzzy causal models and new unan- swered counterfactual questions. But at this point, subsequent experiments or hybrid lab-field studies can take over by providing observational information about which parameter combinations have to be held fixed and which have to be tweaked. Thus, a common manipulation-based and counterfactual rationale leads to heuristics for further model specification and testing.15 3. Choosing (non)model organisms. Using manipulation as a common rationale also helps to settle on suitable model or nonmodel organisms used in both lab and field studies. For example, it suggests choosing short-lived organisms when testing causal relationships also under physical manipulation is feasible.

There are other methods for trying to address the topic of connecting observational and experimental epigenetic research. For example, Omri Tal, Eva Kisdi, and Eva Jablonka (2010) offer a “nonmanipulative” account of the epigenetic contribution to covariance between relatives, which considers both experimental and observational data.16 However, as shown above, manipulative methodologies are an excellent choice. Such methods enable ecologists and evolutionary biologists to establish reciprocal transparent research efforts across disciplinary barriers with greater simplicity. As a consequence, the young field of epigenetics can benefit from causal knowledge coming from lab experiments and thus thrive. The above case of lab and field epigenetics exemplifies potential ways to integrate different sorts of data in epigenetics. Sharing a common manip-

▸ 53 CHALLENGES OF EPIGENETICS

ulative rationale facilitates what Sabina Leonelli (2013) has called interlevel and cross-species integration. This means, first, relating data about different features of a common model organism—data that were collected on different levels of organization and in situations that vary with respect to their degree of control over the organism. Second, data from studies of different organisms can be related and compared (in lab-lab, lab-field, or field-field settings) in order to understand both variation based on similar processes and mechanisms across the tree of life and, more generally, the relevance of epigenetics for biology as a whole. Finally, this case also shows us that there is little need to be worried about the level of artificiality in epigenetic studies when it comes to addressing evolutionary explananda. Instead, the field seems to be heading in the right direction by going out of the lab and, at the same time, developing research methodologies that are heuristically useful because they allow the development of reciprocal transparent models in ecological epigenetics and molecular biology. In addition, even at this early stage, the results of epigenetic field studies offer sufficient causal information to, by themselves, guide further research—even across disciplinary boundaries—on the phenomena under study.

MIXING “WHYS” AND “HOWS”— PROXIMATE VERSUS ULTIMATE CAUSE EXPLANATION

In 2014 Frank Johannes, from the Groningen Biomolecular and Biotech- nology Institute, and his colleagues have presented a greenhouse study in Science, and in it they tied DNA methylation patterns to heritable variation in flowering time and root length in epigenetically different strains of isogenic Arabidopsis thaliana (Cortijo et al. 2014; see also Pennisi 2013). These experimentally induced strains show across their genome differentially methylated regions that causally contribute to the complex traits and are stably inherited independent of DNA sequence changes. Since 30 percent of these differentially methylated regions overlap with the naturally occurring methylation pattern in wild populations of Arabidopsis, these findings are suggested to be of interest for evolutionary biologists. The publication of Johannes’s work was advertised under the banner of theoretical revolution by the University of Groningen (2014): in a press release

54 ◂ CHALLENGES OF EPIGENETICS the results were summarized as a revolutionary new view of inheritance in plants, a view that will likely prompt a revision of genetics textbooks. Johannes, supporting this idea of a substantial theoretical change, notes, “This is a breakthrough, because it changes the way we view genetics” (quoted in University of Groningen 2014). As this case exemplifies, findings that epigenetic factors make causal contributions to the development and inheriting of complex traits in population or ecological epigenetics are often presented and reviewed as if they invoke a major theoretical shift in biology. However, with respect to evolutionary biology, this is due not just to the fact that new models of population dynamics have to be developed that are able to consider previously unrecognized epigenetic causal influences, as discussed above. There is also something about the conceptualization of these new epigenetic causes that makes scientists, as well as the wider public, discuss them within the context of a major quantitative expansion or even qualitative, revolutionary shift in biological theory. Let me explain this issue in detail. When Johannes emphasizes in the same review (University of Gronin- gen 2014) that epigenetic factors “cause variation on which natural selection can act,” this implies that these factors are not only new to population studies on natural selection but that they are explicitly prohibited to be of evolutionary relevance. This assumption is supported by a historically influential conceptual distinction in biology: the distinction between proximate causes and ultimate causes. According to this dual scheme of biological causality, biologists studying proximate causes ask how questions about developmental processes making up the individual organism whereas biologists studying ultimate causes ask why or what for questions about evolutionary function. Thus, as Ernst Mayr (1961) emphasized, proximate causes, on the surface, resemble Aristotelian efficient causes while ultimate causes resemble Aristotelian final causes.17 According to Mayr, functional biologists are interested in proximate causes, because they seek to investigate how systems work. In contrast, evolutionary biologists are interested in ultimate causes, because they seek to investigate how phylogenesis has produced one system rather than another. He illustrates this distinction by means of an example of avian migration. Migration can be investigated by asking why birds migrate (i.e., what is the selective advantage of migration) and how they migrate (i.e., how they develop skills like navigation and timing). Although these two investigations are under-

▸ 55 CHALLENGES OF EPIGENETICS

stood to be complementary, and thus both are important, they should be treated as distinct from one another. This means thathow questions cannot be addressed by explanantia citing ultimate causes (i.e., telling a story of adaptation) and, more important, that why questions cannot be addressed by explanantia citing proximate causes (i.e., telling a story of trait development). According to the latter argument against the explanatory relevance of development to evolution, “evolution is a change in the genetic composition of populations,” as Theodosius Dob- zhansky (1951, 16) emphasizes (see also Mayr 1984). Today this view is still prominent in evolutionary biology. For example, Deborah Charlesworth and colleagues (2017, 10) have argued that “allele frequency change caused by natural selection is the only credible process underlying the evolution of adaptive organismal traits.” In other words, change in genotype frequencies can and should be studied through population genetics alone. This view of explanatory asymmetry conflicts with Johannes’s claim that nongenetic epigenetic markers “acquired” during development are relevant for explaining evolutionary processes. The proximate-ultimate formulation was introduced into evolutionary biology in the late 1930s by John Baker (1938), although even at that time the underlying idea had been around for a while (see below). Baker, who was interested in the evolution of breeding seasons, states that ultimate and proximate factors regulate the timing of reproduction. Ultimate factors, like food availability, determine reproductive success and survival, while proximate factors determine how organisms actually time reproduction. For example, in boreal regions proximate factors include day length; in the trop- ics they include rainfall. The reception of Baker’s distinction did not gather much speed until Mayr’s 1961 article in Science and Nikolaas Tinbergen’s (1963) methodological analysis of ethology, which supports a similar parti- tioning of biological questions.18 As a consequence, the distinction made by both Mayr and Tinbergen reflected mainstream causal reasoning for a long time in evolutionary biology—even for those evolutionary biologists interested in developmental processes: “We can progress towards understanding the evolution of adaptations without understanding how relevant structures develop. Hence, if the complaint against the ‘adaptationist programme’ is that it distracts attention from developmental biology, I have some sympa- thy. Development is important and little understood, and ought to be studied. If, however, the complaint is that adaptation cannot (rather than ought

56 ◂ CHALLENGES OF EPIGENETICS not) be studied without an understanding of developmental constraints, I am much less ready to agree” (Maynard Smith 1982, 6; original emphasis). As John Maynard Smith and other evolutionary biologists emphasize, the distinction of biological causes is backed up theoretically by Weismann’s concept of the separation of germ line and soma. It acts as a demarcation line between two distinct classes of causes—in Dawkinsian terms, causes of the replicator and causes of the vehicle.19 There has been constant criticism of this distinction since even before Mayr and Tinbergen’s time. For example, as early as 1916 Julian Huxley (1916, 161) criticized the parting between ultimate cause and, as he called it, “immediate cause.” Even today, both biologists and philosophers of science interested in the divisions of problems in terms of “how-why” and “proximate- ultimate” have not reached a consensus on how ultimate and proximate cause explanations relate to each other or what kind of explanations they actually are.20 Despite this lack of consensus, however, Mayr’s view on biological causality and explanation did not die out. In contrast, it somehow managed to stay alive in contemporary biology.21 Especially in light of the debate on how to expand the framework of evolutionary theory by more developmentally oriented accounts, such as epigenetics, evo-devo and niche construction theory, the proximate-ultimate debate is hotter than ever before (see, e.g., Thierry 2005; Laland et al. 2011, 2013a, 2013b; Haig 2011, 2013; Scott-Phillips et al. 2011; Dickins and Rahman 2012; Guerrero-Bosagna 2012; Calcott 2013a, 2013b; Dickins and Barton 2013; Gardner 2013; Mesoudi et al. 2013; Martínez and Esposito 2014; and Scholl and Pigliucci 2014). According to Kevin Laland and colleagues (2011, 1512), it stands “at the center of some of contemporary biology’s fiercest debates” about the role ofepigenetic inheritance and, more generally, developmental responsiveness or plasticity in evolution. Most prominently, Laland and colleagues (2011, 2013a, 2013b) have argued for replacing the proximate-ultimate dichotomy with a concept of “reciprocal causation,” which would better grasp the feedback between causal factors in evolving systems. They refer to the capacity of phenotypic plasticity or, more specifically, the activities of organisms to alter selection pressures. Their paradigmatic cases are those of niche construction, in which organisms—subject to selection—modify their environments, an activity that, in turn, shapes natural selection pressures. More generally, they seek to enrich

▸ 57 CHALLENGES OF EPIGENETICS

evolutionary biology by newly conceptualizing phenomena in which developmental plasticity triggers evolutionary novelty and biases evolutionary trajectories. As a consequence of the growing interest in ontogenetic agency and developmental (or proximate cause) explanations in evolutionary biology and, more specifically, because of Laland et al.’s concept of reciprocal causation, more orthodox biologists currently feel the need to highlight again the prevailing standards of how to construe evolutionary explanation: “There is a conflation in the EES [i.e., extended evolutionary synthesis] literature between ultimate and proximate causation, and, as a consequence, a failure to address the issues of levels of organization within biology in ultimate terms” (Dickins and Rahman 2012, 2919). With respect to epigenetics, Thomas Dickins and colleagues in particular have argued that although “epigenetic phenomena are often said to challenge natural selection (an ultimate process) as a source of explanation, . . . these are better understood as proximate mechanisms” (Scott-Phillips et al. 2011, 39; see also Dickins and Dickins 2007, 2008). In addition, if these phenomena are considered from an evolutionary perspective, they should be understood to exclusively serve the ultimate function of calibrating organisms to environmental stochasticity or uncertainty. This means that plastic phenotypes can fine-tune adaptations during development but never “drive, or co-cause, evolution, through generating innovation, biasing variation, or imposing directionality on evolutionary trajectories,” as Laland et al. (2013b, 731) argue. In other words, Dickins and colleagues argue that questions asking why epigenetic inheritance (or nongenetic inheritance in general) has evolved can only be answered by listing those causally relevant genetic variants as explanantia that have been selected during phylogenesis. Thus, their criticism rests on a thesis about the sufficiency of explanatory resources offered by the modern synthesis. This debate on whether to keep or reinterpret the proximate-ultimate distinction or even to replace it with Laland et al.’s concept of causal reciprocity (or some other account) is in full swing. Some argue that even Laland et al.’s conceptualization does not capture all causal dependency relations of interest for evolutionary biology (Martínez and Esposito 2014; Scholl and Pigliucci 2014). Moreover, some hold that, first, the concept of reciprocity itself relies on the dichotomy between development and evolution (Dickins and Barton 2013; Martínez and Esposito 2014); that, second, it is not condu-

58 ◂ CHALLENGES OF EPIGENETICS cive to successful biological science, since it allows proximate and ultimate explanations to bleed into each other until their distinction is meaningless (Gardner 2013); and that, third, in contrast to the prevailing dichotomous framework, it does not lead to falsifiable questions (Dickins and Rahman 2012).22 I will not—at least not directly—contribute to these discussions but will take up a more fundamental issue about the proximate-ultimate distinction, which hinders successful integration of epigenetic explanations into the prevailing evolutionary framework.

PROXIMATE CAUSE EXPLANATION REVISITED

Brett Calcott (2013a, 779) argues that Ernst Mayr’s distinction is too simple, because in biology there are questions being asked and explanations being given that do not fit in either of the two categories—“proximate” and “ultimate.” I will show that his simplicity claim is right, but he errs about the reasons. Mayr’s picture of explanatory dualism in biology is not too simple because there are other, still unrecognized explanations besides the already described ones, but because the predominant reading of the proximate cause explanation offers a simplifying picture of what (and how) developmental explanations actually explain. Calcott identifies two contrasts to be essential in Mayr’s distinction: the contrast between explanations relating things synchronically and diachronically—addressing how something works at a particular time and how it changes over time—and between explanations tracing differences on the level of individual mechanisms and on the population level. Based on this analysis, he argues that Mayr’s proximate cause (PC) explanations answer the question

(PC) How do individuals work at a particular time? and ultimate cause (UC) explanations answer the question

(UC) How do populations change over time?

According to Calcott, explanations answering these two questions can be defined in philosophical terms: a (PC) explanation is a constitutive mecha-

▸ 59 CHALLENGES OF EPIGENETICS

nistic explanation, and a (UC) explanation is a causal explanation (see also chaps. 3 and 4). Thus, the (PC) explanation constitutively relates differences in properties (of component parts) of individual mechanisms at different levels of organization, while the (UC) explanation causally relates differences in evolutionary events and population dynamics. For example, the (PC) explanation addresses how, in a migrating bird, navigation skills are “brought about,” and (UC) explanations address why navigation skills are advantageous in birds (and why we find changes in relevant genotype frequencies in bird populations). Then, Calcott shows that this dichotomy does not consider those explanations that trace diachronic change in individual mechanisms and that address questions of the form

(PUC) How do individuals change over time?

These explanations—he calls them “lineage explanations”—are considered to be most relevant in evolutionary developmental biology (see also Calcott 2009).23 Here, lineage explanation describes a pattern of evolutionary explanation that invokes neither populations nor natural selection. According to Calcott, in evo-devo, answering how the development of, for example, the tetrapod limb and its ancestors’ fins changed over time means offering an explanation that traces differences between the developmental mechanisms of individuals that produce the relevant morphological shapes and that existed at different times (without including information about evolutionary forces and population processes). However, in contrast to what Calcott thinks, this explanation of mechanistic change over time is not restricted to evo-devo. Rather, it is an essential feature of developmental proximate cause explanation. While introducing functional biology as an explanatory enterprise tracing “static” synchronic dependencies, Mayr (1961, 1502) also described functional biologists to be “vitally concerned with the operation and interaction of structural elements.” This latter, more dynamic perspective on developmental processes has been echoed by advocates and critics of the proximate-ultimate distinction ever since—resulting in a reading of Mayr’s dichotomy as one between developmental and evolutionary processes, between ontogenesis and phylogenesis: “One apparently plausible interpretation of this [i.e., Mayr’s] dichotomy is that proximate causes concern processes occurring during the

60 ◂ CHALLENGES OF EPIGENETICS life of an organism while ultimate causes refer to those processes (particularly natural selection) that shaped its genome” (R. Francis 1990, 401). And Laland et al. (2013b, 731) state that “the primary problem . . . , reinforced by the proximate-ultimate causation dichotomy, for evolutionary biology itself is that it has hindered the full recognition of the pathways by which organisms’ ontogenetic processes affect evolutionary processes.” When philosophers of science and biologists talk about, as Donald Dewsbury (1999, 190) describes it, “causes of behaviour that occur within an animal’s lifetime and those that preceded its life,” they seek to emphasize how traits evolve within a developmental or evolutionary time frame. Note that the developmental frame is extended in time and thus is not identical to (PC)’s time frame of synchronicity. Accordingly, as emphasized by Fabrizzio Mc Manus (2012) and Petri Ylikoski (2013), developmental explanations are not straightforward constitutive explanations. They trace dependencies between causal capacities of (component parts of) a system at different levels of organization at different points in time: “While a constitutive explanation relates the system’s capacities synchronously, a developmental explanation involves both time and significant changes in the system’s causal capacities. The aim of such explanations is to figure out the pathway by which the system’s initial causal capacities are realized and transformed into new causal capacities that characterize the system’s later phases” (Ylikoski 2013, 292–93; emphasis added). For example, in order to understand how a fertilized egg develops into an adult organism, developmental biologists consider the changes of mechanistic parts of the developing system as well as their relations. Developmental explanations thus relate the causal capacities of a system’s parts or their organization at earlier phases of development with capacities of the whole system (usually a phenotypic trait) at a later time in development (usually in adult organisms). Consider, for example, changes in the epigenetic makeup of a pluripotent stem cell that are related to differences in the phenotypic appearance of a cell (linked with the stem cell through a cellular lineage) in the terminal state of full differentiation. Investigating trait development means, in this case, asking which differences in the fully differentiated cell’s phenotype would be due to a change in a particular gene’s activity early in differentiation. It does not mean asking which differences in the phenotypic appearance of the stem cell itself would be due to a change in a particular gene in this stem cell early in differentiation (although this question is interesting

▸ 61 CHALLENGES OF EPIGENETICS

from an nondevelopmental point of view). The latter case is an instance of Calcott’s (PC). The former necessitates a reconceptualization of what is commonly described as a developmental proximate cause question. It takes the form

(PC*) How do individuals change over time?24

Since (PC*) and Calcott’s (PUC) are identical, introducing the latter is redundant. It is the relevance of question (PC*), not of (PC), in evolutionary biology that has been at the heart of the proximate-ultimate debate in recent decades. This means that it has not been debated whether straightforward constitutive dependencies that relate things synchronically in individuals should be interlinked with causal dependencies in populations. Rather, the bone of contention has been whether dynamic shifts, constraints, and interrelations between different levels of organization during the course of individual development should be considered in evolutionary research. Con- cepts like developmental responsiveness, phenotypic plasticity, and niche construction all refer to such diachronic dependency relations. Following Petri Ylikoski, developmental explanations tracing these particular dependencies can be labeled hybrid explanations: “Developmental explanations . . . combine in an interesting way both causal and constitutive elements. The explanation is about a causal process, but both the explanans and explanandum are causal capacities. Furthermore, the explanans refers to parts (or their organization) of the developmental process, while the explanandum refers to the causal capacities of the system that is an outcome of this process” (Ylikoski 2013, 294; original emphasis). In other words, developmental explanations often differ from straightforward causal explanations in their relata—they often do not relate events but causal capacities. In addition, developmental explanations differ from straightforward mechanistic explanations in relating their explanans and explanandum diachronically. Once this hybrid character of developmental explanations is understood, the distinction between proximate and ultimate cause explanations can be reassessed. The real distinction is not one, as both Calcott and Mayr suggest, between answering (PC) and (UC) but between answering how individuals change over developmental time (PC*) and how populations change over evolutionary time (UC). Revealing the more complex form of developmental explanation not

62 ◂ CHALLENGES OF EPIGENETICS only sheds light on the proximate-ultimate distinction actually debated but also helps to (1) reveal some conceptual obscurity in the ongoing debate and (2) clarify the explanatory role of developmentally oriented accounts, like epigenetics, within the framework of an extended evolutionary synthesis. As for the first item above (revealing conceptual obscurity), a number of authors have described the proximate-ultimate distinction as one between (explanations tracing) causes acting immediately (in the present) and those acting in the past:

[Ultimate causes] are causes that have a history and that have been incor- porated into the system through many thousands of generations of natural selection. . . . Proximate causes govern the responses of the individual (and his organs) to immediate factors of the environment. (Mayr 1961, 1503)

Proximate explanations focus on causes in the present; evolutionary explanations focus on how the present has been shaped by events in the past. (Hochman 2013, 593–94)

A proximate cause is an immediate, mechanical influence on a trait. . . . Ultimate causes are historical explanations; these explain why an organism has one trait rather than another, often in terms of natural selection. (Laland et al. 2011, 1512)25

In light of the structure of developmental explanation, these authors obscure the dynamic complex phenomena they aim to discuss and the strategies used to investigate them. Take, for instance, one example of niche construction discussed by Laland et al. (2011, 2013a, 2015). Earthworms change the structure and chemistry of the soils and, as a consequence, modify the selection pressures acting on them in their environment (Odling-Smee et al. 2003). In order for an earthworm’s fitness to be a function of environmental, organism-modified factors, the environment has to be changed, in fact, through the lifetimes of individual earthworms. What is investigated in this case is whether changes in the soil-processing behavior of an earthworm, for example, early in development, would make a difference on the evolutionary level (i.e., change of selection pressure and fitness) later on. In short, the mechanistic characteristics of soil-processing are not changed at a particular time but are changed over time. Moreover, what niche con-

▸ 63 CHALLENGES OF EPIGENETICS

structionists usually emphasize is that developmental and evolutionary processes coevolve, with the former acting as a constructive (constraining and/ or creative) force on the latter over the course of time. Focusing entirely on time slices of ontogenetic processes counteracts this idea of dynamic construction. As for the second benefit of revealing the more complex form of a developmental explanation (clarifying the explanatory role of developmentally oriented accounts), if we adopt Mayr’s (1984) explanatory asymmetry thesis, the (PC*)-(UC) distinction claims that explanations of how individuals change over time do not convey information for explanations of how populations change over time. It is exactly this view that advocates of an extended evolutionary synthesis in general and evolutionary-minded epigeneticists in particular oppose. Take, for instance, the case of transgenerational epigenetic inheritance. What explanations of epigenetic inheritance actually offer is information on how a change in nongenetic developmental processes of one organism in the parental generation can make a difference in trait development in the offspring. This is a (PC*)-like explanation.26 However, in contrast to straightforward developmental explanations, the time frame in which dependencies are traced is slightly expanded to several generations. Dickins and Rahman (2012, 2916) have claimed, against the explanatory value of these explanations in evolutionary biology, that “to simply outline other inheritance systems and describe how they work is not to provide an account at the ultimate level.” In contrast, an ultimate explanation should provide information about how natural selection, as the result of trait variation, competition, and inheritance, shapes populations. This critique, again, is based on conflating (PC) and (PC*) information. By focusing on asymmetric dependency relations across organisms’ lifetimes, epigenetics has developed a theory of inheritance that accounts for how variants appear both in individuals and in populations. In contrast to random genetic mutations, novel heritable epigenetic variants are understood to be environmentally induced simultaneously in multiple individuals in a population (see Jablonka and Raz 2009; and Laland et al. 2013a). Thus, explanations of epigenetic inheritance do not entirely function on the level of individuals. What they offer instead is information on asymmetric dependencies in between developmental and evolutionary time and different levels of organization. Thus, epigenetics can easily develop population models, like the ones discussed above, that provide answers to how changes in epigenetic inheri-

64 ◂ CHALLENGES OF EPIGENETICS tance systems (i.e., in proximate causes) in individuals make a difference in populations, for example, through the induction and inheritance of a certain novel trait. Notice that such explanations do not answer (UC) questions in the pure form: they do not offer information per se on why, in the sense of “what for,” a population changes over time or which evolutionary forces (natural selection, drift, etc.) are responsible for the population changing. Instead, they offer valuable information on how dynamic changes in developmental processes across generations and between individuals and environment can affect the rate, availability and specific characteristics of variations in populations, which could then be selected. Thus, epigenetic explanations assist and constrain adaptationist (UC) explanations of evolutionary change.27 However, as Laland and colleagues emphasize, despite the growing community of developmentally minded evolutionary biologists in the fields of epigenetics, evo-devo, and niche construction theory, the calls of these scientists for expanding evolutionary theory and overcoming the prevailing explanatory framework in evolutionary biology (which is biased by population genetics) should still be understood as a disorganized protest movement: “It is probably fair to say that these various lobbies currently more resemble a disorganized protest movement than a viable alternative govern- ment. There is not yet a clear statement as to precisely what any extended evolutionary synthesis (EES) entails, what its core assumptions are, and how they differ from the modern synthesis” (Laland et al. 2013a, 807; but see Laland et al. 2015). At least two problems currently hinder formation of a unified so-called “EES movement.” First, issues with respect to concepts and the structure of explanations have to be settled, that is, we must approach, as Laland and colleagues (2013a, 807) emphasize, a “constructive conceptualization of development” and developmental explanation. As argued above, the developmental perspective of an extended evolutionary synthesis that is introduced into the orthodox neo-Darwinian framework has to be understood better. In contrast to Calcott’s position, the EES movement should not be considered a bundle of explanatory approaches transcending Mayr’s traditional conceptual distinction. Rather, the predominant reading of Mayr’s distinction offers a simplifying picture of what (and how) developmental explanations actually explain. Once this mistake has been corrected, the appropriateness of developmentally oriented approaches, such as epigenetics, in

▸ 65 CHALLENGES OF EPIGENETICS

evolutionary biology becomes apparent. In addition, this conceptualization allows using the proximate-ultimate distinction in a way that does not suppress the explanatory potential of epigenetics for the theory of evolution. This means that we do not have to replace Mayr’s distinction, as suggested by a number of biologists and philosophers of science (see, e.g., Laland et al. 2011, 2013a; Mesoudi et al. 2013; and Martínez and Esposito 2014). Instead, it can be kept, although in a revised form.28 Therefore, we need to overcome previous attempts to make sense of this conceptual parting as one between constitutive and causal explanation. In addition, we need to reject the idea of explanatory asymmetry between proximate cause and ultimate cause explanation. This reinterpretation of the proximate-ultimate distinction not only prevents conceptual obscurity but also permits clarifying how developmentally oriented explanations can be related to evolutionary ones in an extended evolutionary synthesis framework. A second challenge for the EES movement concerns issues about theoretical integration and explanatory power. A number of authors understand an expanded theory of evolution as a more pluralistic framework in which various orthodox and progressive explanatory approaches are theoretically integrated (West-Eberhard 2003; Thierry 2005; Calcott 2013a; Laland et al. 2013b). If, however, all of these approaches rightfully explain something within this “patchwork,” as Calcott (2013a, 779) calls it, biologists are in need of a theory that tells them which account is superior to an alternative explanation in a certain situation and why. For example, since Laland and colleagues concede that the proximate-ultimate distinction has a heuristic value at least in explaining some evolutionary phenomena, they, as Dick- ins and Barton (2013, 754) correctly note, “need to adjudicate on when it is acceptable to use the proximate-ultimate distinction.” This problem and similar ones do not refer to the appropriateness or, as one might say, explanatoriness of the explanations included in a pluralistic extended evolutionary synthesis (i.e., to the question of whether and how these novel explanations, in fact, explain) but to their explanatory power. That is the question of how good these new explanations are and what constitutes their value. We will return in chapter 5 to this crucial second issue hindering the current expansion of evolutionary theory. In the preceding pages, conflicts have been discussed that arise due to differences in the explananda and explanantia chosen by epigeneticists and evolutionary biologists, respectively. These interdisciplinary conflicts con-

66 ◂ CHALLENGES OF EPIGENETICS cern issues of whether (and how) epigenetic explanations can address evolutionary explananda even though they draw on highly artificial scenarios and focus on proximate rather than ultimate causes. However, the theoretical integration of epigenetics into the theoretical framework of modern biology in general and evolutionary biology in particular is hindered not only by interdisciplinary hurdles like these. There are also intradisciplinary challenges hindering epigeneticists from coming of age. They concern issues about how to understand the field of epigenetics methodologically and conceptually, as well as about how epigenetic complexity should be explained. Let us now turn to these issues.

WILL THE REAL EPIGENETICIST PLEASE STAND UP!

Usually, discussions about whether to expand the modern synthesis focus on methodological, explanatory, or conceptual conflicts that arise (or may arise) between “old” and “new” theoretical approaches (see, e.g., Laland et al. 2009; Craig 2010; Pigliucci and Müller 2010b; Griesemer 2011c; and Haig 2011). Conflicts that emergewithin new approaches or between new approaches are, on the other hand, rarely addressed. This includes, for instance, issues concerning how to conceptually integrate evo-devo and niche construction theory (Laland et al. 2008) or how to link methodologies like selection of model organisms in evo-devo and epigenetics (Minelli and Baedke 2014). In addition, the plurality of scientific “cultures” in these young disciplines has to be taken into account (Winther 2015). Thus, we might ask several questions: Does the field of epigenetics constitute a new coherent scientific paradigm? Has it developed into a novel theoretical framework powerful enough, for example, even to reframe philosophical debates on the concept of the identity of living beings, as argued by Giovanni Boniolo and Giuseppe Testa (2012)?29 Or is it more of a heterogeneous array of highly diverse approaches that are only loosely connected to each other by the term “epigenetics”? In order to come up with an accurate answer to these questions, it is essential to understand how those biologists investigating epigenetic phenomena understand their own highly dynamic research field and how they, as opposed to other fields of research, define the term “epigenetics.” Therefore, I will review self-

▸ 67 CHALLENGES OF EPIGENETICS

definitions of their research focus and, in particular, methodological reflections. More specifically, I will focus on the methodologies of a particular, often overlooked group of epigeneticists: epigeneticists in stem cell biology who investigate cell differentiation processes or cell “reprogramming.” The latter is the experimental reversal of cell differentiation. This characterization of the methodological focus of today’s stem cell epigenetics exposes “another” epigenetics, one whose representatives have developed a conception of their field that separates it from the specific definitions of epigenetics crystallized in recent years. What is revealed is a methodological diversity that makes it difficult to localize a core identity of the epigenetic research program. In particular, contrary to the standard experimental methods of molecular epigeneticists, some biologists in stem cell and systems biology seek to methodologically reframe epigenetics by adopting a program of mathematical modeling developed by epigenetics’ founding father, Conrad Hal Waddington.30 This program is closely linked to his model of the so-called “epigenetic landscape.” While some historians and philosophers have claimed that Waddington was not able to establish a consistent legacy, that the reception of his work never really gathered speed in mainstream biology, and that the term “epigenetics” is of course used in a quite different way today (see, e.g., Jablonka and Lamb 2002; Van Speybroeck 2002; and Peterson 2011), I argue that his methodological work is very much alive today and in fact drives intradisciplinary conflicts about what “real” epigenetic research is and how it could be organized as a discipline. Interestingly, this Waddingtonian “old-school epigenetics” that is currently reemerging in stem cell research also calls for expanding the framework of the modern synthesis but on other terms. In addition, this field cannot easily be assigned to one of the seemingly contradictory poles of “neo-Darwinism and gene-centrism” versus “neo- Lamarckism, post-Darwinism, and post-genomics.”

WADDINGTON’S EPIGENETICS AND HIS “EPIGENETIC LANDSCAPE” MODEL

The idea, as suggested, for example, by Boniolo and Testa (2012), of simply dividing epigenetics into intragenerational epigenetics, which is concerned with how environmental stimuli are connected to the regulation of chromatin and DNA methylation, and intergenerational epigenetics, which is con-

68 ◂ CHALLENGES OF EPIGENETICS cerned with the very same regulatory phenomena but across generations of cells or multicellular organisms, seems to have problems living up to the intradisciplinary diversity and conflicts we find in modern epigenetics. For example, some epigeneticists in stem cell research have voiced misgiv- ings about how the term “epigenetics” is used, in particular by molecular biologists: “Such DNA and histone modifications were given the attribute ‘epigenetic’—an onomasiologically unfortunate choice . . . —to distinguish them from genetic, DNA sequence-based mechanism[s] of inheritance” (Huang 2009b, 549). Such intradisciplinary objections to descriptions of nongenetic inheritance as “epigenetic” require some clarification, since, according to the above definition, both stem cell research on cell differentiation processes and research on nongenetic intercellular or organismic transmission focus on the very same phenomenon: regulation of gene activity through chromatin modification. However, this phenomenon is approached from different perspectives. While the former line of research is interested in the role of nongenetic factors in embryo- and morphogenesis, the latter focuses on the heritability and (partly on the) evolutionary relevance of these factors, as well as on their potential independence from genetic determinants. Both areas are called “epigenetics,” although the latter epigenetics has gained prominence in public and scientific discourse. However, in addition to these differences in contextualizing the processes-to-be-explained, the above critical remark results from a more fundamental conflict about how to methodologically approach developmental phenomena—through mathematical modeling or experimental manipulation. In an approach addressing “Lillie’s paradox,” which concerns the issue of how to reconcile the view that all cells of a higher organism have the same genes, with the differences observed in developing cells during embryogenesis, Conrad Hal Waddington (1942b) introduced the concept of the “epigenotype” in the early 1940s (see also Gilbert 2012). The epigenotype can be understood as a web of context-sensitive developmental processes that jointly links the genotype with particular phenotypic traits. It summarizes Waddington’s astonishingly prescient idea that sets of genes act together in guiding the complex mechanics underlying development trajectories. In order to model the complexity of these developmental processes and reconcile genetic stability with phenotypic flexibility, Waddington chose to topographically represent development with his model of the epigenetic

▸ 69 CHALLENGES OF EPIGENETICS

A ▸

◂ B

FIG. 2.3. A pictorial version of Waddington’s EL (A) and the complex system of interactions underlying the EL (B). Original caption of (B): “Th e pegs in the ground represent genes; the strings leading from them the chemical tendencies which the genes produce. Th e modelling of the epigenetic landscape, which slopes down from above one’s head towards the distance, is controlled by the pull of these numerous guy-ropes which are ultimately anchored to the genes” (Waddington 1957, 36) (A: Waddington 1957, 29; B: Waddington 1957, 36; reproduced with permission from Taylor and Francis).

70 ◂ CHALLENGES OF EPIGENETICS landscape (EL). The EL model contains a set of developmental choices that a differentiating cell in the embryo faces. This process is represented by a ball rolling down a landscape that contains several valleys. In the course of differentiation from a pluripotent to a unipotent cell state, the “ball” faces a number of branching points or “choices” in the landscape. Waddington’s drawings of this landscape subsequently became well known in biology.31 His most famous EL image is depicted in figure 2.3A. The pathways, also called “chreodes,” are understood as buffered or canalized pathways. According to Waddington, “canalization” means that up to a certain threshold neither external nor internal perturbation, like environmental noise and/or genetic variation, will affect the pathway. Thus, highly canalized valleys are deep and perturbation can barely push the cell/ball over the valley’s walls into an adjacent developmental path. Consequently, the same phenotype will come into being even though the genotype or the environment might have changed to some extent. The idea to develop the visual metaphor of the EL likely emerged in the 1930s in the discussions of the Theoretical Biology Club in Cambridge—a lively gathering of biologists and philosophers interested in conceptual issues related to biological phenomena (Peterson 2016). Many members of the group, such as Joseph Henry Woodger, Waddington, and Joseph Need- ham, supported Alfred North Whitehead’s “process philosophy” or “organicism,” which represents an antireductionist, systemic view of the organism emphasizing the complex interrelatedness of its developing parts (see, e.g., Waddington 1970a, 1977c; see also Haraway 1976; and Peterson 2011).32 Fig- ure 2.3B illustrates Waddington’s idea that only complexes of genes (and their products), not single genes, affect the course of any developmental pathway. Waddington (1940, 93) commented on the introduction of his EL images as follows: “I think it expresses, formally at least, some characteristics of development [for example, canalization] which are not easy to grasp in any other way.”33 In other words, to “those who tend to think in pictures” these images make available a specific topographical framework, which predeter- mines how to think about the developmental phenomenon under study: in terms of stability and change (1957, 35). During the rise of molecular biology’s reductionist, single-gene view of development, Waddington believed that he had to enhance his EL framework in order to participate in future explanatory approaches in genetics

▸ 71 CHALLENGES OF EPIGENETICS

FIG. 2.4. An EL-attractor surface in state space. For description, see the text (Wad- dington 1977a, 143; reprinted with permission from Elsevier).

and developmental biology. In particular, this was due to the development of Jacob and Monod’s (1961) operon model, which enabled biologists to explain much more explicitly the active status of genes and thus to describe the path a cell was to take in development. His “2.0 version” of the EL was supposed to be able to mathematize and quantify complex phenomena in development without falling prey to a reductionist perspective. Therefore, he reinterpreted the EL in terms of an attractor surface backed up by a group of topological approaches, especially catastrophe theory (see Waddington 1969a, 1974). This EL-attractor surface is illustrated in figure 2.4. According to Waddington, this mathematically more explicit version of the epigenetic landscape represents a developing system by a single point (e.g., point A in fig. 2.4) in a multidimensional state space, the axes of which represent the values of all system variables, like the concentrations of all gene products in a cell. The velocity and direction in which the system develops in state space is represented by a vector (arrows for A and B on the surface in fig. 2.4). Given enough time, the developing system will end up at a certain attractor, in a state of terminal differentiation.

72 ◂ CHALLENGES OF EPIGENETICS

In his late-career book Tools for Thought, which is a discussion of various heuristic tools applicable in research on complex phenomena, Waddington (1977b) described how his landscape images could be used to coordinate and “canalize” methodological strategies and modeling efforts (see also Waddington 1977a). If scientists want to investigate an unknown, developing complex system like an organism during embryogenesis, they should, according to Waddington, ask themselves how they are to find out what the shape of the landscape looks like that represents particular properties of the system. To answer this question he adopts a mathematical modeling technique, called the “ravines method,” developed by the Russian mathematicians Israel Gel’fand and Mikhail L’vovich Tsetlin (1962, 1971, 1973).34 The ravines method should guide the formalization of the complex dynamics of the human nervous system, like neural locomotor development, by mini- mizing a function of many variables that describe this system. Gel’fand and Tsetlin assume that the variables are organized in such a way that they shape a long ravine. The value of the function is regarded as the elevation of a surface, and the minimum lies somewhere along this ravine. They understand this mathematical modeling approach as a tool for organizing a mostly unknown phenomenon-to-be-explained and to derive new hypotheses and methodologies. Waddington reinterpreted the ravines method by drawing on his EL framework. He claimed that the stability characteristics of a complex system can be (experimentally) investigated and quantitatively grasped by means of his late version of an EL attractor surface (see fig. 2.4). At first, the landscape topography is drawn rather fuzzily, and the hills and valleys are hard to distinguish. At this early stage, two exploration methodologies, described by Gel’fand and Tsetlin, have to be adopted by the investigator in order to proceed to a more exact drawing. These strategies are calledlocal exploration and jump in the dark (Waddington 1977a, 113–14). Local exploration. The scientist carefully “scans” the surface of a developmental ravine or valley by slightly altering the system in various ways—or as Waddington (1977a, 113) puts it, “we can think of our actions as going out into the landscape for the same distance in every direction from where we stand, thus describing a circle.”35 Then, different responses of the system can be observed depending on the “directions” of perturbation. For example, pushing the imaginary ball “uphill” is more difficult than pushing it “downhill.” Next, the explorer should take small steps (i.e., time intervals)

▸ 73 CHALLENGES OF EPIGENETICS

in the direction in which the system can be altered most easily (e.g., in fig. 2.4, from point A “downhill” toward Aʹ) and repeat the perturbation procedure. Once the object has reached the bottom of the valley, changes in the system become difficult in all directions. This exploration method guided by an initially less exact EL image enables one to deduce the course and the topography of the valley and thus the stability characteristics of one potential developmental fate of the system. In order to explore other potential developmental slopes one has to adopt a second strategy. Jump in the dark. The scientist explores other valleys by (experimentally) changing an aspect of the system that is significantly different from the previous one. Metaphorically speaking, she makes a large “jump,” which hopefully carries her to the opposite hillside of the developmental track just explored (e.g., in fig. 2.4. we might change from position Aʹ to Bʹ; see Wad- dington 1977a, 113–14). Then, the local exploration strategy comes into play anew in order to trace the course of this new slope. As Waddington’s interpretation of the ravines method shows, the EL may be used as a visual framework to continuously coordinate methodological strategies such as experimental manipulation regarding research on mostly unknown (complex) phenomena. It enables scientists to locate and thus to spatially distinguish different exploration methods. Here, both procedures—drawing a precise and quantified EL surface, which means locating the landscape’s topographical properties in state space and formally describing the attractor surface in a mathematical model, as well as collecting data “somewhere on the surface,” for example, through experimental research— proceed in a deeply intertwined fashion. This methodological vision in Waddington’s late-career work—using the EL model for connecting the analysis of stability properties of dynamic biological systems with the mathematical description of these properties—has been taken up and developed further in stem cell biology.

THE RENAISSANCE OF “OLD-SCHOOL EPIGENETICS” IN TODAY’S STEM CELL BIOLOGY

A number of new EL models have been developed in order to describe cell differentiation processes and reprogramming of differentiated cells in so-called “induced pluripotent stem cells,” or iPS cells (see Zhou and Melton 2008; Yamanaka 2009; Huang 2011; Wang et al. 2011; and Furusawa

74 ◂ CHALLENGES OF EPIGENETICS

A ▸

◂ B

C ▸

FIG. 2.5. EL images showing (A) cell diff erentiation, (B) reprogramming, (C) de- and transdiff erentiation of cells. (B) and (C) show “antigravitational” (i.e., experimentally induced) processes. (B) During pluripotent reprogramming, induced pluripotent stem (iPS) cells (gray ball) are formed. (C) A mature cell is dediff erentiated or rejuvenated into a progenitor stage (gray ball) and directly transdiff erentiated into another mature cell (white ball) (Zhou and Melton 2008, 383, slightly modifi ed; reprinted with permission from Elsevier).

▸ 75 CHALLENGES OF EPIGENETICS

and Kaneko 2012).36 The latter procedure, in which a mature unipotent cell is rejuvenated as a stem cell, is often called “epigenetic reprogramming.” It includes experimentally changing the expression of a number of epigenetic regulatory factors (Takahashi and Yamanaka 2006; Hochedlinger and Plath 2009; Fisher and Merkenschlager 2010). Research on cellular reprogramming goes back to clone research on amphibians in the 1960s and 1970s (see Gurdon 2006), although at that time “reprogramming” was considered to be a controversial metaphor for the suggested reversibility of cell differentiation (Danielli and DiBerardino 1979). More recently, Kazutoshi Takahashi and Shinya Yamanaka (2006) demonstrated the induction of pluripotent stem cells from mouse embryonic or adult fibroblasts by introducing four epigenetic regulatory factors (Oct3/4, Sox2, c-Myc, and Klf4) to an embryonic stem cell culture. Inter- estingly, while stem cell biologists got along without any reference to Wad- dington’s work before 2006, about this time a widespread renaissance of EL models emerged in the debate on reprogramming. In this new and current context, EL models enable one to distinguish intuitively the two subfields of cell differentiation and cell reprogramming (including different experimentally induced processes of the latter) by the movement direction of the ball on an EL surface (downhill, uphill, across hills; see fig. 2.5). Besides this illustrative function of ELs in stem cell biology, in this field they play a crucial role in developing mathematical models of cell differentiation and reprogramming pathways. We might refer to the title of a paper by Jin Wang and colleagues (2011), “Quantifying the Waddington Landscape and Biological Paths for Development and Differentiation,” to outline this usage of Waddington’s model. More precisely, the EL should support quantitatively capturing the dynamic properties of gene regulatory networks (GRNs) underlying the phenomena of cellular development.37 Therefore, simple deterministic mathematical models based on ordinary differential equations (ODEs) are usually constructed. A system of ODEs should be able to represent changes in the number or concentration of each molecule in a network over time. These network dynamics (i.e., changes in gene expression profiles) can be visualized as developmental trajectories on an EL attractor surface, whereas each network state maps to one point in state space (fig. 2.6). However, ELs in stem cell biology should not be understood primarily as

76 ◂ CHALLENGES OF EPIGENETICS

FIG. 2.6. Illustration of network dynamics of a two-gene circuit in an EL-attractor surface image. (Left) wiring diagram of a two-gene network (mutual inhibition of genes A and B); (right) epigenetic landscape in two-dimensional state space. The gene expression profiles map to points in the state space, leading to an EL surface with two stable attractors (i.e., two distinct cell types). The arrows on the landscape depict flow vectors in cellular development determined by the gene network. The landscape potential (often referred to as “quasi potential”) is depicted as the surface’s elevation being inversely related to the likelihood of occurrence of a particular state in phase space. (Huang et al. 2009, 872, slightly modified; reprinted with permission from Elsevier).

precise representations of known GRNs but as “programmatic” toy models of dynamic cellular systems: “Although biologists not used to the concepts of network dynamics, state space and generalized potentials may mistake the intuitive landscape picture for an overstretched cartoon, it should be stressed that the landscape has a formal basis in the theory of dynamical, non-equilibrium systems . . . even if the specific details are not known yet” (Huang et al. 2009, 875; emphasis added). This passage suggests that (this group of) stem cell biologists, up to this point, have aimed at developing an approach to constructing formal models—generic or toy models—that manage stability characteristics of dynamic systems in similar, yet quantitative fashion as EL images. Often, a (preliminary) representative mechanistic description does not yet exist. In times of limited empirical input and little

▸ 77 CHALLENGES OF EPIGENETICS

knowledge of mechanisms in cell differentiation, quantification of specific dynamic phenomena has yet to come. Again, this procedure of quantification and formalization using basic work data from a few simple gene regulatory circuits is guided by Waddington’s landscape model as a visual corrective device: “To predict the actual, specific epigenetic landscape of the human genome, the detailed knowledge of the actual wiring diagram of the genomic GRN is required. . . . As we continue to gather the pieces of specific information needed to construct the GRN’s wiring diagram, we shall hence be guided by the broad vision of systems dynamics and the epigenetic landscape” (Huang et al. 2009, 875).38 At this stage, stem cell biologists’ EL images may be understood as placeholders having heuristic functions in the construction of models of real-world phenomena that at this point cannot precisely be represented. The role of ELs as guiding devices in stem cell epigenetics can be described in more detail by means of the modeling techniques developed by Sui Huang and colleagues (2009), illustrated in figure 2.6. Similar to Waddington’s interpretation of Israel Gel’fand and Mikhail L’vovich Tsetlin’s ravines method, Huang et al.’s approach to exploring and quantifying the topographical features of the landscape of a two-gene network involves testing multiple starting points of cell trajectories and/or modifications in gene expression profiles (i.e., changes of the network state) at different points in time. In both approaches, the network dynamics of the GRNs bring about two stable pathways of development, leading to two attractors (see fig. 2.6). Accordingly, these attractor surfaces are mathematically derived directly from the dynamics of the GRN. The Waddingtonian procedure of exploring or scanning and formally describing the landscape’s topographical features is usually completed by aligning the observed trajectories. This reveals the quantified valleys of the ELs—here, the two valleys leading to two attractors. These new, EL-based modeling efforts in stem cell biology are geared toward developing quantitatively mapped landscapes, which in the long run should help to predict the efficiency of cellular reprogramming experiments and to identify optional pathways of reprogramming. In other words, by visually grasping spatial distinctness of induced pathways and hills in ELs, methodological guidelines for reprogramming experiments can be intuitively derived.

78 ◂ CHALLENGES OF EPIGENETICS

Besides this heuristic function in model building, according to Wad- dington, ELs can be applied in order to unite different biological research areas. For example, Waddington was convinced that developing a consistent representational framework, which enables biologists to relate genetic data (the underside of the EL) to embryological data (the surface of the EL), helps to establish communication across disciplinary boundaries (see S. Gilbert 1991). According to Melinda B. Fagan (2012b), similar transdis- ciplinary communication efforts can be found in recent stem cell and systems biology approaches using ELs. These disciplines overlap in explaining cell development. Since both explanatory projects are in their early stages, joining forces seems to be worth a try. However, they differ in methods used. While stem cell biologists’ predominant method is experimental manipulation of cells and tissues, systems biologists prefer mathematical modeling. Against this backdrop, Fagan showed that the recently developed EL models are considered a consistent framework for establishing transdis- ciplinary research and communication and for comparing research data and methods. For example, new EL models enable researchers to visualize shared background assumptions, like the assumptions of an undifferenti- ated starting point of stem cells or the potential to develop along multiple trajectories. Moreover, they make it possible to correlate measurements of molecular state defined by epigenetic determinants of gene expression and developmental potential, such as totipotent or unipotent states, in stem cell experiments (see, e.g., Hochedlinger and Plath 2009, 510). This analysis of the usage of Waddington’s EL images in contemporary stem cell research shows that Waddington’s legacy is much broader than usually recognized. Accordingly, studies in the field of stem cell epigenetics, as opposed to the molecular research that focuses on epigenetic inheritance, have balked at adopting a novel and distinct understanding of “epigenetics.” Instead they closely tie in with the usage of the term as introduced by Wad- dington in cell differentiation and developmental genetics. Thus, Wadding- ton’s epigenetics is the conceptual and—more surprisingly—methodological origin of an influential and flourishing branch in modern epigenetics. This branch not only shows special interest in its own disciplinary roots and tries to preserve its historical continuity but actively seeks to take up traditional methodological approaches and develop them further.

▸ 79 CHALLENGES OF EPIGENETICS

INTRADISCIPLINARY CONFLICTS IN EPIGENETICS

In the recent debate on whether and how epigenetics expands the prevailing theoretical framework of biology, the overall field of epigenetics is often described as a field that consistently opposes the reductionist spirit of molecular biology and primarily focuses on (transgenerational) transmission of nongenetic information. Paradigmatically, Sylvana Barros and Steven Offenbacher (2009, 401; emphasis added) note that “epigenetics, as the term suggests, can be seen as a major turn away from molecular biology’s Central Dogma, recognizing that there are epigenetic inheritance systems through which non-sequence-dependent DNA variations can be transmitted in cell, tissue, and organismal lineages. Thus, current epigenetics not only offers new insights into gene regulation and heredity, but it also profoundly challenges the way we think about evolution, genetics, and development.” This view, according to which findings on epigenetic inheritance challenge the neo-Darwinian research program, has often been juxtaposed with the concept of the inheritance of acquired characteristics (Jablonka and Lamb 1995, 2008; E. Richards 2006; Sano 2010; Gissis and Jablonka 2011; see also chap. 1). However, the protest movement of epigenetics is far more heterogeneous and multifaceted than this epigenetic inheritance view suggests. Let us for this reason compare two different epigenetic standpoints currently calling for an expansion of the theory of evolution: the epigenetics of Eva Jablonka (see Jablonka and Lamb 1995, 2005, 2008; and Jablonka and Raz 2009) and of Sui Huang (2009b, 2012a, 2012b). While Jablonka can be described as the grande dame of molecular epigenetics, Huang, a stem cell and cancer biologist at the Institute for Systems Biology in Seattle, may be one of the most influential figures (perhapsthe most influential figure) in current research on the epigenetic foundations of cell differentiation. Interestingly, the differences in the two views of epigenetics by Huang and Jablonka, as well as the conflicts and challenges arising from them, have not yet been widely recognized in the field. So far, only one side—Huang—is paying attention and criticizing the other. Table 2.1 offers a comparison of the two views in tabular form. Huang’s epigenetics departs in at least three ways from Jablonka’s conception of epigenetics and molecular epigenetics. First, it differs in its explanandum phenomena. Huang (2012a, 153, S1; 2012b, 73, 76) does not focus on revealing (heritable) chromatin modifications as the (or an) essential driv-

80 ◂ CHALLENGES OF EPIGENETICS

TABLE 2.1. JABLONKA’S EPIGENETICS VERSUS HUANG’S EPIGENETICS

JABLONKA’S EPIGENETICS HUANG’S EPIGENETICS

Research focus Nongenetics molecular factors Gene networks Interest in inheritance + 0/− Interest in evolutionary issues + + Neo- or quasi- Lamarckian dimension + 0 Call for an extended synthesis + + in terms of including developmental biology + 0 in terms of a methodological renewal − + Network thinking 0 + Uniformity of methods used − +

Annotations: + = “is central to the research field”; 0 = “is taken into account in the research field”; − = “is of little to no relevance to the research field.” ing force of gene regulation. Instead he seeks to reestablish a Waddingto- nian concept of epigenetics, which is basically an offshoot of developmental genetics that focuses on the etiology of the complex genotype-phenotype map. Thus, Huang (2012a, S1) criticizes, “They [i.e., molecular biologists] employ ‘epigenetics’ to describe a distinct and narrow set of phenomena [i.e., chromatin modifications] that is only loosely related to the original meaning intended by Waddington” (see also Huang 2012b, 73). Second, Huang’s “old-school epigenetics” departs from molecular genetics not only conceptually but methodologically. He, among others, utilizes

▸ 81 CHALLENGES OF EPIGENETICS

and refines Waddington’s methodological framework of the EL and criticizes the common metaphorical usage of the model in epigenetics: “since the term ‘epigenetic’ is used in molecular biology to describe covalent modifications—which does not do justice to Waddington’s original ideas . . . —his landscape is continuously being (mis)interpreted in a loosely meta- phoric manner” (Huang 2009b, 554). In other words, he blames those areas of epigenetics focusing on nongenetic processes of inheritance for, onomasiologically speaking, neither investigating the correct phenomena—namely, instead of entire GRNs, just a part of these networks, that is, nongenetic markers—nor adopting a formalized methodology and a true systemic network perspective in order to approach these phenomena. In particular, Huang asks epigeneticists to distance themselves from a monocausal manner of explanation, like “methylation of gene X caused a change in phenotypic trait Y.” 39 Rather, epigenetic complexity should be explored and explained by means of a mathematical framework. Based on this methodologically related criticism, his approach departs from simplistic views of the genotype-phenotype map, both in neo-Darwinism and epigenetics: “Changes in epigenetic marks, or ‘epimutations,’ are actually the conceptual cousins of genetic mutations and thus, share with the central dogma the limitations of linear thinking: they affect individual gene loci and thus, by invoking them in reasoning about inheritance[,] one implicitly follows the scheme of a mono-causal mapping from (epi)genotype to phenotype of the central dogma. Therefore, epimutations cannot come to the rescue of the Neo-Darwinian framework in view of the non-linear genotype- phenotype relationship” (Huang 2012a, S2). Instead of propagating a monocausal and linear relationship between genotype or epigenotype and phenotype, a more complex Whiteheadian picture of these connections should be established. Following Waddington, mediated and nonpredictable changes on the level of phenotypic traits should now be understood as effects of changes in components of complex underlying gene networks. Third, Jablonka and Huang have a different understanding of the role that epigenetics plays in evolution. Similar to Jablonka and other evolutionary- minded molecular epigeneticists, Huang has pursued a research program that investigates the evolutionary relevance of nongenetic variation above DNA, for example, at the level of enzymatic regulation, which is brought about by gene networks. However, the emphasis does not lie on the phenomenon of independent inheritance of this epigenetic variation (see Huang

82 ◂ CHALLENGES OF EPIGENETICS

2012a, 156). Rather, the main thesis he defends is that, as a result of emergent nongenetic characteristics of the interaction of genes in networks during development, phenotypic evolution can be given a certain direction. Pro- vided that this process leads to an adaptive trait, this developmentally determined direction may then be fixed by genetic variation in the genome over the long run. In this process, epigenetic information stored “outside” the DNA is understood as a product of systemic interaction in gene networks: “But what ‘epigenetic’ apparatus can universally provide information encoding and storage without altering the DNA sequence? The answer is: the GRN (. . . or briefly, ‘gene network’)—the network established by the fact that genes influences sic[ ] the expression of other genes via a web of molecular regulatory interactions encoded in the genome” (Huang 2012a, 150–51; emphasis added). To be more precise, for example, stochastic variation— so-called gene expression noise—in one and the same gene network may bring about a number of variations in gene expression profiles. The stability characteristic of these variations can then be formally described and topographically represented by an EL approach. In short, the attractor surface of an EL represents an “intermediate phenotype,” as Huang (2012a, S4) calls it. In order for this intermediate epigenetic phenotype (and the interacting genes underlying this gene expression profile) to be selected, however, it also has to be stable across generations. Interestingly, Huang’s stem cell epigenetics shares some features of the work of orthodox-minded population geneticists. Both question whether nongenetic (epigenetic) inheritance patterns are stable enough to affect population dynamics (see, e.g., Huang 2012a, S2; 2012b, 73). According to this view, what is needed is the inclusion of genes’ networklike patterns of interaction in order to coordinate expression profiles in the genome—the how, when, and where of gene activity—thus bringing about intra- and intergenerationally stable epigenetic patterns. Epigenetic markers alone are nonspecific with respect to a particular DNA sequence or gene locus. They are considered “blind” units that suffer from an intrinsic lack of function and therefore require genetic guidance (Huang 2009b, 549; 2012a, S2). Moreover, reflecting somewhat Richard Dawkins’s (1976, 1978) concept of the gene as a replicator, Huang (2012a, 154) defines gene networks according to the criteria of continuity, stability, and invariance: “The network structure has been wired by evolution and hard-coded into the genomic

▸ 83 CHALLENGES OF EPIGENETICS

sequence. It is intrinsic to and, in the absence of mutations, invariant for a given genome (organism). Thus, each cell in metazoa has the same GRN.” Thus, the structure of the network can be changed only over phylogenetic time, not ontogenetic time.40 This strict Mayrian parting between two different time frames, an evolutionary one and a developmental one, contradicts Jablonka and Lamb’s (2005, 102) idea of overcoming the dichotomies of ontogenesis versus phylogenesis and of proximate causes versus ultimate causes. Again, although Huang’s approach admits that preexisting genetic variations alone, without any new mutations, may bring about emergent nongenetic variation that can be selected, he does not explicitly argue for the independent evolutionary relevance of the heritability of this variation. A new adaptive gene expression profile has to be brought about in every generation anew through genetic susceptibility (to gene expression noise) or through persistent environmental conditions until the genes take the helm and fix the new trait and the new EL topography, respectively. To explain this process, Huang’s systemic stem cell epigenetics needs only Mendelian genetics, not transgenerational epigenetic inheritance, such as maternal effects or germ-line mediated inheritance. Accordingly, once the importance of epigenetic inheritance in epigenetics is downgraded, discussing the field’s Lamarckian dimension becomes far less important.41 In summary, juxtaposing Jablonka’s and Huang’s approaches reveals the conceptual and methodological heterogeneity of the field commonly called epigenetics. This intradisciplinary diversity and its associated conflicts are in stark contrast to the commonly propagated view of epigenetics as an innovative young science that consistently rebels against neo-Darwinian conservatism and gene-centrism by means of a novel theory of nongenetic inheritance. In fact, contemporary epigeneticists asking for uniting approaches in systems and stem cell biology often favor gene-centered explanations that form a contrast to catchy slogans such as “Today’s Molec- ular Epigenetics: Turning Away from Gene-Centrism,” offered by Linda Van Speybroeck (2002, 78). However, this interest in genes is often guided by a systemic Waddingtonian view of gene networks that distances itself from linear, monocausal models of gene regulatory networks and genotype- phenotype relationships. In contrast, these studies highlight the nonlinear and nonadditive effects of genes on the phenotype.42

84 ◂ CHALLENGES OF EPIGENETICS

Like Jablonka, Huang calls for expanding the theoretical framework of the modern synthesis, although the latter call has not yet resonated with the advocates of an extended evolutionary synthesis. Interestingly, as the prin- cipal plaintiff of evolutionary epigeneticists interested in cell differentiation and reprogramming, Huang presents a research framework for investigating complex developing systems that rejects reductionist tendencies and linear thinking not only in neo-Darwinian mainstream epigenetics but also in molecular epigenetics. From a historical point of view, Huang seeks to legitimize his intradisciplinary criticism by building on Waddington’s classical work and thus main- taining a conceptual and methodological continuity of epigenetic research. On that account, advancing to a more nuanced understanding not only of the concept of contemporary epigenetics but also, and especially in the extended evolutionary synthesis debate, of epigenetics’ capacity for innovation in modern biology necessitates considering the variety and diversity of the epigenetic research program, its historical development, and the resulting internal conflicts that hinder theoretical integration of epigenetics.

THEORETICAL INTEGRATION AND PLURALISM

The intradisciplinary tensions described above concern the questions of how one should understand the field of epigenetics conceptually and methodologically, as well as how epigenetics should explain. To conclude this chapter, I sketch some potential solutions for integrating systemic stem cell epigenetics and molecular epigenetics of inheritance phenomena. From a pragmatic perspective, it is worth mentioning that such an integration would be an asset for molecular epigenetics. It would not only provide a strong formal framework for investigating how environmental cues, on the one hand, and the stochastic nature of gene expression as well as the dynamic stabilities during (cell) development, on the other hand, work together. Moreover, it would also facilitate epigenetics’ overall integration with the theoretical corpus of the modern synthesis. Those who are still skeptical about the evolutionary role of transgenerational epigenetic inheritance can easily embrace (another part of) epigenetics, as stem cell epigenetics, which, in contrast to “Jablonka’s epigenetics,” lacks a distinct, less neo-Darwinian and gene-centered theory of inheritance. Therefore, stem

▸ 85 CHALLENGES OF EPIGENETICS

cell epigenetics in turn has to accept more seriously that the nonlinear and stochastic nature of GRNs affects not only developmental pathways but nongenetic inheritance processes in an emergent manner. In order to outline a way for integration in this case, a look at similar cases might be helpful. Philosophers of science have offered a couple of solutions suggesting how to better integrate studies on GRNs and cell differentiation in systems biology with other studies in evolutionary and developmental biology (see Green et al. 2015; O’Malley et al. 2015).43 They argue that integration is promoted by labor division of explanations. Formal methodological accounts, such as the one of Huang, offer mathematical models that allow for predictions of developmental pathways (especially cell fate). These quantitative descriptions of GRN dynamics (i.e., changes in gene expression profiles) usually come in the form of ordinary differential equations that follow some general principles, which are said to apply to all developmental dynamics, such as bifurcating patterns and metastability of network states that guarantee multiple developmental trajectories. For example, Huang (2011, 2249) states that “whatever stem cells do, the fundamental laws governing the underlying regulatory systems must be obeyed.” According to Sara Green, Melinda B. Fagan, and Johannes Jaeger (2015), this law-based explanatory account can, in principle, be complemented by experimental methodologies that offer more detailed mechanistic models of GRNs’ properties and an experimental test of theoretical predictions. In line with this idea of integrating complementary explanatory and methodological programs, one may add with respect to epigenetics that experimental molecular methodologies that focus on the inheritance of nongenetic variation produced by GRNs can provide an interesting corrective for stem cell epigenetics. The former offers the latter valuable insight on whether the principles taken to be fundamental for modeling developmental dynamics also work with the dynamics of epigenetic inheritance. Following Melinda B. Fagan (2012b), one could argue that this kind of labor division can be facilitated through the framework of the epigenetic landscape. As outlined above, the landscape images allow for mediating between distinct methodological approaches that can plot and compare their results on these images. For example, experimental manipulation can specify the underside of the landscape (i.e., Waddington’s pegs), while formal modeling makes it possible to describe quantitatively the dynamics of a particular GRN as pathways of a developing system on the landscape

86 ◂ CHALLENGES OF EPIGENETICS surface over time. Huang (2009b, 547) describes the value of these images as follows: “Using permissively simplifying and pedagogically intuitive pictures, we hope to open the eyes of experimental biologists not familiar with dynamical systems to a set of principles that will benefit our thinking about the sources of stability and flexibility of cellular states.” In other words, sharing a common visual framework may help researchers to agree on basic principles and assumptions, as well as important physical entities and interactions of interest, and thus facilitate communication about their research results. In this manner the epigenetic landscape may again unfold its heuristic potential in epigenetics. While Waddington used it as a tool for integration at the interface of genetics, embryology, and evolution, today it may again integrate different approaches—causal-mechanistic and mathematical—under the broad idea of “epigenetics.” However, such representational similarities should not suggest that epigenetics can, in the end, be fully theoretically unified or even that certain research results and perspectives can be reduced or substituted by others, such as the idea that epigenetic inheritance will finally be explained in terms of laws only.44 Instead, similar to what we find in other multifaceted research frameworks in the biosciences, such as the research on protein folding, in epigenetics the distinct perspectives on different levels of organization, such as in silico simulations of the physics of GRN dynamics, as well as in vitro and in vivo manipulations of chemical components in networks, cells, and tissues, provide input to answer each other’s research questions in a complementary manner.45 In other words, they seek to inform rather than determine each other. For example, causal explanations of the stochastic patterns of transgenerational epigenetic inheritance can be enriched by mathematical models of the role of developmental noise in GRNs, but the former cannot be reduced to the latter. Every single perspective provides only a partial account of epigenetic processes. According to the more pluralist perspective above, the current discussion on which of these different research perspectives is thereal epigenetics seems misguided. In epigenetics, there cannot be one clear-cut question with one definite answer, nor can the different accounts be simply unified to support an overarching theory of epigenetics. Rather, coordinating tools such as the epigenetic landscape have to be further developed to more suc- cessfully interlink data that have been gathered from different levels of organization and thus increase crosstalk among different epigeneticists.46

▸ 87 CHALLENGES OF EPIGENETICS

This result is in line with the view described earlier—that epigenetics is a Wittgensteinian cluster concept, according to which epigenetics has no clear-cut boundaries and shows some features not relevant for all of its single instantiations. In other words, epigenetics’ various accounts, such as systemic stem cell epigenetics and molecular epigenetics of inheritance phenomena, only show family resemblance. This view is similar to the way Rasmus G. Winther (2015) describes the explanatory and methodological plurality of the related field of evo-devo. He understands evo-devo as a “trading zone” with various combinations of research paradigms and styles.47 In a similar manner, in the trading zone of epigenetics, differentparadigms —neo-Dar- winism, Lamarckism, gene-centrism, and postgenomics—as well as styles— experimental, observational, and mathematical—coexist simultaneously and are loosely interlinked. As in evo-devo, the relationships among these paradigms and explanatory styles are complex, because the latter are realized in the former in multiple forms. ▶ ▶ ▶

The integration of epigenetics into the theoretical framework of modern biology in general and of evolutionary biology in particular faces several conceptual and methodological challenges. First, in order for the developmentally oriented view of epigenetics to become involved in evolutionary biology, two interdisciplinary issues have to be addressed. First, how can molecular epigenetic explanations that are based on highly artificial experimental setups address evolutionary explananda? Second, how can epigenetic explanations that focus on developmental proximate causes rather than on ultimate causes address evolutionary explananda? The first issue has been addressed by showing that studies of epigenetic processes that are experimentally induced in the lab (in molecular epigenetics) and those observed in natural populations in the field (in ecological or evolutionary epigenetics) are not that different after all. They share a similar methodological framework, one that allows them to pose heuristically fruitful research questions and to build reciprocal transparent models. The second issue above becomes far less fundamental if one understands the predominant reading of Mayr’s classical proximate-ultimate distinction as offering a simplifying picture of what (and how) developmental explanations actually explain. Once the nature of developmental dependencies has been revealed,

88 ◂ CHALLENGES OF EPIGENETICS the appropriateness of developmentally oriented approaches, such as epigenetics, in evolutionary biology is secured. As for the second issue, besides these interdisciplinary conflicts between epigenetics and evolutionary biology, there are also intradisciplinary conflicts about how to understand the field of epigenetics conceptually or methodologically. The heterogeneity of epigenetics and its inner conflicts become apparent when one juxtaposes the approaches of prominent advocates of stem cell epigenetics and molecular epigenetics. In addition, this analysis shows that Waddington’s “old-school epigenetics,” including its comprehensive methodological program, is still very much alive in epigenetics. These different perspectives of epigenetics need to be integrated more thought- fully in order to complement each other, rather than having one simply be substituted for the other.

▸ 89 THREE CAUSAL EXPLANATION

Man’s mind cannot grasp the causes of events in their com- pleteness, but the desire to find those causes is implanted in man’s soul. LEO TOLSTOY, 1868–69

Manipulation of the variables in a model, or simulation, is intended to provide answers to the critical question, “What would happen if . . .” CONRAD HAL WADDINGTON, 1977

SO FAR WE HAVE seen that epigenetics will have to travel a rocky road to be integrated into mainstream biology. The difficulties on this path, -how ever, exist not just for historical reasons, such as the methodological traditions addressed in the previous chapter. As this chapter and the next show, a more detailed philosophical investigation brings to light a number of additional features in epigenetics that set it apart from the way many scientists understand molecular and cell biology. These features and their associated challenges refer to the concept of scientific explanation. CAUSAL EXPLANATION

Since the mid-1990s, the nature of biological explanation has been the focus of various philosophical analyses. In particular, philosophical accounts of explanation have turned out to be inadequate for explaining biological phenomena, and the existence and explanatory role of laws in biology have been questioned. As a reaction to this period of uncertainty, a very influential position has been developed. This position, adopted by many philosophers of science, draws heavily on James Woodward’s interventionist account of explanation (Woodward 1997, 2003, 2004, 2007a; Woodward and Hitchcock 2003). It serves as a blueprint for more detailed accounts of both causal explanation and mechanistic explanation in biology (see, e.g., Waters 2007; and Craver 2007). While the former kind of explanation relates biological events, the latter relates causal properties of biological entities, usually across different levels of organization. The former traces diachronous causal relations and the latter, synchronous constitutive relations. This chapter focuses on causal explanation in epigenetics before turning to constitutive explanation. I show here that the interventionist account elucidates and justifies the way that epigeneticists in molecular biology infer causality in complex living systems. Therefore, experimental work on epigenetic inheritance phenomena, as well as the comprehensive methodological framework that rests upon manipulation of inducing environmental variables and multifactorial experimentation, is reviewed. In addition, I discuss whether the explanatory practices of nonmanipulative field studies in evolutionary epigenetics, described in the previous chapter, fit the interventionist schema. Then, against the popular belief that interventionist explanation is in urgent need of supplementary information gathered from lower levels (so-called “mechanistic detail”) that addresses why a certain causal dependency relation exists, I argue that tracing interventionist causality forms an exhaustive explanatory approach in biology. This argument has two components. First, more fundamental factors such as genes do not have a unique ontic or epistemic status, and this means that it is not necessary to list them as a mechanistic detail in every explanans addressing biological complexity. In other words, genes are considered to be causal factors that often have a mere potential, non-actual difference-making effect on phenotypic traits. Second, the populations in which higher-level nongenetic factors are identified as actual difference-making causes are distinct from the actual populations in which genes make a difference. These arguments imply the autonomy of

▸ 91 CAUSAL EXPLANATION

higher-level epigenetic explanation, which requires little or no mechanistic detail in molecular and cell biology.1

THE INTERVENTIONIST ACCOUNT OF CAUSAL EXPLANATION

The idea that causal relationships are somehow exploitable for purposes of control and manipulation has intuitive appeal—especially, but not exclusively, to biologists who are experimentally acting upon phenomena- to-be-explained. Francis Bacon ([1620] 1994) famously advocated this idea in his Novum Organon. According to Bacon, we can only uncover and explain the hidden causal structures of nature by interacting with them. In other words, we have to manipulate putative causes in order to reveal them. This view of causal explanation was subsequently developed further in the twentieth century by philosophers as well as scientists (Collingwood 1940; von Wright 1971; P. Holland 1986; Menzies and Price 1993; Freedman 1997; Pearl 2000). It is commonly called the manipulationist or interventionist account of causal explanation. The discussion of causal explanation in epigenetics presented below focuses mainly on James Woodward’s contribution to the interventionist account. In contrast to the other notions of manipulation, Woodward refers to a special kind of manipulation relevant to causal explanation, which he calls interventions. Interventions are not based on any kind of human agency, which is often thought to be necessary for them to be carried out.2 In other words, Woodward’s interventionist account offers a non-anthropocentric account of causal explanation. According to Woodward, the invariance of a generalization under interventions on the value of its variables is what matters in scientific explanation, in contrast to its lawlikeness, for example.3 The basic idea of his interventionist theory of causation (ITC) can be described as follows:

(ITC) A variable X causes variable Y if and only if {(i) there is a possible manipulation (intervention) of some value of X that will change the value of Y (or the probability distribution of Y) and (ii) this relationship would remain stable}.

92 ◂ CAUSAL EXPLANATION

This formula can be summarized under the slogan “invariance under interventions.” Type-causal claims (the type of causal level that Woodward’s account best suits) relate variables that can take more than one value. In this sense, X causes Y, if X, as a result of an intervention, takes one of the values, for example, {ingestion of aspirin, no ingestion of aspirin}, and if Y, following this intervention on X, would take one of the values {relief from headache does occur, relief from headache does not occur}, and if interventions on X (i.e., changes in the value of X), are accompanied by systematic changes in the value of Y (or the probability distribution of Y). In other words, every time we tweak the X cause, the effect Y will accordingly shift. In addition, the following criteria have to be met in order to derive a generalization correctly, according to principle (ITC):

(iii) An intervention I does not directly change Y via a route that does not go through X (iv) I is the only cause of X; the value of X is entirely set by I (v) I does not affect other causes of Y such as an off-path variable, say Z, which is not on the pathway X→Y, but connects these variables through another causal pathway. Only those variables that are intermediate causes of Y on the pathway X→Y (if there are any) are probabilistically dependent on I. (Woodward 2003, 94–99; 2009, 247)

For instance, consider an experiment designed to determine whether eating bananas (variable B that takes one of the two values 1 and 0) causes recovery (R) from a hangover by replenishing the potassium (P) lost during last night’s party. To determine the validity of the experiment we need to rule out the possibility that there are causes of R that are probabilistically dependent on an intervention I on B and that affect R independently of the causal pathway I→B→P→R. As shown below, such an experimental design—fixing off-path variables by additional manipulations—provides the methodological basis for causal explanation in molecular and evolutionary epigenetics. As (ITC)’s condition (ii) above indicates, the interventionist account is a variant of the counterfactual account of causation, since the test for sys- tematically changing invariant relationships between variables relies on a counterfactual notion of invariance. As Woodward notes, “It has to do with whether a relationship would remain stable if, perhaps contrary to actual fact, certain changes or interventions occur” (2003, 279, emphasis added;

▸ 93 CAUSAL EXPLANATION

see also Woodward 1984). The creation of these hypothetical, counterfactual situations is guided by asking so-called what-if-things-had-been-different- questions, or w-questions. He argues that any successful causal explanation ought to be able to answer w-questions by presenting a “hypothetical or counterfactual experiment that shows us that and how manipulation of the factors mentioned in the explanation . . . would be a way of manipulating or altering the phenomenon explained” (Woodward 2003, 11). Following this explanatory approach, scientists derive causal generalizations from counterfactual situations that would be invariant—they would hold—under some appropriate set of interventions. For the experiment described above, an interventionist reading of the causal generalization “Eating bananas causes recovery from hangover by replenishing the potassium level” is invariant under a set or range of interventions that could be performed on the cause variable (eating bananas) to change the value of the effect variable (recovery). Concerning this set of manipulations, this generalization has explanatory power. The interventionist account includes three important ideas: difference making, change-relatedness, and contrastive explanation. Conceiving of the explanans in causal explanations as a difference maker means thatX explains why Y, because if X had not happened, then Y would not have happened either. Here, the cause X is the difference maker of values of Y (see Mackie 1974; P. Lipton 1993; Woodward 2003; Menzies 2007; Waters 2007; Strevens 2008). In other words, if the explanans had been different, the explanandum would have been different as well. In order to perform a difference-making intervention, one has to have a well-defined notion or idea of what it would mean to change a certain generalization. Moreover, this generalization should describe how changes in the value of some of its variable(s) are related to one another. If this criterion is met, this generalization is labeled change-relating. In contrast, a generalization is non-change-relating and thus non-explanatory if there is no well-defined notion of what it would mean to change a value (or values) of a term (or terms) or what it means to represent the generalization’s terms as variables. Moreover, a generalization might preclude changing the properties of its own terms, as many laws, like “all noble gases are chemically inert,” do. We will return to change- and non-change-relating generalizations in chapter 4. Change-relatedness and difference making are natural partners. Change-

94 ◂ CAUSAL EXPLANATION relating generalizations link changes in the value of different variable(s). In addition, difference making is closely linked with the idea of a contrastive explanandum according to which our explanations do not simply address questions like “Why this?” but more specific ones, like “Why this rather than that?” (see Woodward 2003, 67, 145–46). This contrastive structure naturally forces us to be more specific about what the explanandum is and in what respect it could have been different. If we combine the ideas of difference making and contrastiveness, we obtain a powerful heuristic: “First, you create, find, or imagine the difference to be explained, and then you proceed to find the differences between the cases. Then you test whether these -can didates can really make the difference, by testing whether they can bring about the difference to be explained. If the procedure does not work, you can try to be more precise about the explanandum, and try again” (Ylikoski 2013, 290; original emphasis).4 This account of causal explanation has some advantages over other philosophical accounts of explanation, especially Carl Hempel’s (1965) covering- law account. Hempel’s basic idea is that a phenomenon is explained if some law covers it as a special case. Besides the influence of the covering-law account on philosophers of science (primarily dealing with the exact sciences) in the twentieth century, both the applicability of this account to the biological sciences as well as the existence of laws in biology more generally have been questioned (see Salmon 1984, 28–32; 1989, 46–60, 68–80; Kitcher 1989, 411–14; Beatty 1995). For example, the covering-law account does not allow for explanations of phenomena that have low probabilities of occurrence. In contrast, according to the interventionist view, for a treatment against a low potassium level (hypokalemia) in human cells to be a cause of recovery from hangover headaches by restoring the functioning of nerve cells, this treatment does not have to raise or lower the probability of the occurrence of the effect in every background condition. It merely should do so under some of the possible interventions on the cause and some background conditions. Thus, this new interventionist account of explanation serves well the idea of explaining dependencies in complex biological systems that very often lack stable probabilities.5 In addition, the interventionist account can be used to explicate what it is about causal laws that makes them explanatory. Explanations are no longer considered to be explanatory due to their argumentative structure but due to their ability to correctly describe invariant relationships. Thus,

▸ 95 CAUSAL EXPLANATION

lawlike invariant generalizations can now be understood as a special case of invariant interventionist ones. The former are simply more invariant (and/ or extrapolable) than the latter. Given these advantages of the interventionist account, the indispensability and necessity of laws for explanation in biology can, at least, be questioned.

CAUSAL EXPLANATION, FROM EXPERIMENTAL PHYSIOLOGY TO EPIGENETICS

Does the interventionist account of causal explanation presented above suitably describe causal reasoning in biology in general and experimental and observational epigenetic research in particular? Moreover, does it allow us to better understand and perhaps even justify certain explanatory practices in these fields? Among others, C. Kenneth Waters (2007) and James Woodward (2001, 2010) have argued that the interventionist view accords well with the intuitions and practices of many biologists with regard to experimentation and explanation. In fact, biologists like to tweak things. Since the seventeenth century, experimental biology has tried to uncover the secrets of nature with increasing precision by using physical manipulation. This methodology in turn has had a far-reaching impact on the way questions were asked and problems were solved in a variety of experiment-based and even non-experimental disciplines. To elucidate the growing influence of manipulation in biology, leading directly to modern epigenetics’ methodology and causal explanation, let us first consider a historical case of interventionist explanation by Claude Bernard.

EXPERIMENTAL PHYSIOLOGY

Bernard laid the foundation for nineteenth-century experimental physiology by developing unique techniques to manipulate entities in living systems. Additionally, he is known for coining the term milieu intérieurs (the internal environment), better known as homeostasis, which can be understood as the sum of any process in living organisms that actively maintains stable internal conditions necessary for survival (see Gross 1998; and Noble 2008a).

96 ◂ CAUSAL EXPLANATION

Let us take a closer look at one of his studies on the relationship between the color of the blood in the renal vein upon exciting the active kidney and the urine flow through the urethra (urination), described in his famous Introduction à l’étude de la médecine expérimentale. Bernard ([1865] 1949, 155) began this investigation by asking, “What could be its [i.e., the correlation’s] cause?” He proposed a hypothetical causal structure—the coloring of the blood is causally related to the functional state of the kidney—and developed an experiment that was meant to answer this question: “It occurred to me that the red coloring of the venous blood might well be connected with the secreting active state of the kidney. On this hypothesis, if the renal secretion was stopped, the venous blood should become dark: that is what happened; when the renal secretion was reëstablished, the venous blood should become crimson again; this is also succeeded in verifying whenever I excited the secretion of urine. I thus secured experimental proof that there is a connection between the secretion of urine and the coloring of blood in the renal vein” (Bernard [1865] 1949, 155–56; emphasis added). Bernard’s method is governed by the idea of deducing from a causal hypothesis what would happen to the observed association if particular variables are controlled and fixed at a certain value, respectively. Accordingly, we can articulate a w-question guiding his investigation: “What difference would it make for the color of the venous blood leaving the kidney if the renal secretion of the kidney was stopped?” To answer this question, Bernard performed a number of innovative experimental interventions on the kidney variable. For instance, one such intervention entailed changing the value from {active kidney} to {inactive kidney}, thus stopping the secretion. This caused the blood-color variable to change its value. In fact, we often do not find such w-questions being asked explicitly in manipulative experimentation literature. But often, as in the case of Bernard’s experiment, we can extrapolate from the counterfactual structure of hypothetical explanations that coming up with these explanations is based on implicitly asking w-questions. Ber- nard’s hypothetical causal explanation is, “If the renal secretion was stopped, the venous blood should become dark” ([1865] 1949, 155). To sum up, Bernard generated a counterfactual pattern of association, in which some variables do not vary naturally, and then compared it with the initial state and the predicted outcome. Assuming other confounding variables are controlled for as well, this contrastive approach enabled him to articulate a difference-making explanans (secretion in the kidney, i.e.,

▸ 97 CAUSAL EXPLANATION

the oxyhemoglobin content of the renal blood) that is able to address the explanandum (blue- or crimson-colored blood).

CLASSICAL AND MOLECULAR EPIGENETICS

Subsequently, the manipulation-based approach to causal reasoning has been influential beyond the field of physiology. For example, Conrad Hal Waddington’s (1942a, 1956) work on canalization and genetic assimilation was founded on experimentally induced changes in embryogenesis via interventions on developmental factors. In a series of experiments on genetic assimilation, he induced phenotypic changes by exposing newly laid Dro- sophila eggs to ether. He then selected for the so-called bithorax phenotype (conversion of the meso- and metathorax), created by some individuals in response to the ether treatment. After several generations, this new phenotype occurred even in the absence of the initial environmental influence. In addition, when he defined the term “epigenetics” as the science investigating developmental causality, he located causal explanation in the center of epigenetics right from its beginning. However, it can be questioned whether his experimental methodology and his view of causality in development are, in fact, built on counterfactual reasoning and the idea of answering w-questions. There is strong initial evidence that this was indeed the case, as Waddington (1977a, 207; emphasis added) himself clearly emphasized that “manipulation of the variables in a model, or simulation, is intended to provide answers to the critical question, ‘What would happen if. . . .’” For now, let us turn to an investigation of epigenetic inheritance in modern molecular biology in order to show that the basic tenets of interventionism can make sense of, and moreover justify, how causal models are built and how causality is inferred from experiments in contemporary epigenetics. In order to estimate the theoretical impact epigenetics could have on the notion of heredity as used by neo-Darwinians, the molecular processes leading to intra- and transgenerational epigenetic effects (see chap. 1) have to be rendered more precisely. A first step toward this will be to estimate the degree of sensitivity and responsiveness to environmental influences that these systems exhibit, before turning to inheritance phenomena. Manipulating environmental variables and observing the corresponding effects in the development of traits is perhaps the most straightforward way to achieve this goal. In fact, many molecular studies on epigenetic depen-

98 ◂ CAUSAL EXPLANATION

A ▸

◂ B

FIG. 3.1. Two path diagrams representing the controlled experiment by Vastenhouw et al. (2006) on transgenerational gene silencing in the nematode C. elegans through manipulation of environmental, epigenetic, and genetic variables. (A) The hypothetical causal system before experimental manipulation; (B) the causal system during (and after) manipulation. The question mark indicates the relationship under investigation

(i.e., inheritance of phenotypically relevant epigenetic variation). Environmentwt rep-

resents the environment before manipulation, and variable environmenti represents

changed nutrition only in the parental generation F0. In path diagram (A), epigenetic as well as genetic variation is assumed to show wild-type values (wt); the green transgenerational phenotypewt can be induced by environmentwt in every generation

anew without inheritance. In (B) intervention 1 (I1; changing nutrition in F0) changes the phenotype of the progeny to nongreen phenotypei; causal off-path variables (envi-

ronment and genetic variation) are controlled for by I1 and I2 (i.e., using genetically identical organisms) (A, B: Baedke 2012, 159).

▸ 99 CAUSAL EXPLANATION

dencies intervene on environmental variables by, for example, using phar- macological or toxicological agents, by providing particular nutrition, or by exposing organisms to an enriched environment that includes novel objects and elevated social interactions to induce certain (heritable) phenotypic effects. These methods take into account the high degree of developmental responsiveness and variability in complex epigenetic systems. In addition, genetic and environmental off-path variables have to be controlled for while tweaking epigenetic processes to preclude the causal influence that any genes or environmental factors may have on the observed (transgenerational) phenotypic effects. Thus, multifactorial experimentation, a tool well known from genetics, takes the next step in epigenetics: epigeneticists manipulate three different types of causal factors—epigenetic, environmental, and genetic. But are these manipulations really interventions in the sense described by Woodward’s account? To address this question, I point to an experimental study done by Nadine Vastenhouw and colleagues (2006) that exhibits a commonly used design in experimental epigenetics. As noted in the previous chapter, they investigated long-term transgenerational gene silencing in the nematode Caenorhabditis elegans mediated by a process known as RNA interference (RNAi).6 They used a mutant strain with a gene (gfp) expressing green fluorescent protein (GFP) under the control of a germ-line-specific promoter; worms that expressed GFP in the germ line fluoresced green when exposed to ultraviolet illumination. They then fed the animals bacteria that express double-stranded RNA (dsRNA) homologous to the gfp-DNA sequence. This dsRNA triggered silencing of the gene gfp in the germ line and led to a heritable weaker green phenotype with reduced or no GFP expression. By selecting for this new phenotype, it is possible to produce gene silencing that could be inherited for at least eighty generations. Figure 3.1 shows two path diagrams of the assumed causal system: before (fig. 3.1A) and after (fig. 3.1B) the change in nutrition. Figure 3.1A depicts the green worm line, while figure 3.1B represents the line from green worms to ordinary worms. This experiment by Vastenhouw and her colleagues can easily be given an interventionist reading. The following causal generalization

(G) Environmentally induced gene silencing (i.e., epigenetic variation) produces a new phenotype that is heritable over many generations of sexual reproduction.

100 ◂ CAUSAL EXPLANATION is invariant under intervention on the value of the environmental variable (cause) intended to change the value of the transgenerational phenotype (effect). According to Woodward (2003), this dependency relation is an explanatory relation, because the manipulations performed in this experiment show the specific kind of causal structure described by the notion of intervention:

Intervention 1 (I1) on the environment variable avoids affecting other environmental and genetic causes of the phenotype. By feeding green worms

bacteria expressing double-stranded RNA only in the F0 generation, the new phenotype cannot be induced by the environment in every generation anew. As a consequence, the only way for nongreen nematodes to develop in the next generations is by inherited variation. In addition, by using nonmutagenic substances, a change in the genetic material can be prevented. This is depicted in figure 3.1B as two arrow-breaking events on

the pathways “environment (F0) → environment (F1+n)” and “environment → genetic variation.”

Intervention 2 (I2) on the genetic variation variable represents the technique of using a strain of genetically identical (isogenic) organisms. Thus, the genetic variation variable is fixed to value {wt}, the green worm value. This procedure is depicted in figure 3.1B by two arrow-breaking events on the pathways “genetic variation → epigenetic variation” and “genetic variation → phenotype.”

By means of these interventions Vastenhouw and colleagues were able to come up with the interesting causal explanation (G). (G) holds that by a single environmental change, one can change a trait not only in a single worm but also in its offspring, even though this event does not directly influence the worm’s offspring genome. The dependency relation traced in (G) between RNAi-induced epigenetic variation and the new heritable phenotype remains stable during changes made upon the inducing event. It is invariant. Although the authors do not explicitly state their interventionist w- questions and/or counterfactually articulate their causal explanations, the scholarly literature contains various similar papers on epigenetic inheritance that do so. These articles contain (hypothesized) counterfactual explanations in the form of “if the value of an environmental/epigenetic/genetic variable

▸ 101 CAUSAL EXPLANATION

X would be changed, the value of the (transgenerational) phenotype Y would change accordingly” (see, e.g., Anway et al. 2005, esp. 1467; Holmquist and Ashley 2006, various pages; Rassoulzadegan et al. 2006, 472–73; Kelly and Aramayo 2007, various pages; and Molnar et al. 2010, esp. 874). This suggests that interventions performed in epigenetic experiments are often backed up by a counterfactual explanatory approach that guides development of experimental design applicable to answering Woodwardian w-questions. The above case can be understood as an exemplar of experimental investigations of epigenetic causality. Most of the studies on epigenetic soma- to-soma and germ-line transmission described in chapter 1 use the same or similar techniques of multifactorial manipulation on systems of genetic uniformity in order to explain how epigenetic factors causally mediate environmental input in development and heredity. Moreover, as the above analysis exemplifies, the interventionist account provides a suitable framework to elucidate the manipulative and counterfactual manner in which epigeneticists conceptualize causation and causal explanation in the lab. But does the same hold for those field studies in epigenetics that do not or only rarely manipulate the object under study? In other words, what kind of causal reasoning supports epigenetic investigations outside the lab?

EVOLUTIONARY EPIGENETICS

At first glance, focusing on interventionism might seem strange in observational contexts simply because we are not able to manipulate purely observational data. This has been well known at least since Robert Lucas (1976) argued that performing interventions in, for example, policy making on a system under study always disturbs the stability of the systems. If we nevertheless claim that in observational studies it is useful to think about causation as being potentially exploitable for purposes of manipulation, it is therefore necessary to present a notion of invariance under intervention that meets two criteria:

(i) It accounts for statistical methods used in the relevant field. (ii) It is a counterfactually defined notion in the sense defined above.

Federica Russo (2011, 2012) and colleagues (Russo et al. 2011), focusing on structural equation modeling techniques in observational studies

102 ◂ CAUSAL EXPLANATION in the social sciences, have argued against condition (i) by claiming that Woodwardian invariant generalizations are not central to causal inference in observational contexts, since testing the condition of invariance does not regard manipulation as methodologically fundamental: “rather than manipulation, the basic idea or rationale underpinning causal analysis [in observational and in experimental contexts] is that some form of joint variation between variables of interest has to be evaluated” (Russo et al. 2011, 52; original emphasis). According to Russo, the notion of (co-)variation of variables acts as a precondition to the notion of manipulation and not vice versa. Therefore, in nonexperimental contexts the first question to be asked is whether a data set reveals meaningful co-variation between putative cause and effect variables. Subsequently, scientists can use further tests, like the one for invariance, to determine whether the observed co-variation is in fact causal.7 Against condition (ii) Russo argues that even if statistical techniques that manipulate values of variables in a data set are used in observational studies to determine whether associations are causal, such invariance tests are not counterfactual ones. We do not test whether generalizations would remain stable if we were to intervene but rather whether they are in fact stable across subpopulations or different partitions of the data set. Thus, according to Russo, it is rather a noncounterfactual concept of invariance that establishes causal reasoning in observational contexts. To show that (i) and (ii) can actually be defended, the natural experiment in epigenetics discussed in chapter 2 should be revisited. The community ecologists Carlos Herrera and Pilar Bazaga (2011) discovered that epigenetic variation in a wild population of the plant Viola cazorlensis is significantly correlated with long-term differences in herbivory but only weakly with herbivory-related DNA sequence variation. As we saw, in that study structural equation modeling (SEM) heavily relies on statistical manipulations. In particular, the process of building a causal model is closely linked to manipulating variables on paper. In SEM this method is called “statistical conditioning.” In the case of Herrera and Bazaga’s study, they intervened on a part of the assumed causal structure by changing a particular environmental variable (substrate type) to a certain non-actual value (see fig. 3.1). This should isolate the variables of interest, namely, epigenetic variation and herbivory damage. Statistically holding this environmental variable, be it {rock}, {cliffs}, or {sandy soil}, fixed at a par-

▸ 103 CAUSAL EXPLANATION

ticular value stabilizes the total effect on the potential output variable of herbivory damage. Are these manipulative model building procedures guided by an interventionist view on causal explanation similar to the one we find in experimental epigenetics? As mentioned earlier, the method of statistical manipulation used in this evolutionary epigenetics study, can, in fact, easily be understood as being guided by interventionist w-questions like “What would happen if the plants would not be on their natural substrate type but on, for example, rocks?” Answering this question means creating a coun-

terfactual, artificial population—populationi (depicted in fig. 3.1B)—for which the value of this environmental variable does not change naturally.8 According to Herrera and Bazaga (2011, 1679; emphasis added), “Because we were interested in the herbivory-genetic-epigenetic causal structure after controlling for ecological, substrate-related effects, the residuals after controlling for the average effect of substrate on herbivory were used . . .

rather than the raw herbivory data.” This counterfactual populationi may be understood as a population of plants living in an environment that does

not interact with them through substrate types. Populationi is compared with the natural population to determine whether the observed relationship between epigenetic (and genetic) and herbivory variation would remain stable. This procedure is closely linked to Woodward’s concept of “weak invariance” (Woodward 2003, chaps. 6.15, 7.8): referring to observational data drawn from an epidemiological study on the causal relationship between smoking and lung cancer done by Cornfield and colleagues (1959), the notion of weak invariance means, broadly speaking, that the generalization “smoking causes lung cancer” would remain stable across different partitions of the data set, for example, across groups of people with different genetic backgrounds or socioeconomic conditions. These subpopulations of the data set are brought about by controlling for potentially confounding variables. However, Federica Russo (2011, 2012) has argued that this notion of weak invariance, essential for SEM in observational studies, is not the standard Woodwardian notion of counterfactual invariance but rather a kind of factual invariance dealing only with observed, “real-world” data. But it is in fact a counterfactual invariance, since, as shown above, testing (weak) invariance means testing whether the generalization “epigenetic variation causes difference in herbivory damage” would remain stable if we were to bring

104 ◂ CAUSAL EXPLANATION about a certain counterfactual situation by controlling for the substrate type variable. It is this notion of counterfactual invariance that is crucial both for Woodward’s own account on causal inference in SEM and for SEM model building techniques in ecological or evolutionary epigenetics, where testing for weak invariance provides information about which variables should be included or excluded from a causal model. What does this tell us about the explanatory status of causal models in epigenetic field studies that exhibit weak invariance? Looking more closely at this study, one can argue that such models in fact do causally explain, although in a weak sense, because there is a statistical intervention during which a particular association remains invariant. Such a manipulation spec- ifies the processes underlying the observed association and suppliesauxil - iary causal evidence, for example, about which variables should be omitted from an explanation.9 As a consequence, the analysis of weak invariance helps to minimize the total effect of the negligible causal background on the relevant variables. In this sense, proving Woodwardian counterfactual invariance offers crucial information for model building in observational data analysis. As the above cases from molecular and evolutionary epigenetics exemplify, the interventionist account provides a suitable framework for elucidating the manipulative and counterfactual manner in which epigeneticists conceptualize causation and causal explanation, both inside and outside the lab. Moreover, it can support the correctness of explanatory claims by epigeneticists in light of different explanatory standards in biology. The next section focuses on this latter issue.

THE MECHANISTIC THEORY OF CAUSATION

Woodward’s interventionist conception of explanation seems to accord well with the desire of the special sciences (including, e.g., biology, neuroscience, and the social sciences) to describe causal dependencies in their own disciplinary realm. This means that causal relationships do not exist only at some fundamental level and that they cannot be explained only by physicists.10 Strictly speaking, this idea of higher-level causation argues against so-called process or conserved quantity theories of causation, which hold that causality necessarily includes a transfer of mass-energy, linear momentum, or

▸ 105 CAUSAL EXPLANATION

charge via interaction (see Salmon 1994, 1997; and Dowe 1995, 2000; see also Quine 1973, 5). This physicists’ view of causality is usually exemplified by referring to causal processes, such as billiard balls moving across the table or atoms decaying, and by differentiating such real processes from pseudo processes like moving shadows. The interventionist view, however, does not make such a commitment to the level where causality is said to take place. As long as the conditions of (ITC) are satisfied, any relationship between physical or even nonphysical objects can be deemed causal if these objects have clearly defined and changeable values and the relationship shows invariance. This openness has led to various discussions concerning the applicability of the interventionist framework to nonphysical phenomena in fields like historiography (Leuridan and Froeyman 2012), sociology (Russo et al. 2011; Mateiescu 2012; Russo 2012), and psychology (Schulz et al. 2007; Woodward 2007b). Especially in the field of psychology, the topic of mental causation plays a major role. It has to stand its ground against Hermann von Helmholtz’s (1847) idea of physical closure of the world.11 If his claim that every physical event already has a sufficient physical cause is correct, there is no room left for the mind to cause anything. In contrast to this debate, the issue with regard to biology in general and epigenetics in particular is slightly different. Here, we operatewithin the physically closed world. Nonetheless, we are still dealing with issues like causal autonomy (Menzies and List 2010) as well as emergence (Broad 1925; Emmeche et al. 2000) and downward causation (Campbell 1974). The first two concepts explicitly entail some acknowledgment of the irreducibility of certain higher-level entities. For example, Alan Garfinkel (1981) defends the causal autonomy of the special sciences by arguing that these scientists’ claims about higher-level dependencies are true regardless of how they are physically realized. In a similar manner, emergent entities are understood, for example, by John Stuart Mill (1843), to be novel and irreducible, although they arise out of more fundamental entities.12 If emergent entities are said to have causal influence on more fundamental levels, we are dealing with the third concept, “downward causation” (Hendry 2006; Noble 2012). In the epigenetic realm of molecular and cell biology, we face first and foremost the related problem of the extent to which dependencies located on a more fundamental level can legitimately be ignored in explanation. Usually, this means questioning whether one is allowed to exclude information about genes that may be coding for epigenetic molecules or enzymes,

106 ◂ CAUSAL EXPLANATION like DNA methylase. This problem is particularly acute in light of Crick’s central dogma, which suggests that the latter epigenetic molecules are merely “downstream” causal factors with respect to phenotypic traits. In other words, every organism is constituted by a one-way street of causal processes leading from genotype to phenotype, and thus the level of the gene is the one at which all biological explanations bottom out.13 Therefore, we may ask, “Can epigenetic causality be subsumed by genetic causality?” Or in more general terms, “Can epigenetics be subsumed by genetics?” This issue is closely linked with the so-called mechanistic account of explanation. Some philosophers of science have argued that the description of mechanisms, rather than invariant dependencies, adequately accounts for the explanatory practice of some scientists, such as molecular biologists (see, e.g., Machamer et al. 2000; Bechtel and Abrahamsen 2005; Craver 2007; and Craver and Darden 2013). According to those philosophers of science, identifying (components of) mechanisms is usually thought to show why something works. I will not discuss in detail this kind of explanation— that is, identifying and elucidating mechanisms—at this point, as this is the topic of chapter 4. In this chapter, I define mechanistic explanations in a tentative and rather broad way, as bottom-up explanations relating causal capacities at different levels of organization. The only idea related to this mechanistic view that I discuss here, with respect to causal explanation, is the idea that information about mechanisms (henceforth referred to collectively as “mechanistic detail”) underlying a dependency relation elucidates why this relation holds. Accordingly, in the case of epigenetics, mechanistic detail about genes is thought to elucidate why epigenetic factors are causally involved in complex systems of development and heredity. Does this general mechanistic strategy contradict Woodward’s claim that causal explanations uncover patterns of systematic counterfactual dependence? As some have argued, describing the mechanism of a phenomenon is consistent with the causal-interventionist way of explaining (see Woodward 2002; Glennan 2005; and Craver 2007).14 At the same time, interventionist explanations tracing invariance are often considered to be rather weak or superficial compared to mechanistic ones, because they often are lacking in conveying how (or why) an observed dependency relation produces a phenomenon. This is the view of mechanistic proponent Carl Craver (2006, 2008).15 He states that explanations incorporating mechanistic detail—by citing somewhat more fundamental or causally more “upstream” factors—

▸ 107 CAUSAL EXPLANATION

can be understood as a necessary addition to the interventionist strategy in terms of clarifying the internal causal structure underlying a certain systematic dependency relation. Invariance under intervention is thus, in contrast to Woodward’s (2003) position, understood as insufficient for there to be causation or to figure in explanation.16 Given the above, we may call an account of explanation, which holds that a causal relation necessitates more fundamental factors to be considered in order to testify to its causal status, a mechanistic theory of causation (MTC). This account makes the following claim:

(MTC) Two events X and Y are causally related if and only if their connectedness results from a more fundamental mechanism, due to which X, if changed, would make a difference toY . Changes in the underlying causal-mechanistic structure would make a difference with respect to Y, via changes in X and/or any intermediate variable between X and Y.

According to (MTC) and the tentative definition of a mechanism above, a mechanism can be considered to be more fundamental if it entails more upstream causal factors than the dependency relation X–Y (including its intermediate variables). If these more fundamental factors are cited in an explanans, they are thought to contain more explanatorily relevant information and mechanistic detail, respectively, about the dependency relation under study. This suggests that the concept of mechanistic detail is a contrastive concept comparing different factors located at more or less fundamental levels of organization.17 Accordingly, the distinction between causal explanations tracing invariance but omitting mechanistic detail and those that do consider such detail is not a clear-cut one. At first glance, this view on the relation between the interventionist and mechanistic approaches to causation and the insufficiency of the former seems to be predominant in molecular epigenetics as well. Although the interventionist account suitably describes the way molecular epigeneticists think about causation and causal explanation (see above), at the same time it remains unclear whether these scientists regard invariance as a necessary and sufficient condition for a generalization to represent a causal relation, since they sometimes also seek to articulate the underlying mechanisms explaining why (or how) an observed dependency relation arises.18

108 ◂ CAUSAL EXPLANATION

One example of such an investigation, which highlights mechanistic “why” questions, is given by Minoo Rassoulzadegan and colleagues (2006), as described briefly in chapter 2. They investigated transgenerational RNA-mediated transmission of epigenetic information in mice. They showed that the injection of small RNAs (microRNAs) into fertilized eggs causes a heritable epigenetic modification, a so-called “paramutation” (which means that one allele induces a heritable change, for example, a change in a DNA methylation pattern, in another allele of a single locus) in male mice. This results in white tail-tips and paws. Thus, they established that “injection of microRNAs in a one-cell mice [sic] embryo causes a transgenerational phenotype in male mice.” They also state, however, that “the molecular mechanisms involved would remain to be established” (Rassoulzadegan et al. 2006, 472). In other words, they seem to feel the need to characterize the mechanism(s) underlying this dependency relationship, which “hopefully lead[s] to a more complete and better defined [causal] picture” (473). They are particularly interested in mechanistically clarifying the genotype- phenotype relationship leading to the paramutated progeny. Given this and similar studies in molecular epigenetics seeking to provide mechanistic detail (see, e.g., Costa 2008, 9, 14; Bonasio et al. 2010, 612; Dunn and Bale 2011, 2229; and Nelson et al. 2012, E2766), one may ask if causality in complex epigenetic systems necessitates mechanistic detail on why it exists. Or is invariance under intervention a necessary and sufficient condition for causation? More specifically, can only the gene turn epigenetics into a science concerned with causality in living systems?

DIFFERENCE MAKING AND HIGHER-LEVEL EXPLANATION

It is not surprising that in molecular studies on epigenetic processes in development and inheritance we find mechanism to be a topic of discussion. As I show here, however, explaining a phenomenon by describing a mechanism for that phenomenon is a second (and not necessary) explanatory step for epigeneticists that is distinct from interventionist explanation. Tracing invariance alone suffices to establish so-called “nonmechanistic” causal explanations. In a similar manner, explaining how eating bananas causes recovery from hangover by replenishing the potassium level is an

▸ 109 CAUSAL EXPLANATION

explanation in its own right. The dependency traced does not necessitate information on why, for instance, a change in the potassium level influences the functioning of nerve cells, potentially leading to headache, in order to be explanatorily relevant. In the case of epigenetics, this sufficiency is based on the fact that more fundamental genes have neither a special ontic nor an epistemic status, in contrast to less fundamental non-DNA factors.19 In addition, the latter higher-level causes can be identified as the sole actual difference makers in populations distinct from those in which genes make an actual difference. This autonomy of interventionist causal explanations, however, does not render mechanistic explanation, in general, nor information about genes in epigenetics, in particular, redundant. In order to present this argument in full detail, let us consider a version of the (MTC) principle developed by Kenneth Waters (2007) in his influential paper “Causes That Make a Difference.” His concept of difference making, which draws on Woodward’s interventionist account, highlights the special ontological status of genes as primary causal agents in many biological fields, ranging from classical genetics to molecular biology and developmental biology. In light of (MTC), we can interpret genes as entities providing explanatorily relevant mechanistic detail to suggest why difference making occurs in higher-level dependency relations of complex living systems.20 In contrast to this view, based on recent studies on the diverse causal roles of nongenetic factors, it has been argued that all causes in a cell or during development are on an ontological par. This latter idea is sometimes called the causal democracy or causal parity thesis (Oyama 2000; Oyama et al. 2001; see also chap. 5).21 Let me take these two diametrical positions as starting points for specifying both the role of epigenetic causes in biological phenomena, like heredity phenomena, and the status of epigenetics in the biological sciences. Waters distinguishes among three kinds of difference-making causes of an actual event: a potential difference maker,the actual difference maker,an actual difference maker. He describes these three concepts by referring to classical genetics: Thomas Hunt Morgan’s work in the 1910s on the inheritance of the genes coding for eye color in Drosophila melanogaster. Like other classical geneticists, Morgan was interested in alterations of inheritance patterns of phenotypic traits, which are explained by the transmission of genotypic differences across generations. Many years later this explanatory interest led Richard Dawkins (1982, 92; emphasis added) to state that

110 ◂ CAUSAL EXPLANATION it is “a fundamental truth, though it is not always realized, that whenever a geneticist studies a gene ‘for’ any phenotypic character, he is always referring to a difference between two alleles.” Morgan (1926, 305–6, quoted in Waters 2007, 559; emphasis added) described the genotype-phenotype relationship that is prone to changes as follows: “If now one gene is changed so that it produces some substance different from that which it produced before, the end-result may be affected, and if the change affects one organ predominatingly [sic] it may appear the one gene alone has produced this effect. In a strictly causal sense this is true, but the effect is produced only in conjunction with all the other genes. In other words, they are all still contributing, as before, to the end-result which is different in so far as one of them is different.” As this passage shows, classical geneticists were (still) fully aware of the complexity of the genotype-phenotype relationship. According to this view, many heritable genes can be causally relevant for a particular phenotypic character to develop and to be inherited. This many-to-one relationship is called polygeny or polygenic trait (Mather 1943; Waddington 1943; Dove 1993). For instance, in humans hundreds of genetic variants in at least 180 loci influence adult height (Lango Allen et al. 2010). Other examples of polygeny in humans include pigmentation (Sturm 2009) and the risk of, respectively, schizophrenia and bipolar disorder (Purcell et al. 2009). In D. melanogaster, wing shape is a polygenic trait (Weber et al. 1999). If we are dealing with a many-to-one relationship, as in the case of polygenic traits, every causally contributing gene—or to paraphrase Morgan, every gene working in conjunction with others—is a potential difference maker. It should be noted that difference making always presupposes that causes are situated in an actual population of at least two entities, like a number of fruit flies, which differ with respect to a certain effect, such as eye color.22 In addition, classical geneticists developed experimental methods that allowed them to narrow the class of potential difference makers, so that the genotype-phenotype relationship under study could be investigated as a one-to-one relationship. This includes using standardized environmental conditions and particular breeding regimens that permit the development of distinctive inheritance patterns (see Kohler 1994). For example, Morgan created an experimental setup in which only one of the D. melanogaster genes (the purple gene coding for eye color) actually differed; the new mutant fruit flies—and if crossed, their offspring—showed red eyes in

▸ 111 CAUSAL EXPLANATION

contrast to the eye color of the wild-type flies in the population. Thus, in contrast to all the other genes coding for eye color, only this single purple gene actually caused the difference in the phenotype in the experimental settings. Generally speaking, while potential difference makers are defined as variables that exhibit uniform values in the actual population and thus do not make an actual difference, the value of a single variable does actually differ. Waters calls such a variable the actual difference maker. In contrast to the former causal concept of a potential difference maker, it is the actual difference maker that, according to Waters, Morgan understood to be a cause in a “strictly causal sense.” As Waters (2007, 559–60; emphasis added) summarizes, “[Morgan’s] science entailed, as do biological sciences in general, identifying one or a few elements as the ‘actual cause(s)’ in situations that necessarily involve many causes.” As this quote suggests, the strategy of narrowing the class of potential difference makers to a single actual difference-making cause may fail (see emphasis in the above quote). In such cases we are dealing with a number of actual difference makers, each of which is an actual difference maker. Here, for example, a number of genes jointly cause a difference in the phenotype. Evaluating quantitatively the causal influence that every single actual difference maker exerts on the phenotype has become the major research effort in quantitative genetics. In addition, unraveling webs of large numbers of different actual difference makers has turned into a methodological cornerstone of the complexity science of epigenetics. In particular, it is highly important for determining “how much” causal influence a nongenetic factor has on a difference in a phenotypic trait, in contrast to the DNA-sequence(s) coding for the trait. To back up the idea that genes are more important causal factors for traits than epigenetic regulatory factors, one has to show that genes have a unique ontological or epistemic status compared to the epigenetic factors.

THE ONTOLOGICAL STATUS OF GENES

Waters claims that distinctions between more or less equally important causal factors are not simply ad hoc and fully pragmatic. Instead, he states that the causal status of a particular factor as the actual difference maker is fixed by ontology once given an actual difference in a given population. Thus, according to Waters, many biologists, including developmental biologists, often rightfully center their attention on genes and DNA, despite the

112 ◂ CAUSAL EXPLANATION fact that there exist multiple causes of phenotypic differences. This is the case since, in many explanatory contexts, these particular entities possess the ontological status of actual difference makers. Although there are a few non-DNA entities that, according to Waters (2007, 575–76), hold this status as well, like splicing agents during RNA synthesis in eukaryotes, “the idea that all causal factors . . . are on a causal par . . . is false.” In other words, in many actual situations and, accordingly, given a number of explananda, there are variables that exhibit uniform values and that do not make an actual difference, while there are some other variables, mostly DNA, whose values do actually differ in the actual population. Thus, Waters’s view on causal explanation in biology can be summarized as follows:

(W) In a number of situations, explanations primarily cite information on DNA as explanans variables due to DNA’s special ontological status as primary causal agent, that is, actual difference maker.

In addition, given that there are often other nongenetic agents that make an actual difference to protein synthesis, such as posttranscriptional editing or splicing, Waters claims that only DNA shows a particular specificity. This causal specificity refers to the amount of control that interventions on a putative causal factor can exert on the effect. However, as many studies in quantitative epigenetics, such as epiRILs (epigenetic recombinant inbred lines), suggest, there are a number of developmental and inheritance situations in which the causal specificity of nongenetic factors seems to match those of nucleic acid.23 At any rate, there are at least three further problems related to (W) and the ontological status of genes as actual difference makers, respectively. First, given the increasing knowledge of phenomena like overlapping DNA-coding regions or nontranscripted DNA sequences, the ontological foundation of the concept of “gene” has become less intuitively clear (see, e.g., Pearson 2006). Thus, it is anything but clear how the cases presented by Waters, ranging from classical genetics to today’s quantitative genetics, relate to one another. Second, Waters puts considerable effort into carefully analyzing highly selected cases that already support an explanatory emphasis on DNA (e.g., by excluding posttranscriptional regulation of gene expression and environmental influences). He assumes that his analysis can disprove the positions

▸ 113 CAUSAL EXPLANATION

expressed by authors working on developmental systems theory (DST), who argue for the parity of DNA and other factors in much broader genetic, epigenetic, and environmental settings (see, e.g., Oyama 1985; and Griffiths and Gray 1994). In fact, advocates of causal parity do not usually object to the result Waters draws from his cases (see Griffiths and Stotz 2013, 81). However, they criticize his position, claiming that he neither addresses their more complex cases of causal parity nor describes interesting explanatory cases of contemporary molecular biology, especially in the context of development. Third, there is a more fundamental fallacy in (W): specifically, the claim that a biological explanation picks out, by any means, more important causes by citing those causes with more causal power than others. This ontological claim rests on the dubious assumption that all causes relevant for a phenomenon under study share objectively measurable or comparable properties. As Elliott Sober (1988) correctly emphasizes, however, in most cases there exists no common currency in which the relative causal importance of different causal factors can be evaluated. For example, in statistical investigations using multiple regression, regression coefficients should not, as one might presume, be understood as objectively measurable, separable causal influences and properties of the modeled system, respectively. Usually, separability and additivity of causal influences result from modeling assumptions (see Pearl 2000; and Ylikoski and Kuorikoski 2010). This means that in a model of the genotype-phenotype relationship(s) determining eye color in D. melanogaster, the actual difference-making causal effect of the purple gene is not due to its special ontic status. Rather, the difference observed is relative to the variation in all the other causal factors actually present in the experimental population, where, as we know, there is very little variation, due to breeding regimens and standardized environmental conditions. In a similar manner, even if a model of headache recovery exhibits equal causal influences and difference-making effects, respectively, of the variables “eating bananas” and “changing the behavior of potassium ion channels in cellular membranes,” this result does not offer information on the causal powers of these factors. This analysis renders false not only Waters’s account but also the thesis of “causal democracy” or “causal parity.” Both err in their belief in the objective measurability of causal importance.24 In Waters’s claim (W), this fallacy leads to the quite general but erroneous assumption that there are

114 ◂ CAUSAL EXPLANATION primary causal factors—usually more “upstream” or fundamental causes, like genes—that, if cited in many explanatory contexts, entail the mechanistic detail necessary to explain a phenomenon.

THE EPISTEMIC STATUS OF GENES

Although, according to the above, genes do not have a unique ontic status, maybe they have a unique epistemic one. Thus, let us now investigate the somewhat weaker, non-Watersian claim that, due to the DNA’s special epistemic status as the focus of attention in biological explanation, citing genes rather than less fundamental causes as explanans variables is better. Let us call this claim (W*):

(W*) In a number of situations, explanations primarily cite information on DNA as explanans variables, due to DNA’s special epistemic status as primary causal agent, that is, actual difference maker.

(W*) holds that the techniques developed by biologists to investigate complex systems have enabled them to describe a fundamental set of actual difference makers: genes. However, this claim of genes as unique, epistemic, actual difference-making causes is also wrong. Especially in epigenetics, we find various cases of explanations of trait development and inheritance in which causally relevant genes merely hold the status of potential difference makers while there are a large number of nongenetic (in Waters’s sense, non-DNA) factors. This set of factors ranges from extra- and intraorgan- ismic environmental factors to various molecules, which are identified as causes that bring about an actual difference in a population. To exemplify these cases of difference making, let us consider a paradigmatic explanation of nongenetic inheritance in epigenetics: a number of genes G1–G10 jointly cause a particular epigenetic pattern E to occur, for example, a particular DNA methylation or histone acetylation pattern, which then brings about a heritable phenotypic trait P. Additionally, the development and inheritance of P are the product of other genes G11–G20 via a route that does not go through E. This results in the following schema: 25 G1–G10 → E → P ← G11–G20. As with Morgan’s case of eye color inheritance in D. melanogaster, this example of epigenetically influenced development and inheritance has to

▸ 115 CAUSAL EXPLANATION

deal with the problem of polygeny. In other words, given non–one-to-one relationships, epigeneticists are faced with the challenge of having to find out how to pick out one (or a few) cause(s) when in fact there are many. However, their solution to this problem differs from what Morgan and Waters offered. Following Waters, let us specify the actual populations of variables in order to understand which causes are the actual difference makers in our example. Suppose there are two different experimentally established, stan-

dardized, and reproducible populations p1 and p2. Population p1 is a proper

subset of p2. Waters understands the boundaries-to-be-drawn of these

actual populations as follows: p1 entails (in the best case) a single gene,

say G10, which is the actual difference maker with respect to P (via E). The

other genes, G1–G9 and G11–G20, are merely potential difference makers with

respect to P in p1. In addition, this means that G10 is at least an actual differ-

ence maker with respect to P in the wider population p2.

FIG. 3.2. Different populations in a scenario of epigenetic causality. Epigenetic factor

E makes a difference with respect to traitP in the actual population p1. The variables of

genes G1–G20 exhibit uniform values in the actual population p1 and thus do not make an actual difference in E. For further explanation, see the text.

116 ◂ CAUSAL EXPLANATION

In contrast, epigeneticists use a radically different strategy of boundary drawing (fig. 3.2). They seek to show that non-DNA causal factors, such as DNA methylation or histone modification patternE , can act as actual difference makers as well. For that reason, they identifyE —rather than a gene—as the actual difference maker with respect toP in p1; all genes G1–G20 are considered as merely potential difference makers with respect to P in p1. This solution turns E into at least one actual difference maker with respect to P in p2.

In population p2, there might be other genes that are also actual difference makers with respect to P, but as long as the less fundamental and more “downstream” cause E also makes a difference with respect to P in this population, then we can try to narrow the more complex population p2 to pick out a non-DNA factor as the sole actual difference maker with respect toP .

This narrower population is p1. Interestingly, the strategy of identifying higher-level epigenetic factors, like E, as sole actual difference makers of invariant dependencies in complex systems is legitimated by the view on the genotype-phenotype relationship presented by Waters himself: Morgan’s many-to-one view of genetic causation, known as polygeny. Of course, this view of the complexity of the genotype-phenotype map is also supported by the opposing phenomena, called pleiotropy, polyphenism, and phenotypic plasticity, which state that a single gene, through one to many relationships, causes multiple discrete phenotypes (Hodgkin 1998; Stearns 2010; Wagner and Zhang 2011; see also chap. 1). If many genes jointly bring about differences in a heritable phenotypic trait or if one gene brings about differences in multiple phenotypic traits, this means that if the genotype-phenotype relationship is understood as a highly complex web of interactions rather than a set of linear, causal, “button-up” highways, epigenetics offers the following methodological rationale: consider genes as potential difference makers in an actual (natural or experimental) situation and find other factors exhibiting a “direct” effect on phenotypic variation. This does not mean ignoring the causal influence of genes but instead considering them as causal factors that have merely a potential, non-actual difference-making effect on the phenotypic trait. I call this kind of nongenetic actual difference makingepigenetic actual difference making. This explanatory strategy, of course, may include not only turning a DNA methylation state or histone tail acetylation state but also an environ-

▸ 117 CAUSAL EXPLANATION

mental event, like an intruding toxic agent or rising temperature, into the actual difference maker. For example, in the above case we can easily imag-

ine an actual population p3 in which the variation in the value of E is fully determined by an environmental variable, say EV, and not by the genes

G1–G10. This turns EV into the actual difference maker with respect to P in

p3, given that the variables G11–G20 do not differ either. This kind ofenvi - ronmental actual difference making matches the causal scenario described not only in the above study of Vastenhouw and colleagues (2006) on epigenetic inheritance in the nematode C. elegans but also various other experimental studies in epigenetics, including investigations on changes in mRNA splicing in response to environmental stress in yeast (Pleiss et al. 2007), heritable epigenetic and phenotypic changes induced by exposure to endocrine disruptors, such as vinclozolin, in rats (Skinner et al. 2011), and high population density or predation causing heritable changes in the development of aphids (Srinivasan and Brisson 2012). Identifying actual difference-making environmental factors is central for not only epigenetics but also the related field of ecological and evolutionary developmental biology, known as eco-evo-devo (see Gilbert and Epel 2009; and Ledón-Rettig and Pfennig 2011). As the above shows, (W*), which highlights the special epistemic status of the DNA as the primary causal agent and thus its role as primary explanans in biological explanation, is, like Waters’s principle (W), defective. Both epigenetic actual difference making and environmental actual difference making, which are widely used explanatory strategies in epigenetics, present counterexamples to (W*). Thus, compared with higher-level non-DNA factors, genes show neither a unique ontic nor an epistemic status, which necessitates paying special attention to them in any causal explanation of complex living systems. In addition, this analysis supports Jablonka and Lamb’s (2010, 138–39) empirically based causal autonomy thesis, which states that there are phenotypic “variations that are independent of variations in DNA sequences, and they have a degree of autonomy from DNA variation.” Finally, this analysis offers a solid explanatory footing for a common Sunday morning ritual: in order to recover from a hangover we merely have to eat a banana, rather than ask for gene therapy that changes the biochemical makeup of potassium channels in such a way that they allow for easier flow of potassium ions into the extracellular space. Lucky us.

118 ◂ CAUSAL EXPLANATION

THE ROLE OF GENES AND “CAUSES WITHOUT MECHANISMS”

Now that we have circumscribed the idea that actual difference making is not metaphysically or epistemically restricted to tracing genes as causes of traits in a cell or organism, we are able to specify the “causal nature” of epigenetics—a science that adopts this very idea as one of its core principles of explanation. To be more precise, we can define epigenetics as the field at the intersection of molecular and cell biology, on the one hand, and of ecology as well as evolutionary biology, on the other; in this intersection epigenetic and environmental factors are identified asactual difference makers. By acknowledging that and understanding how epigeneticists focus their attention on this less fundamental subset of causes as actual difference makers, we can see that a special research program comes to light. This program seeks to evaluate the causal influence that higher-level nongenetic factors have on the development and inheritance of phenotypic traits. Therefore, the experimental strategies (including standardized setups) developed by classical geneticists and molecular biologists do not have to be modified. In contrast, biologists have to adopt the idea that less fundamental and more causally downstream factors are full-fledged actual difference-making causes compared with more fundamental, lower-level ones. Additionally, the historical development that some call the “epigenetic turn” can now be given a more sophisticated reading. Guido Nicolosi and Guido Ruivenkamp (2012) have described the epigenetic turn as a historical shift from a gene-centric paradigm into a view that highlights both the complexity of the genotype-phenotype relationship as well as the interrelationship between organism and environment. In contrast, this analysis of causal explanation in epigenetics suggests that, above all, epigenetics has induced a shift in applying the concept of actual difference making. This shift allows molecular biologists in particular to trace invariably higher- level causes as explanantia that make a difference in complex explanandum phenomena. At this point, the question comes to the fore as to whether epigeneticists, by focusing on “epigenetic actual difference making” and “environmental actual difference making,” willingly exclude mechanistic detail on more fundamental causes—especially on genes—in those cases in which a higher-level non-DNA causal factor brings about a change in an actual pop-

▸ 119 CAUSAL EXPLANATION

ulation. In other words, “Do epigeneticists really fully reject the mechanists’ principle (MTC)?” To answer this question we first have to dispel a myth about epigenetics, which seems to have steadily been spread, especially in popular discourse: epigeneticists are actually very interested in the causality of genes. As the study of Rassoulzadegan and colleagues (2006) on epigenetic transmission of white tail-tips in mice shows, these scientists often seek to investigate the genetic mechanisms underlying epigenetic dependencies and differences in phenotypic traits. As I will show, epigeneticists do not seek to eliminate genes from the “causality landscape” of complex living systems, which would be rather absurd. Instead, they integrate genetic dependencies into epigenetic explanations. What does this mean? Somewhat irritatingly, at least at a first glance, even our paradigmatic case of higher-level explanation, the “green worm study” of RNAi-induced transgenerational epigenetic effects in C. elegans by Vastenhouw and colleagues (2006), seems to have an (MTC) dimension to it. First, they present their interesting non-Weismannian causal generalization (G): “environmentally induced gene silencing produces a new phenotype that is heritable over many generations of sexual reproduction” (2006, 882). More precisely, they show that (G), the dependency relation between diet-induced epigenetic variation and nonfluorescing nematodes, is invariant under changes introduced in the inducing event. In short, they applied the strategy of environmental actual difference making. Second, however, they also asked, “Is RNAi the mechanism behind the initial silencing [or the inheritance]?” (882). This means that the invariant relationship between the induction event and the transgenerational effect observed does not elucidate whether RNAi is the mechanism producing the initiation of gene silencing (see the pathway “environment → epigenetic variation” in fig. 3.1 above) and/or the inheritance or maintenance of silencing (see the pathway with a question mark in fig. 3.1). In other words, “Why is greenness suppressed in the offspring of green worms?” To answer this question, they knocked down canonical RNAi genes, like rde-1, but after changing the inducing event to

environmenti in the parental generation. This knockdown of RNAi genes we may call “intervention 3” (fig. 3.3A). It nonetheless does not change the maintenance of the nonfluorescing phenotype of the nematode. The -off spring of green worms remain nongreen. In other words, the C. elegans rde-1 mutant with an inactive gene coding for the functioning of RNAi still shows

120 ◂ CAUSAL EXPLANATION

A ▸

◂ B

FIG. 3.3. Two path diagrams representing a second controlled experiment by Vast- enhouw et al. (2006) on transgenerational gene silencing in the nematode C. elegans. (A) and (B) show the causal system during (and after) manipulation. The hypothetical causal system before experimental manipulation is depicted in figure 3.1A. In path diagram (A) gene rde-1, involved in RNAi, is knocked down after the environment is changed by I1. This manipulation is called I3. Given I3, the inheritance of phenotypei (induced by RNAi mediated initiation of gene silencing) maintains. In (B) gene hda-

4, coding for a histone deacetylase, is knocked down. This manipulation is called 4I .

Given I4, the inheritance of phenotypei is abolished and the wild-type phenotype is reestablished in the offspring.

▸ 121 CAUSAL EXPLANATION

inheritance of the induced transgenerational phenotype. This means that rde-1 is a dispensable mechanistic causal factor for answering the question “Why (G)?” Other genes, such as hda-4, coding for a histone deacetylase, however, do affect the maintenance of gene silencing.26 When hda-4 is knocked out by “intervention 4,” inheritance of the green phenotype is abolished (fig. 3.3B). Thus, hda-4 is an explanatorily relevant factor for answering the question “Why (G)?” This leads to the following (MTC)-like explanation:

(G*) Environmentally induced gene silencing produces a new phenotype that is heritable over many generations of sexual reproduction, given the expression of certain genes (e.g., hda-4) that are involved in the inheritance of the trait.

(G*) shows that the gene hda-4, among others, is causally involved in whether or not worms’ loss of greenness, induced through a particular environmental cue, is maintained across generations.27 With respect to the issue at stake here, we may ask, “What does the example of (G) and (G*) teach us about the way epigeneticists investigate genetic causalities underlying epigenetic dependencies?” A higher-level dependency, like the one traced by (G), contains at least one non-DNA causal factor that makes an actual difference with respect to the effect in a population and is invariant under intervention. (G)’s explanans entails information about how—in the sense of “due to which actual difference maker(s)”—the heritability of a phenotypic trait is influenced over many generations by causal factors other than genes. In (G), genes are downgraded to potential difference makers due to their non-actual polygenic or polyphenic influences. In other words, (G) fits an actual causal niche that is left open by the complexity of genetic causation. One may seek to establish independently of (G) an additional (MTC)- like investigation leading to (G*). This new explanation presents information addressing why the very same higher-level dependency relation traced by (G) exists. Note that (G*) is not a necessary addition to or replacement for (G), as argued by a mechanistic theory of causation. Explaining “Why (G)?” does not yield the same explanation that (G) offers. This is the case, because (G) and (G*) hold with respect to different actual populations. The actual population in which the epigenetic dependency is traced

122 ◂ CAUSAL EXPLANATION

A ▸

◂ B

FIG. 3.4. Different populations in a scenario of epigenetic causality. (A)G 1 (via E)

makes a difference with respect toP in the actual population p1*. (B) G10 (via E) makes

a difference with respect to P in the actual population p1**. The variables G2–G20 (in A)

and G1–G9, G11–G20 (in B), respectively, exhibit uniform values in the actual populations and thus do not make an actual difference in P.

▸ 123 CAUSAL EXPLANATION

by (G) resembles the above population scenario of an epigenetic factor E, which is the actual difference maker of a phenotypic traitP in population

p1, as depicted in figure 3.2. In p1 all causally “upstream” genes, like rde-1 and hda-4, are potential difference makers. In contrast, (G*) refers to a number of scenarios, in which epigeneticists use a distinct kind of boundary

drawing (fig. 3.4). Here,p 1* and p1** include a number of genes as (potential) actual difference-making causes of P via E. In other words, different genes,

like G1 representing rde-1 or G10 representing hda-4, are treated as (potential)

difference makers with respect to distinct actual populations (see p1* in fig.

3.4A and p1** in fig. 3.4B). Thus, (G*) is a distinct explanation, which places the dependency relation between E and P within an (or a number of) actual population(s) dif-

ferent from that of (G), namely p1. Sometimes this strategy of permuting different actual populations reveals that some genes, like rde-1, assumed to make an actual difference, do in fact leave the value ofP in one of these

actual populations unchanged. This is the case inp 1*. Such genes are thus identified as merely potential difference makers with respect toP in this actual population. These “negative” cases help to narrow both the set of genes that do make an actual difference with respect to P, as well as the number of actual populations in which such difference making takes place. In contrast to the explanatory strategies of “epigenetic actual difference making” and “environmental actual difference making,” let us call this strategy of epigeneticists genetic actual difference making. Often this latter kind of investigation is cumbersome. For example, the attempt of Vastenhouw and colleagues (2006) to keep green worms from manifesting the green fluorescence was quite laborious. They had to test 164 genes known to be involved in RNA silencing processes and dozens of candidate genes with the potential to cause gene silencing (via histone [de]acetylation and chromatin remodeling) in order to identify a genetic actual difference maker and its associated population. “Genetic actual difference making” is the core principle in revealing polygenic or polyphenic causation in epigenetics. It offers explanatorily relevant information that helps to quantitatively assess the causal influence that genes have on E → P. However, given the fact that an actual population also

exists in which E is the sole actual difference maker, namely p1, the actual causal influence of genes might, in principle, be very low or even zero. This may, for instance, be the case if identifying genes as actual difference makers

124 ◂ CAUSAL EXPLANATION is not very successful or if it fails completely. The latter occurs, if, for example, the value of P via E is fully determined by an intra- or extraorganismic environmental factor such that we cannot find a single population in which a gene could act as an actual difference maker of P. Often, however, genes and epigenetic molecules jointly make an actual difference in populations. With respect to Vastenhouw et al.’s (2006) explanations (G) and (G*), this result, again, shows that explaining which genes are causally involved in (G) is not a necessary condition for (G) to hold, as claimed by the mechanistic view on causality (MTC).28 In more general terms, tracing invariance of a nongenetic dependency, as described by principle (ITC), is a necessary and sufficient condition for an epigenetic explanation to be rendered causal. Finally, one may argue that the idea of using different populations to test actual difference making of more fundamental causes has not been introduced into biology by epigeneticists. Rather, it should be understood as a feature of a general reductionist strategy of biologists who seek to fill black boxes by identifying underlying mechanisms. An example of such a strategy, discussed by Lindley Darden (2005), is the accumulation of mechanistic detail gathered in molecular genetics since the mid-twentieth century. During this period, new data were used to fill the black boxes in Mendelian geneticists’ explanations of trait inheritance. With respect to epigenetics, this “black box interpretation” may be backed up by statements of epigeneticists calling for overcoming the present limited knowledge about and understanding of epigenetic mechanisms. For example, Gregory Dunn and Tracy Bale (2011, 2229) state, “An examination of mechanisms required to epigenetically program an organism is critical in furthering our understanding of how population-wide epidemics occur, and ultimately how they can be intervened upon. [Still,] mechanistically, little is known about the transgenerational epigenetic perpetuation of traits such as obesity or metabolic syndrome.” Advocates of the “black box argument” could claim that such remarks hint at the fact that epigenetic dependencies, like the one between E and P, can be explained somewhat better by means of a reductionist strategy clarifying genetic mechanisms and the genotype-phenotype relationship, respectively. In other words, “genetic actual difference making” in epigenetics can be deemed identical to the strategy of filling in black boxes. This black box argument is, however, wrong. If we understand the gradual molecularization of classical geneticists’ explanations of trait inheri-

▸ 125 CAUSAL EXPLANATION

tance as a shift similar to that from the actual populationp 1 to, for example,

p1** , this would include fully replacing epigenetic factor E with gene G10 in p1** . In other words, E is fully determined by G10. But, in fact, in epigeneticists’ “genetic actual difference making” the higher-level intermediate cause E does not lose its status as an actual difference maker with respect toP

when shifting from p1 to p1** . Although “genetic actual difference making”

may uncover more fundamental causal factors, like G10, which are causally involved in bringing about P, this does not make E lose its actual difference-

making status. Thus, E is not understood as fully determined by G10 but as a

causal factor of equal value in p1** . In addition, and even more importantly, it is not the not yet acquired knowledge of mechanistic detail that provisionally secures the causal autonomy of E, as in the black box case in classical genetics. As the example of

p1 shows (see fig. 3.2),E ’s causal autonomy is secured in due consideration and with knowledge of the causal influences of genes onE → P. Epigeneti- cists thus claim that we can explain the complex interrelationship between environmental and genetic influences in development and heredity only if we concede E’s special causal status as a nongenetically determined actual difference maker. Additionally, this new strategy enables evaluation of the causal quantities of the various different factors involved in these complex phenomena. This result is distinguished from the time- and knowledge-dependent account on causal inference offered by Alex Broadbent (2011). With respect to causal inference in epidemiology, he describes two distinct views. First, the mechanistic stance describes the idea that discovering mechanisms is necessary for the inference of a causal claim, and second is the black box stance, which is the idea that discovering mechanisms is not necessary for the inference of a causal claim. This distinction resembles my distinction between the mechanistic theory of causation (MTC) and interventionist theory of causation (ITC). However, Broadbent’s distinction is introduced in order to explain the differences between the two views in situations in which we already know or do not yet know the relevant mechanisms. The above analysis of higher-level causal claims in epigenetics does not adopt such a framework that is sensitive to time and the state of knowledge at any given time. Henceforth, I will refer to those less fundamental factors that—independent of the amount of supplementary mechanistic detail available—

126 ◂ CAUSAL EXPLANATION are identified by epigeneticists as full-fledged actual difference makers in molecular or cellular dependency relations as causes without mechanisms.29 I borrow this term from Marcel Weber (2008). Weber uses the Hodgkin- Huxley electrophysiological model of action potentials in neurons to show, pace Carl Craver (2008), that the model offers a causal explanation of action potentials despite false mechanistic assumptions made by A. L. Hodgkin and A. F. Huxley (see also Levy 2014). In particular, Weber argues that tracing invariance under intervention suffices for there to be causation and to figure in explanation. Accordingly, he states that mechanisms “may enrich our understanding of specific systems, but they are by no means necessary for giving causal explanations of natural phenomena” (M. Weber 2008, 1006). Usually, such anti-(MTC) accounts of explanation back up their position by emphasizing the increase in causal knowledge gained through pure, nonmechanistic interventionist explanation. Although I fully support this argument, my analysis follows a different strategy. It suggests that the idea of explanations that offer mechanistic detail about underlying causal relations is not as straightforward as it seems. If mechanistic detail is added to a higher-level causal explanation, what matters in order to evaluate its explanatory relevance is its difference-making status. To be more precise, as explanation in molecular epigenetics exemplifies, information about more fundamental factors may in fact be added to an explanation as a set of mere potential difference-making causes. In a similar manner, an explanation of a car accident caused by fatigue can, of course, include a description of the car’s mechanical makeup, like its lack of a fatigue warning system, though that lack did not actually cause the accident. In this sense, mechanistic detail may figure as non-actually relevant or actually irrelevant information in causal explanation. Thus, a less fundamental factor can, in fact, be the only full-fledged cause (as the actual difference maker) and figure as such in an explanans of a causal explanation, even if mechanistic detail is added. What is more, even adding other lower-level causes and mechanisms as actual difference makers to this picture does not refute the above result, since adding such causes implies shifting the explanandum or the population in which differences are traced. In short, the more precise concept of the difference-making status of mechanistic detail developed here justifies tracing causes without mechanisms.

▸ 127 CAUSAL EXPLANATION ▶ ▶ ▶

The interventionist account of causal explanation elucidates and legit- imizes the way many molecular and evolutionary epigeneticists conceive of causation and causal explanation. In line with this explanatory framework, a comprehensive methodological toolbox that rests upon manipulation of environmental variables and multifactorial experimentation has been developed. In particular, epigeneticists in the fields of molecular and cell biology trace invariant dependencies, including variables located at less fundamental, nongenetic levels, in order to describe their influence on phenotypic development and inheritance. In contrast to a mechanistic theory of causation, this epigenetic higher- level explanation does not require additional mechanistic information on why a dependency relation under study exists in order to be deemed causal. There are two reasons for this. First, more fundamental factors, paradigmatically genes, do not have a unique ontic or epistemic status, which would necessitate listing them in every explanans of causal explanation. Second, the explanatory realm in which nongenetic factors are actual difference-making causes is distinct from the realm of genes. Although many studies in the field of epigenetics do not ignore the latter kind of genetic explanation, they focus their attention on the former. Here, explanations are sought that legitimately trace what I call “causes without mechanisms.”

128 ◂ FOUR MECHANISTIC EXPLANATION

The machinery of protein synthesis is elaborate, but the individual steps are, in principle, simple. BERNARD D. DAVIS, 1980

For God’s sake, let us be men not monkeys minding machines or sitting with our tails curled while the machine amuses us, the radio or film or gramophone. Monkeys with a bland grin on our faces. D. H. LAWRENCE, 1929

THE CONCEPT OF MECHANISM has been widely talked about in molecular biology ever since its rise as a discipline. For example, when François Jacob and Jacques Monod (1961, 353; emphasis added) tried to address the phenomenon of information transfer from DNA to protein, they described a mechanism taken to be responsible for the molecular explanandum phenomenon—the messenger RNA mechanism: “The property attributed to MECHANISTIC EXPLANATION

the structural messenger of being an unstable intermediate is one of the most specific and novel implications of this scheme. . . . This leads to a new concept of the mechanism of information transfer, where the protein synthesizing centers (ribosomes) play the role of non-specific constituents which can synthesize different proteins, according to specific instructions which they receive from the genes through M-RNA.” This interest in mechanisms can be traced back at least to the classical period of molecular biology in the early 1950s. James Watson and Francis Crick (1953, 737) refer to the organizational structure of DNA as a mechanism that is (in part) responsible for the phenomenon of DNA replication: “It has not escaped our notice that the specific pairing [of bases in the DNA helix] we have postulated immediately suggests a possible copying mechanism for the genetic material.” Since epigeneticists are primarily molecular biologists, it is thus not surprising that the word mechanism is used by them as well—despite the fact that nonmechanistic causal explanations are also widely used in this new field, as shown in the preceding chapter. For example, Eva Jablonka and Gal Raz (2009, 131; emphasis added) refer to four “mechanisms that underlie [cellular and transgenerational] epigenetic inheritance”: self-sustaining feedback loops, structural inheritance, chromatin marking, and RNA- mediated inheritance (see chap. 1). Moreover, this talk of mechanisms also appears in the few more philosophical considerations of epigenetics, unfortunately, without clarifying the concept of mechanism (see, e.g., Lamm 2013; and Lux 2013). Against this backdrop, the question remains whether the word mechanism we find in this newly emerging research on the molecular basis of epigenetic control systems means the same as its linguistic equivalent in classical studies in molecular biology. A prominent idea not only among traditional molecular biologists but also among philosophers of science in the so-called “new mechanism movement” is that a causal mechanism and its role in explanation are best elucidated by drawing on the properties of machines. Against this view, I argue that many epigeneticists introduce a concept of mechanism that draws on different ontological background assumptions. In addition, I show that in some epigenetic models mechanisms have to be labeled “noncausal,” since they contain non- difference-making component parts.

130 ◂ MECHANISTIC EXPLANATION

THE NEW MECHANISTS AND CONSTITUTIVE MOLECULAR EXPLANATION

The subject of mechanistic explanation, or whether causal explanations are in need of mechanistic information from lower levels of organization that back them up, was already broached in the previous chapter. Let us now discuss in more detail the concept of mechanism and the structure of constitutive mechanistic explanation. Since the publication of William Bechtel and Robert Richardson’s (1993) book Discovering Complexity: Decomposition and Localization as Strategies in Scientific Research, philosophers of science in general and philosophers of biology in particular have shown increasing interest in analyzing what mechanisms and mechanistic explanations are. This project was begun in an effort to make up for shortcomings of law-centered accounts of explanation in the special sciences dealing with regularities in multilevel complexity (see, e.g., Bechtel and Richardson 1993; Bechtel 2006; and Craver 2007). It has come to be labeled the “new mechanism movement” or “new mechanistic philosophy.” It was driven by the idea of making sense of the fact that biologists often refer to causal mechanisms when investigating living phenomena-to-be-explained (see Kauffman 1971a). According to the standard philosophical reconstruction of this scientific practice, mechanisms are thought to identify causal relations taken to be responsible for bringing about the explanandum phenomenon. In the new mechanistic philosophy we find an increasing number of definitions of the notion of “mechanism.” Here are some widely used examples:

Mechanisms are entities and activities organized such that they are productive of regular changes from start or set-up to finish or termination conditions. (Machamer, Darden, and Craver [hereafter, MDC] 2000, 3; emphasis added)

A mechanism for a behavior is a complex system that produces that behavior by the interaction of a number of parts, where the interactions between parts can be characterized by direct, invariant, change-relating generalizations. (Glennan 2002, S344; emphasis added)

A mechanism is a structure performing a function in virtue of its component parts, component operations, and their organization. The orchestrated

▸ 131 MECHANISTIC EXPLANATION

functioning of the mechanism is responsible for one or more phenomena. (Bechtel and Abrahamsen 2005, 423; emphasis added)

Following Bechtel and Richardson (1993), many so-called new mechanists in philosophy of science explicitly or implicitly take mechanisms to be real entities or, as Bechtel (2006, 33) puts it, “real systems in nature.” Accord- ingly, mechanistic models are explanatory because they allow us to grasp the causal structure of the world. By thinking about mechanisms as systems (Bechtel and Richardson 1993; Glennan 1996, 2002, 2009; MDC 2000) or structures (Bechtel and Abra- hamsen 2005) with particular causal capacities, more recent mechanistic approaches depart from older causal process accounts that describe mechanisms as a network of interacting processes.1 This second generation of new mechanists seeks to present a conception of (biological) explanation that not only breaks the mold of logical empiricists’ deductive-nomological (DN) accounts but also departs from other non-DN “physicalized” accounts, like those of Peter Railton (1978), Wesley Salmon (1984), or Phil Dowe (2000). These latter causal process accounts take a nexus of continuous physical processes to be a mechanism. This view usually rules out the possibility for causation to happen at higher levels of organization. Let us illustrate how mechanistic explanations work by returning to the example of information transfer from DNA to protein and Jacob and Monod’s (1961) description of the messenger RNA mechanism, respectively. Notwithstanding the differences expressed in the definitions of a mechanism and its parts, listed above—Bechtel and Abrahamsen as well as Glen- nan speak of parts and interactions, while MDC focus on entities and activities—Jacob and Monod’s description (JM) can be reconstructed as follows:

(JM) DNA (an entity) is transcribed (an activity) into mRNA (an entity). This new part of the mechanism is organized in such a way that it can bind itself to the ribosomes (other entities) that specify (an activity) the amino acid sequence of the protein product (a newly formed entity).

According to philosophical orthodoxy, (JM) is thought to explain an observed macro behavior—the phenomenon of protein synthesis—in terms of the behavior and organization of an underlying mechanism’s microcom-

132 ◂ MECHANISTIC EXPLANATION ponent parts, like mRNAs, and their causal capacities or properties.2 Thus, in the absence of an overarching law of protein synthesis, the mechanism described in (JM) can explain higher-level dispositions and properties of the embedding molecular system by the lower-level causal capacities of its parts. This view of mechanistic explanation through decomposing phenomena is inspired in various ways by earlier works, by Robert Cummins (1975) and Herbert Simon (1962), on functional relations and decomposition. Accord- ing to Cummins, functions occur in the framework of parthood. To offer a functional explanation means showing how causal capacities of particular components contribute to the global capacity of a system as a whole. The explanandum is the global capacity of the system, while the explanans is the organization of the system as well as the capacities of the system’s parts. In the new mechanistic philosophy, this so-called “causal role” account of functional explanation was deeply influential for conceptualizing the relations between parts and wholes located at different levels of organization (see, e.g., MDC 2000, 2007; Bechtel and Abrahamsen 2005; Menzies 2012; and Levy and Bechtel 2013). In addition to building on Cummins’s view of function, the new mechanistic movement has drawn inspiration from Herbert Simon’s so-called “near-decomposability theorem.” For Simon, the behavior of a hierarchically organized system is brought about by a mechanism whose parts have a stronger causal relation to one another than to things outside the mechanism. Thus, the mechanism’s relative “independence”—its so-called modularity—is warranted by the degrees of internal interaction among its parts.3 In some cases, the system is not fully decomposable because the internal interactions are strong, while in other cases these interactions between parts are more or less negligible and the system is near decomposable. This framework is used by new mechanists to distinguish between systems that are organized, such as a mechanism, and mere aggregates that do not exhibit such a property (see Bechtel and Richardson 1993; Craver 2002b; Bechtel and Abrahamsen 2005; and Levy 2013; see also Wimsatt 1974, 1986). Against this background of parthood and decomposability, many new mechanists state that elucidating a mechanism through the identification and manipulation of its component parts should provide understanding, prediction, and control. Usually these epistemic features are thought to be directly derivable from knowledge about which causal role a part plays in

▸ 133 MECHANISTIC EXPLANATION

a mechanism. According to the influentialinterventionist view of mechanisms, such a causal role can be attributed to a part if it makes a difference in the higher-level properties of the whole system. More generally, this means that not only lower-level behaviors—intralevel relations—but also relations between the lower and the higher level—interlevel relations—can be characterized by means of James Woodward’s (2003) criterion of invariance. In other words, interventionism is thought to elucidate and justify the structure of not only causal explanations in biology but of mechanistic explanations as well. For those adopting this view, like Carl Craver (2007), mechanistic explanation may be described as tracing change-relating, interlevel dependencies between variables, whereas a change in lower-level variables (i.e., parts of a mechanism) would change higher-level variables (i.e., the whole).4 Although this Woodwardian understanding of explanation in terms of mechanisms makes it a close relative to causal explanation, causal and mechanistic explanations are not the same. They address related but different explananda. The answer to “What synthesizes a protein?” or “What makes a cell contain a protein?” is different from the response to “How is a protein synthesized?” While the first two questions address those molecular entities and their interaction, which constitutes the molecular system of a cell to have proteins, the second one is an inquiry about the causal history or process leading to the appearance of proteins. The former explanation relates causal capacities or properties of things across different levels of organization, while the latter refers to events (see Ylikoski 2013).5 The former is constitutive, while the latter is causal (see also chap. 3). This chapter deals with explanations in terms of causal mechanisms tracing constitutive molecular dependencies. In the case of Jacob and Monod’s investigations, as in many other scientific inquiries, both explanatory projects are interlinked in order to better explain the phenomenon. This two-faceted research agenda has led philosophers of science to overlook its inherent differences. In much of the recent mechanism literature using ambiguous notions like “systems’ behaviors” and “activities of mechanisms,” the explanandum of a constitutive explanation is mistakenly considered to refer to both causal capacities and events. In contrast to causal dependency relations, however, constitutive relations are synchronous (they do not occur over time) and asymmetric (the parts’ causal capacities constitute the system and not vice versa).6 In addition,

134 ◂ MECHANISTIC EXPLANATION their relata cannot be conceived as independently existing. While the DNA (the cause) and the protein (the effect) can exist independently of each other along the causal chain DNA → mRNA → protein, in the case of constitution, being a cell having the causal capacity to synthesize proteins is to be specifically constituted by entities like DNA and mRNA. These entities in turn are themselves organized in a certain way and perform certain activities. Of course, the mechanism’s component parts themselves might be organized spatiotemporally, but their interlevel relation to the system’s properties is not successive. These mechanistic constitutive dependencies are depicted in figure 4.1.

FIG. 4.1. Consensus view of a multilevel mechanism. Three causal processes are located on different levels of organization. The whole system (S with behavior Ψ) can be decomposed into a set of component parts (e.g., x3 with property Φ3) on the next lower level. These parts figure themselves as wholes that can be decomposed into parts

(e.g., z1, with property µ1) on the next lower level. Vertical relations between parts and wholes are constitutive. Horizontal interactions are causal. For further description, see the text (after Craver 2007, 194).

▸ 135 MECHANISTIC EXPLANATION

In the case of Jacob and Monod (1961) the behavioral-level explanandum phenomenon S can be understood as a cell synthesizing proteins (under genetic control and at different rates), which is behaviorΨ .7 At the lowest level of organization we find the messenger RNA mechanism containing

(among other things) the molecular entities z1, . . . , z3 and their properties

µ1, . . . , µ3 involved in transcribing, transporting, translating, and degrading mRNA. This mechanism is itself a component part of a larger, higher-level mechanism, summarized, for example, by the central dogma. It contains

the entities x1, . . . , x4 and properties Φ1, . . . , Φ4. Both mechanisms together constitute S’s Ψ-ing (in Craver’s terminology). S’s Ψ-ing can change over time through a diachronous alteration of the causal properties of the mechanisms’ parts and/or their organization on different levels of organization.

The “vertical” interlevel relations between properties of parts (e.g., x3 being an mRNA molecule) are, however, constitutively related to other properties of the system (like the amount of proteins synthesized). That is, a change in the parts’ properties on different lower levels would bring about a change in S’s Ψ-ing in a synchronous, nonsuccessive manner. In other words, mechanisms are “vertically” static or fixed. There is a strong consensus among new mechanists that the description of causal mechanisms adequately accounts for the explanatory practice in a particular field in biology: cell biology and molecular biology (see Bechtel 2006; Darden 2005, 2006; and Darden and Tabery 2010). In fact, those two disciplinary domains are linked to nearly all paradigmatic cases in contemporary philosophy of biology used to illustrate mechanistic explanations: of protein synthesis (MDC 2000, 18–21; Darden 2005, 359–61, 365–66), photosynthesis (Tabery 2004, 4–8), synaptic plasticity and action potential (Craver 2002a, S85–S88), synapse transmission (Bogen 2005, 400; Tabery 2009, 655–57; Andersen 2012, 423–25), Mendelian genetics (Darden 2005; Glennan 2005, 446), cellular metabolism (Bechtel and Abrahamsen 2005, 428–29), and Krebs cycle (Perini 2005, 260–65), as well as stemness and cell differentiation (Fagan 2012a).8 Given the fact that epigeneticists are basically molecular biologists interested in tracing mechanisms of gene regulation, cellular inheritance, and differentiation, one should expect their explanations to meet the basic criteria of the mechanistic molecular explanation delineated above—(1) parthood, decomposability, and “static” interlevel relations, as well as (2) describability by means of interventionism. In the next two sections, I show that this assumption is fallacious.

136 ◂ MECHANISTIC EXPLANATION

MACHINES IN BIOLOGY AND PHILOSOPHY, AND “DEMACHINIZATION”

The concept of mechanism and mechanistic explanation is usually introduced in a quite natural manner into the world of epigenetics and the debate on whether to extend the modern synthesis. For example, Werner Callebaut (2010, 467; original emphasis) emphasizes that, in this context, “true explanations ought to provide us with the mechanism, the how and why at work in any particular pattern or process we attempt to explain.” This unquestioned adoption does not address, however, whether the new mechanists’ description(s) of a mechanism and mechanistic explanation suitably grasp(s) what is at stake when investigating epigenetic mechanisms at the molecular and cellular level. In addition, the potential contemporary change in modern biological theory—a change in the content or focus of explanations—is usually not considered to include a conceptual change with regard to the notion of mechanisms. In order to address these issues, let us first focus on how biologists use the notion of mechanism.

MACHINES IN BIOLOGY

Throughout the centuries the notion of mechanism has been a common “playmate” for biologists. Usually it has been linked to reductionist investigations of biological phenomena or, more generally, to a deterministic or atomistic understanding of nature. This traditional view is usually labeled mechanicism. Inspired by René Descartes’s ([1667] 1998, 171) physiological work on the “machine of our body,” this view invoked human-made machines, like combustion or steam engines, as models of living systems. With respect to humans, for example, Giovanni Borelli (1680) in his De motu animalium explains the actions of bones, joints, and muscles by means of the lever rule and illustrates this explanatory approach with corresponding pictures of “human machines.” In a similar way, Fritz Kahn (1926) depicts the biochemical processes of the human body as processes in a factory. With certain similarities to nineteenth-century reductionist biologists such as Hermann von Helmholtz, Emil du Bois-Reymond, Ernst Brücke, and Carl Ludwig, biologists at the dawn of molecularization studies in the 1930s and 1940s were fascinated with elucidating molecular mechanisms and their components’ interactions with increasing resolution and at lower and

▸ 137 MECHANISTIC EXPLANATION

lower levels of organization (see Galaty 1974). Despite growing criticism of mechanistic reasoning in biology (see Hertwig 1900, 4; and Hartmann 1937, 37), the availability of new techniques, such as X-ray crystallography, paper chromatography, ultra centrifugation, and electron microscopy, provided the means to characterize molecular structures both physico-chemically and quantitatively. These techniques induced the molecularization of other fields as well. In cell biology, for example, it turned descriptive cytology into molecular cell biology (see Alberts et al. 2002; and Bechtel 2006). After Warren Weaver (1938) coined the term “molecular biology” and William Astbury (1941) picked it up, this initially rather diffuse field, which had been united merely by a common interest in a class of tiny macromolecules, gradually developed into a more coherent interdisciplinary, reductionist research program including various experimental practices (see Olby 1990). This program was famously described by Francis Crick (1966, 10): “[The] ultimate aim of the modern movement in biology is to explain all biology in terms of physics and chemistry.” A common, recurring motif shared by many molecular biologists soon became the rediscovered mechanicists’ concept of machine mechanism, as Daniel Nicholson (2012) calls it. The machine metaphor was crucial, especially in cybernetics, as in John von Neumann’s theory on self-reproducing and evolving cellular automata (Neumann 1966; see also Kay 2000, 73–127). In addition, analogies to human-built machines were drawn to aid in grasping what kind of particular ontological class comprises molecular mechanisms, like cells, that is, how they differ from or correspond to artificial machines. For example, François Jacob (1974, 271; emphasis added) argues that “if the bacterial cell is to be considered as a factory [i.e., an assemblage of various machines] it must be a factory of a special kind. The products of human technology are totally different from the machines that produce them, and therefore totally different from the factory itself. The bacterial cell, on the other hand, makes its own constituents; the ultimate product is identical with itself. The factory produces; the cell reproduces.”9 Such analogies were echoed outside the sciences, in the work of the Polish writer Stanisław Lem, for example. Being deeply inspired by cybernetics and the discovery of DNA, in his collection of essays titled Summa Technologiae he argued that DNA-based evolution and computational machines, understood as Turing machines, actually show evolutionary processes of the same kind: “Just as the birds conquered the sky and the herbivorous mammals the steppe, the combustion engine

138 ◂ MECHANISTIC EXPLANATION vehicle took mastery over the roads, thus giving rise to ever more special- ized varieties. In the ‘struggle for life,’ the automobile not only pushed out the stagecoach but also ‘gave birth’ to the bus, the truck, the bulldozer, the fire engine, the tank, the off-road vehicle, and dozens of other means of transport” (Lem 1964, 15–16). Lem even argues that in technological evolution we also find “sexual selection,” leading, for example, to aerodynam- ically less efficient automotive designs, fantastically shaped cooling pipes, and so on, although, he admits, nature cannot alter the models it produces each year, as, for example, the fashion industry can. In addition to this, molecular mechanisms were considered by many not only to behave as if they were machines or to resemble machines but to be machines themselves.10 For example, the cell has been ontologically characterized by Jacques Monod (1977, 108; emphasis added) to be a machine: “By its properties, by the microscope clockwork function that establishes between DNA and protein, as between organism and medium, an entirely one-way relationship, this system obviously defies ‘dialectical’ description. It is not Hegelian at all, but thoroughly Cartesian: the cell is indeed a machine” (see also Nicholson 2012, 154). This second, metaphysically loaded way to use machine language caught on rapidly and became a common linguistic motif as well as part of the professional terminology in contemporary molecular biology. For example, scientists have called molecular entities like chaperones and cilia molecular machines (Clare and Saibil 2013, 846; Raeker et al. 2012, 11241). In addition, they have described activities like the bacterial outer membrane’s β-barrel folding to be facilitated by the “Bam machine [i.e., β-barrel assembly machine]: A molecular cooper” (Ricci and Silhavy 2012, 1067).11 Most recently, this linguistic development has been driven by the field of synthetic biology, in which the metaphorical border between human- made and natural machines has been crossed in an unprecedented manner. Synthetic biology is a field that combines engineering with molecular and cell biology. For example, every year the synthetic biology community organizes iGEM, the International Genetically Engineered Machines competition, in which participants build simple biological systems from standardized, Lego-like building blocks called BioBricks. These systems are then surgically placed into living cells. The aim is to use engineering techniques to create semiartificial molecular machines—usually cells with synthetic genomes that have new and often quite unusual properties, like cells that

▸ 139 MECHANISTIC EXPLANATION

produce a light show by fluorescing in synchrony. This engineering science investigates and modifies living systems under the slogan “Life is a DNA software system and . . . the information contained within that DNA code contains all the information necessary for life,” as J. Craig Venter (2014, 7), the field’s “godfather,” puts it.12 Classical and contemporary molecular biologists’ fascination with investigating living systems by reducing them to their underlying cellular and subcellular molecular structures and by conceiving of these entities, as well as their interactions, as (assemblages of) machine mechanisms has brought about two developments. First, it facilitated the development of a mechanistic view of how evolution, through blind natural selection, opti- mally designs living systems on the organismic and molecular level (see Dawkins 1976, 1986; Dennett 1995). This includes the idea that organisms themselves have been turned into machines (Jacob 1974). The ontological assumptions of classical mechanicism underlying this use of machine language lead to a number of dangerous misconceptions and constraints in biological research.13 As Nicholson (2013, 2014a) has shown in a series of papers, machine conceptualizations hit the wall when one tries to investigate properties of living systems that machines do not share, such as self- maintenance and reproduction, as well as intrinsic purposiveness (i.e., the internal operations of living systems are directed toward the maintenance of their own organization rather than being designed to fulfill an external agent’s functional ends, as is the case of machines).14 Second, and more importantly with respect to the issue discussed in this chapter, molecular biologists’ fascination with machines has fundamentally shaped the way philosophers of science have conceived of mechanisms and mechanistic explanation in biology. In fact, the notion of mechanism among new mechanists is not as clearly distinguished from the classical concept of machine mechanism as is often assumed (Nicholson 2012; see also Woodward 2013). Nicholson argues that the concept of mechanism (he calls it “causal mechanism”), used in the new mechanistic philosophy to conceptualize explanatory practices in the biosciences, and the classical metaphysical thesis on the nature of life, known as mechanicism, are regularly conflated by new mechanists. As a consequence, causal mechanisms would be better understood epistemically instead of ontically in order to disabuse them of the idea that mechanisms and machines are similar. By building on these claims by Nicholson and offering examples of mechanism

140 ◂ MECHANISTIC EXPLANATION language in epigenetics, I will highlight below the limits of mechanicists’ central ideas—decomposability, discreteness, and fixity—when used in order to conceptualize and explain complex living systems. In addition, I sketch some potential solutions for how this atomistic conceptual framework can be replaced by a more accurate ontology of mechanisms.15

MACHINES IN PHILOSOPHY

The new mechanist movement’s foundation pillars were set by Bechtel and Richardson (1993). Their examination of how mechanisms are discovered is heavily influenced by cell and molecular biologists’ assumption that we can legitimately investigate molecular mechanisms by functionally and structurally decomposing complex systems into locatable things at lower organizational levels.16 In addition, as Nicholson (2012) has correctly noticed, Bechtel and Richardson (1993, 17; emphasis added) integrate the machine jargon of molecular biologists into their definition of a mechanism and mechanistic explanation: “By calling the explanations mechanistic, we are highlighting the fact that they treat the systems as producing a certain behavior in a manner analogous to that of machines developed through human technology. A machine is a composite of interrelated parts, each performing its own functions, that are combined in such a way that each contributes to producing a behavior of the system. A mechanistic explanation identifies these parts and their organization, showing how the behavior of the machine is a consequence of the parts and their organization.” Until today, the idea, suggested by classical molecular biologists, that mechanisms have something in common with machines—be it that mechanisms are indeed machines or that mechanisms can at least be epistemically grasped and conceptualized as machines and thus better manipulated and controlled—is widely shared among new mechanists. Not surprisingly, we find machine language in metaphysical investigations carried out by those authors interested in clarifying the nature of a causal mechanism. For example, for Stuart Glennan, biological mechanisms are akin to combustion engines and watches (see Glennan 1996, 52–53; 2002, S345; and Kuhlmann and Glennan 2014, 340–41). However, also in those accounts by new mechanists who regularly stress that they do not take a metaphysical perspective on mechanisms, the classical mechanicists’ machine framework has survived. This includes nearly all recent epistemic

▸ 141 MECHANISTIC EXPLANATION

views of mechanisms. For example, Craver, Bechtel, and Darden regularly use analogies from engineering or machines, such as mousetraps, steam engines, windmills, cars, radios, and so on, to describe how mechanisms work (see, e.g., Craver and Bechtel 2006; Bechtel 2006, 47–49; and Craver and Darden 2013, 16).17 However, most authors interested in the mechanistic explanation usually downplay the conceptual continuity of mechanicism in today’s theories or consider it to be of minor importance. They usually state that the term “mechanism” has simply changed over time and that there is no serious danger of conflating the traditional with the modern view (see MDC 2000, 3; and Craver 2007, 3).18 However, this danger is, in fact, a serious one. Take, for instance, this statement of Craver and Darden (2005, 234) in which they treat mechanicists’ machine mechanisms and biological mechanisms as natural partners: “From the perspective of biology . . . one might tell a triumphal story of the success of mechanism [i.e., mechanicism] over various forms of vitalism, as well as over biological theories appealing to intelligent design. Indeed, one cannot open a journal in any field of contemporary biology without encountering appeals to the mechanism [i.e., what new mechanists call a causal mechanism] for this or that phenomenon” (see also Nicholson 2012, 154). Other hints as to the tacit background ontology of epistemically interested new mechanists can be found by considering the conceptual framework they drew on when starting to think about what a biological mechanism actually is and how its description can explain (1) Robert Cummins’s (1975) view that functions appear as instantiations or constitutive relations (not extended in time) between causal capacities of parts and wholes and (2) Herbert Simon’s (1962) idea of (near-)decomposability according to which different parts and wholes can be distinguished according to the degree of their internal interactions. From this conceptual framework one easily ends up with a machinelike concept of a mechanism. It states, in line with the causal role view of functional explanation, that (i), parts do affect wholes constitutively and in a synchronic, nondynamic manner, respectively. Thus, mechanisms show a temporally static and fixed vertical relationship between levels. As a consequence, new mechanists are usually not interested in understanding the history of a system and/or a part of a system. Both are usually treated as pregiven. In other words, it is not questioned why there is a particular part-

142 ◂ MECHANISTIC EXPLANATION whole relation or how the different levels of organization in a mechanism emerge, evolve, or develop. Like in a machine, they are simply there. An additional assumption is (ii), that wholes, parts, and levels of organization show discreteness. This is the case because the criterion of parthood, according to which wholes are at a higher level than the parts of which they are composed, suggests a strict hierarchical organization of living systems as mechanisms. Thus, mechanisms must consist of discrete levels that can be precisely demarcated from one another. These levels of organization have determinate boundaries, and their properties (i.e., the behaviors of parts and wholes) are distinguishable, intrinsic properties. These ideas are supported by a general antiholistic view shared among new mechanists according to which not every part of a mechanism is connected to every other part (see Kuhlmann and Glennan 2014). Even though most new mechanists agree that there is a general perspec- tivism underlying their views of mechanisms (e.g., a system can be decomposed in various ways), they usually accept a realist account of explanation in which descriptions of mechanisms are explanatory because they refer to real things. Thus, we may ask how the ontological background assumptions (i) and (ii), which are perfectly compatible with those of classical mechanicists, influence the way new mechanists conceive of mechanistic explanation. The examples usually chosen by new mechanists to explicate how mechanistic explanation works are basic, machinelike textbook examples from classical cell and molecular biology, such as photosynthesis and synapse transmission. In these cases, parts usually do have clear molecular sizes and shapes and interact in a local manner. Interestingly, however, new mechanists also seem to believe that their conceptual framework can be employed to explain more complex systems with fewer discrete parts and less space- and temporally restricted interactions, such as cellular metabolism, regulatory networks, and cell differentiation. In such cases, it is acknowledged that machinelike conceptualizations may fail sometimes. For example, Bechtel and Richardson (1993, chap. 4) note that the decomposition strategy and the machine analogy, respectively, may fail when parts are difficult to isolate and their causal roles do not show sufficient independence. At the same time, however, they seem to be convinced of the general (although not universal) usefulness of the machine view when applied to complex systems with nonlinear interactions. Take, for instance, Bechtel’s (2011, 2013a, 2013b) and Bechtel and Abrahamsen’s (2010, 2011, 2013) more

▸ 143 MECHANISTIC EXPLANATION

recent work on mechanistic models in systems biology and chronobiology. Here, they describe the new mechanists’ concept of mechanism applied to complex systems as a continuation of a tradition that can be traced back to Descartes’s natural philosophy. For example, Bechtel (2011, 536) states that new mechanists operate with a merely “extended Cartesian conception of a mechanism” (see also Bechtel 2006, 20–24; and 2008, chap. 1). There has, however, been an important expansion to the new mechanist framework, as highlighted, for example, by Bechtel and Abrahamsen (2010). They argue that the standard framework neglects the epistemic virtues of other tools, especially dynamic models, for explaining the temporal organization of mechanisms and their interlevel dynamics. They emphasize that the epistemic roles of these mathematical models in, for example, investigations of oscillatory phenomena, are not recognized sufficiently in the standard framework. However, at the same time, they claim that this so-called dynamic mechanistic explanation is compatible with the consensus machine mechanism view. It supplements (not replaces) the program of decomposition by conveying how a mechanism is situated and how its parts work together—a procedure they call recomposition. This means that even though Bechtel and colleagues acknowledge the limitations of the new mechanists’ conceptual framework and its need for complementary enhancement by other explanatory practices, they continue to draw on basic assumptions underlying machine views of mechanisms. They argue that in order to understand complex systems, like the circadian rhythm, we not only have to take the machine—in this case, a clock—apart but also put it back together again. We have to “recompose the clock” (Bech- tel 2013a, 223). Thus, the overall story stays the same: explaining is engineering.19 There is no investigation into whether the internal organization of a mechanism (i.e., its inter- and intralevel relations) could be understood using criteria other than parthood and composition, respectively. On such criterion is scale—for example, time scale (see below). In general, new mechanists’ fascination with applying old mechanicist ideals to today’s systemic view of biology is puzzling. If one takes a realist’s stance, it becomes quite confusing to discern what the target system should look like, as it is anything but straightforward for one to imagine how complex living systems should, for example, be made up of nondiscrete parts with non-intrinsic properties in a machinelike manner. Surprisingly, in those cases in which new mechanists seem to recognize the limitations of

144 ◂ MECHANISTIC EXPLANATION their conceptual framework—once parts are less well defined and interactions less discrete and local—they keep on using it nevertheless. Often this persistence is legitimized by assuming the superiority of the heuristic value of their framework for the target system.20 Most important, even though the limitations of the classical mechanicist framework are increasingly acknowledged by new mechanists, they usually do not reflect on the general biases their program imposes on our understanding of what mechanistic explanation in modern biology actually is. In addition, they have difficulty with moving away from their basic ontological assumptions that have led to these biases. For them, mechanisms always have to be conceptualized within the narrow framework of parthood, discreteness, and fixity. By drawing on epigenetic mechanisms, I show in the next section that this assumption is erroneous. In fact, epigenetics might offer stimulating examples for developing new ways of conceptualizing what a biological mechanism is.

EPIGENETIC MECHANISMS, “DEMACHINIZATION,” AND PROCESS ONTOLOGY

Before proceeding, it is important be clear about one thing: the rather narrow, machinelike view of mechanisms has been a powerful tool in biology for a long time. It allowed molecular and cell biologists in the twentieth century to come up with astonishing explanations for how causal processes at different levels of organization are related to one another. This was possible since the macromolecular structures of molecular mechanisms and/ or their parts that were investigated readily met the criteria of localizabil- ity and decomposability. For example, electron microscopy enabled these researchers to draw a fascinating picture of a cell, a membrane, a chaperone, or a cilium. These ostensibly accessible systems were usually studied in such a way that it—the machine—can be broken down into distinct, context- independent, sufficiently stable structural parts, to which we can precisely attribute well-defined output functions. However, this molecular atomism is in stark contrast to the growing ideal of systemic thinking in modern cell and molecular biology and the fact, stressed by Raphael Falk (2011, 382), that over the last decades “biology became very much the science of complex systems” and that it grew more interested in many-to-many rather than one-to-one or one-to-many relationships (see also Woese 2004).

▸ 145 MECHANISTIC EXPLANATION

According to this notion of interactivism, molecular explanations can incorporate functions of (sub)cellular assemblies only if the assemblies are understood to be collectively interdependent and if the inner- and extraorganismic contexts that make their functions possible are taken into account. Accordingly, Bruce Alberts et al. (2002, 473; emphasis added) note with respect to epigenetics that “although studies of the epigenome are in early stages, the idea that environmental events can be permanently registered by our cells is a fascinating one that presents an important challenge to the next generation of biological scientists” (see also Davies 2012, 42). Today, many epigenetic researchers have adopted this systemic perspective as one of the most crucial principles guiding their investigations. This holds true for various cases in which the traditional antiholistic, ontological assumptions underlying the traditional view of mechanism seem to have been overcome. Let us first consider how epigeneticists react to the idea that mechanisms show machinelike discreteness and parthood or, to be more precise, that mechanisms are made up of discrete entities with intrinsic causal properties located on different levels of organization and connected through compositional relationships to one another. It is actually not surprising that the heuristics of decomposition and localization that build on these assumptions regularly fail when applied to complex systems. Take, for instance, the basic difference between the gene and the “epigene.” In line with Watson and Crick’s (1953) reading, the mechanism bringing about a protein Y to be expressed by a gene X is hard to decompose structurally by itself since we cannot structurally decompose the DNA into distinct genes. In other words, it is difficult to define X’s boundaries by referring to its physical properties alone. However, genes can be characterized more precisely in functional terms.21 This means that despite the complicated phenomena of nontranscribed DNA sequences, overlapping coding regions, and nonprotein coding DNA sequences, gene X’s boundaries can be described as corresponding to specific boundaries in a DNA sequence showing the particular function to express protein Y. This procedure of cutting a molecular mechanism at its joint is aimed at identifying constitutive subunits and their causal capacities. Let us now take a look at epigenetics. As Adrian Bird (2007, 396) correctly emphasizes, “Geneticists study the gene; however, for epigeneticists, there is no obvious ‘epigene.’”22 First of all, with respect to the above system of gene X and protein Y, this means that it is simply not possible to decom-

146 ◂ MECHANISTIC EXPLANATION pose the system and thus indicate what epigenetic subunits are involved in bringing about the phenomenon—at least if the target system is considered to be a machine or machinelike. One may argue that here we are merely dealing with linguistic differences and that we could, in principle, introduce the term “epigene” as a placeholder for a finite class of nongenetic entities that one could in fact identify as constitutive parts of a molecular mechanism. However, I think that this linguistic feature is crucial since it gives us a hint about what a large number of epigeneticists seek to explain by describing mechanisms. If we replace the “epigene” with, for instance, a methylation mark on gene X, would this help epigeneticists solve the decomposition and localization dilemma described above? No. The reason is that although it is important for epigeneticists to locate where in the genome DNA methylation occurs, pairing the system’s macro-level operations and behaviors with the regulatory capacities of these potential component parts is not as straightforward as it seems. More generally, there are as many epigenomes as there are cells, and every cell’s epigenotype fluctuates over time depending on the intra- and extracellular context. Thus, the causal capacities of both epigenetic mechanisms and their parts are considered to be essentially relational (or global) and not intrinsic. As Paul C. W. Davies (2012, 42; emphasis added) notes, “The genome is associated with a physical object (DNA) with a specific location, whereas the epigenome is a global, systemic, entity. Furthermore, genomic information is tied to specific coded molecular sequences stored in DNA. Although epigenomic information can be associated with certain non-DNA molecular sequences, it is mostly not. Therefore, there does not seem to be a stored ‘epigenetic programme.’”23 Thus, explaining—on a more general level—how epigenetic regulation and/or inheritance mechanistically constitutes a cell’s or organism’s phenotype that exhibits certain properties is possible only by incorporating all relevant environmental, organismic, cellular, subcellular, genetic, and other contextual factors and their interactions into a mechanistic model. This shift in explanation toward the idea that all properties the epigenome exhibits are essential relational properties can be grasped only by adopting a truly decentralized, non-atomistic view of molecular dependencies. It may be labeled “epi-geneticization,” as Linda Van Speybroeck et al. (2007) suggest, or “demachinization.” Yair Neuman (2007, 625) describes this general systemic view as follows: “Study- ing biological systems means more than breaking the system down into its

▸ 147 MECHANISTIC EXPLANATION

components and focusing on the digital information encapsulated in each cell. Studying biological systems also involves examining how digital information turns into biological meaning and how biological components are orchestrated through various interactions to constitute the living whole. The idea that living systems should be studied as wholes is the basic axiom of what is known as the systems perspective in biology.” This systems perspective, shared by many epigeneticists, expresses a general skepticism toward the idea of chopping the system under study into ontologically distinct units with discrete causal capacities. It may be understood as directly resulting from the difficulties that arise when seeking to conventionally investigate complex molecular assemblies, their properties, and (sub-)cellular interactions and, at the same time, trying to consider these assemblies’ inner- and/or extraorganismic context. In short, this view suggests that epigenetic mechanisms, unlike machines, cannot be easily taken apart. More generally, we see that this holistic stance clearly rejects the basic ideas of parthood and decomposition underlying the new mechanists’ concept of mechanism. But do epigeneticists at least keep the ontological assumption defended (explicitly or implicitly) by most new mechanists— that, in a mechanism, interlevel relations between entities located at different levels of organization are synchronically organized. In other words, are epigenetic mechanisms vertically static or fixed? As already discussed with respect to the distinction between proximate and ultimate cause explanations (see chap. 2), the idea of interlevel synchronicity becomes difficult in developmental mechanisms in general. For example, to explain how a fertilized egg develops into a multicellular adult organism, one has to describe a mechanism that relates, across levels of organization, the causal capacities of a system’s parts or their organization at an earlier phase in development with capacities of the whole system (e.g., a particular trait) at a later phase in development (e.g., in the adult organism). Petri Ylikoski (2013) labeled this kind of mechanistic explanation “hybrid explanation,” as it replaces the traditional idea of interlevel constitution with interlevel causality. This means that in mechanistic explanations the explanans and explanandum are related diachronically rather than synchronically, as they are in the usual case. One may argue that this special explanatory focus in developmental investigations is backed up by a more dynamic ontology of how developmental mechanisms in general and epigenetic mechanisms in particular are

148 ◂ MECHANISTIC EXPLANATION built up. Because epigenetic mechanisms cannot vertically set up an organism synchronically, they have to build up the organisms diachronically and over time. In other words, in a process from single cells to multicellular organisms, different levels of organization—cell, tissue, and organism— arise over time. What exactly does this mean? The new mechanists usually treat the levels of organization of a target system as pregiven. Due to this ahistorical vertical perspective, the states of a pregiven system change over time, but the system remains the same. A mousetrap stays a mousetrap, come what may. It will never be a car engine. As a consequence, and in line with Cummins’s (1975) concept of functional explanation, organization is always treated as part of the explanans, not the explanandum. In other words, it is never questioned as to how the different levels of organization in a mechanism emerge, evolve, or develop. In contrast, in many fields of epigenetics the levels of organization, as well as the entities and causal processes located at these levels, are not conceptualized as already existing and gradually changing over developmental and evolutionary time but as emerging during plastic development. Take, for instance, the epigenetic mechanisms of cellular differentiation. They do not build up tissues and organisms the way one puts together the components of a car in order to construct the vehicle. Instead, in dynamic living systems the entities and their properties (as well as the levels of organization at which those properties are located) arise in a nonlinear manner. For example, as Sui Huang (2012a) argues, during ontogenesis the causal properties of cells emerge in a spontaneous way from the collective and nonlinear interactions in gene networks (see also Strohman 1995; Wagner 1996; Bronfman et al. 2014; and Burggren 2015). As a consequence, lower-level entities always have the collective capacity to develop quite different higher- level entities. Only against this backdrop does the idea that “coexistence of multiple stable phenotypes within one genotype,” which allows for environmental responsiveness, developmental plasticity, and “sudden, broad evolutionary changes” (Huang 2012a, 149), make sense. In contrast, the components of a mousetrap usually do have different capacities. Once assembled, they may change (e.g., they may show signs of wear), as might the mousetrap along with them (e.g., the wear allows mice to survive and escape), but the mousetrap’s components, no matter how they are put together, usually do not allow one to build a car engine, for instance. An epistemic program that incorporates these fundamental

▸ 149 MECHANISTIC EXPLANATION

differences also embraces a less mechanicist and preformationist but more dynamic view on mechanisms that understands biological organization not only as an explanans but as an explanandum that needs to be addressed mechanistically itself, notwithstanding the view of most new mechanists. So far we have seen that the holistic and dynamic perspective taken by many epigeneticists rests on the insight that neither the decomposition- and-localization framework nor the idea of vertical fixity in mechanisms is the gold standard of conceptualizing biological mechanisms and their interlevel dependencies. In other words, epigenetics allows one to become aware of erroneous and outdated ontological assumptions underlying philosophers’ (and unfortunately many biologists’) views of mechanisms. Now that we have outlined how many epigeneticists do not understand their target system, can we more precisely grasp how they in fact conceive of it? As is apparent, the limitations of traditional machine-inspired metaphysics for conceptualizing epigenetic mechanisms have led some authors to avoid making any ontological assumptions about the target system at all. For example, Paul C. W. Davies (2012) suggests that we should take the epigenome not to be a real object but a “virtual object.” Others have suggested that the epigenome does not exist in isolation. For example, Ji-Hoon Joo and colleagues (2011, 289; emphasis added) are interested in the metabolic processes producing the assemblies of the DNA methylation mechanism during development state: “The metabolic pathways that generate the primary methyl donors needed for the de novo establishment [for example, during meiosis] and maintenance [for example, in cell lineage] of the DNA methylation profile are complicated anddo not exist in isolation from other essential biochemical reactions necessary for cell survival, division, and differentiation.” What does this mean? Do epigenetic mechanisms and their internal relations not exist in the real world? One could say that they do not exist as machines. In a world that is fundamentally built up on substances— organisms, cells, genes—and clear compositional hierarchies among these substances, then epigenetic mechanisms cannot exist. However, if we reject this substance-based view of the living world, which underlies not only classical molecular biology and a substantial part of today’s biology but also the new mechanistic philosophy, and instead embrace a more processual view of reality, then epigenetic mechanisms might come to life. A process-centered metaphysics suggests that all living things are fundamentally made up of numerous interconnected processes. All things, includ-

150 ◂ MECHANISTIC EXPLANATION ing organisms, cells, and genes, are not spatio-temporally discrete units but ultimately processual. Inspired by the works of Alfred North Whitehead (1929), this notion of a processual metaphysics laid the conceptual and experimental foundations of Conrad Hal Waddington’s classical epigenetics (Waddington 1970a, 1977c; see also Peterson 2011).24 For example, Wadding- ton understood the organism as a process of development from a zygote to death in which every time-slice of this sequence can only be understood in light of its history and future trajectory. In other words, this view focuses on the becoming of entities rather than on their being. It characterized how entities, like organisms, emerge, are stabilized, and thus maintain themselves, and it did so by describing the various interconnected processes underlying every stage in the history of the entity. Waddington’s epigenetic landscape images are meant to express this very idea (see chap. 2). The tradition of substance metaphysics has dominated philosophers’ as well as biologists’ views of reality so far. But it seems that times may be changing. Given the trend toward systemic investigations of complex dynamic systems in biology, as well as a growing number of philosophers of science, first and foremost John Dupré and colleagues (see Bapteste and Dupré 2013; Dupré 2013, 2014; Dupré and Guttinger 2016; and Nicholson and Dupré 2018; see also Henning and Scarfe 2013), who emphasize the need to adopt a processual perspective for studying the complex dynamics and close connectedness of living systems, maybe the time is ripe to give process ontology a serious chance. A big advantage of the processual view is that it can more easily conceptualize the fuzzy boundaries we find in complex systems. While parts and wholes are thought to have determinate and sharp boundaries, living systems rarely show this kind of discreteness. Examples of this include the permanent interconnectedness of developing systems with their environment and the omnipresence of symbiosis.25 We can more easily conceptualize these entities as processes, because they are more flexible when we seek to know when exactly an entity begins or ends. This view fits nicely with comments we find in the field of epigenetics, such as that of Warren W. Burggren (2015, 80): “Epigenetic phenomena should not be regarded as ‘digital’ (on-off), in which a modified trait necessarily suddenly disappears between one generation and the next. Rather, dynamic epigenetic phenomena may be better depicted by graded, time-related changes that can potentially involve the ‘washout’ of modified phenotype both within and across

▸ 151 MECHANISTIC EXPLANATION

generations. Conceivably, an epigenetic effect might also ‘wash-in’ over multiple generations, and there may be unexplored additive effects resulting from the pressures of environmental stressors that wax, wane and then wax again across multiple generations.” More generally, the process view no longer necessitates distinguishing levels of organization in mechanisms by the criterion of parthood, which is, as we have seen above, usually understood to be a synchronic and noncausal relationship. Instead, it allows there to be not only intralevel (horizontal) but also interlevel (vertical) dynamics, as processes on different levels in a mechanism can now be linked vertically through other causal processes. In order to distinguish horizontal from vertical causal relations (e.g., processes occurring between genes and those occurring between genes and the phenotype) and to locate processes at particular levels in a mechanism, the criterion of parthood can be replaced by a number of other criteria (see Wimsatt 1994; Eronen 2013; and DiFrisco 2017). One such criterion is the scale of time or rate at which processes occur. The rate of a process can refer to its frequency or the time the process takes to overcome perturbations (relaxation time). This idea of ordering nature vertically by means of the criterion of time scale also grew out of classical epigenetics. For example, Waddington (1957) developed a hierarchical model of time scales that includes both processes on lower levels of organization with a faster rate, such as processes at the molecular level, and processes on higher levels with a slower rate, such as evolutionary processes. Brian Goodwin (1963), a student of Waddington, used the rate criterion to demarcate the levels of genetic, epigenetic, and metabolic causal processes in cells, as well as to describe their dynamic interactions. He located different cellular variables at the same levels if they shared a similar relaxation time. As Goodwin notes, parameters and mean values in equations describing faster lower-level processes become variables in equations describing slower higher-level processes, and vice versa. This procedure of “smoothing” rate characteristics is supported by the idea that if two processes differ with respect to their rates, changes in these rates affect each other less. Thus, given higher rate differences, their effects on each other will be buffered more. Today, similar ideas seem to underlie contemporary epigenetic studies on how the rate with which genes are expressed in gene networks (i.e., their gene expression profiles) mechanistically affect the stability characteristics of developing phenotypes. Interestingly, dynamic

152 ◂ MECHANISTIC EXPLANATION models of these networks, which Bechtel and colleagues tried to conceptualize by means of the traditional parthood view of mechanisms (see, e.g., Bechtel 2011; and Bechtel and Abrahamsen 2010, 2013), in fact operate with variables such as rates of production, diffusion, and decay of protein products. This makes them good candidates for further investigations of the noncompositional, nonmachinelike view of biological mechanisms.26 These rather brief comments on the holistic and dynamic perspective taken by many epigeneticists, as well as on their alternative processual ways of conceptually grasping and reasoning about the target system, show two things. First, we come to realize that neither the decomposition-and- localization framework nor the idea of vertical fixity in mechanisms has to be accepted as the gold standard of conceptualizing biological mechanisms and their interlevel dependencies. Second, we understand what it could mean for epigeneticists to “look down” when they seek to mechanistically explain. As we learned earlier (see chap. 3), epigeneticists in fact look down, for example, to the gene just as geneticists do. Both acknowledge the heuristic value of methodological reductionism (Ayala 1989)—the strategy by means of which the behaviors of systems on a higher level can be studied in terms of the properties of objects located on a lower level—and apply it as a substantial part of their research program.27 This has led Alfred Tauber (2011) to claim that epigenetics is not that new after all, in contrast to standard molecular biology. However, it is a different kind of reductionism we find here, as many epigeneticists do not share the ontological commitments of mechanicism still guiding in large part mechanistic explanations in today’s molecular genetics. As a consequence, many systemic-thinking epigeneticists neither implement the concepts of decomposability and parthood nor a static view on interlevel relations as cornerstones in their research program (as one should expect according to the new mechanistic philosophy). In other words, epigenetics integrates reductionist methodologies into a less narrow mechanistic framework, one in which ontological machine views lose their explanatory autocracy and methodological guidance potential. More generally, this shows that methodological reductionism does not necessarily include reductive part-whole explanation. ▶ ▶ ▶

▸ 153 MECHANISTIC EXPLANATION

In this part of the chapter I have attempted to delineate the historical process of the “demachinization” of molecular mechanisms. Although the increasing interest in epigenetic molecular mechanisms has shaped the view on mechanisms in molecular biology in general, this conceptual shift is not driven exclusively by epigenetics. For example, parts of systems biology take a similarly non-atomistic view toward molecular mechanisms. In addition, the related idea that biological systems should not be investigated in isolation but as an intact whole is not new either (see, e.g., Levins and Lewontin 1985). Moreover, as emphasized persuasively by Alexander Powell and John Dupré (2009), molecular biologists have been gradually turning toward issues of complexity and systemic perspectives since the mid- twentieth century (see also chap. 1). It is important to mention that this long trend toward complexity and systemic views does not necessarily go hand in hand with the rejection of mechanicists’ views. For example, a substantial part of today’s work on gene regulatory networks (GRNs) fully endorses machine and engineering analogies. This perspective has been defended, especially by the late Eric Davidson, the leading GRNs researcher. He understood development as the realization of information stored as a “genomic regulatory code” (Peter and Davidson 2011, 635) in a “vast delocalized computer” (Davidson 2006, 188). Against this background, mathematically describing developmental pathways as a sequence of “on” or “off” switches of genes means understanding how the computer works. Due to the deterministic character of the computer program, for Davidson GRNs allow for very limited flexibility in embryonic development. This view is criticized by those systemic-minded researchers who emphasize that gene interaction in GRNs is often not deterministic but stochastic in nature and thus cannot easily be computed through a sequence of algorithmic functions. For example, it has been critically remarked that nongenetic heterogeneity in cell differentiation (like cell-to-cell variability of cell phenotypes) resulting from interactions of GRNs should no longer be described as “noise” that ought to be ignored (see Chang et al. 2008; Huang 2009a; and Pujadas and Feinberg 2012; see also chap. 2).28 Another trend in genomics that also counteracts demachinization is the development of the so-called CRISPR gene editing system (see Jinek et al. 2012).29 Supported by a growing media hype, this cheap and easy to cus- tomize tool allows for editing of the genome and even the epigenome with

154 ◂ MECHANISTIC EXPLANATION unprecedented precision. The general perspective on molecular causality taken by many scientists working on CRISPR is (still) one that resembles Davidson’s view. They believe that the changes made during “genome engineering” are discrete and local, like the ones made while changing a computer program. In other words, they believe that things are not as complex as expected. After all, it is only machines they are dealing with. As these lines show, the conceptual change of demachinization does not constitute a “revolutionary”—that is, a sharp, rather than gradual—development. Moreover, it does not separate distinctly epigenetic and orthodox molecular explanation in terms of mechanism, as it is neither an exclusive feature of epigenetics nor is it present in all epigenetic investigations. At the same time, however, the above discussions show that epigenetics offers prime examples of the modern systems view in molecular and cell biology, which allows biologists and especially new mechanists to become aware of erroneous and outdated ontological assumptions underlying their narrow views of mechanisms. In other words, becoming aware of epigeneticists’ ontology helps in overcoming the uncritically accepted metaphysics of mechanicism, as well as its associated constraints and biases, when reasoning about complex living systems. Let us now turn to another, less metaphysical feature of mechanistic explanation in epigenetics that also challenges the consensus view of new mechanists on what mechanistic explanation is.

ROBUSTNESS AND “MECHANISMS WITHOUT CAUSES”

Within the community of new mechanists we find an influential group of philosophers of science who claim that the lower-level interactions of a mechanism’s parts are potentially exploitable for the purposes of manipulation and control and that they can be described by means of the concepts of intervention, invariance, and difference making. Advocates of a theory of mechanisms that incorporates this Woodward-style (2002, 2011) interventionism include Glennan (2002, 2005), Steel (2008, chap. 3), Craver (2006, 2007, 93–106), Leuridan (2010), and Kuorikoski and Ylikoski (2013).30 Glennan, for example, claims that intralevel relations can be characterized by direct, invariant, change-relating generalizations (see his definition

▸ 155 MECHANISTIC EXPLANATION

above). Craver (2007, 94; emphasis added) describes the causal nature of the activities of mechanisms’ parts and the parts’ “horizontal” relationships, respectively, as follows: “The central idea is that causal relationships are distinctive in that they are potentially exploitable for the purposes of manipulation and control. More specifically, variable X is causally relevant to variable Y in conditions W if some ideal intervention on X in conditions W changes the value of Y (or the probability distribution over possible values of Y).” This means that the investigation of “horizontal,” lower-level causal dependencies between component parts of a mechanism is best performed by applying a manipulationist approach. However, despite its application on intralevel causal relations between variables located at one level of biological organization, Woodward’s basic framework is also often thought to be compatible with “vertical” or interlevel constitutive relations between variables located at different levels. Jaakko Kuorikoski (2012, 374) summarizes this view as follows: “a system-level property depends constitutively on the properties of its constituent parts in the sense that if the properties of the parts had been suitably different . . . , the system-level property would have been different as well (in some systematic way).” Mechanistic explanations can thus be described as tracing change- relating, interlevel constitutive dependencies between variables, whereas, for example, a change in the lower-level parts of a mechanism would change the higher-level system’s behavior. Nevertheless, in line with the metaphysical distinction between causal and constitutive dependencies and between causal and constitutive explanation drawn above, Craver and Bechtel (2007, 547; emphasis added) emphasize that there is no such thing as interlevel causation: “We argue that intelligible appeals to interlevel causes (top-down and bottom-up) can be understood, without remainder, as appeals to mechanistically mediated effects. Mechanistically mediated effects are hybrids of causal and constitutive relations, where the causal relations are exclusively intralevel.” Craver and Bechtel claim that no constitutive relation between levels of organization can also be a causal relation, since the former, in contrast to the latter, is synchronous, symmetric, and its relata “part” and “whole” are not logically independent, in contrast to a Humean understanding of the causal relata “cause” and “effect.” By means of this distinction, Craver defines the bottom-up and top-downmutual manipulability of constitutive relata, a component part’s activity Φ and a mechanism’s or system’s general behavior Ψ, in two principles of constitutive relevance (CR). (CR1)

156 ◂ MECHANISTIC EXPLANATION describes bottom-up constitutive relevance, while (CR2) describes top- down constitutive relevance:

(CR1) When Φ is set to the value Φ1 in an ideal intervention, then Ψ takes

on the value f(Φ1). (Craver 2007, 155)

(CR2) If Ψ is set to the value Ψ1 in an ideal intervention, then Φ takes on 31 the value f(Ψ1). (Craver 2007, 159)

Craver’s (and Bechtel’s) Woodwardian mutual manipulability account of constitutive relevance has been targeted by several critics. For example, it has been argued that, according to this mechanistic framework, constitutive relevance implies causal relevance despite the claim that they are strictly distinct (Leuridan 2012), that this account unintentionally entails the nonexistence of downward causation (Fazekas and Kertész 2011), and that the distinction between constitutive and causal dependency relations cannot easily be drawn in developmental biology (Mc Manus 2012; Ylikoski 2013). As an expansion of this critique, by drawing on examples of epigenetic mechanisms I will now argue that Craver’s understanding of constitutive relations does not match those explanatorily relevant dependencies bringing about the phenomenon of robustness in complex evolved systems, such as in epigenetic systems. More generally, this problem casts doubt on whether Woodward’s account correctly describes all biological mechanisms. Craver’s (2007) idea that an ideal intervention that changes a lower-level mechanism’s part constitutively changes the overall system S’s behavior (i.e., S’s Ψ-ing) is expressed in his principle (CR1). (CR1) has been supplemented by the principle (CR1a), holding “that the intervention, I, leaves all of the other dependency relations in S’s Ψ-ing unchanged” (2007, 156–57; original emphasis). This supplement should exclude problems such as recovery and reorganization, also known as robustness, which are compensatory effects that prevent changes to S’s Ψ-ing.32 However, robustness is a common and important phenomenon in biological investigations of complex living systems, such as regulation in genomes, cells, or organisms. This amounts to the question of whether (CR1a) allows for suitably characterizing robust mechanisms.33 According to Bert Leuridan (2012, 405–6), (CR1a) simply is too strong. Robustness of a system S’s behavior means that some, or perhaps most, possible effects of interlevel interventions are prevented or buffered. Such buffering through phenomena like reorganization may occur immediately or

▸ 157 MECHANISTIC EXPLANATION

may be delayed, as in the case of drug tolerance. Accordingly, this means that in some robust systems we might indeed be able to perform an ideal intervention that avoids the buffering of that system’s behavior. Craver’s principle (CR1a), however, metaphysically demands that an ideal interlevel intervention is possible in all, not just some, robust systems. This leads to two options: we could drop (CR1a) or try to weaken the principle. Since we do want the elaborated manipulability account of constitutive relevance to be able to deal with robustness and thus to be of value for biological explanations, we should avoid dropping (CR1a) prematurely. Thus, we can weaken the principle and turn it into the following:

(CR1a*) In some of the systems S1, . . . , Sn, there should be a system S1 in which an ideal interlevel intervention, I, on the component

x’s Φ-ing changes the value of S1’s Ψ-ing under the condition

that I leaves all of the other dependency relations in S1’s Ψ‑ing unchanged.

Unfortunately, even this weaker characterization of constitutive dependencies does not help in grasping what makes a description of a mechanism in

the case of S1 explain S1’s robustness. To make this clear, let us take a look at robustness analysis in epigenetics—to be exact, in stem cell epigenetics. While epigenetics is, of course, not the only field in which the robustness of biological systems is investigated, it is ideal for philosophical investigations of robustness, as its long history is closely related to this issue.34 In fact, besides epigeneticists’ developing interest in inheritance phenomena, the ideas of stability or robustness and flexibility have been a cornerstone of epigenetic investigations ever since Waddington linked the field of developmental genetics to his concept of canalization. His research was in large part dedicated to the question of why development is not affected by certain genetic or environmental perturbations. Waddington was convinced that this property allows developing organisms to buffer perturbations up to a certain level through the buildup of hidden genetic variability that is not expressed phenotypically. Among others, Waddington’s student Brian Goodwin (1963) further developed this idea of how different levels of organization are buffered against each other (see above). Until now, this research focus had not changed much. For example, in a review in Cell, Aaron Goldberg and colleagues (2007, 637; emphasis added)

158 ◂ MECHANISTIC EXPLANATION defined one of the most central questions in epigenetic investigations as follows: “What mechanisms enable epigenetic stability in a defined cellular lineage while allowing epigenetic flexibility during cellular differentiation and development?” In fact, this question about robust mechanisms has become the center of attention for those epigeneticists working in the fields of cell differentiation and reprogramming. While the former describes the process by which a pluripotent stem cell develops into a mature, fully differentiated cell, for example, a muscle cell, during embryogenesis, the latter character- izes the experimentally induced process by which a mature unipotent cell is rejuvenated into a stem cell (the so-called induced pluripotent stem cells, iPS) or dedifferentiated into another cell type. Cell reprogramming basically is conducted by forcing the expression of a number of epigenetic regulatory factors, which makes some call this procedure of induced pluripotency “epigenetic reprogramming.” To understand what is so special about these robustness mechanisms and their role in the explanation of evolved complex systems in epigenetics, we need to recall the tenets of the manipulationist account of mechanisms and mechanistic explanation.35 To wit, if we want to mechanistically explain a phenomenon, according to Craver and other new mechanists, we have to trace change-relating, constitutive dependencies between components with causal capacities at different levels of biological organization. In these change-relating dependencies, according to (CR1), (CR1a), and (CR1a*), respectively, a change in the lower-level constitutive parts of a mechanism would change the higher-level system’s behavior. Thus, by analogy to the idea that the explanans of causal explanations is a difference-maker, Craver (2007) in particular conceives of component parts of a mechanism as interlevel difference-makers in constitutive explanation. A number of stem cell epigeneticists seek to understand the nature of robustness in development mechanistically, by trying to answer this question: Under which set of constitutive factors does the biological system under study display stability or robustness? In other words, they investigate why we sometimes observe maintenance of specific functionalities of living systems despite perturbation. Thus, this explanandum resembles precisely the problem of failed interlevel difference-making in some robust systems, as described in (CR1a*). For these epigeneticists, however, addressing this very explanandum means listing in an explanans the explanatorily relevant lower-level parts that do not make a difference or have a very low constitu-

▸ 159 MECHANISTIC EX PLANATION

tive relevance. Such an explanans is considered to address an explanandum only on the supposition that the former also cites those component parts and their activities that, if manipulated, do not bring about a change in the latter. As claimed by Fridolin Gross (2015), what is explanatorily relevant in these cases of robustness are non-change-relating relationships.36 To clarify this claim, let us briefl y consider a common type of investigation of dynamic stability in stem cell epigenetics: Sudin Bhattacharya and colleagues (2011) seek to understand how gene regulatory networks determine the stable or robust diff erentiation of pluripotent stem cells into unipotent cells during embryonic development. Th ey compute the dynamics of a bistable two-gene regulatory network—a network of the two genes x and y with two attractors,

FIG. 4.2. Computed attractor surface of a two-gene circuit’s network dynamics. Mul- tiple trajectories converge to the two attractors A and B (arrows). A quasi-potential surface (i.e., the elevation represents a path-integral quasi potential) is derived directly from deterministic rate equations describing the dynamic behavior of the two genes (Bhattacharya et al. 2011, e4, slightly modifi ed; reproduced with permission from Bhattacharya et al).

160 ◂ MECHANISTIC EXPLANATION

A and B—with mutual inhibition of the genes (fig. 4.2). In Craver’s terminology, a system S’s Ψ-ing (a cell differentiating to A or B) is modeled by describing a mechanism that contains two components x and y (genes) with a mutual causal relationship, like x’s Φ1-ing and y’s Φ2-ing. The biologists use stochastic simulations to show that the elevation of the computed landscape and its topography, respectively, correlate to stable steady states or attractors of the network and thus to the likelihood of occurrence of particular cell fates. In this model, the network dynamics of the gene regulatory network bring about two stable pathways of development leading to the two attractors A and B, which may be formally described. What is considered to be an explanatorily relevant constitutive dependency within this mathematical model? The answer is: primarily those factors on which manipulation or perturbation—a change in the gene expression profile in one of the genes x or y at a certain point in time—does not change the fate of a differentiating cell. Manipulations on the gene network are depicted in figure 4.2 as changes in the starting points of trajectories. While very few of these changes or perturbations make a difference in the explanandum phenomenon (i.e., they lead to a switch in the “direction” of a developmental pathway, for example, from attractor/cell type B to A), most changes do not. Here, the basic idea is that maybe all possible manipulations performed on one component part (for example, on gene x) alone are not able to make a difference in the overall system (for example, a change in the end point of a pathway from one cell type to another). To understand a complex system’s dynamic stability, tracing such non-change-relating dependencies of constitutive parts that do not bring about a change in the whole system is considered to be highly explanatorily relevant. In order to specify which explanandum phenomenon non-difference- makers are thought to address in this case, we have to replace the toy model presented earlier with a more realistic scenario. Imagine the mechanism of a gene regulatory network containing a thousand genes, which determine a bistable cell differentiation process. nI Craver’s terminology, the system S’s Ψ-ing is determined by a mechanism containing the lower-level component parts x1, . . . , x1000 and their causal capacities Φ1, . . . , Φ1000. Now add to this picture another one million potentially perturbing environmental or epigenetic regulatory factors, for example. Additionally, imagine that in some states at a particular point in time during cellular differentiation it is easy to switch a system from one cell type to another (low robustness), whereas

▸ 161 MECHANISTIC EXPLANATION

in other situations it takes strong multifactorial perturbation to change the system’s state (high robustness). Epigeneticists working on robustness claim that, in order to grasp the stochastic nature of a dynamic stability phenomenon, one has to “embrace” the complexity of its underlying network. Mechanistic models that consider only a small set of strong difference-makers—for example, a few genes and some additional factors influencing the mapping between genes and the system’s behavior—often struggle to explain why the system’s stability changes so radically over time under apparently quite similar circumstances. Thus, in order to solve this problem, additional non-change-relating relationships and weak difference-makers are included in the model.37 The strategy described here is to consider those small-effect factors that may merely lead to changes on some intermediate levels of organization. Although they do not (or only very weakly) influence the overall behavior of the system under study, they may mediate other, stronger, and (more) direct causal dependencies, leading, in total, to phenomena like recovery and reorganization. Changing these weak mediating component parts’ activities in isolation one by one (often) does not make a difference to the overall dynamic behavior of the system. However, they are crucial for understanding the system’s dynamic behavior. This systemic perspective on causes may be visualized using the metaphor of an orchestra being able to play an overture only if all of the musicians contribute, no matter how small their individual parts are or whether the sound of the individual instruments can be discerned by the amateur ear. This result does not mean that these component parts’ activities are not able to make any difference at all. For example, changes in the expression profile in one of the one thousand genes of the gene regulatory network may lead to a change on the protein level, although not to a change in the dynamics of cell differentiation at the system level. In other words, with regard to the level of the explanandum phenomenon, these factors have to be considered as explanatorily relevant non-difference-makers. How can we characterize this relevance more precisely? We may say that non-difference-making offers explanatorily relevant information about how the system under study internally secures dynamic stability through nonmodularity of some of its underlying processes. It helps biologists to gain non-interventionist information about which causal context suppresses modularity, buffers potential changes, and thus maintains the behavior of the system overall.38

162 ◂ MECHANISTIC EXPLANATION

As the above explanation of robust complex living systems shows, some scientific investigations in epigenetics need to trace not only information about difference-making but also constitutive dependency relations containing non-difference-makers. However, this information about which components under which circumstances do not make a difference does not fit with the manipulationist view on mechanisms and constitution. In other words, rather than focusing on how to circumvent robustness phenomena by trying to establish change-relating constitutive dependencies in every robust system (see CR1a) or at least in some robust systems (see CR1a*), we have to address the deeper problem of failed interlevel changes. This, however, is not possible from within the manipulationist framework. As epigenetics teaches us, the interventionist criteria for explanatorily relevant information to be considered in biological mechanistic models must be expanded. Restricting the concept of mechanism to component parts that constitutively make a difference is insufficient to address robustness in complex evolved systems. Based on her investigation of how systems remain robust in knockout experiments, Sandra Mitchell (2009, chap. 4) comes to a similar, but slightly different conclusion.39 She argues that understanding how robustness is brought about necessitates focusing on the nonseparability and nonmodularity of causal factors. Unlike Woodward, she holds that in complex robust systems a single putative causal factor can carry explanatory weight, not in isolation but only in the system. Components thus should not be studied in isolation but through multilevel, multicomponent integrated approaches. While this is certainly true, there is another conceptual tradition that makes grasping the mechanisms underlying robustness difficult. That tradition is not necessarily linked to the idea that causal factors are separable and modular, respectively. It is the classical distinction, picked up by Woodward, between noncausal coexistence laws—that is, non-change-relating regularities—and causal succession laws—that is, change-relating regularities. Let me describe this issue in more detail. Within the interventionist framework, only change-relating regularities act as proper scientific explanations. They allow for the possibility of manipulation. According to Woodward, a generalization is non-change-relating and thus fails to be explanatory if its terms suffer, for example, from being too imprecise to be represented as variables with well defined and manip- ulable values. However, some generalizations do not purport to represent a

▸ 163 MECHANISTIC EXPLANATION

change-relating relation in the first place but instead express a fact about the

constitution of the system, like the law of radioactive decay, “water is H2O,” and many claims in taxonomy. In traditional terms such non-change-relating generalizations were labeled “coexistence laws,” in contrast to change-relating ones or so-called “succession laws.” For example, following a Humean view of causality, John Stuart Mill (1843, bk. 3, chap. 22) and Carl Gustav Hempel (1965, 352) stated that causal laws should be conceived only as uniformities of succession that describe regular ways in which events or states of a system follow one another in time.40 These laws necessarily include change. In contrast, coexistence laws offer noncausal information about how the system is constituted. In particular, they show how two or more magnitudes are related to one another in a system under study and thus which regions of an n-dimensional state space the system is able to occupy at a time. These laws usually do not include change. An example of such a coexistence law is the competitive exclusion principle in ecology (see Raerinne and Baedke 2015). It states that n number of ecologically similar species with lower than n number of lim- iting resources cannot coexist in a stable manner. In other words, there is a state of the system (a similarity/resource scenario) that cannot be maintained with another state of the same system (coexistence) without a change in the system.41 The usefulness of this conceptual parting and the idea that causality has little to do with coexistence laws has been defended by a number of authors. Not least through Woodward’s influential account, laws of succession and change-relating regularities have become, respectively, the prototype of causal explanation and of scientific explanation more generally. How- ever, the above analysis of mechanistic explanations of robustness suggests that we should be skeptical about this conceptual tradition, as robustness explanations closely link non-change-relatedness to explanatory relevance. As in coexistence laws, such non-change-relating regularities offer information on what regions of an n-dimensional state space a particular system is able to occupy without undergoing change (e.g., in the case above, without switching from cell type A to B). Robustness explanation considers this information causally and explanatorily relevant, because one cannot causally understand a change in the system—that is, when robustness breaks down—without understanding which mechanisms and causal factors suppress this change and thus enable stasis. For the latter, we need

164 ◂ MECHANISTIC EXPLANATION non-change-relating regularities and non-difference-making factors as part of our explanantia. Therefore, I suggest that we should overcome the traditional conceptual distinction between change/non-change-relatedness, as it constrains our understanding of robustness. Non-change-relating regularities also carry valuable information for mechanistic explanation. In other words, causes can also work as such in non-change-relating relationships. We gain the causal information relevant for understanding how (dynamic) stability is secured only if we step out of this influential but narrow conceptual framework. This result casts doubt on the applicability of Woodward’s account for describing biological mechanisms, since the explanatory dependencies between the parts and the overall system in which, for example, many epigeneticists are interested, are not covered by an account focusing solely on change-relating relationships. In addition, the overemphasis on change- relatedness and difference-making violates our commonplace intuition that when dealing with mosaiclike, multifactorial phenomena one should not necessarily expect to change the phenomenon by changing a single constitutive part. Removing a blade of grass from a stack of hay does not make us expect the stack to change its overall shape. Nevertheless, the absence of a change in the shape tells us something about the inner structure of the stack of hay. It offers information about the causal properties of contextual blades of grass suppressing the change and thus maintains the stack’s shape. Henceforth, I refer to those mechanisms relevant in biological explanation that entail constitutive dependencies that cannot be described adequately in terms of the new mechanists’ philosophical orthodoxy as mechanisms without causes. This means that although biological mechanisms can still be considered causal mechanisms, they cannot be understood entirely by “close analogy” (Craver 2007, 152) to what is usually considered to be a causal dependency relation: a change-relating, difference-making relationship. ▶ ▶ ▶

It has been argued that epigenetics plays an important role in overcoming the mechanicists’ outdated ontological assumptions that underlie not only classical molecular biologists’ views of what a mechanism and mechanis-

▸ 165 MECHANISTIC EXPLANATION

tic explanation are but also the views of many new mechanists. In short, mechanisms become less machinelike, less atomistic, and more dynamic. In addition, some epigenetic mechanisms—so-called robust mechanisms— appear to be incompatible with the interventionist concept of causation advocated by many new mechanists. These cases are labeled mechanisms without causes. ▶ ▶ ▶

The analysis of the causal and mechanistic structure of epigenetic explanations in this and the previous chapter elucidates how epigenetics, besides related fields such as systems biology, is involved in establishing a new explanatory agenda in molecular and cell biology. First and foremost, a number of epigeneticists distinguish their approach from prevailing standards of explanation and philosophical conceptualizations of scientific explanation by tracing causes without mechanisms and mechanisms without causes. Both of these strategies seek to better address epigenetic complexity. The former strategy of causes without mechanisms enables one to account for the causal autonomy of less fundamental molecular factors. Speaking more generally, the idea of lowering the amount of underlying mechanistic detail in explanantia addressing epigenetic complexity can be seen as a reaction to the general “representation problem” in complexity sciences (see chap. 1). This problem refers to the fact that even with complete knowledge and full consideration of all components and interrelations of a complex system, it is not possible to offer a fixed, complete, and formal description. The latter strategy, of mechanisms without causes, helps us to better grasp interlevel relations underlying the dynamics of complex phenomena. Again, speaking more generally, the underlying idea of this strategy—to be more liberal about some of the properties of a mechanism listed in an explanans—can be seen as a moderate solution to the so-called “Borges problem” (see chap. 1), which is the attempt to investigate complex wholes as wholes. These two strategies, causes without mechanisms and mechanisms without causes, are summarized in table 4.1.

166 ◂ MECHANISTIC EXPLANATION

TABLE 4.1. TWO STRATEGIES OF EPIGENETICISTS ACCOUNTING FOR THE COMPLEXITY OF EPIGENETIC SYSTEMS CAUSAL MECHANISTIC STRATEGY EXPLANATION EXPLANATION

Nonmechanistic causal explanation • higher-level explanation Causes X • identification of relevant without (interventionist) factors (difference-makers) mechanisms for phenomena with less consideration of underlying mechanistic detail

Mechanistic explanation • interlevel explanation Mechanisms X • noncausal mechanisms: without (partially non- some relations are non- causes interventionist) interventionist ones (with non-difference- making parts)

▸ 167 FIVE ASSESSING THE EXPLANATORY POWER OF EPIGENETICS

Science is built up with facts, as a house is with stones. But a collection of facts is no more a science than a heap of stones is a house. HENRI POINCARÉ, 1902

A scientist . . . who wants to understand as many aspects of his theory as possible, will adopt a pluralist methodology, he will compare theories with other theories rather than with “experience,” “data,” or “facts,” and he will try to improve rather than discard the views that appear to lose in the competition. PAUL FEYERABEND, 1975

IN THE PREVIOUS CHAPTERS a number of central features of epigenetic explanation have been described. I showed how some of them differ from still widely accepted explanatory standards of traditional molecular biology and evolutionary biology. For example, in evolutionary biology, epigenetic explanations depart from orthodox explanation by emphasizing proximate causes rather than ultimate causes (see chap. 2). In molecular ASSESSING THE EXPLANATORY POWER OF EPIGENETICS biology, a number of epigeneticists focus on “nonmechanistic” causal explanations. Such explanations provide less information about mechanisms that underlie causal dependencies (see chap. 3). There are two issues that have to be distinguished here: first, determining what it means for these epigenetic explanations to count as explanatory at all, and, second, addressing how good the explanations offered are. The former issue has already been addressed by invoking criteria like the invariance or heuristic value of explanations. Now I turn to the latter. This issue concerns the concept of explanatory power or value, not explanatoriness (or appropriateness) of explanation per se. For example, given the previous analysis here, various questions arise, such as “Are epigeneticists’ explanations better?” and “How can we evaluate the explanatory power of epigenetic explanations in contrast to standard molecular explanations entailing (more) mechanistic detail and prevailing evolutionary (i.e., ultimate cause) explanations?” This chapter answers these questions first by presenting a general contrastive framework suitable for evaluating the value of scientific explanations and second by applying this framework to the issue at hand. I show that the contrastive approach adopted here is able to give precise guidance about why and when epigenetic explanations are legitimately chosen (i.e., have more explanatory power) over other prevailing molecular and evolutionary explanations. Thus, it represents a crucial tool for an extended evolutionary synthesis seeking to integrate epigenetic explanations into a field of other explanatory accounts.

WHAT EXPLANATORY POWER IS AND IS NOT

Nowadays epigenetic explanations of development, inheritance, and evolution seem to be ubiquitous in modern biology. However, biologists have not reached any consensus on whether they should prefer these rather new explanations over other, more orthodox approaches addressing the same explanandum phenomena. This problem is one related to the concept of the value or power of explanations. This issue should not be confused with the so-called demarcation problem. The latter refers to the issue of how to distinguish scientific explanation (or scientific theory) from non- or pseudoscientific explanation (see Pigliucci and Boudry 2013). Popper (1935) famously argued that only scien-

▸ 169 ASSESSING THE EXPLANATORY POWER OF EPIGENETICS

tific explanation can make testable and falsifiable predictions.1 Others have claimed that the status of a scientific theory is (or should) depend on the amount of empirical evidential support or the formal structure of evidential confirmation (see, e.g., Carnap 1928) and that it is, at least to a certain degree, consistent in certain cases with the preexisting background knowledge and experimental results in a particular field.2 Following the idea of “Occam’s razor,” one may even argue that scientists (should) seek the most parsimonious or economical explanation (Sober 1975).3 These criteria of predictive capability, evidential support or structure, coherency or compatibility, and parsimony or simplicity should first and foremost be understood as referring to the scientific character of explanation. In other words, they constitute the (logical) structure of scientific explanation or, to put it in still other terms, the explanatoriness of science. However, the distinction between this concept of explanatoriness and that of explanatory power is not a clear-cut one. One might consider there to be a particular threshold above which explanations satisfying these criteria can legitimately be labeled “scientific.” Accordingly, those statements passing the threshold might be called “better” if they differ qualitatively or quantitatively in one or more of the above criteria. This problem of grasping the core of the concept of explanatory power and its distinctness from (simply) being an explanation becomes even more difficult, as the value of an explanation is sometimes also described as its “explanatory depth” (Hitchcock and Woodward 2003; Weslake 2010).4 Below I defend a concept of explanatory power, which moves away from the above criteria by treating these criteria as background conditions for there being a scientific explanation. Then, following Woodward (2003, 257–65), Hitchcock and Woodward (2003), and especially Ylikoski and Kuorikoski (2010), I present an interventionist framework in which the value of explanations can be evaluated along various parameters, like precision, robustness, or cognitive salience. The basic idea is to assess how these criteria affect the (degree of) invariance of an explanation. In biology and philosophy of biology, issues concerning explanatory power have recently been discussed with respect to evolutionary explanation. In particular, the debate has centered on how much—in the sense of what kind of facts—evolutionary explanations referring to natural selection can explain. For example, while it is widely accepted that they can address explananda like dynamics of trait frequencies and organisms’ survival (see

170 ◂ ASSESSING THE EXPLANATORY POWER OF EPIGENETICS

Sober 1984), it is still unsettled as to whether natural selection can also explain specific traits on the level of the individual.5 This debate is somewhat related to the issue at stake here, as epigeneticists ask other biologists, like more orthodox-minded evolutionary biologists, to accept their statements not only as valid scientific explanations—their statements’ explanatoriness or appropriateness—but even as superior (and complementary) alternatives. This is a claim about the power or value of epigenetic explanation. What is more, issues of explanatory power become eminently important in pluralist theoretical approaches like an extended evolutionary synthesis, which claim to interlink novel epigenetic explanations with various other available explanations. In order to develop, in this theoretical context, reliable criteria with which we can evaluate why and when we should choose epigenetic explanations over competing alternatives, we have to be as precise as possible about the concept of explanatory power being applied. Therefore, let us first demarcate this concept from several other related yet distinct concepts that have been discussed in philosophy of biology with respect to epigenetics. Many of them resemble the above traditional criteria of what it means for a scientific explanation to be explanatory. These include likeliness (or evidential power), causal power (or importance), attractiveness, and riskiness. Likeliness. The issue of assessing the value of epigenetic explanations can easily be mixed up with the evaluation of evidential power or likeliness of explanations. For example, some authors have advocated revising the prevailing explanatory focus of evolutionary biology because of the high and ever-increasing evidential support of evolutionarily relevant epigenetic explanations of developmental phenomena (see Jablonka and Lamb 2005, 2010; and Jablonka and Raz 2009).6 The empirical question of whether one of two competing explanations is more likely (i.e., supported by more evidence) than the other differs, however, from the non-empirical, philosophical analysis of the value of explanations addressed here. What matters with regard to the issue at stake is to determine which explanations in biology should be favored once we have achieved a status of evidential parity between epigenetic and orthodox explanations. Thus, likeliness may be understood only as a criterion for securing explanatoriness in scientific explanation. In other words, since evidential support often leads to standoffs between two alternative explanations, there must be another way of distinguishing the value of explanations.7

▸ 171 ASSESSING THE EXPLANATORY POWER OF EPIGENETICS

Causal power or importance. Some philosophers of biology have recently conflated the assessment of explanatory power with the question of whether a biological explanation picks out causes that are (by any means) more important than others. This ontological idea that some explanations cite causes with more causal power rests on the dubious assumption that all causes relevant for a phenomenon under study share objectively measurable or comparable properties.8 This idea is central to so-called “causal democracy” or “causal parity” claims in developmental systems theory (see Oyama 2000; Moss 2001; and Oyama et al. 2001).9 They question the idea of privi- leging some causes, notably genes, in development and evolution over others on a priori grounds. Instead, they claim that biological research should be guided by the strategy of “parity of reasoning.” This means that every causal (epigenetic, environmental, genetic, etc.) factor must be approached as if it might be causally important in development. In other words, developmental systems theorists argue for a compensatory modeling assumption that should lay the foundation for explanations citing as explanans variables only those causes with objectively measurable high causal power. Although I support the general criticism of developmental systems theorists against gene-centric methodologies, these authors do not explain why picking out the most important causes of a phenomenon automatically increases the value of explanations citing only those causes. They simply take for granted that this is the case. It seems dubious, however, to assume that two explanations containing different sets of causes with equal causal importance with regard to a common explanandum phenomenon are always equally good.10 Attractiveness. David Haig (2011, e204) claims that choosing orthodox or progressive explanatory agendas in evolutionary biology, which he calls neo-Darwinian and neo-Lamarckian explanations, respectively, is, from the perspective of epigeneticists, mainly a personal preference, and also reflects “emotional and aesthetic reasons.” In particular, Haig claims that distinct anthropological outlooks underlie the two alternative explanatory agendas in modern biology:

It seems to me that arguments between supporters and critics of the Mod- ern Synthesis often are based on differences of preference and thinking style rather than matters of substance. . . . I will give my subjective impressions of the reasons why many people I have talked with, both in the general public

172 ◂ ASSESSING THE EXPLANATORY POWER OF EPIGENETICS

and the scientific community, have a visceral attraction to Lamarckism and a visceral dislike of Darwinism. . . . A neo-Darwinian view, with its emphasis on chance and randomness in the origin of variation, is perceived as positing a world without meaning that is less attractive than a Lamarckian view in which organisms have agency in shaping their evolutionary destiny. Natural selection, with its reliance on differential survival and reproductive competition, is also perceived as bleak and harsh. . . . It is a view from which many recoil. Phenotypic plasticity and the inheritance of acquired characters seem to hold hope that we all can improve without processes of selection. (2011, e204; emphasis added)

Haig thinks that epigeneticists turn away from the “matters of substance” and reject the possibility of assessing the evidential power of the two evolutionary explanations discussed. Thus, the procedure of selecting explanations turns into a highly irrational process of comparison in which explanations are chosen regardless of whether they are supported by strong evidence. According to Haig, for epigeneticists the two alternative explanations are no longer more or less good but rather more or less attractive. Accordingly, if this assumption is correct, the current theoretical development in modern biology must be understood as a highly irrational “anything goes” process, as Feyerabend (1975) would have called it. In contrast, I show that indeed the value of different orthodox and epigenetic explanations can be evaluated and compared rationally. Riskiness. James Griesemer (2011a, 23–24; 2011c) suggests that whether or not epigenetic explanations are accepted and research investments in epigenetics are made can be measured, among other criteria, in terms of how risky it would be for a biologist accepting this particular novel explanation in a certain theoretical and/or disciplinary context. He claims, for instance, that for molecular biologists explanations citing epigenetic mechanisms of DNA regulation, like DNA methylation, are less risky and can thus be adopted more easily, since these new mechanisms are not considered, qualitatively speaking, as being novel compared to previously known regulatory mechanisms. In other words, adding these epigenetic mechanisms merely means expanding the list of known regulatory molecular mechanisms. In contrast, evolutionary biologists interested in heredity are usually cautious about whether they should accept explanations of nongenetic inheritance citing the very same epigenetic mechanisms. This is the case, according to

▸ 173 ASSESSING THE EXPLANATORY POWER OF EPIGENETICS

Griesemer, because these mechanisms (involved in both gene regulation and heredity) might introduce qualitatively novel population dynamics into the field of evolutionary biology—dynamics that cannot be modeled by population geneticists by simply adding a new parameter of epigenetic inheritance. This approach seems to give an intuitively plausible answer to why rather few evolutionary biologists are willing to fully embrace epigenetics. This result, however, must be assessed in light of the traditional assumption that it is, first and foremost, the riskiness of hypotheses and models that, if con- firmed, advance science. As Popper (1963, 36) has noted, “Confirmations should count only if they are the result of risky predictions. . . . A theory which is not refutable by any conceivable event is non-scientific.” Accord- ing to this thesis, (falsifiable) epigenetic explanations in evolutionary biology—a field with qualitatively distinct background knowledge—rather than in molecular biology would fit the essence of a real scientific explanation.11 There is an important point in which my view differs from that of Griese- mer. His idea of evaluating risk is directed toward future problems, possibilities, and achievements created by novel explanations. To be more precise, what is evaluated is the riskiness of potential future adoptions of epigenetic explanations in light of the current explanatory focus and/or strategies in a particular field. With respect to the issue of theoretical integration and epigenetics, scientists do not primarily seek to understand potential future development but the current challenges of epigenetics. Therefore, I defend a “here and now” approach to explanatory power that assesses the current value of an explanation entirely in light of the current background knowledge and standards of explanation. Focusing on these present achievements of explanations means not asking questions such as “What might be the future impact of epigenetic explanations on evolutionary biology or their future achievements with regard to the modern synthesis?” but rather “How good are contemporary epigenetic explanations compared to other (more or less competing) ‘non-epigenetic’ explanations?” Let me develop these ideas in more detail. The notion of explanatory power or value is distinct from the concepts of likeliness, causal power, attractiveness, and riskiness discussed above. It can be specified by adopting an interventionist framework of contrastive explanation (Ylikoski and Kuorikoski 2010).12 Based on this counterfactual framework, explanatory power can be evaluated by comparing the ranges of inferences to potentially

174 ◂ ASSESSING THE EXPLANATORY POWER OF EPIGENETICS new counterfactual situations and thus the number and (theoretical and/ or practical) importance of answers to Woodwardian w-questions (what-if- things-had-been-different questions) that rival explanations make possible. For example, if an explanation of a particular population dynamic in mice that considers epigenetic factors as explanans variables enables biologists to answer a greater number of questions, as well as more novel questions on what would happen to (single mice in) this population if (part of) it was changed than would a second alternative explanation not considering epigenetic factors, the former explanation carries more explanatorily relevant information about the relation between possible values of explanans variables and explanandum variables. Thus, it is a better explanation. In other words, a good explanation is able to provide information on patterns of counterfactual dependency and difference-making of the form “Given two variables X and Y, X explains why Y if Y depends on X in the sense that if X had not happened, then Y would not have happened either, or: if X would have been different, Y would have been different as well.” Within this framework of counterfactual dependence and difference- making, understanding of an explanandum phenomenon comes in degrees. It can be measured by comparing the range of inferences to (potentially new) counterfactual situations (i.e., answers to “what-if” questions) made possible by alternative explanations. Take, again, the explanation of a transmission pattern of a phenotypic trait in mice that cites epigenetic factors as difference-making explanans variables. If this explanation enables scientists to infer more, and if, compared to the present background knowledge in the field, it offers novel answers on what would happen to the trait and/or the pattern of transmission if one (or more) epigenetic factor(s) were to be tweaked, this explanation improves the understanding of trait transmission in mice compared to an explanation omitting these epigenetic factors. In addition, explanatory power can be more easily assessed if we try to be as precise as possible about both explanantia and explananda as well as their explanatory relationship. Therefore, we can adopt an account of explanation in which both explanandum and explanans are contrastive (see Woodward 2003; Schaffer 2005; Ylikoski 2007; Ylikoski and Kuorikoski 2010; and Raer- inne and Baedke 2015). According to this account of double contrastiveness, the claim “x rather than x* explains y rather than y* in context U” can be represented as “x [x*] explains y [y*] in context U,” where x and y represent the explanans and explanandum, respectively, and x* and y* represent

▸ 175 ASSESSING THE EXPLANATORY POWER OF EPIGENETICS

their contrast classes. A contrast class can, in principle, contain an indefi- nite number of variables, such as x*, x**, x***, and so forth. The contextU describes the system’s other properties and environment. To underline the usefulness of this contrastive focus, let us consider a well-known fictional example. In Douglas Adams’s 1978 book The Hitchhik- er’s Guide to the Galaxy, the supercomputer Deep Thought was constructed to answer the “ultimate question of life, the universe, and everything.” It takes the computer 7.5 million years to come up with the highly unsatisfying answer “42.” After all this time, it turns out that the ultimate question originally posed—the “ultimate explanandum”—has to be specified in order to compute a more precise and more satisfying answer. In order to avoid such a fatal mistake, we should require an explanation to provide as much information about the contrast class of the explanandum as possible. With regard to epigenetics, this strategy makes new epigenetic explanations more precise in terms of specifying what exactly their explananda (and consequen- tially their explanantia) are about and what they are not about in contrast to competing prevailing explanations. Consider, for instance, a biological explanation taking these two forms:

(E) A set of causal factors and mechanisms x explains the trait y of an organism o, given a set of other properties of o and o’s habitat z, called U. (E*) A set of causal factors and mechanisms x [set x*] explains the trait y [y*] of an organism o given a set of other properties of o and o’s habitat z, called U.13

The explanation of the form (E) entails less explanatorily relevant information about the dependency relation between x and y than explanation (E*). Because of the contrastive nature of (E*)’s explanans and explanandum, biologists are able to make more and (possibly) more important inferences about counterfactual situations and thus to answer new w-questions, like “What would happen to y or y* if we change x in U?” This turns (E*) into the explanation with more explanatory power that enables scientists to better understand the explanandum phenomenon y. In other words, this strategy seeks to increase the number and importance of counterfactual inferences made possible by an explanation by sharpen- ing the explanandum phenomenon and thus asking for more mechanis-

176 ◂ ASSESSING THE EXPLANATORY POWER OF EPIGENETICS tic detail to be entailed in the explanans in order to address the refined explanandum. Let us call this strategy the Deep Thought strategy. It is widespread in biology. At the same time, however, it is not true that supplying mechanistic details alone produces better explanations. For example, Wesley Salmon (1984, 1989) has emphasized that the addition of irrelevant information to the explanans is fatal to the power of an explanation. In a similar manner, Marco Nathan (2012) has noted with respect to explanation in molecular biology that increasing the number of (lower-level) causes and mechanisms included in an explanans is often not only irrelevant but detrimental for the value of biological explanations. Neither author, however, presents an account of explanatory power that tells us in detail why this is so, beyond referring to the criteria of “relevance” or “necessity,” as well as in which explanatory context adding mechanistic detail is either a virtue or fatal. In contrast, the account of explanatory power by Ylikoski and Kuorikoski (2010) does. It argues that precision is not the sole factor determining the value of an explanation. Instead, explanatory power can be evaluated on various different dimensions that show important trade-offs between each other. For example, the virtue of precision and thus the usefulness of the Deep Thought strategy is constrained by at least two other virtues: nonsensitivity and cognitive salience.14 First, an explanandum characterized too precisely can, for instance, make an explanation too hard for us to understand. In other words, precision reduces cognitive salience. This is a crucial problem for sciences that, like epigenetics, are dealing with complex phenomena. Often, the more complex the phenomena are, the more complicated are the explanations and the harder it gets to infer counterfactual situations and to translate them into interventions. Second, a highly detailed explanandum makes an explanatory relationship more sensitive to changes in values of explanans variables and/or to changes in values of background conditions. This means that it continues to hold under a small number of these changes. For example, in the case of population dynamics in mice discussed above, the dependency relation between an explanans and a rather coarse-grained description of the changing mice population considering, for example, only the distribution pattern of mice is harder to disrupt than one of an explanans with a fine-grained alternative explanandum including additional details on genetic and epigenetic makeup of mice as well as behavior and environmental information

▸ 177 ASSESSING THE EXPLANATORY POWER OF EPIGENETICS

on predator abundance or distribution and climate. The latter explanatory relationship can be said to be more sensitive to changes. If an explanation like the former one is less sensitive, it is more powerful in that it enables inferences to more counterfactual situations in which variables included in the explanans and/or background conditions take non-actual values without breaking the explanatory relationship. Thus, regularities with less sensitivity carry more explanatory power since they give correct answers to a larger set of w-questions. However, as Ylikoski and Kuorikoski (2010, 218) note, it “remains to be shown whether [their] conceptual apparatus is actually useful in making sense of scientific controversies.” We will see that, indeed, this concept of explanatory power is highly useful. In the present debate on epigenetic explanation it helps to elucidate why and when epigenetic explanations are legitimately chosen over orthodox molecular and evolutionary explanations.

CHOOSING THE BETTER EXPLANATION

The above framework of explanatory power allows us to compare both sufficiently similar, directly competing, and alternative explanations, as well as indirectly competing ones addressing different explananda. The latter kind of comparison is made possible if two explanations share (or should share) similar standards of explanation that can then be evaluated.15 Let us consider these two kinds of comparisons as direct and indirect comparison, respectively. Both are important for assessing the value of epigenetic explanations. Given the number of new features epigenetics adds to molecular and evolutionary explanation (see chaps. 2 and 3), what follows is restricted to evaluating the value of molecular epigenetic explanations that trace “causes without mechanisms” and evolutionary epigenetic explanations that cite proximate causes. Accordingly, two crucial questions have to be answered in order to assess whether or not, why, and when to favor these new explanations in the field of molecular and evolutionary biology over prevailing explanations. Given a scenario of evidential parity,

178 ◂ ASSESSING THE EXPLANATORY POWER OF EPIGENETICS

(ME) Do molecular epigenetics’ causal explanations with less detail (i.e., higher-level explanations omitting explanatory information about genes) have more explanatory power than standard mechanistic molecular explanations? and (EE) Do proximate cause explanations (answering how a character evolved) have more explanatory power in evolutionary biology than standard ultimate cause explanations (indicating what a character evolved for)?

According to the clause “given a scenario of evidential parity,” I seek to compare biological explanations (or explanatory agendas) that are considered to be equally likely or supported by equal evidence. More generally, this means, following Henri Poincaré’s wise words introducing this chapter, that there is in fact more to science, and thus theoretic improvement, than merely accumulating evidence alone. This “more” is not only, as Poincaré argued, the methodologies of scientists but also, as should be emphasized, their standards for what makes good explanations. Therefore, let us make a “pretend play” of evidential parity that can address the two questions on explanations’ value in molecular biology (ME) and evolutionary biology (EE). (ME) points directly toward the conflict between mechanistically detailed molecular explanation—especially those explanations with a strong emphasis on genetic mechanisms—and those that entail little or no mechanistic detail (see chap. 3). (EE) addresses the debate over choosing evolutionary proximate cause explanations or ultimate cause explanations (see chap. 2). Before we turn to questions (ME) and (EE) in some detail, let me emphasize that both represent the most extreme poles to be integrated. In contrast, more moderate cases have been discussed in the previous chapters: autonomous higher-level explanations that include information about more fundamental levels, as well as proximate cause explanations that address the population levels of evolutionary phenomena. Here, integration is a result of reconceptualizing explanatory dependency relations. However, since theoretical integration is a case-by-case issue, more extreme cases, like the ones discussed below, might actually appear, especially when there is a standoff between explanations in terms of evidence and when researchers are rather reluctant to make commitments to one another. Currently, these cases are maybe the biggest obstacles for an extended evolutionary synthesis.

▸ 179 ASSESSING THE EXPLANATORY POWER OF EPIGENETICS

MOLECULAR EPIGENETICISTS’ EXPLANATION VERSUS MOLECULAR BIOLOGISTS’ EXPLANATION

In contrast to standard mechanistic molecular explanations, many molecular epigenetic explanations provide little or no explanatory information on genetic mechanisms. Can these higher-level explanations nevertheless rightfully claim to be the better explanations in molecular biology? Let us focus first on the virtue of (non)sensitivity. As described above, an explanation becomes less sensitive if we can increase the range of change in background conditions and/or in values of explanans variables without breaking the explanatory relationship between the explanans and the explanandum. To put it differently, an explanatory relationship is less sensitive if it continues to hold under a larger set of interventions on values of explanans variables and/or background variables. The first type of nonsensitivity against changes in background conditions can be called robustness. It can be characterized more precisely as follows: “Explana- tion A is strictly less sensitive than explanation B if and only if there is at least one change in background conditions with respect to which A is less sensitive than B and there are no background conditions in which B is less sensitive than A” (Ylikoski and Kuorikoski 2010, 209; emphasis added).16 The second type of nonsensitivity against changes in explanans variables is called invariance (Woodward 2003). Both increased robustness and invariance make an explanation more powerful in that they permit the making of inferences to more counterfactual situations in which background variables or variables included in the explanans take non- actual values. Although Ylikoski and Kuorikoski mention this distinction, their examples focus exclusively on robustness. But, as I show below, this distinction is crucial and needs more attention when explanatory power is evaluated. Take, for instance, two paradigmatic explanations of transgenerational trait heritability or persistence in molecular genetics and molecular epigenetics:

(Emg) Trait X of organism A is inherited by A from generation F0 to Fn (given certain background conditions) due to a set of heritable

genetic causal factors Gn determining X (and thus X can be selected).

(Eme) Trait X of organism A is inherited by A from generation F0 to Fn (given certain background conditions) due to a set of heritable epi-

180 ◂ ASSESSING THE EXPLANATORY POWER OF EPIGENETICS

genetic causal factors (i.e., regulatory factors) En determining X (and thus X can be selected).17

For example, (Eme) can be an explanation of nongenetic transgenerational inheritance of obesity (see Waterland et al. 2008; Dunn and Bale 2009, 2011; Pentinat et al. 2010; and Li et al. 2012). As epigenetics research on risk factors of chronic diseases such as obesity in humans suggests, obesogenic phenotypes can be understood as the result of a mismatch between the developing organism’s “prediction” of the future (generations’) environment and the actual environmental conditions of the adult organism/progeny (see, e.g., Gluckman and Hanson 2004; and Gluckman et al. 2008). According to this model, epigenetic alterations brought about by changes in the maternal (in utero and/or early postnatal) environment of the organism affect how the organism and its progeny will respond to later environmental challenges. For example, the organism can show altered tolerance to high-fat diets.

In contrast, (Emg)-like explanations can be found in genome-wide association studies that seek to identify heritable changes in DNA as risk factors of common human diseases, such as obesity (see Risch and Merikangas 1996; Rankinen et al. 2006; and Do et al. 2013). In this field, risk factors are commonly described as disease-causing genetic variants (see, e.g., Manolio et al. 2008). The set of genetic causes nG in (Emg) is thought to represent component parts of a mechanism that brings about the overall behavior of the epigenetic regulatory factors En, as well as the causal dependency between these factors and the inherited trait, like obesity. Thus, the explanation (Emg) entails more mechanistic detail than the molecular epigenetics alternative

(Eme), with Gn accounting for why the causes had the effect they did in (Eme). According to molecular and evolutionary biology’s orthodoxy, the explanatory relationships between genetic mechanisms and phenotypic traits are considered to set the standard of how much sensitivity an explanatory relationship is permitted in order for it to be explanatory at all. As a result, explanations like (Eme) are often subject to criticism. For example, evolutionary biologists claim that, compared to explanations of genetic inheritance of traits, epigenetic inheritance explanations cannot address the phenomenon of trait heritability or persistence with sufficiently low sensitivity. To be more precise, they argue that the determining epigenetic factors are highly sensitive to any relevant changes and thus exhibit no sufficient stability (i.e., they are highly reversible due to, for instance, high

▸ 181 ASSESSING THE EXPLANATORY POWER OF EPIGENETICS

epimutation rates) in order to be evolutionarily relevant (see, e.g., Pál and Hurst 2004; and Deichmann 2016a). We can call such a claim a “hypersen- sitivity accusation” (HSA). An “explanation is hyper-sensitive if any change in the background conditions [considered to be relevant] would break the dependence between the explanans and the explanandum” (Ylikoski and Kuorikoski 2010, 209; emphasis added). Note that, with respect to the example above, this HSA of orthodox-minded biologists refers to the robustness

of the explanation (Eme) given changes in inner- and/or extraorganismic environmental background variables; this includes genetic changes as well. As a result, the HSA does not hold or make (at least in a straightforward

way) (Eme) a worse explanation than (Emg), since (Eme) in fact scores (sufficiently high) on the invariance dimension of sensitivity. Let me develop this argument in more detail. The epigenetic research agenda is focused on explaining (partial) autonomy or nonsensitivity of epigenetic factors to genetic causes. Once these genetic factors are fixed to a certain value in, for example, inbred lines, and once other confounding variables are controlled

for via cross-fostering, for example, the invariance of explanations like (Eme) is secured. This means that genetic factors (in addition to other environmental factors) are not, as mistakenly understood by critics, treated as background conditions against which to test the robustness of transgenerational epigenetic inheritance claims. Instead, they are understood as important explanans variables against which to test the invariance of the explanations. This means that the scientists test under which combination of values of

these explanans variables the relationship under study, En → X, holds. Due to epigeneticists’ awareness of the low robustness of their explanations, their methodological and explanatory strategy is to incorporate more previously omitted background variables into their models and then to test the invariance of the dependency relations under study. To claim otherwise means to ignore the genetic dimension of epigenetics. Therefore, epigenetic explanations usually do not come in the form of

(Eme) but in the following form:

(Eme*) Under a certain combination of values of a set of genetic and (other)

environmental variables Zn, trait X of organism A is inherited by A

from generation F0 to Fn (given certain background conditions) due

to a set of heritable epigenetic causal factors En determining X (and thus X can be selected).18

182 ◂ ASSESSING THE EXPLANATORY POWER OF EPIGENETICS

As this formulation shows, the HSA of orthodox-minded biologists resulting from prevailing standards of explanations’ sensitivity is based on conflating different notions of (non)sensitivity, namely, robustness and invariance, in order to estimate the explanatory value of epigenetics. In addition, molecular epigeneticists’ explanations do not necessarily display high sensitivity. They contain higher-level invariant dependencies contextualized by specific genetic and (other) environmental settings. Thus, both molecular genetic and molecular epigenetic explanations consider genetic causal factors, but they serve different roles in each approach according to the dependency relations of interest (see also chap. 3). But what about the precision of these explanations? One could intuitively argue that saying a genetic narrative describes why a certain trait, like obesity, occurs in development and is inherited across generations is more precise than saying an epigenetic narrative describes why less distal causes (in a particular inner- and/or extraorganismic environment) bring about and transmit the trait. In fact, neither epigenetic explananda nor their contrast classes can be said to be in any way sharper than those of standard molecular explanations. In molecular biology, an explanandum is usually formu- lated in such a way that it directly restricts the class of potential explanans variables by which it can be addressed. This means that molecular biologists often do not ask which causal factors bring about the phenotypic characteristic under study but instead ask which genes code or otherwise influence (under particular circumstances) the effect observed and the relevant intermediate causes. This explanatory restrictedness is crucial for all milestone discoveries in the history of molecular biology, from the description of the mechanism of protein synthesis (1958), to the discovery of the “operon” (1961), the central dogma (1970), and the Human Genome Project (2000). In the last case, focusing on genes rather than on other causal factors was considered to be an explanatory strategy suitable not only for uncovering the causality of developmental and heredity phenomena but also, as Walter Gilbert (1991, 99) describes it, for gaining “total [causal] knowledge of the human organism.” Epigenetics pursues a different strategy. Similar to the approach of developmental systems theorists, it questions the usefulness of the standard restrictive account and asks for an unbiased approach toward living systems. As a consequence, epigenetic explananda look different. They exhibit a lower sharpness—as in fact they have to—in order to suitably grasp the

▸ 183 ASSESSING THE EXPLANATORY POWER OF EPIGENETICS

phenomenon-to-be-explained without presupposing a certain hierarchical order of explanatory power between possible explanans variables. In other words, the decreased level of fine-grainedness of an epigenetic explanandum necessitates less detailed (lower-level) information on the explanans. This enables epigeneticists to explore the complex web of different inner- and extraorganismic causal factors and multifactorial causation on various levels of organization with respect to the less sharp explanandum. For example, in the case of obesity, genome-wide association studies are methodologically constrained to assess a member of a particular class of macromolecules, deoxyribonucleic acids, as the cause of the phenotypic effect. In contrast, epigeneticists such as Robert Waterland and colleagues (2008, 1373), in a rather unrestrictive manner, ask how “maternal obesity before and during pregnancy and lactation impairs developmental establishment of body weight regulatory mechanisms in the fetus or infant, causing transgenerational amplification of obesity prevalence and severity.” This explanandum is addressed, although “the biological mechanisms underlying such processes [are expected to] remain unknown” (1373). Waterland and colleagues used agouti viable yellow (Avy) mice, which are mice with a mutation in the Avy allele. They are prone to overeating and show yellow fur color. Some heterozygote Avy/aa females and their progeny were weaned either on a methyl-supplemented or a standard diet and then mated with a/a males. This procedure was repeated for several generations. The tendency for obesity was transgenerationally exacerbated in the

group not fed a methyl-supplemented diet. In generation F3, these animals showed increased body weight compared with their methyl-supplemented counterparts. In other words, Waterland et al. (2008, 1373) traced a causal dependency, which shows preemption: the “transgenerational amplification of body weight was prevented by a promethylation dietary supplement.” Interestingly, the effect of methyl supplementation on body weight was independent of epigenetic changes at the Avy locus, as the “methyl group” of animals did not show a change in yellow coat color. According to this result, the increase in methylation that prevents obesity across generations does not happen at the Avy locus but at some other place in the genome. To put it another way, there is a previously unrecognized environmentally induced and epigenetically mediated causal effect on obesity across generations, although the underlying mechanisms and their component parts have not been specified.19

184 ◂ ASSESSING THE EXPLANATORY POWER OF EPIGENETICS

One major disadvantage of this methodological step toward a more liberal way of seeking possible explanantia of such phenomena is that epigenetic explanations do not proceed by the Deep Thought strategy. This would grant them more explanatory power, since adding information on how an underlying (genetic) mechanism is involved in bringing about an observed phenomenon like obesity would permit the answering of more w-questions about factors originally omitted from the explanation. Nevertheless, there are several objections to be made against the claim that the less precise nature of epigenetic explanations determines their overall value: First, a less detailed explanandum does not need an explanans that is mechanistically very detailed. As a consequence, cognitive salience is increased. Omitting, for example, the complex interplay between genes and epigenetic regulatory factors from an explanans and instead considering merely the latter to be explanatorily relevant under certain environmental situations (e.g., in a certain inbred line exposed to a set of environmental factors) makes it easier to grasp an epigenetic explanation of multifactorial dependencies. Second, nonsensitivity or stability is secured, since an explanatory relationship containing less mechanistic detail will more likely continue to hold under a larger set of changes in values of relevant variables. This result is in line with the previous finding that epigenetic explanations can (at least) claim high invariance. Third, often a more detailed explanation means more detail about other causal factors than the ones already known (as well as their relationships to the already known ones). In other words, more detail usually means more information about how robust the relationship under study is. This leads to no improvement of the higher-level original explanation, but it offers answers to w‑questions concerning a new (broader) explanandum through interventions on additional causal factors not known before. To put it differently, a more detailed mechanistic explanation of a causal relationship between X and Y should not be understood as a better explanation of that relationship per se but rather as a better explanation of the causal embeddedness of the dependency between X and Y: an explanation of the complex network of factors on multiple levels (potentially) relevant for X → Y. In other words, it is a better explanation of relationships between originally unknown factors and X and/or Y.

▸ 185 ASSESSING THE EXPLANATORY POWER OF EPIGENETICS

The concept of causal embeddedness enables us to distinguish two complementary explanatory approaches with different explananda. Adding information on causal embeddedness to an epigenetic explanatory relationship means shifting or broadening this explanation rather than directly improving it. This strategy leads to a new, yet related explanation addressing a different explanandum, not necessarily to a more fine-grained explanation of one and the same explanandum.20 For example, discovering why obesity, determined by a DNA methylation pattern, is transgenerationally inherited in mice is different from determining why obesity is inherited in

mice, given this pattern and a number of other factors Zn, which are causally more upstream than the cause and the effect and which underlie the dependency between “DNA methylation pattern → obesity.” The second approach does not by itself necessarily present more explanatory information about the original dependency relation. If such an increase in causal embeddedness information is wrongly understood as an improvement of the original explanandum rather than a creation of a new explanandum as a separate complementary explanatory strategy, this can easily lead to misguided hierarchy thinking in explanation and to conflicts about an explanatory “sover- eign right.” The latter refers to the problem of which of the two explanatory approaches is rightfully allowed to explain a phenomenon. Fourth, regarding in particular the challenge of integrating epigenetics with evolutionary biology, it is necessary to mention another rather historical argument, one that underlies the advantage of choosing less precise but also less sensitive and more comprehensible explanations. The strategy of valuing explanatory stability over precision was a major criterion for Dar- win in developing his theory of evolution. Although he did not open the mechanistic black box of heredity, Darwin was nevertheless able to establish highly stable or insensitive explanations of trait persistence and change. This explanatory scaffold of the theory of evolution turns nonmechanistic epigenetic explanation into an easily integrable close relative. Finally, and somewhat related to the defense of the heuristic value of nonmechanistic causal explanation in ecological epigenetics (see chaps. 2 and 3), these explanations (directly or indirectly) lead to the development of further experiments and medical treatment. For example, as Peter Gluck- man and colleagues (2011, 246) note, medicine profits from epigenetics’ nonmechanistic knowledge about chronic diseases like obesity, since it “can explain the clinical and experimental evidence about the consequences—

186 ◂ ASSESSING THE EXPLANATORY POWER OF EPIGENETICS from the age of puberty to predisposition to disease—of prenatal and early postnatal environment in terms of subtle adjustments in the adult phenotype.” According to the above, the initially posed question (ME) “Do molecular epigenetic explanations with less detail have more explanatory power than standard mechanistic molecular explanations?” can be answered as follows: in the virtual state space of explanatory value, epigeneticists found a niche in which their explanations exhibiting low-precision but low-sensitivity and high-cognitive salience claim high explanatory power. Crediting molecular epigeneticists’ explanations with low explanatory power is thus not as straightforward as assumed by some biologists who stick to prevailing standards of explanation, since these explanations display explanatory goodness on some dimensions. In fact, epigenetic higher-level explanations providing little or no information about underlying genetic mechanisms are a valid alternative to mechanistic causal explanation in molecular biology, because the former makes possible a large number of theoretically and/or practically important inferences in counterfactual situations. In other words, in cases where nonsensitivity and cognitive salience are valued over precision, epigenetic low-precision but low-sensitivity and high-cognitive salience explanations are better than orthodox explanations for helping us understand epigenetic phenomena on the molecular level and for interrelating molecular and evolutionary explanations.

EVOLUTIONARY EPIGENETICISTS’ EXPLANATION VERSUS EVOLUTIONARY BIOLOGISTS’ EXPLANATION

In contrast to standard ultimate cause explanations being able to determine, according to Ernst Mayr (1961), what a phenotypic character evolved for, epigenetic explanations cite developmental proximate causes as explanans variables in order to answer how a character evolved. Can these new, rather unorthodox explanations rightfully claim to be the better explanations in evolutionary biology? Before answering this particular question, we have to address another more general one: with regard to phylogeny, how can we compare epigenetic explanations and evolutionary explanations at all? At first glance, it seems they address the same explanandum: the role of heredity in evolution. However, as emphasized by various authors (see Pennisi 2008; S.

▸ 187 ASSESSING THE EXPLANATORY POWER OF EPIGENETICS

Gilbert and Epel 2009, 441–57; Pigliucci 2009; and Gissis and Jablonka 2011), epigenetics seeks instead to determine what the standard explanatory agenda in evolutionary biology usually leaves out: developmental processes and questions pertaining to the developmental origins of variations, including heritable variations. This means introducing various new explananda into evolutionary biology. Thus, epigeneticists—if they are interested in evolutionary explanation, as ecological epigenetics is—address different non-neo-Darwinian explananda, although they deal with evolution.21 But if an epigenetic explanans is presented for an evolutionary explanandum phenomenon, it is usually called up to meet the standards of explanation set by the prevailing research field—neo-Darwinian evolutionary biology—of this explanatory domain. Because of this, an intense conflict between these two kinds of explanations might emerge, although they address very different explananda and

FIG. 5.1. The explanatory programs of epigenetics (left) and orthodox evolutionary biology (right) (after Guerrero-Bosagna 2012, 287).

188 ◂ ASSESSING THE EXPLANATORY POWER OF EPIGENETICS are thus not directly competing with each other. This conflict is based on the fact that the orthodox evolutionary research agenda, following Mayr (1984), claims to have developed the only legitimate explanatory tool to address evolutionary explanandum phenomena: adaptationist ultimate cause explanations. According to this explanatory program, only those causes and mechanisms that “fit” the ultimate cause of adaptation by natural selection are legitimately included in an explanans. Thus, as long as epigenetics refuses to address an evolutionary explanandum phenomenon by means of an ultimate cause explanans—a “benefits story” about what the phenomenon under study evolved for—it is considered to be lacking in explanatory excellence. Because of this, Thomas Dickins and Qazi Rahman (2012) argue that epigenetic explanations that seek to explain how a specific evolutionary phenomenon has evolved in its development are not valid evolutionary explanations. Figure 5.1 offers a diagrammatic representation of these two distinct explanatory programs. This controversy leads us to these central questions: “Is it possible to evaluate whether often rejected epigenetic proximate cause explanations have more explanatory power in evolutionary biology than standard ultimate cause explanations?” and “What are the explanatory advantages of proximate cause explanations in contrast to orthodox adaptationist explanation and when (or given which particular phenomena-to-be-explained) are the former advantageous over the latter?” The contrastive framework of explanatory power presented above will help us to settle these issues. Carlos Guerrero-Bosagna (2012, 296; original emphasis) is in favor of epigenetics, claiming that ultimate cause explanations are a recipe for speculation: “focusing on ‘what is the mechanism built for’ (i.e. focusing on ultimate causes) implies only speculations with no relevant evolutionary implications.” However, he does not consistently link this criticism of the speculative nature of ultimate cause explanations to his idea of evaluating what he calls “explanatory power.” Rather he conflates this notion several times with the notion of evidential power or likeliness. For instance, he compares two alternative, ultimate, and proximate cause explanations of the very same explanandum phenomena—the evolution of bird wing morphology—by listing evidence for both.22 He then concludes that if epigenetic explanations are supported by an equally large amount of evidence compared to ultimate cause explanations, this dismisses the need for the latter. This is because he considers the speculative nature of adaptationist “benefit”

▸ 189 ASSESSING THE EXPLANATORY POWER OF EPIGENETICS

explanations to be an a priori fallacy committed by orthodox evolutionary biologists; it acts as the decisive factor in evidential parity situations. However, Guerrero-Bosagna’s negative assessment of ultimate cause explanations does not itself rest on a detailed evaluation of their value. In contrast to his argument, which is based on the estimation of evidential parity and a priori assumptions, we can elaborate his skepticism against standard evolutionary explanation by means of the evaluation of explanatory power. Let us for this reason consider two paradigmatic evolu-

tionary explanations, an ultimate cause explanation (Euc) and a proximate

cause explanation (Epc), addressing the same explanandum, “the evolution of obesity in humans”:

(Euc) Obesity in current human populations evolved because in times of rapid fluctuations between food surplus and famine, which have been frequent in human history, this character has been advantageous or under positive selection.

(Epc) Obesity in current human populations evolved because it has been induced by exposure to specific environmental conditions in early life and was nongenetically mediated and inherited.

Explanation (Euc) is called the “thrifty gene” hypothesis, developed by James Neel (1962). More recently it has been advocated by, among others, Andrew

Prentice (2005) and colleagues (Prentice et al. 2008). Explanation (Epc) is supported by research on environmental induction in early development via exposure to maternal high-fat (or low-protein) diets (see Zambrano et al. 2006; Parente et al. 2008; Shankar et al. 2008; White et al. 2009; and Benkalfat et al. 2011) and on nongenetic transgenerational inheritance of induced obesogenic phenotypes (see discussion above).23 Again, let us assume evidential parity in this case, which means that both alternatives are equally well supported by evidence. With this “pretend play” as a basis, we

can start comparing the explanatory power of (Euc) and (Epc).

Is (Epc) more precise than (Euc)? Although the two explanations do not differ in precision or in the sharpness of their shared explanandum, they do differ in the sharpness of the contrast class of their explanandum:

Explanandum of (Euc) {present incidence of obesity in humans} [no obesity in humans]

190 ◂ ASSESSING THE EXPLANATORY POWER OF EPIGENETICS

Explanandum of (Epc) {present incidence of obesity in humans} [different grades of obesity in humans]

Thus, they take the forms

* (Euc ) Obesity in current human populations [no obesity in humans] evolved because in times of rapid fluctuations between food surplus and famine, which have been frequent in human history, this character has been advantageous or under positive selection. * (Epc ) Obesity in current human populations [different grades of obesity in humans] evolved because it has been induced by exposure to specific environmental conditions in early life and was nongenetically mediated and inherited.

Evolutionary ultimate cause explanations, like (Euc), should enable the researcher to account only for a very limited range of the contrast space relative to the “evolutionary” explanandum. This means the presentation of an explanans that should simply answer why the character under study evolved, rather than not evolved.24 In our case, that question is “Why did obesity evolve in humans rather than not evolve?” For example, Prentice and colleagues (2008, 1609; emphasis added), who support Neel’s thrifty * gene hypothesis, introduce (Euc ) in order to answer questions of this particular kind: “First, . . . we would not expect hunter-gatherer populations to display a thrifty genotype, as evidence suggests that they havenot been subjected to famines. Second, there is now ample verification to show that sub- sistence agriculturalist populations gain weight rapidly in harvest seasons . . . and readily become obese when relieved of food shortages by moving to an urban cash-based economy.” They add that since “famines and seasonal food shortages have been virtually a norm in sedentary agricultural societies in the past 10000–12000 years” instead of not having occurred frequently during this period, there has been “more than ample time for significant natural selection to have occurred” (1609, emphasis added). The potential non-occurrence of the explanandum phenomenon is provided by a counterfactual community scenario whose non-actuality has to be addressed by the explanation. In this case, the non-occurring scenario includes other forms of cultural traditions and/or constant food supply in human populations.25 Now the explanans makes possible addressing both the explanandum itself

▸ 191 ASSESSING THE EXPLANATORY POWER OF EPIGENETICS

as well as its contrast class by proving natural selection occurs in one case but not in the other.26 This explanation takes the following form:

** (Euc ) Obesity in current human populations [no obesity in humans] evolved because in times of rapid fluctuations between food surplus and famine, which have been frequent in human history [have not been frequent in human history], this character has been advantageous or under positive selection.

However, such an ultimate cause explanatory approach that primarily seeks to grasp the reason for the occurrence rather than the non-occurrence of a certain (evolutionary) phenomenon shows relatively little precision. In other words, if the explanandum, through a rather narrow contrast class, demands little sharpness, consequently the explanans will have little precision. * In contrast, developmental proximate cause explanations, like (Epc ), addressing how something has evolved in development, should be able to address a much broader contrast class. In our case of obesity in humans, this means, for example, answering the question “Why did the grade of obesity observed in humans evolve rather than other (non-actual) grades of obe- * sity?” Thus, although explanation (Epc ) has the same sharpness as (Euc) with regard to the explanandum itself, its much broader contrast class enables the proximate cause research agenda to gradually refine the sharpness of * the original explanandum of (Epc) and (Epc ) {present incidence of obesity in humans} and turn it into this:

** Explanandum of (Epc ) {present incidence of obesity grade x1 in humans}

[different obesity grades x1+n in humans]

Accordingly, the explanation takes a new form:

** (Epc ) Obesity grade x1 in current human populations [different obesity

grades x1+n in humans] evolved because it has been induced by exposure to specific environmental conditions in early life and was nongenetically mediated and inherited.

In order to answer how or under which circumstances this particular obesity

grade x1 or another grade x1+n is brought about by the complex interplay

192 ◂ ASSESSING THE EXPLANATORY POWER OF EPIGENETICS of genetic, epigenetic, and environmental factors in human development, developmental biologists and epigeneticists use animal models.27 Christy White and colleagues (2009), for example, use rats to investigate under which environmental and developmental factors obesity grade x1, which is defined by a particular body fat percentage, serum leptin levels and insulin tolerance levels are more likely to occur in the offspring of maternally obese dams rather than under other circumstances.28 Their explanans is summarized as follows:

The work described here supports the concept of nutritional programming and indicates that offspring of maternally obese dams weigh more and exhibit increased body adiposity compared with offspring of nonobese dams. In addition, the present work extends this observation by effectively distinguishing between the direct effects of a high-fat diet and effects that are secondary to maternal obesity [leading to different body fat percentages, leptin levels, and insulin tolerance levels in the offspring]. Because the offspring of HF[high-fat]-fed nonobese . . . dams were similar to the offspring of LF [low-fat] dams, these data clearly indicate that consumption of a high-fat diet itself (i.e., in the absence of obesity) is insufficient to predispose offspring to obesity. It is instead maternal obesity that is required for this programming effect, indicating that some feature of the obese maternal environment induces a permanent structural or functional change in the offspring. (White et al. 2009, R1471; emphasis added)

The different character of this explanans compared with the explanation by Prentice and colleagues (2008) of what made organisms “become obese” in evolution, rather than “not become obese,” should be noted. White and colleagues (2009, R1471) seek to explain how—under which circumstances in development—an organism might end up “weigh[ing] more and exhibit[ing] increased body adiposity [i.e., different obesity grades] compared with” an organism that develops under different circumstances and shows other grades of obesity, for example, different leptin and insulin tolerance levels, or no obesity at all. Other studies underlining the transgenerational influence of programming during gestation for the development of obesity in the offspring present similarly precise explanatory relationships between explanandum and

▸ 193 ASSESSING THE EXPLANATORY POWER OF EPIGENETICS

explanans. For example, Kartik Shankar and colleagues (2008, R528) show that feeding different high-fat and low-fat diets to adult female rats over a period of three weeks not only causes substantial differences “in body weight gain, adiposity, serum insulin, leptin, and insulin resistance” in those female rats but that “offspring from obese dams gained remarkably greater . . . body weight and higher % body fat when fed a high-fat diet” (fig. 5.2). In addition, systematic changes in endocrine and metabolic parameters, like insulin and leptin levels, have been documented in the offspring. As in the example above, epigeneticists and developmental biologists in this case explain how, under which environmental and/or organismic circumstances,

an organism develops a particular obesity grade x1 in contrast to other pos-

sible grades x1+n. These similar “developmentally based” explanatory approaches that focus on proximate causes in contrast to ultimate causes easily lead to a variety of

answers to w-questions, like “What would happen to obesity grade x1 of the

FIG. 5.2. Body weight of male offspring of lean or obese dams from weaning through postnatal day 130. Offspring were fed low-fat or high-fat diets to mimic a postnatal non-obesogenic or obesogenic environment. *P < 0.05 (two-way ANOVA followed by Student-Neuman-Keuls post hoc analyses) (Shankar et al. 2008, R532).

194 ◂ ASSESSING THE EXPLANATORY POWER OF EPIGENETICS offspring if one or many explanans variables (i.e., causal factors determining the developmental process) were changed to non-actual values by intervention?” Seeking to answer such questions is a fruitful scientific enterprise when trying to understand the occurrence, transgenerational persistence, and characteristics of a trait in ontogeny and phylogeny. Let us now return to Guerrero-Bosagna’s (2012) claim that proximate cause explanations are better than ultimate cause contenders. As the explanations of humans’ obesity exemplify, his critique of the speculative nature of ultimate cause explanations can be addressed with a more differentiated philosophical approach by means of the concept of explanatory power. His claim is correct in those cases where an orthodox adaptationist “what for” explanandum’s contrast class shows little sharpness (compared to a devel- opmentalist “how” perspective on the very same explanandum) and thus demands an explanans with little precision. The explanatory information offered by such an imprecise explanans makes possible only a limited range of inferences to counterfactual situations wherein variables included and excluded in the explanation take non-actual values. In this sense, pure adaptationist explanations make us understand the explanandum less thoroughly compared to epigenetic explanations. However, it would be too early to conclude that ultimate cause explanations are always or simpliciter a worse explanation compared to epigenetic proximate cause explanations. I offer here two arguments supporting this position. First, there is in principle no reason why the contrast classes of explananda addressed by ultimate cause explanations cannot be sharpened. A combination of ultimate and proximate cause explanations could address this more precise class, in which the latter explanation addresses especially those parts of the class that cannot be tackled by adaptationist explanatory information. Second, as discussed above, there is an important trade- off between precision and the dimensions of nonsensitivity and cognitive salience, whereas the latter two are usually decreased when precision of the ** explanandum is increased. Once we turn (Epc) into (Epc ), it becomes easier to disrupt the dependency between the new more “fine-grained” explanans, which includes a great deal of epigenetic developmental information, and the explanandum. This dependency is more sensitive to changes in background conditions (it is less robust) and to values of explanans variables (it is less invariant). In addition, a model of such a complex dependency relation is often harder to grasp mentally. Thus, according to these criteria,

▸ 195 ASSESSING THE EXPLANATORY POWER OF EPIGENETICS

a less precise adaptationist explanation is more powerful in that it enables one to make inferences to more counterfactual situations in which variables included and excluded in the explanans take non-actual values. To put it differently, based on the dimension of nonsensitivity and cognitive salience, a developmental proximate cause explanation more likely scores less well with respect to evolutionary phenomena than a traditional, less precise “benefit” explanation. Therefore, in many cases where the theoretical and/or practical context values nonsensitivity over precision, rather abstract but less sensitive adaptationist explanations may rightfully claim more explanatory power than competing epigenetic explanations. More generally, the above considerations stress the importance of case-by-case assessment of explanatory value in evolutionary biology based on a diverse catalog of criteria. Thus, neither “orthodox” nor “progressive” explanations can be said to have more explanatory power per se. However, against Mayr’s criticism of proximate cause explanations (see chap. 2), the latter can be credited with some explanatory power with respect to biological evolution. Guerrero-Bosagna comes to a radically different conclusion from his comparison of standard evolutionary and epigenetic developmental explanation. He admits that, given the case of evidential parity, “the same evidence can be interpreted in a finalist or non-finalist manner depending on the epistemological position adopted by the researcher in terms of associat- ing a specific feature to a purpose (predefined end) or alternatively considering efficient [i.e., proximate] causes” (Guerrero-Bosagna 2012, 288). But then he concludes that proximate cause explanations outperform ultimate cause contenders in nearly every situation. He claims that since benefit is hard to measure (due to its strong dependence on the scientist’s view about what should be considered beneficial), every explanans hypothesized in order to demonstrate benefits for organisms or populations has to be considered to be an ad hoc, “a priori speculation on future ecological-organism relationships” (296; emphasis added). Therefore, if possible, these explanations should always be avoided in evolutionary biology. According to the assessment of explanatory power in evolutionary explanation presented above, such criticism overshoots the mark. In order to argue for overcoming Mayr’s fallacy of abandoning proximate cause explanations in evolutionary biology, one does not have to completely substi- tute ultimate cause explanations with developmental ones, pace Guerrero- Bosagna. We can instead show that both explanations are, in a certain

196 ◂ ASSESSING THE EXPLANATORY POWER OF EPIGENETICS way—with respect to their particular profiles of precision, nonsensitivity, and cognitive salience—good explanations.

EXPLANATIONS IN THE EXTENDED EVOLUTIONARY SYNTHESIS

In their pathbreaking book Evolution in Four Dimensions, Eva Jablonka and Marion Lamb (2005) describe the nature of a new, extended theoretical approach to inheritance and evolution. They argue that more attention should be paid to epigenetic processes and developmental phenomena in general. They state, “We should not expect a single, universal model for . . . all the dimensions of heredity and evolution” (2005, 378). Thus, the new extended framework, suggested by recent research in fields like epigenetics, is considered to be more pluralist by nature. Similarly, Massimo Pigliucci and Gerd Müller introduce their edited volume Evolution—The Extended Synthesis by emphasizing that “evolutionary theory is no longer confined to the explanation of the increase in frequency and maintenance of favorable variants, but also becomes a theory of the mechanistic [developmental] conditions for the origin and innovation of traits” (2010a, 13; emphasis added). Note the word “also” in this quote. It suggests a similar pluralist perspective on how to develop an extended evolutionary synthesis, a synthesis of old and new explanations, as advocated by Jablonka and Lamb. In other words, these authors hold that old and new explanatory approaches coming, for example, from epigenetics, (eco-)evo-devo, niche-construction theory, and systems biology are not contradictory but (in principle) complementary. I believe that this complementary assumption is insufficient for an extended evolutionary synthesis to start flourishing. Simply highlighting the idea of integrating, for instance, ultimate and proximate cause explanations in evolutionary biology does not directly or by itself develop into a new unified, pluralist research agenda in biology. In addition, historical analysis of how the novel synthesis is differentiated from traditional concepts and views like molecular Weismannism or gene-centrism (Jablonka and Lamb 1995, 2005) cannot fulfill this demanding integrative work on its own either. What is needed instead are reliable criteria that help to clarify why and with respect to which explananda certain new explanations are more valuable than older ones, and vice versa, and how to combine the two

▸ 197 ASSESSING THE EXPLANATORY POWER OF EPIGENETICS

within a general agenda. The advocates of an extended evolutionary synthesis, both biologists and philosophers of biology, say virtually nothing about how to solve this issue (see, e.g., Jablonka and Lamb 2005, 2007; Müller 2007; Pigliucci 2007, 2009; Pigliucci and Müller 2010b; and Laland et al. 2014, 2015). Against this backdrop, I argue that in order to set the diversity of new and old explanatory approaches, including those coming from epigenetics, to work under a joint framework, this framework must have a contrastive approach of explanatory power that shows why both kinds of explanations are, in a certain way, good explanations. The above framework brings to mind the specific values, like precision, nonsensitivity, and cognitive salience, of distinct biological explanations, the trade-offs between them, and thus how they should be combined in order to jointly lead to a better understanding of the phenomena-to-be-explained. It is a framework suitable to support any account reaching for explanatory integration, not for substitution or reduction. Understanding these criteria also means becoming aware of the fact that novel explanations are not necessarily any weaker. They just exhibit different explanatory virtues than more orthodox explanations that have become entrenched as the paradigmatic exemplars of good explanation in biology. This kind of entrenchment of explanatory virtues considered worthy of pursuing might actually be detrimental. It can prevent biologists from searching for explanations that are located in regions of a state space of explanatory virtues, such as in the region of low precision and low sensitivity, which might go along with unforeseen advantages, including biomedical advantages. Take, for example, the case of epigenetic reprogramming. The technique of reprogramming cells relies entirely on causal explanations that entail little or no mechanistic detail—they show low precision and low sensitivity—indicating why, from a molecular geneticist’s point of view, the four regulatory factors, Oct3/4, Sox2, c-Myc, and Klf4, are crucial for inducing pluripotent stem cells. However, these explanations and the techniques based on them are of high medical value in stem cell therapy and cancer treatment. In the context of epigenetics and an extended evolutionary synthesis, this approach of explanatory power helps in settling various recent conflicts. First, it helps in abandoning the commonplace dichotomy between proximate and ultimate explanations, by integrating the two. Even in those cases

198 ◂ ASSESSING THE EXPLANATORY POWER OF EPIGENETICS where the explanans contains mixed causes—cases of “ultimate-cause– proximate-cause explanations”—explanatory value can still be assessed. This is crucial not only for epigenetic explanations but for evo-devoists who ask to reconsider developmental mechanisms as constraining elements in biological evolution. Second, given the contrastive framework of explanatory power, other crucial yet completely unrecognized conflicts hindering the integration of epigenetics’ explanatory agenda with other fields, like molecular genetics, can now be addressed. This refers to the various explanatory challenges described in previous chapters, such as interlinking explanations that trace “causes without mechanisms” and those that seek to describe the genetic basis of phenotypic traits, such as in genome-wide association studies. Another challenge might be evaluating which mechanistic explanations are better for helping us to understand certain complex phenomena, whether it is those that include non-change-relating dependency relations as explanans variables (the so-called “mechanisms without causes”) or those that consider merely change-relating dependencies. However, the concept of explanatory power applied here should not be considered the holy grail of an extended evolutionary synthesis. For one thing, in the young field of epigenetics there are intradisciplinary debates about how to historically, conceptually, and methodologically conceive of the research conducted (see chap. 2). Similar debates can be found in other rather young biological fields, such as evo-devo.29 These inner conflicts are unlikely to be settled just by the evaluation tool presented above. Nevertheless, this heuristic framework could be a useful device for guiding us to new cross-disciplinary approaches linking molecular biology with genomics and systems biology, development and morphology with evolution, and even Lamarckian with Darwinian views. Most important, it can act as a tool for theoretical improvement. If we know with respect to which dimensions an explanation or a whole explanatory agenda has less to offer than a competitor, and if, in line with Paul Feyerabend’s note introducing this chapter, we do not want to discard or falsify those explanations that lose in the competition, the contrastive evaluation of explanatory value helps us to directly improve the defeated explanation. In other words, explanatory approaches are improved methodologically by reinforcing them with knowledge about how to conduct them and by looking to other explanations that can compensate for their shortcomings. With respect to the

▸ 199 ASSESSING THE EXPLANATORY POWER OF EPIGENETICS

comprehensive explanatory framework of the extended evolutionary synthesis, which is the sum of various more or less good biological explanations, adopting this multilayered concept of explanatory power thus means making good on a promise made by several biologists and philosophers of biology: “The extended synthesis will be better than the modern synthesis.” ▶ ▶ ▶

A general contrastive framework suitable for evaluating the goodness of scientific explanations has been presented. It is able to give precise guidance for why and when epigenetic explanations are better than prevailing molecular genetics and evolutionary explanations, even though they entail less mechanistic detail and favor proximate causes. Thus, this concept of explanatory power represents a helpful tool for a new pluralist approach in biology: an extended evolutionary synthesis, which seeks to integrate epigenetic explanations into a field of other, more orthodox explanatory accounts.

200 ◂ CONCLUSION A PHILOSOPHY OF EPIGENETICS

Philosophy begins in wonder. And, at the end, when philosophic thought has done its best, the wonder remains. ALFRED NORTH WHITEHEAD, 1938

SEVERAL OF THE ISSUES discussed in this book have a common ancestor that can even be located in space and time: Villa Serbelloni, Bel- lagio, Italy, 1966 to 1970. At this Rockefeller Foundation villa overlooking the beautiful Lake Como, Conrad Hal Waddington, the grandfather of epigenetics, held four symposia to which he invited several internationally renowned theoretical, developmental, and molecular biologists, as well as physicists, mathematicians, linguists, and philosophers. Among others, the participants were Ernst Mayr, Francis Crick, Lewis Wolpert, Richard Lewontin, John Maynard Smith, Ruth Sager, Stuart Kauffman, René Thom, Christopher Zeeman, and Brian Goodwin. At the advent of philosophy of biology, the focal point of these meetings was to sketch a research program of theoretical biology. The proceedings of these meetings were entitled Towards a Theoretical Biology (Waddington 1968, 1969b, 1970b, 1972).1 CONCLUSION

According to Waddington, this biological discipline should be located where theoretical and philosophical issues of biology intersect. It can be understood as an expansion of attempts of the organicists, especially in the 1920s to 1930s, to investigate the conceptual foundations of biology. How- ever, while these earlier discussions have been focused primarily on the question of how biology can be located between mechanicism and vitalism, Waddington understood theoretical biology to be concerned, first and foremost, with those issues related to the complexity of living systems, as well as the relations between the systems’ micro- and macrostates. During these four meetings, various approaches were discussed regarding how to model complex systems, their dynamics, and stability characteristics. Some- what anticipating recent theoretical developments, Richard Lewontin even gave a talk titled “On the Irrelevance of Genes” and called for developing a “‘geneless’ theory of population genetics” (see in Waddington 1970b, 71). Moreover, conceptual models like Mayr’s distinction between proximate and ultimate causes have been discussed.2 These meetings may be understood as Waddington’s last attempt to enrich his work in developmental and evolutionary biology, primarily on what he coined “epigenetics,” by a wider, more interdisciplinary, and more philosophically informed research program. This attempt has served as a model and stimulus for the present book. I have tried here to trace his idea that there exists a channel of reciprocal “fertilization” between the various phenomena, the conceptual, explanatory, and methodological characteristics of epigenetics, and the current debates in philosophy of science—a channel that is waiting to be joined by philosophers of biology and biologists. Thus, the issues discussed in this book may be understood, again in a Waddingtonian way, as a methodological skeleton of concepts and topics pointing toward a philosophy of epigenetics. Guided by this motive, this book has offered a discussion of philosophical issues emerging from the interplay between scientific explanations, methods, and the complexity of living systems investigated in modern epigenetics. By building on Waddington’s original definition of epigenetics as a complexity science concerned with the causal analysis of development, it has been investigated how causality or causal mechanisms are traced in modern epigenetics and how they are conceptualized in, for example, experiment-based causal and mechanistic models, mathematical models,

202 ◂ CONCLUSION or models inferring causality from observational data. More such analyses are urgently needed, because epigenetic agents gradually seem to take over genes’ causal and explanatory supremacy. Previously, notions like the “epigenetic turn” and “epi-geneticization” were introduced to circumscribe in a rather fuzzy way the feeling that contemporary epigenetics seems to introduce something new to modern biology. One common answer seeking to specify this “something” was the idea that epigenetics reintroduces the seemingly outdated concept of Lamarck- ian inheritance of acquired characteristics into biological theory. This claim has provoked a wide discussion in academia, as well as in the wider public. While the Lamarckian dimension of epigenetics is a historically interesting debate, in this book I have tried to assess epigenetics’ particular theoretical character as well as the structure of the current “epigenetic turn” from a different philosophical perspective. By focusing on conceptualizations of complex living systems and scientific explanation in epigenetics, this book has defended the view that, as a consequence of their interest in grasping a special layer of biological complexity, labeled “epigenetic complexity,” epigeneticists play an important role in developing a novel explanatory agenda. Let me briefly summarize this argument: epigenetics has been shown to investigate molecular complexity in an expanded manner, both structurally and dynamically (see chap. 1). In other words, it is a complexity science concerned with nonsimple relations between nongenetic factors and their role in plastic development and heredity. As an analysis of the methodological and explanatory framework invoked to investigate these relations shows, epigeneticists explain and conceptualize living systems in a manner different from a number of other biologists, especially in molecular or cell biology (see chaps. 3 and 4) and evolutionary biology (see chap. 5). Most important, they consider other dependency relations as explanatorily relevant and develop different standards of explanation. These demarcations should not be understood as exclusive and clear- cut. Throughout this book epigenetics has been treated as a Wittgensteinian cluster concept. This takes into consideration that, of course, epigeneticists’ practices and theories are used by other scientists as well. It is not a field with sharp disciplinary, conceptual, or methodological boundaries. Despite this diversity, it was possible to identify a bundle of crucial features of epi-

▸ 203 CONCLUSION

genetics and epigenetic explanation that allow us not only to better understand current trends in biology but also to fill gaps in our current philosophical theories of scientific explanation. This bundle of epigenetic features refers, first and foremost, to causal explanation (see chap. 3) and mechanistic explanation (see chap. 4). First, epigeneticists in molecular and cell biology question how much mechanistic detail—especially on genes—is necessary for a causal relation traced in a complex system to act as an explanans. In short, this explanatory strategy in epigenetics can be labeled “causes without mechanisms.” This means that although the genetic basis of epigenetic phenomena is not neglected, genes no longer carry with them a special causal and explanatory weight. In other words, in contrast to traditional standards of molecular explanation and modeling, genes no longer have a unique ontic or epistemic status, which would necessitate listing them in every explanans of causal explanations. Moreover, genes in fact do not share the explanatory realm in which epigenetic causes actually make a difference. From a more general perspective, the idea of downgrading (but not neglecting) the explanatory value of underlying mechanistic detail in explanantia of epigenetic complexity may be understood as a reaction to the general “representation problem” in complexity sciences (see chap. 1). According to this problem, it is not possible to offer a fixed and complete description of a complex system, even when its components and their interrelations are completely known and fully considered. Interestingly, this investigation shows that epigenetic research does not only differ from what a number of other biologists are doing. It also elucidates that epigenetics has developed standards of explanation that challenge common theoretical frameworks in philosophy of science. For example, as the above analysis shows, epigenetic causal explanation is a highly instruc- tive case showing how, contrary to a widely accepted mechanistic theory of causation, the legitimacy and autonomy of higher-level generalizations can be secured without the need for supplementary mechanistic information from more fundamental levels of organization. In addition, mechanistic explanation shows a particular conceptual foundation in epigenetics, as the concept of biological mechanism has been decisively demachinized in this field. This conceptual orientation challenges not only biological standards but still widely held philosophical positions as well, as it cannot be grasped in a satisfactory way by the prevailing con-

204 ◂ CONCLUSION ceptual frameworks used by philosophers of science. In particular, this concept of epigenetic mechanism cannot easily be reconciled with a reading of mechanisms as machines, as defended (explicitly or implicitly) by a number of philosophers of biology in the new mechanism movement. Moreover, there are models of (robust) epigenetic mechanisms that address some microproperties of a complex system as explanatory causal properties, although they do not, in fact, causally influence the macrostate of the system. I have summarized this strategy as explanations tracing “mechanisms without causes.” Interestingly, this explanatory strategy cannot be fully grasped by the very theory of scientific explanation—interventionism—currently invoked by many philosophers of science who are interested in biological mechanisms. In particular, it contradicts the standard philosophical view that non-change-relating dependencies in mechanisms are noncausal and non-explanatory and should thus be omitted from a model. From a more general perspective, epigeneticists’ idea of being more liberal about some of the properties of a mechanism listed in an explanans may be understood as a solution to the complexity problem that I have called the “Borges problem” (see chap. 1). It refers to the challenge of understanding complex wholes as wholes. The above issues concerned the explanatoriness of epigenetics—whether and how epigeneticists explain—including the explanatory standards and criteria in biology. A central underlying idea of this investigation of the conceptual and explanatory characteristics of epigenetics was to better understand the various conflicts hindering the field’s theoretical integration, especially into evolutionary biology. Such conflicts arise due to the explananda and explanantia chosen by epigeneticists, to their interrelations, and to the methodological strategies applied to establish explanatory relations. As a response to these conflicts, I discussed solutions to problems pertaining to whether and how evolutionary explananda can be addressed by epigenetic explanations, although the latter are often based on highly artificial experimental setups and, in contrast to the traditional conceptual framework of evolutionary theory, focus on proximate causes rather than on ultimate causes (see chap. 2). Moreover, intradisciplinary conflicts in epigenetics have been historically reconstructed (see also chap. 2). They basically have an impact on the problems of how to understand the field of epigenetics methodologically and conceptually and how epigeneticists should explain, for example, by modeling epigenetic complexity mathematically

▸ 205 CONCLUSION

through a traditional Waddingtonian framework. This discussion may lead to a better understanding of the internal diversity of the field called epigenetics. The investigation of epigenetics’ explanatoriness and its associated inter- and intradisciplinary challenges has been accompanied by an assessment of the field’s explanatory power (see chap. 5). The basic idea was that knowing the rules by which these scientists rightfully explain biological complexity is not enough for them to function as a novel and important voice in the theoretical choir of modern biology. What is crucial is assessing not only whether epigeneticists can “sing” but whether they sing better than long-serving members of the choir. As we have seen, both on the molecular and evolutionary level, epigenetic explanations are, with respect to certain parameters, better than their long-serving competitors. Knowing these parameters enables us to allocate to epigenetics a particular vocal part and thus to place it within the choir most effectively. The presented framework of explanatory power offers the advocates of a more pluralist extended evolutionary synthesis view a useful heuristic tool that shows why and when epigenetic explanations are legitimately chosen over orthodox molecular and evolutionary explanations. In sum, this book has pursued the goal of not only enriching our understanding of the theoretical shifts taking place in modern biology but also contributing to recent debates in philosophy of science, particularly as they concern scientific explanation and the conceptual framework invoked to understand biological research. Moreover, it has tried to convey insight into particular historical developments in biology that are crucial for understanding the current status of epigenetics. Finally, some ideas or concepts developed in this book, for example, the concept of explanatory power, should be considered an attempt to contribute to a “philosophy for biology,” as James Griesemer (2011b, 2011c) calls it. Following Griesemer, such an approach could contribute to epigenetic research by offering concepts that help in organizing phenomena and articulating the research agenda, by clarifying heuristic research strategies, and by describing theory structures and explanatory strategies. I believe that philosophical observations about conceptual, theoretical, and explanatory differences in particular can help biologists overcome intra- and interdisciplinary issues and develop novel research questions and agendas. Besides the issues addressed in this book, however, there are various

206 ◂ CONCLUSION other philosophical, historical, and even anthropological topics on which I have merely touched in this book or that I have completely omitted. I will now briefly describe some of these loose ends that also concern a philosophy of/for epigenetics.3 First, how does the concept of organism function in epigenetic explanation? According to Richard Dawkins, organisms are mere survival machines programmed by selfish genes. Thus, they are explanatorily relevant for evolution only as targets of selection. This view has come under attack through a renewed interest in the plasticity and agency of the developing organism as evolutionarily relevant phenomena in fields like epigenetics and niche construction theory (Bateson 2005, 2012; Sultan 2015). This development has to be accompanied and maybe even guided by philosophers of science who focus their attention on the structure of organism- centered explanation and on challenges to theoretically reintegrating the concept of organism and related concepts, like “organismic organization” and “biological purposiveness,” into (evolutionary) biology (see Mossio et al. 2009; Huneman 2010; Mossio and Moreno 2010; Henning and Scarfe 2013; Nicholson 2014b; and D. Walsh 2015). Moreover, the conceptual relation between organisms and machines has to be reassessed (Nicholson 2013, 2014a). Second and somewhat related, the role of model organisms in epigenetics has to be considered. As Ernst Mayr (2002, xiv) once noted, “There is more to biology than rats, Drosophila, Caenorhabditis, and E. coli.” In other words, turning to new species means getting to know new phenomena and thus avoiding methodological and explanatory bias. More specifically, if epigenetics wants to focus on the previously neglected flexibility of developmental processes and nongenetic inheritance of complex traits, the number and diversity of model organisms have to be increased. This idea of extending the lists of traditional model organisms in order to better grasp the developmental mechanisms underlying biological diversity has started to spread (see V. Lloyd et al. 2012). Today epigeneticists increasingly focus on new species, like the tiny freshwater crustacean Daphnia (Harris et al. 2012), horned scarab beetles in the genus Onthophagus (Valena and Moczek 2012), and aphid species (Srinivasan and Brisson 2012). This development, however, is accompanied by various philosophical problems that concern the foundation of heuristic functions of new model organisms and the possible criteria to be adopted for their selection. This includes, for example, how the selection of model organisms guides and/

▸ 207 CONCLUSION

or limits our understanding of epigenetic complexity. In addition, there are issues related to the reliability of extrapolation from model to target species, given the complexity and diversity of the relationships between genotype and phenotype and especially their changeability in evolution. Moreover, there are sociodisciplinary issues on how to integrate epigenetic explanation based on new model organisms with other biological disciplines using traditional model species, like genetics, or even with other newly emerged disciplines, like evo-devo, which face similar methodological challenges.4 Third, due to the interrelatedness of epigenetic inheritance with other genetic and nongenetic inheritance systems, a philosophy of epigenetics has to conceptually specify and distinguish these various modes of heredity. This includes, for example, developing a conception of homology, which can be applied across two different inheritance systems (Powell and Shea 2014). This problem relates to the question of whether a trait inherited in two different lineages can be called homologous even though different processes are involved in its development and transmission. Other subjects and notions of relevance for philosophers of science interested in epigenetics include biological information; genetic, nongenetic, and environmental information, as well as its forms of representation (Jablonka 2002; Shea 2007, 2013); and the distinction between externalist and internalist or constructionist explanations (Godfrey-Smith 1996). The externalist view—the orthodox view in evolutionary theory—explains “properties of organic systems in terms of properties of their environments” (Godfrey- Smith 1996, 30). In other words, the organism’s properties are explained as a result of natural selection. In contrast, the internalist view addresses “one set of organic properties in terms of other internal or intrinsic properties of the organic system” (30). Whether this distinction is accepted in principle, how it is interpreted, or whether it is dropped by epigeneticists is anything but clear. Another issue to be investigated in the realm of philosophy of epigenetics is whether the popular ideas of gene-centrism and genetic determinism, which, as Paul Griffiths and Russell Gray (2005) show, resolutely refuse to die, are currently making way for a view of epigenetic determinism (Waggoner and Uller 2015). According to such a view, epigenetic factors are, in contrast to genes, not only considered to be explanatorily superior but causally superior in trait development and transmission. Fourth, a philosophy of epigenetics should strongly rely on historical studies in biology. As Norwood Russell Hanson (1962, 580) and Imre Laka-

208 ◂ CONCLUSION tos (1971, 91) both knew very well (although they took slightly different views), “Philosophy of science without history of science is empty.” Accord- ingly, as I have tried to show in this book, understanding the contemporary challenges of epigenetics, like the challenges of theoretical integration or methodological orientation, necessitates understanding the history of epigenetics as well as the development of conceptual frameworks and explanatory standards in various biological fields more generally. These historical studies should not be restricted to investigations tracing the Lamarckian history of contemporary epigenetics, like the debate on whether the neo-Lamarckian Paul Kammerer discovered epigenetic inheritance in the early twentieth century (Vargas 2009; Gliboff 2010; Svardal 2010; Vargas et al. 2016; see also Pennisi 2009). There are various other topics relevant for understanding the historical development of epigenetics. Examples include how David Nanney’s (1958) work on non-DNA systems of cellular heredity influenced the history of epigenetics (see Haig 2004, 2007, 2012) or how experimental research on cloning and reprogramming in amphibians, which started in the 1960s and 1970s (see Gurdon 2006), turned into stem cell research on the epigenetic mechanisms of reprogramming—studies that strongly rely on a Waddingtonian tradition of mathematical modeling. Somewhat related to the latter point, Waddington’s impact on modern biology is far from adequately understood. Erik Peterson (2011) lists several reasons why the reception of his work never really gathered speed in mainstream biology. However, when one focuses less on Waddington’s conceptual legacy (see, e.g., Hall 1992; and Gilbert 2000) and more on his methodological legacy, what is revealed is that his work had a widespread impact not just in contemporary biology (see chap. 2) but throughout the life sciences in general (see Baedke 2013). Moreover, to really understand Waddington’s legacy, one has to consider his more diverse later work on, for example, theoretical biology, his attempt to mathematize epigenetics, and his influence on scholars like Brian Goodwin (see, e.g., Goodwin and Saunders 1989) and mathematicians like René Thom. In addition, the account of organicism, including its underlying Whiteheadian process ontology, defended by Wad- dington (1970a, 1977b) and other members of the Theoretical Biology Club in the 1930s, including Joseph Henry Woodger (see Nicholson and Gawne 2014), has to be reassessed in light of the currently reemerging interest in systems, webs, and developmental processes in modern biology (see Peter-

▸ 209 CONCLUSION

son 2011, 2016). In other words, one may ask whether there exists a genuine metaphysics of epigenetics. Fifth, a philosophy of epigenetics may also address anthropological, bioethical, and sociological issues. This includes issues on how epigenetic research and associated ideas like phenotypic flexibility influence the anthropological concepts of human being, innateness, autonomy, self- determination, fate, and even (multigenerational) responsibility. Moreover, we have to ask, given full knowledge of epigenetic mechanisms, to what extent and why humans should (not) mold their organismic destiny and that of their descendants. Inspired by a growing number of epigenetic studies on humans, such as on epigenetic inheritance in athletes (Ehlert, Simon, and Moser 2013), there is yet a vast body of popular science literature dealing with these issues (see, e.g., Lipton 2005; Church 2009; Blech 2010; and Bell 2013). These works include The Genie in Your Genes: Epigenetic Medi- cine and the New Biology of Intention, Gene sind kein Schicksal (Genes are not destiny), and “Epigenetics: How to Alter Your Genes.” As these titles illustrate, epigenetics affects human life on more fundamental anthropological and ethical dimensions. This phenomenon is also gaining increasing attention in philosophical anthropology, bioethics, science and technology studies, and social theory (see Gravlee 2009; Kuzawa et al. 2009; Niewöhner 2011; Boniolo and Testa 2012; Hedlund 2012; Pickersgill et al. 2013; Scarfe 2013; Meloni 2014a, 2014b, 2016; and Baedke 2017b). Additionally, this recent discourse inside and outside of academia may trigger investigations of the sociological impact of epigenetics on folk biology, which encompasses the common nonscientific understandings of biology (see Griffiths 2011). One particular anthropological issue that I would like to describe in further detail is whether epigenetic research results may lead to a new understanding of identity and individuality. For example, Giovanni Boniolo and Giuseppe Testa (2012; see also Boniolo 2013) argue that epigenetics helps to solve old questions about synchronic biological identity—“Who is that living being?”—and diachronic identity—“How does it persist from one period of time to another?” They state that epigenetics teaches us that in order to solve these problems we need not focus on single phenotypic (or mental) properties but may look to the whole phenotype, which is developed in an environmentally sensitive process of epigenetic gene regulation and maintained over time as a robust developmental system. Thus, for example, diachronic identity can be understood as the continuity of the

210 ◂ CONCLUSION whole phenotype over time, guaranteed by the continuity of epigenetic processes (e.g., in cell lineages) that allow for the robustness of the organism in interaction with the environment. In short, the continuity of individuals’ epigenetic history constitutes the identity of single- and multicellular organisms. This view bears some interesting consequences. Given the transgenerational continuity of epigenetic processes and maintenance of phenotypes, it seems to imply that living beings can no longer be understood as (at least not strictly) delimited over time by mitotic or meiotic cycles. This means that we are no longer only biologically similar to our parents or children. Rather, we are them in a way. A related radical view has recently been defended by Scott Gilbert and colleagues (S. Gilbert et al. 2012; S. Gilbert 2014; Chiu and Gilbert 2015) in their work on holobionts (i.e., multicellular eukaryotes with multiple species of persistent symbionts). One such unit is formed through the close epigenetic interaction between the mother and fetus that is established through symbiotic microbial communities. These results not only question the standard human birth narratives of the origin of a new individual but, more generally, state that we are never individuals. One is always a supraindividual and even interspecies unit.5 Sixth and finally, a philosophy of epigenetics has to locate both itssub - ject area and itself within scientific discourse. This means, first, not only understanding how epigenetics differs from more orthodox accounts in the framework of the modern synthesis but how it differs from other progressive biological accounts, like evo-devo, niche-construction theory, and systems biology. A lack of historical and philosophical understanding of the latter kind of difference hinders the integration of epigenetics with other novel theoretical accounts. One such problem concerns the differences between epigenetics or, more particularly, epigenomics and systems biology. For example, it is anything but clear whether they share the same concept of “system.” Another, more crucial problem refers to epigenetics’ relationship to evo-devo. A growing number of evo-devoists have adopted the notion of “epigenetic(s).” Often they do so when discussing which epigenetic mechanisms of developmental plasticity might trigger evolutionary novelty. In such debates on how so-called “epigenetic innovation” leads to evolutionary novelties, the term “epigenetics” is usually defined in a rather classical manner by excluding epigenetic inheritance (see, e.g., Müller 2010, 322).6

▸ 211 CONCLUSION

This is not surprising, given that Waddington’s epigenetics and, in particular, his investigations of major shifts in morphological body plans are considered to be a major influence for contemporary evo-devo. However, this link between evo-devo (especially evolutionary novelty) and epigenetics becomes much less clear if one gives epigenetics a modern reading. This is the case because in evo-devo the idea of evolutionary novelty arises from a comprehensive comparative research program (Minelli 2009, 2015) that rests on the concepts of morphological form (Müller and Newman 2003; Laubichler and Maienschein 2009), body plan (Raff 1996, 30; Jenner 2006), and, most important, homology (Rutishauser and Moline 2005; G. Wagner 2014). Homology in particular is necessary to show that a new trait—usually a novel form, such as bird wings, bipedalism, eyes, and so forth—is, as Mary Jane West-Eberhard (2003, 198) calls it, “qualitatively distinct” from existing traits.7 Unfortunately, in contrast to evo-devoists, contemporary epigeneticists do not focus on morphological forms and body plans. Even more important and at least thus far, they seem to be little interested in carrying out comparative analyses in order to trace homologies (and taxonomic distribution of species’ traits). Against this backdrop, it becomes conceptually ambiguous as to how epigenetic variation can be understood as qualitatively novel in evolution and thus, more generally, how modern epigenetics can contribute to evo-devo research on phenotypic novelties. In addition, it is anything but clear if the concepts of (epigenetic) innovation and novelty are synonymous and whether they rest on the same assumptions of similarity. Besides these issues related to identifying the subject area of a philosophy of epigenetics, the challenge of locating the field itself within scientific discourse is also highly important if it is to thrive. While the field’s limits relative to other approaches and debates in philosophy of biology seem to be as fluid and diffuse as those between epigenetics and other disciplines, the field’s limits relative to biology itself can be defined more clearly. Although philosophy of epigenetics may refer to and even partake in biological debates, it should not be measured by biologists’ standards but should remind itself of its very own conceptual and methodological traditions. For example, it should seek to enrich and enlighten biological debates on competing research agendas by drawing on traditional conceptual frameworks and theories of scientific explanation and change, to uncover conceptual challenges and inconsistencies, and to point to unquestioned methodolog-

212 ◂ CONCLUSION ical, epistemic, and even anthropological premises. But, again, although all of these issues may be of interest for epigeneticists as well, they should be addressed by making use of the philosopher’s toolbox, not the empirical scientist’s. Accordingly, a philosophy of epigenetics should first and foremost seek to go beyond biology by enriching philosophical discourse. This book has offered various examples of how epigenetics can be utilized in order to achieve this very goal, as it has focused primarily on those issues, like the structure of scientific explanation, modeling, and the concept of causality in the biological sciences, which recently have engaged the attention of philosophers of science more generally. In short, the research program of a philosophy of epigenetics should not merely be understood as a philosophy for biology that only clarifies epigenetics’ history and its conceptual, methodological, and theoretical challenges. Such a framework would be in danger of gradually merging with its research subject and thus losing track of philosophical research questions and methodologies. Rather, I am calling for a biologically informed yet genuine philosophical domain participating in relevant current discourses in both philosophy of science and biology. By focusing on epigenetics as a case study (or a series of case studies), this new field can stimulate more general philosophical debates concerning biology and even other special sciences and, in return, act as a mediator between these general debates and particular theoretical and/or practical issues in epigenetics and related fields. For example, by concentrating on the appropriateness and value of epigenetic explanations, as suggested in these pages, it can elucidate the limits of other forms of scientific explanation, like genetic explanation. In other words, it can make sense of the still insufficiently investigated “nonsense part” highlighted in the title of Sheldon Krimsky and Jeremy Gruber’s (2013) edited book Genetic Explanations: Sense and Nonsense. If we aim at locating a philosophy of epigenetics within philosophy, we also have to specify its relatedness to other accounts in philosophy of biology, like developmental systems theory. Similar to epigenetics, this theory emphasizes the significance of nongenetic factors in development and evolution. Developmental systems theory has become particularly crucial for developing a view of causality as a reciprocal influence in niche- construction theory (see Sterelny 2001; and Laland et al. 2009, 2011). How- ever, although developmental systems theorists often refer to findings in epigenetics in order to visualize and/or underline their thesis of causal par-

▸ 213 CONCLUSION

ity of genetic and nongenetic factors in development, it remains an unsettled issue whether philosophers of science and biologists interested in epigenetic phenomena share, for example, developmental systems theory’s rejection of the notion of genetic information (Oyama 1985). In addition, it can be asked whether they really adopt the developmental systems theorists’ view of evolution as the construction of entire developmental systems and whole ecosystems, respectively.8 At this point in time, epigenetics has a lot of catching up to do compared to other rather new biological fields, like developmental niche construction and evo-devo. In contrast to epigenetics, research in both of these two fields is already deeply intertwined with philosophical debates on their conceptual and methodological foundations. This might be due to the fact that central ideas, concepts, and methodologies in these fields had already been developed by researchers during or (in the case of evo-devo) before the 1980s. While developmental niche constructionists primarily share the philosophical framework of developmental systems theory on how to conceptually and explanatorily approach development, heredity, and evolution, evo-devo’s historical and philosophical foundations have been carefully examined by philosophers of biology and biologists since about 2000.9 If epigenetics wants to continue to thrive and become a research program fully integrated into the theoretical potpourri of modern biology, it has to move more seriously toward a philosophy of epigenetics. Let me conclude by locating this new philosophical domain in a historical context. From 25 to 28 November 2001 a conference, Contextualizing the Genome: The Role of Epigenetics in Genetics, Development, and Evolution, was held in Ghent, Belgium, by the “Evolution and Complexity” research community. Its participants included, among others, biologists, historians, and philosophers such as Eva Jablonka, Marion Lamb, Linda Van Speybroeck, Vincent Colot, Gerd B. Müller, Csaba Pál, James Shapiro, Scott Gilbert, James Griesemer, William Wimsatt, and Michel Morange. This conference resembled the interdisciplinary ideal of Waddington’s Lake Como meetings, as the subject of epigenetics was to be approached from an experimental, historical, and conceptual perspective (for a report, see Jablonka et al. 2002). However, the conference’s comprehensive proceedings, published as the December 2002 issue of the renowned Annals of the New York Academy of Sciences, did not invoke a lasting shift toward philosophical issues of epigenetics. So, what went wrong in the early 2000s?

214 ◂ CONCLUSION

It seems that at the turn of the millennium the time had not yet come for a philosophy of epigenetics. In the years since, the knowledge of causal relations and mechanisms underlying epigenetic complexity has radically increased as epigenetics has started to boom and the number of articles published has mushroomed. In addition, biologists’ and philosophers’ common call for expanding the neo-Darwinian framework of biological theory, by incorporating epigenetics as well as other research programs, has gradually increased in volume. In this extended evolutionary synthesis, movement biologists show growing interest in discussing with philosophers the conceptual foundations of their research (see Baedke 2017a). Even more important, during this time philosophers of science have gradually turned toward issues concerning investigations of complex phenomena in the special sciences. In particular, the structure of biological explanation, as well as conceptualizations and methodologies in biology, have increasingly aroused philosophers’ interest. In light of these recent developments, today the time is ripe to give philosophy of epigenetics a new chance—a golden opportunity it surely must not miss.

▸ 215

NOTES

INTRODUCTION 1. For a historical overview of the different roles the gene played in twentieth- century biology, see the work by Rheinberger et al. (2015). For the disappointing results of the so-called Human Genome Project, see chapter 1. 2. Throughout this book the terms “dependency (relation)” and “relation (ship)” are used interchangeably. They can refer to a (degree of) connectedness between the properties of entities (e.g., DNA, cells, organisms) or events (e.g., transcription, mitosis, reproduction) in the world. For example, these dependencies can be said to be causal (see chap. 3) or constitutive (see chap. 4). In addition, the terms above are also used to describe explanatory dependencies or relations expressed, for example, in a model, equation, generalization, law, and so forth, between an explanandum (a phenomenon that is to be explained) and an explanans (usually sentences or equations that should explain the phenomenon). 3. For example, model organisms may be considered exemplary of higher taxons (Bolker 2009; Ankeny and Leonelli 2011). Due to this feature they can facilitate surrogate reasoning and extrapolation in contrastive explanation. NOTES TO PAGES 12–22

4. For further historical “stimuli” (and also some sociological and anthropological ones), see the conclusion of this book.

1. HOW EPIGENETICS DEALS WITH BIOLOGICAL COMPLEXITY 1. For early studies in philosophy of biology, see, for example, the works of Schaffner (1967a, 1967b, 1969), Wimsatt (1974, 1976), and Hull (1974). For an even longer tradition of philosophy of biology starting in the early twentieth century, see the concluding chapter in this book. 2. Far-from-equilibrium open systems, like organisms, have to work against the second law of thermodynamics, which says heat cannot be perfectly transformed into work. Thus, it holds that the amount of free energy is constantly decreasing while entropy (as a measure for the amount of dissipated energy) is constantly increasing. Against this tendency the organization of open systems allows them to import matter rich in free energy to stay at a low entropy state far from equilibrium, for example, in organisms at a state far from death (Schrödinger 1944). Note that also some nonliving systems, such as flames, are far from-equilibrium open systems (Nicolis and Prigogine 1977). 3. Among these authors, abstraction refers to the amount of detail omitted from an explanation. In what follows, I treat abstraction as following this line of thought (see esp. chap. 3). I do not address the idea, prominent especially in the history of empiricism, that properties, concepts, and numbers result from abstraction from concrete instances (i.e., they are abstract objects). More- over, abstraction should be distinguished from idealization, which describes the idea as including in a model assumptions that are known to be false with respect to a given phenomenon (see Jones 2005). 4. A gene’s genomic context may be understood as the network of regulatory genes involved in regulating the expression of the particular gene, for example, through coding for repressor and activator proteins or microRNAs (Jacob and Monod 1959; He and Hannon 2004). The latter may modulate gene activity on the posttranscriptional level. 5. For Waddington’s research and biographical writings, see, for example, works by Robertson (1977), Hall (1992), S. Gilbert (2000, 2012), Slack (2002), Van Speybroeck (2002), and Peterson (2011, 2016). 6. Notice that canalization is, in fact, not the exact opposite of phenotypic plasticity, since, on the molecular level, developmental robustness of a trait can

218 ◂ NOTES TO PAGES 24–29

be understood as a product of the plastic responses of developmental processes compensating for environmental or genetic variation. Thus, both concepts are relative to the level of description chosen. 7. On the history of the notion of epigenetics, see, for example, works by Wu and Morris (2001), Jablonka and Lamb (2002), Van Speybroeck (2002), Holliday (2002, 2006), Haig (2004, 2012), Ptashne (2007), and Deichmann (2016a). 8. For example, Wu and Morris (2001, 1104) define epigenetics as “the study of changes in gene function that are mitotically and/or meiotically heritable and that do not entail a change in DNA sequence.” On definitions of epigenetics in the 1980s and 1990s, see especially the work of Jablonka and Lamb (2002, 86–88). 9. For a more detailed discussion of these regulatory epigenetic processes involved in development and heredity, see works by Jablonka and Lamb (2005, 2010) and Lamm (2014). 10. For the ecological-evolutionary dimension of epigenetics, see also a special issue in Genetics Research International (for editorial, see Schrey et al. 2012) and chapter 2 in this volume. 11. In other words, genetic assimilation narrows the range of variation from a plastic trait to a fixed or canalized one. If, contrarily, plasticity is stabilized through selection of hidden variants that can direct the phenotype in more than one direction, this process is called “genetic accommodation.” Accord- ing to the related so-called “Baldwin effect,” the environmentally induced phenotype is genetically fixed by a new mutation, not by preexisting genetic variation. For a review of both concepts, see the work of S. Gilbert and Epel (2009). 12. On the “genes as followers” view, see also works by Gissis and Jablonka (2011) and Piersma and van Gils (2011). For a discussion of this view, see the article by Schwander and Leimar (2011). 13. On the extended evolutionary synthesis, see also chapter 2. The less common term “post-Darwinian synthesis” is used by some authors in a historically imprecise manner, as they understand it as “postmodern synthesis” (e.g., Huang 2012a). 14. On Lamarck’s comprehensive theory of inheritance, as well as the historical context in which Lamarck’s then widely shared view of the inheritance of acquired characteristics arose, see, for example, the works of Bowler (1983) and Burkhardt (1995).

▸ 219 NOTES TO PAGES 30–37

15. Note that Weismann’s view on heredity was much more complex than these lines might suggest; see, for example, the work of Rheinberger and Müller- Wille (2012). Thus, the discussion that follows should be read as dealing primarily with ideas defended by advocates of Weismannism (e.g., Mayr 1961), rather than by Weismann himself. 16. For a survey of studies of (different forms of) inheritance of acquired characteristics from the twentieth century, see works by Landman (1991), Jablonka et al. (1992), and Jablonka and Lamb (1995). 17. Skinner (2008) and Skinner et al. (2010) have argued that one should conceptually distinguish between a “multigenerational phenotype” and a “transgenerational phenotype.” Only those environmentally induced phenotypes should be considered as epigenetically inherited, transgenerational pheno-

types that (in the absence of repeated exposure in F1–F3) are transmitted to

at least the F3 generation. Effects in F1 and F2 should not be treated as inherited traits but as directly induced traits, because environmental exposure of a

gestating female includes direct exposure of the F0 female, the F1 generation

embryo, and the germ line that will generate the F2 generation (i.e., primordial germ cells). Although this conceptual distinction is crucial for understanding the difference between multigenerational plasticity and transgenerational epigenetic inheritance, it is of minor importance to the issue at stake, that is, epigenetically mediated developmental directionality in evolution. 18. For more examples, see, for example, works by Jablonka and Raz (2009), Villota-Salazar et al. (2016), and Wang et al. (2017). 19. For discussions of epigenetics’ “Lamarckian dimension” in popular science, see, for example, works by Balter (2000), Young (2008), Kegel (2009), Blech (2009), Carey (2011), and Spector (2012). 20. Note that although some of the above examples (including somatic induction followed by epigenetic germ-line inheritance) cast doubt on the adequacy of the distinction between soma and germ line, that is, Weismann’s model, these examples merely concern changes in the epigenome of the germ line, not changes in the DNA sequence of germ-line cells. Thus, they do not directly oppose the central dogma and molecular Weismannism. Whether induced variations in the epigenome of the germ line can induce genetic mutations in a directed manner is a matter of debate and is not discussed here (see Jablonka and Lamb 2005; and Merlin 2010). 21. At this point, I am merely discussing biological theories, not the related philosophical account of developmental systems theory (see chaps. 3 and 5).

220 ◂ NOTES TO PAGES 37–46

22. Evolvability is the capacity of a system to generate heritable, adaptive phenotypic variation and thus to evolve in evolution. 23. The interacting cascades of constructions and selections investigated by niche construction theorists have also been described as “cycles of contingency” (Oyama et al. 2001). 24. On related ideas, see the work of Piaget (1978).

2. CHALLENGES OF EPIGENETICS IN LIGHT OF THE EXTENDED EVOLUTIONARY SYNTHESIS 1. Some authors understand this theoretical change as a replacement (rather than an expansion) of the modern synthesis; see, for example, the article by Noble (2015). For a comprehensive overview of the extended evolutionary synthesis debate, see the volume edited by Pigliucci and Müller (2010b). 2. For similar views, see, for example, works by Hall (2000) and Minelli (2010). For a critical discussion, see Pigliucci’s (2008) article on the proper role of population genetics in modern evolutionary theory. 3. Müller and Pigliucci (2010) have offered comments on Craig’s position. In addition to the three critiques outlined here, it has been assumed that advocates of an extended evolutionary synthesis (and of epigenetics) are driven by extrascientific motives (see Haig 2011 and Welch 2017). 4. For recent studies that have addressed this issue, see, for example, the work of Herman et al. (2013) and Rivoire and Leibler (2014). 5. For a review of a number of highly stable epigenetic transgenerational effects, see the article by Heard and Martienssen (2014). In addition to the issue of stability, the frequency of transgenerational epigenetic inheritance in nature has been questioned (see, e.g., West-Eberhard 2007). 6. In RNA interference, small RNA molecules bind to other specific messenger RNA (mRNA) molecules and either increase or decrease their activity, respectively (e.g., by mRNA degradation, which prevents mRNA from producing a certain protein). In addition, these small molecules can regulate the activity of genes directly at the DNA level (e.g., by inducing DNA methylation). RNAi is involved in directing development and defending cells against parasitic genes. For a detailed description of RNAi-mediated gene silencing pathways, see, for example, the work of Siomi and Siomi (2009). 7. Chapter 3 addresses these two studies in detail. 8. Of historical interest is the fact that this criticism of the statistical signifi-

▸ 221 NOTES TO PAGES 46–52

cance of results in epigenetics resembles Ronald A. Fisher’s criticism of Gre- gor Mendel’s results with the garden pea. On the Fisher-Mendel controversy, see the monograph by Franklin et al. (2008). 9. It should be mentioned that this distinction between artificial and natural epigenetic systems is, of course, not a clear-cut one. In biology, artificiality comes in degrees. Thus, we also may find it in ecological field studies, for example, in the practices of choosing a suitable population or ecosystem. In this work, the term “artificial conditions” refers to research practices based on physical manipulation. 10. “EpiRILs” stands for “epigenetic recombinant inbred lines.” These lines are near isogenic lines that differ in a few DNA sequences but have contrasting DNA methylation profiles. They are usually produced by using parents with few DNA sequence differences but various differences at the epigenetic level. Then, a specific inbreeding procedure is conducted for multiple generations in order to produce a panel of lines. Using epiRILs permits quantification of the impact of epigenetic variation (in natural and experimental populations) on complex traits by ruling out confounding effects of DNA sequence poly- morphism. The different methodologies used in ecological epigenetics are accompanied by a number of new approaches seeking to model the impact of heritable epigenetic variations on population dynamics (see, e.g., Day and Bonduriansky 2011; and Klironomos 2013). Often these models are modifications of classical population genetic models (see Tal et al. 2010; and Geoghe- gan and Spencer 2012). 11. For philosophical literature on SEM, see, for example, the work of Freedman (1997), Pearl (2000), and Woodward (2003). 12. In plants, epigenetic variation can be induced by environmental stresses, including herbivory damage. This variation is not necessarily heritable. 13. In addition, this modeling procedure helps to minimize the total effect of the negligible causal background on the relevant variables. This is a crucial method for estimating the influence of causally relevant variables (included in the model) on the causal association under study. In the case of Herrera and Bazaga’s (2011) work, it allows showing that only 9 percent of multilocus epigenetic variation could be explained by herbivory-related genetic variation. From this finding, they derived the final causal explanation that some values of the epigenetic variation variable likely are related to herbivory independent of genetic variation. Thus, genetic variation would not be a common cause of the association observed.

222 ◂ NOTES TO PAGES 52–57

14. This list does not present an exhaustive catalog of methodological guidelines. It should instead provide the reader insight into the heuristic spectrum of manipulation-based reasoning. 15. In the case discussed, this method leads to new questions, such as “What would happen if all environmentally labile epigenetic variation were controlled?” This special question could be answered by a removal experiment that fixes the herbivory damage variable at value {no herbivory damage} and

thus eliminates its effect on epigenetic variationafter generation F0 (e.g., by using fences in a field experiment). Such an experiment would be similar to the study of RNA interference in C. elegans by Vastenhouw et al. (2006), described earlier. This experimental approach is discussed in more detail in chapter 3. 16. This account by Omri Tal, Eva Kisdi, and Eva Jablonka (2010) considers trajectory information about the number of opportunities for epigenetic reset between generations and assumptions about environmental induction to determine heritable epigenetic variance and epigenetic transmissibility. 17. Of course, as natural selection has no purpose, Mayr rejects Aristotle’s framework of a teleological metaphysics. Thus, ultimate causes do not work in teleological, end-directed systems but do work, following Pittendrigh (1958), in so-called teleonomic systems, which are only apparently purposeful. 18. Baker’s distinction was used, for example, by Lack (1954). However, Mayr (1961) makes reference neither to Baker nor Lack when discussing the proximate-ultimate distinction. Tinbergen (1963) lists distinct questions he believed should be asked in behavioral studies (see also Tinbergen 1951). 19. Later I discuss the proximate-ultimate distinction as one between different kinds of explanations, following Amundson (2005), Calcott (2013a), and Scholl and Pigliucci (2014) and not as one between ontologically distinct causes, as the Weismannian reading above might suggest. Laland et al. (2013a) have framed this distinction in terms of different ontological classes of causes working in ontogenetic and phylogenetic processes. 20. See, for example, the work of R. Francis (1990), Dewsbury (1992, 1999), Sterelny (1992), Beatty (1994), and Ariew (2003). 21. In more recent decades, the proximate-ultimate distinction has been applied by numerous researchers. For example, in evolutionary biology, see the statement of Wilson (2000, 23) and in behavioral ecology, see that of Morse (1980, 92–95). In evolutionary psychology it has widely been accepted (see, e.g., Daly and Wilson 1978; and Crawford 1998). In addition, the distinction has

▸ 223 NOTES TO PAGES 57–68

been discussed continuously in human sciences (e.g., on human cooperation in the work of Marchionni and Vromen [2009], and on developmental psychology in the work of Lickliter and Berry [1990]). 22. See the article by Laland et al. (2013a) for a defense of their position against some of these critical points. 23. For discussions of lineage explanation, see, for example, works by Hochman (2013), Brown (2014, 2015), Keijzer (2015), and Tebbich et al. (2016). 24. This does not mean that (PC) is of no interest for a (PC*) investigation in developmental biology. Rather, answering (PC) questions can be understood as an additional strategy of mechanistically specifying the developmental pathway outlined by a (PC*) explanation. It offers a series of time slices of the complex network of dependencies that characterize the developmental pathway under study. 25. In addition to the authors quoted here, see, for example, the work of Amund- son (2005) and Haig (2013). 26. Please note that epigenetic explanations may relate things synchronically as well. These (PC)-like constitutive explanations are discussed in chapter 4. However, as described below, it is those epigenetic explanations addressing (PC*) questions and relating things over time that carry information relevant for evolutionary biology. 27. In addition, in contrast to what Dickins and Rahman (2012) suggest, this idea of establishing novel population-level explananda, previously neglected in evolutionary biology and thereby guiding this field in a new direction, is a sensible scientific enterprise that does indeed produce heuristically “fruitful, falsifiable questions” (Dickins and Rahman 2012, 2917); see chapter 3. 28. In a similar manner, Scholl and Pigliucci (2014) and Otsuka (2015), for instance, have tried to make sense of Mayr’s distinction or reinterpret it. 29. Boniolo and Testa (2012) argue that epigenetics helps us to answer two old philosophical questions: “What makes a living being unique?” (synchronic identity) and “What makes it remain the same over time, given that certain of its properties change?” (diachronic identity). I return to this issue of identity in the concluding chapter of this volume. 30. The emerging field of systems biology includes a large number of heterogeneous biological and biomedical approaches that seek to investigate properties of biological systems like cells, tissues, and organisms arising due to complex interactions. Typically these systems and networks, like cell signaling or metabolic networks, are investigated by explicitly taking a holistic stance.

224 ◂ NOTES TO PAGES 71–73

Since this field makes extensive use of mathematical and computational modeling techniques, it is often considered to be promoting a new physical- ization of biology (see, e.g., Calvert and Fujimura 2011). On the differences between systems biology and molecular biology, see the work of Backer et al. (2010). 31. However, Waddington’s drawings of this landscape have only recently been subjected to critical scrutiny in history and philosophy of biology. See, for examples, works by Gilbert (1991), Caianiello (2009), Fagan (2012b), Baedke (2013), and Baedke and Schöttler (2017). 32. On process philosophy, see also chapter 4. 33. This view is founded in Whitehead’s ontology. Waddington (1970a, 113–14; emphasis added) noted that every “event,” that is, the ontological unit in the early Whitehead, “is impossible to grasp, conceptually, in its totality” due to its relatedness “to every other event in the universe” (see also Wad- dington 1977c). Thus, in order to account for this highly complex “spider’s web” (Waddington 1970a, 114) of relationships, Waddington adopts a rather unorthodox form of scientific reference making: pictures. They are introduced as a first point of reference in scientific research, which should help investigators to grasp “events” more accurately than previous conceptual approaches alone. Interestingly, this and similar epistemic virtues of pictures in scientific practice have become the subject of analyses in philosophy of science. Scientific images have been understood as stimuli for visual thinking (Bredekamp 2005) and as tools for scientific reasoning (Kulvicki 2010). Furthermore, the capacity of visual representations to bear truth and thus to justify or confirm scientific hypothesis (Perini 2012), as well as their ability to facilitate (or impede) communication (Doyle 2007), has been emphasized. Some have argued for the indispensability of the visual in scientific argumen- tation (Griesemer 1991; Carusi 2012). 34. Besides pursuing an interest in catastrophe theory (especially in the work of René Thom and Christopher Zeeman), Waddington was always on the lookout for other mathematical accounts suitable for providing a formalized version of his EL images that would be applicable to developing systems. In the late 1960s, Dusa McDuff, Waddington’s daughter and a trained mathema- tician, met Israel Gel’fand in Moscow. She introduced her father to Gel’fand and Tsetlin’s ravines method by translating for him their Russian-language papers (McDuff, personal communication). In his late-career work Wad- dington closely tied this approach to his attractor surface version of the EL.

▸ 225 NOTES TO PAGES 73–84

35. Waddington does not describe in detail how the alterations of a complex system’s dynamics should be carried out. We may, however, understand this procedure as the experimental manipulation (e.g., through environmental induction and/or genetic manipulation) of a biological system. It could also be an exploration of the model parameters, followed by modeling-informed experimentation. 36. On the history and usage of Waddington’s models in the life sciences and for a comprehensive list of EL approaches in stem cell biology, see Baedke (2013). 37. These gene networks comprise all concentrations of interacting transcripts of genes involved, as well as proteins and other molecules somehow related; see also chapter 4. What EL models of GRNs do not consider, and what was unknown in Waddington’s time, are dedicated cell memory systems (e.g., copying of methylation patterns, or replication of small RNAs) that confer potential stabilities and dynamics in GRNs, which are conditions that would not be conceivable without these memory systems. 38. This includes that the guiding “visual corrective” of the EL biases modeling practices. Stem cell biologists, for example, choose to develop toy models of a specific type of gene circuit, that is, two genes, which can be visualized in an EL state space (e.g., circuits with N > 2 genes are not so easily visualized). 39. However, if this criticism is directed against Jablonka’s work in epigenetics in particular, it does not perfectly match, since Jablonka also discusses multifactorial, multicausal network models in epigenetics (see Jablonka and Lamb 2005, 121; Jablonka and Lamm 2012). See also table 2.1. 40. On the time frame in which a GRN can change, see Huang’s (2012a, 152) article. The view of the time-invariant and stable structure of GRNs has been defended in particular by Eric Davidson (e.g., 2006). Davidson claims that GRNs allow for no flexibility in embryonic development, as the latter is “hard- wired” in the genome. Differences between Davidson’s and Huang’s views with respect to the stability characteristics of GRNs, as well as how they affect development, are discussed in chapter 4. 41. Despite calling into question the possibility of transgenerational inheritance of “acquired characteristics” through epigenetic pathways, Huang does, however, admit that environmentally induced innerorganismic (intercellular) epigenetic inheritance has a certain Lamarckian dimension. He calls this process “weak Lamarckism” (Huang 2012b, 72).

226 ◂ NOTES TO PAGES 84–92

42. Not all studies of GRNs take this Waddingtonian perspective. For reductionist and deterministic approaches to investigating these networks, see chapter 4. 43. This new synthesis has been labeled “evolutionary systems biology.” In a related manner, Brigandt (2013) and O’Malley et al. (2014), for example, have described how in systems biology different kinds of explanations, causal- mechanistic, on the one side, and law-based or mathematical, on the other, are integrated in a modeling approach that encompasses a number of different levels of organization. 44. On the traditional philosophical model of theory reduction, according to which knowledge from different fields can be logically deduced from a lower- level theory, see, for example, works by Nagel (1949) and Oppenheim and Putnam (1958). 45. On integration of in silico, in vitro, and in vivo accounts of protein folding, see the work of O’Malley et al. (2014). In this case, folding landscapes similar to epigenetic landscapes are used as representational tools. 46. The pluralist idea of multilevel integration (in contrast to theory reduction) adopted here has been highlighted especially by Mitchell (2003, 2009). For an early defense of this position in contrast to the prevailing model or theory reduction, see the work of Darden and Maull (1977). 47. It seems that both trading zones face the particular challenge of integrating explanations involving physical principles with other traditional causal or mechanistic approaches of explanations. On the assumptions hindering integration of physics-based explanations in evo-devo, see the work of Love and Lugar (2013).

3. CAUSAL EXPLANATION 1. Throughout this chapter and chapter 4, please keep in mind two things. First, given epigenetics’ heterogeneity and our Wittgensteinian view of epigenetics as a cluster concept (see the introduction), explanatory practices discussed should be treated neither as being characteristic of epigenetics as a whole nor as appearing solely in epigenetics. Instead, these explanatory features may arise only in particular subfields of epigenetics, as well as (in the same or a similar manner) in related fields, such as evo-devo or systems biology. Only a comprehensive description of the various sets of these features in different epigenetic studies gives us an idea of characteristic epigenetic explanations. Second, I do not treat molecular biology and cell biology

▸ 227 NOTES TO PAGES 92–104

as disciplines that can be precisely distinguished and thus separated from one another, as studies in the latter commonly investigate the molecular relations and processes underlying cellular phenomena. Following Morange (2000a, 2), I treat molecular biology “not [as] a new discipline, but rather [as] a new way of looking at organisms as reservoirs and transmitters of information”—a view that “opened up possibilities of actions and intervention that were revealed during the growth of genetic engineering.” On the status of molecular biology, see also the work of Olby (1990). 2. According to Woodward, interventions only need to be logically possible. For a critical discussion of the metaphysical foundations of the concept of intervention, see the work of Reutlinger (2012). 3. The concept of invariance has to be distinguished from that of “stability” (Mitchell 1997, 2000, 2002; Woodward 2003, 295–302; 2006, 2010). The stability of a generalization refers to the number of changes in values of variables not included in the generalization, that is, background variables. Stabil- ity thus serves an important explanatory role in explicating the reliability and extrapolability of a scientific explanation. 4. For a similar joint account of contrastive explanandum and difference-making explanans, see the work of Northcott (2013). 5. Moreover, the interventionist account enables one to make sense of omissions as causes (such as when the failure to catch a baseball causes your team to lose) and of absences as causes (when fastening a safety belt causes you not to die during a roller coaster ride), as biologists commonly do. What is more, it resolves the problems of explanatory irrelevance (i.e., when irrelevant information is not excluded from an explanans), which has been a major problem for Hempel’s account. See the monograph by Salmon (1984), as well as that of Woodward (2003, 196–203). 6. On RNA interference, see chapter 2, note 6. 7. For a thorough discussion of Russo’s rationale of variation, see Russo’s (2009) monograph. Russo (2012, 136) claims that this rationale is central to experimental contexts as well: “Variation not only guides causal reasoning in observational settings, but does so also in experimental ones.” This claim is questionable, at least for disciplines such as epigenetics, in which most of the (co-)variation known is brought about in the lab by manipulation in the first place. 8. The notion of population in this case refers to a statistical population, but I understand it as being closely linked to the general notion of (natural) popu-

228 ◂ NOTES TO PAGES 105–108

lation used by biologists. See, for example, the discussion of counterfactuals in the paper by Herrera and Bazaga (2011, 1683, 1685). 9. The need to search for auxiliary causal evidence by accessing knowledge about invariance has been discussed in other observational research fields as well, in epidemiology (Lilienfeld and Lilienfeld 1980, chap. 12), for example. 10. Please note, however, that there has been an ongoing debate, starting with Bertrand Russell (1912/1913), on whether the idea of causation is compatible with fundamental physics at all, since, in some cases, in order to determine the state of a small system we have to include in our model the entire universe as its “cause.” As Pearl (2000, 350) notes, “If you wish to include the whole universe in the model, causality disappears because interventions dis- appear—the manipulator and manipulated lose their distinction.” See also the essay by Woodward (2007a, 93). 11. In his treatise on the conservation of energy, Helmholtz (1847) seeks to rule out the possibility that in muscle metabolism there are vital forces necessary to move a muscle. He shows that no energy is lost in muscle movement. 12. Emergence has often been interpreted as (synchronic) supervenience (see McLaughlin 1997; and Kim 1999, 2006). 13. Kaiser (2011, 463) calls reductive molecular or genetic explanations that follow this view “fundamental-level reduction.” For example, Jacob and Monod (1959) were following this view when they introduced a causal hierarchy of the genome, according to which regulatory genes come first, structural genes second, and developmental phenomena third. 14. As argued in chapter 4, the mechanistic account of explanation is, in fact, not fully compatible with the interventionist explanation, as it cannot conceptualize all explanatory dependencies in epigenetic models of robust dynamic systems. 15. Russo and Williamson (2007, 11) argue in a similar manner that it is necessary (though not sufficient) to describe a mechanism in order to establish a causal claim. 16. Woodward (2003, 250) emphasizes that “invariance under at least one testing intervention (on variables figuring in the generalization) is necessary and sufficient for a generalization to represent a causal relationship or to figure in explanations.” Craver (2006, 373), however, claims that “it is insufficient merely to describe components and their activities,” including their invariant dependencies. For example, he refers to the Hodgkin-Huxley model of the action potential and notes that explaining when (i.e., under which condi-

▸ 229 NOTES TO PAGES 108–110

tions) action potentials (the effect) are not produced necessarily includes the notion that one is also “able to say why they are not produced under these conditions” (Craver 2006, 368; emphasis added; see also Craver 2008, 1031). 17. However, we also find noncontrastive interpretations of the notion of difference making and fundamental mechanistic detail, especially outside philosophy of biology. With respect to physics, for example, Ney (2009) defends the view that changes in causally related events are to be investigated by looking at interactions at the most fundamental level, since it is these interactions that tell us “why these events occur—what is responsible for what in a causal sense” (Ney 2009, 741; emphasis added). Please note that Ney does not explicitly defend a concept of causal mechanism. Rather, she claims that the nature of difference-making relations is to be found by investigating our theories of fundamental physics. These theories’ laws, however, can be interpreted as describing “fundamental mechanisms” that necessarily undergird every difference-making event and, if cited in an explanans, answer mechanistic “how” or “why” questions: “Physics does not just provide us with a comprehensive account of what exists in the universe but an account as well of why these events occur. . . . These laws single out those features of systems that are causally relevant to the production of effects” (Ney 2009, 757). 18. This interest in backing up a causal dependency by a description of a mechanism has a long history in biology. For example, in an investigation on the physiological effect (i.e., death) of curare, Bernard ([1865] 1949, 157) seeks to “get . . . an idea of the mechanism of death by curare.” In other words, knowing what curare does seems to be insufficient to explain. In contrast, knowing why it does what it does is thought to be a necessary supplement. 19. Below I give the notion of “gene” a narrow reading as “DNA sequence” and not a broader one as “DNA plus its domain RNAs, proteins, etc.” I choose this definition even though I am fully aware of the threat of building up a reductionist straw man that is easy to attack. However, if we accept the broader concept of “gene,” the thesis of genetic reductionism (Sarkar 1998; see also Kaiser 2011, 463–64), the one I argue against below, simply turns into a version of molecular reductionism (i.e., all biological phenomena can be reduced to facts about molecular entities). In fact, epigeneticists are primarily interested in the causal effects of variations in non-DNA molecules and their explanatory autonomy from DNA variations, and thus in distinguishing these things from the thesis of genetic reductionism (see, e.g., Noble 2008b). In other words, with respect to the causal autonomy of epigenetics, genetic

230 ◂ NOTES TO PAGES 110–125

reductionism is not a straw man position never realized in practice but one that is a real threat to epigenetic explanation. The general problem of ontologically identifying a gene is discussed further in this chapter. On determining a gene’s and an “epigene’s” boundaries, see chapter 4. 20. In contrast to Craver’s (2006, 2008) work and similar to Ney’s (2009) account (see above, chap. 3, note 17), Waters does not explicitly defend a concept of mechanism. However, as I later show here, his account argues for the primary status of more fundamental causes—above all, genes—in explaining dependency relations in living systems. Henceforth, I will understand this set of primary causal agents, if cited in an explanans, as mechanistic detail that is necessary to address an explanandum. 21. On other criticism against the “genes as primum movens” view, see the work of Keller (2000b), Morange (2001), and Moss (2001). 22. Again, “population” is understood as “statistical population.” The actual effect is the difference of a property in the population. 23. See chapter 2, note 10, for a brief discussion of epiRILs. See also works by Reinders et al. (2009), Johannes et al. (2009), and Cortijo et al. (2014). 24. I return to this critique of the concept of causal power or causal importance in chapter 5, where an alternative, non-ontological account of hierarchizing explanations is presented. 25. This example is based on studies of non-DNA variation in (heritable) complex traits, that is, on (transgenerational) changes in DNA methylation and histone modification states in the absence of changes in DNA sequences (Johannes et al. 2008; E. Richards 2008; Eichten et al. 2011, 2013). For reasons of simplicity I exclude from this case, among others, the causal influence of

environmental variables EV1–EVn on E and P. 26. Histone deacetylases are often involved in gene silencing at the transcriptional level, that is, they regulate the information flow between the DNA and mRNA (Delcuve et al. 2012). Thus, the “place” where silencing occurs via histone deacetylases differs from that of silencing via RNAi; the latter is known to initiate posttranscriptional silencing, for example, sequence-specific inhibition or degradation of mRNAs (Filipowicz et al. 2005). 27. For other studies offering more mechanistic detail on “Why (G),” see the publications by Rechavi et al. (2011) and Houri-Ze’evi et al. (2016); see also Wang et al. (2017). 28. In addition, even if mechanistic information about the actual difference making of genes in a certain population is available, considering this infor-

▸ 231 NOTES TO PAGES 127–134

mation does not necessarily enhance the value of the overall explanation. I discuss this problem in detail in chapter 5. 29. Accordingly, whether a causal factor is considered less fundamental depends on the level from which mechanistic detail can, in principle, be collected. Thus, physical, biochemical, genetic, and even (causally more “upstream”) epigenetic relations may be considered to be more fundamental, mechanistic detail that does not necessarily have to be considered in an epigenetic explanation.

4. MECHANISTIC EXPLANATION 1. Although in the quote from the article by MDC (2000, 3) the authors seem to speak of entities as being organized in linear processes, their view should instead be understood as one describing mechanisms as complex systems with cyclic pathways having a number of discrete “steps.” 2. In most recent mechanistic accounts, the explanatory role of a mechanism’s organization is usually not specified (but see Bechtel 2011; Fagan 2012a; and Kuorikoski and Ylikoski 2013). 3. Roughly speaking, modularity means that distinct causal relations are in principle independently changeable; see, for example, the work of Hausman and Woodward (1999), and see also discussion later in this chapter. For a view of near-decomposability and modularity in mechanisms similar to Simon’s, see, for example, Steel’s (2008) monograph. For an even stronger concept of modularity in mechanisms, see works by Woodward (2002) and Menzies (2012). 4. Later I discuss some limitations of the claim that describing a mechanism of a phenomenon is consistent with this causal-interventionist framework of explanation. 5. This distinction also helps to illustrate the difference between causal mechanisms and mechanical processes à la Salmon (see Glennan 2009, 322–24): the notion of mechanism usually refers to causal capacities of things, that is, organized systems of parts or structures (e.g., a phenotype of a cell). In contrast, the notion of a causal process refers to a series of events spread over time (e.g., a molecule passing a membrane). 6. For example, Johnston (1992) and Ylikoski (2013) argue that the part-whole relationship of constitution (contrary to identity) refers to an asymmetric relation, since the whole does not constitute its parts. This contradicts Crav-

232 ◂ NOTES TO PAGES 136–139

er’s (2007, 153–54; Craver and Bechtel 2007) view of constitutive relevance as mutual manipulability, which is discussed later. 7. Please note that in many accounts of mechanistic explanation the target of the explanation is not fully explicated (see Fagan 2012a, 456–58). This means that the explanandum phenomenon could be a mechanism’s working (S Ψ-ing) or its outcome (i.e., its effects produced by the working mechanism). We have to hold it to be the former if we want to maintain the distinction between causal and mechanistic explanation, since assuming the latter turns interlevel mechanistic explanations into causal explanations tracking causal dependencies. 8. Exceptions, that is, discussions of mechanisms outside of cell and molecular biology, can be found with respect to ecology (Raerinne 2011), the social sciences (Kuorikoski 2009), astrophysical mechanisms (Illari and Williamson 2012), and whether natural selection is a mechanism (Skipper and Millstein 2005; Glennan 2009; Illari and Williamson 2010; Matthewson and Calcott 2011). 9. Other contemporary characterizations of (sub)cellular assemblies as being machinelike can be found in the work of, for example, Huo et al. (2010), Buck and Trus (2012), and Shi and Ha (2011). According to Shi and Ha (2011, 3459; emphasis added), “In many ways, these biological nanomachines work just like a machine we use in our daily lives. They use the energy stored in molecular fuels to perform biologically useful work, for example, making new genetic materials as is the case for the replisome. Like a sophisticated clock that contains many moving parts, biological machines are composed of many minuscule moving components as well.” 10. This development in molecular biology can be understood as one in which the notion of “machine” turned into a “dead metaphor” (Rorty 1989, 18). This means that although some molecular biologists, like Jacob (1974) in the quote already presented, draw analogies between human- made machines and molecular mechanisms (e.g., a factory and a cell), most often molecular mechanisms or particulate entities and/or activities are simply treated as members of a special ontological class of (e.g., non- human-made, self-reproducing, or living) machines. Even though the metaphorical content of this latter kind of redescription is not often recognized by researchers in the field, these dead metaphors can still influence and constrain the ways phenomena are conceptualized and investigated, as will be shown.

▸ 233 NOTES TO PAGES 139–145

11. For a more comprehensive list and samples of machine language that have been used in contemporary molecular biology to characterize similar assemblies as molecular machines, see the work of Nicholson (2013, 676). 12. For a detailed analysis of “machine” and “design” language in synthetic biology, see the work of Boudry and Pigliucci (2013) and Holm and Powell (2013). 13. On how the ontological assumptions underlying metaphors shape and bias the way we perceive objects under study and how we communicate scientific results more generally, see works by Lakoff and Johnson (1980, 1999). For evidence of the widespread use of ontologically loaded metaphors in biology, see, for example, works by Keller (1995, 2000b, 2002). 14. An additional argument is presented in Nicholson (2018), in which he lays out distinctions between the thermodynamic conditions under which living systems and machines operate. 15. Note that traditional atomistic machine views (mechanicism) did not necessarily regard living things as passive machines. Instead, as Riskin (2016) shows, many mechanicist views conceptualize living things as active, self-making machines. The concept of agency in mechanicism is not discussed here. 16. Bechtel and Richardson (1993) claim that phenomena are decomposed functionally and that systems are decomposed structurally. According to this view, biologists seek to localize functional subtasks of the overall phenomena in structural component parts (i.e., at lower levels) of the system. 17. Although Craver and Darden (2013, 16) acknowledge that mechanisms are not synonymous with machines, the authors seem to cherish the usefulness of machine reasoning. 18. As a consequence, Craver and Darden (2013, 15–16) address on a single page in their book In Search of Mechanisms the difference between the concepts of machine mechanism and causal mechanism. 19. Interestingly, in Bechtel’s recent work on mechanisms in complex systems, the most influential metaphor in classical mechanicism to represent living beings—the clock—assumes an important role (see, for example, Bechtel [2013a] and Bechtel and Abrahamsen [2013]). On the metaphor of clocks in mechanicism, see, for example, the works of Descartes ([1664] 1972, 113) and Locke ([1689] 1979, 331). 20. This idea of heuristic superiority has encouraged some authors to apply the machine view of mechanisms even to phenomena that do not seem to be mechanisms at all. For example, Matthewson and Calcott (2011, 754) seek

234 ◂ NOTES TO PAGES 146–153

to represent the ontologically rather diffuse process of natural selection “as though it is a mechanism.” 21. This idea of functional decomposition as opposed to structural decomposition in molecular biology fits with Illari and Williamson’s (2010) reading of decomposition. See also works by Bechtel and Abrahamson (2005, 433) and Craver (2007, 188). 22. Davies (2012, 45) holds a similar view: “We can point to a DNA molecule and say ‘There is the genome!’ Information (e.g., the instruction, ‘make protein X!’) flows outward (and upward in the length scale of its impact) from spatially localized segments of a specific one-dimensional structure—the ultimate source of genetic information. However, we will look in vain for any particular physical object within the cell that we can identify as ‘the epigenome.’” 23. Later Davies (2012, 45) adds, “In the case of epigenetics, there is no physical headquarters, no localized commanding officers issuing orders, no geo- graphical nerve centre where the epigenomic ‘programme’ is stored and from where epigenomic instructions emanate to help run the cell. The epigenome . . . is distributed throughout the cell. To be sure, the epigenome is manifested in particular structures (histone tails, nucleosome patterns, methylation patterns, chromatin packing . . . ), but it does not originate there. The epigenome is everywhere and nowhere.” Thus, Davies notes that the epigenome shows double instability: structural instability of the entities, which leads to a failure of structural decomposition, and, even more important with respect to the “functions-first view” of some new mechanists, instability of the causal activities, roles, or the “program” of the parts that contribute to the phenomenon. The latter complicates explanation in terms of functional decomposition in epigenetics. 24. On processual views in early twentieth-century biology (especially in organicism), see, for example, works by Peterson (2014, 2016) and by Nicholson and Gawne (2015). 25. On the relationship between process ontology, microbial symbiosis, and the issue of biological identity, see the concluding chapter in this volume. 26. In contemporary epigenetics, rate differences are often measured by means of differences in concentration. On explanations tracing concentration, see the work of Nathan (2014). 27. For a detailed discussion of various forms of reductionism in molecular biology, see the work of Sarkar (1992), and in the life sciences as a whole, see the work of Kaiser (2011).

▸ 235 NOTES TO PAGES 154–157

28. For a detailed critique of Davidson’s machine view of GRNs and development, see the work of Nicholson (2014a). 29. CRISPR is short for CRISPR/Cas9, which means “clustered regularly inter- spaced short palindromic repeats/CRISPR associated protein 9.” It is a system that combines a protein that can cut DNA with a guide RNA. On epigenome editing and “programmable transcription” through CRISPR systems, see the work of Thakore et al. (2016). 30. For the sake of the argument, I will adopt in the text that follows the new mechanists’ rather narrow but more influential reading of what a mechanism is—especially the ideas of parthood and interlevel constitution—rather than defend a wider concept built on processual perspectives. 31. (CR) results from the assumption that constitutive relevance can be understood by drawing an analogy to causal relevance. The latter is analyzed by Craver in terms of mutual manipulability: a component part (x Φ-ing) is causally relevant to a mechanism’s behavior (S’s Ψ-ing) if the latter can be manipulated by changing x’s Φ-ing and the component’s behavior can be manipulated by intervening on S’s Ψ-ing. If neither x’s Φ-ing nor S’s Ψ-ing can be changed by intervening on the other, the component is irrelevant to the mechanism. Although Craver’s reading of constitution as an idea of symmetric interlevel dependency (showing mutual manipulability) resembles experimental (bottom-up and top-down) practices in the neurosciences, it is inconsistent with the definition of constitution as an asymmetric relationship above (see also, note 6 above, this chapter). In addition, it contradicts his view that a constitutive mechanistic explanation is an explanation of a higher-level mechanism in terms of its lower-level parts, that is, it shows a bottom-up direction (see also Fagan 2012a). This problem casts doubt on the necessity of (CR2), which makes me focus on principle (CR1) below. 32. Bechtel (e.g., 2011; Bechtel and Abrahamsen 2011) also addresses this phenomenon of stability despite environmental change, calling it “autonomy.” It is defined as systems’ “ability to maintain themselves as systems distinct from their environment by directing the flow of mattes and energy so as to build and repair themselves” (Bechtel 2011, 535). 33. Robust mechanisms do not show modularity as a crucial feature. As the complexity of living systems reveals, modularity may fail so that the different parts of a mechanism cannot be altered independently through surgical interventions. For example, dynamical systems theory and systems biology seek to describe such nonmodular systems. On problems arising for the

236 ◂ NOTES TO PAGES 158–163

invariance account of causality and/or the concept of mechanism due to the requirement of modularity, see works by Cartwright (2001, 2002), Mitchell (2008), Fagan (2012a, 459–62), Kuorikoski (2012), and Leuridan (2012, 407– 9). 34. For the history of robustness research from classical to modern epigenetics, see chapter 2. 35. My discussion on the limits of this manipulationist account of constitutive relevance in epigenetics is relevant for investigations of robustness in other fields as well, like explanation of developmental stability in evolutionary developmental biology (evo-devo; see Hallgrímsson et al. 2002) and dynamic modeling of equilibrium in systems biology (e.g., Gunawardena 2010); see also below. 36. Gross’s (2015) claim primarily refers to constitutive explanation in systems biology by means of presenting a causal mechanism able to produce a system- level phenomenon that shows robustness. Some component parts of this mechanism display non-change-relating relationships or very weak dependency relations with the explanandum phenomenon. I (Baedke 2014) understand this view as a general—causal and constitutive—explanatory strategy of citing causal events and causal properties of entities as explanans variables that do not (or do, but very weakly) make a difference in a phenomenon under study. 37. Please note that the concept of robustness in this more realistic case differs slightly from the phenomenon of robustness described in the more simple case above. The simple model developed by Bhattacharya et al. (2011) cannot account for sudden changes (i.e., “chaotic” behavior) of the system. One could thus argue that by switching from an explanation of the latter to one of the former (the more complex scenario), we witness a shift of the explanandum as well, from explaining robustness to dynamic stability. However, this shift does not reject my main argument. Also, explanations of dynamic stability include (at least in part) explanans variables carrying explanatorily relevant information about non-difference-making and non-change-relating dependencies. Excluding this crucial information from an explanans still makes it impossible to completely address the explanandum. 38. On modularity, see also note 3 above, this chapter. 39. In fact, the view of constitutive dependencies and mechanistic explanation presented here also shows some similarity to Fagan’s (2012a) concept of jointness. According to Fagan, mechanistic explanations do not trace sin-

▸ 237 NOTES TO PAGES 164–170

gle causes but show how complementary components jointly interact (i.e., they show a joint causal activity) to constitute a mechanism’s behavior. Thus, she is also interested in constitutive interdependencies in complex systems. However, the view of explanations of robust mechanisms presented above differs slightly from Fagan’s explanations of “jointness mechanisms,” since both accounts trace different counterfactual dependencies. Describing joint constitution means, according Fagan’s (2012a, 464) account, tracing a counterfactual “relation, holding between causal partners that together produce some effect.” This effect can be changed by intervening on the multiple causal factors separately or together. In contrast, my view of robust mechanisms holds that the relevant counterfactuals are those that show under which interventions on the multiple causal factors—separately or together—a mechanism’s behavior is not constitutively changed. 40. When modelers and mathematicians refer to laws of succession, they are referring to rules governing how the states of a system succeed one another in time. In general, laws of succession, such as Galileo’s law and Newton’s second law, are described by differential or difference equations. Given such an equation, and some initial conditions, one can calculate how the values of a magnitude change. 41. Another example of a coexistence law is Pauli’s exclusion principle: given two electrons, A and B, in an atom, A can only occupy the subspace of the state space where it has a quantum number that is not identical to that of B.

5. ASSESSING THE EXPLANATORY POWER OF EPIGENETICS 1. Similarly, Hempel (1965, 695–705) has argued that scientific explanations and predictions are structurally identical. 2. The latter coherentist idea can be found in, for example, Hempel’s (1965) covering-law account, since it describes scientific explanation as the “reconciliation” of (already known) general laws with particular cases. For a critical account of the relationship between theoretical congruence and scientific progress, see the work of Kuhn (1962). 3. Sober (1975) links the simplicity of a scientific explanation to its informative- ness. 4. According to Hempel (1965, 345), “explanatory depth” basically refers to the accuracy and generality of an explanatory model.

238 ◂ NOTES TO PAGES 171–176

5. For those authors denying that evolutionary explanations can explain the traits of individuals, that is, to explain why a particular individual has a certain trait rather than another trait, see, for example, works by Sober (1984) and Stegmann (2010). For those claiming that natural selection can explain individual traits, see, for example, works by Neander (1995) and Forber (2005). 6. Critiques of an extended evolutionary synthesis also link the two concepts of explanatory power and likeliness, as they emphasize missing evidential support for certain novel explanations. See, for example, the article by Wray et al. (2014). 7. Please note that the concept of likeliness does to a certain degree underlie the concept discussed next—causal importance—because the measurability and comparability of causes’ influences rest upon the idea that one can collect empirical evidence for different levels of influence. 8. With respect to epigenetic inheritance, Noble (2008b) seems to notice this shortcoming of the concept of causal power or importance as well. He states, “If by genetic causation we mean the totality of the inherited causes of the phenotype, then it is plainly incorrect to exclude the non-DNA inheritance from this role, and it probably does not make much sense to ask which is more important, since only an interaction between DNA and non-DNA inheritance produces anything at all” (2008b, 3009; emphasis added). 9. On developmental systems theory, see also works by Oyama (1985) and Grif- fiths and Gray (1994). 10. In addition, this approach raises the question of how one can ontologically assess the causal power of entities cited as explanans variables. See also chapter 3 on this issue. 11. This traditional conceptualization of riskiness also shows that this criterion is usually concerned with the issue of explanatoriness, not explanatory power. In fact, what Popper was interested in was, above all, clarifying what a scientific explanation is (or should be), that is, revealing its inner logical structure and thus determining how it differs from nonscientific statements. 12. For similar accounts of contrastive explanation, see, for example, works by van Fraasen (1980), Sober (1986), Lipton (1990), Hitchcock (1996), Schaffer (2005), and Ylikoski (2007, 2013). See also chapter 3 for more on contrastive explanation. 13. As this example suggests, the general contrastive framework adopted here is applicable to both causal and mechanistic explanation (see Ylikoski 2013).

▸ 239 NOTES TO PAGES 177–186

14. Ylikoski and Kuorikoski (2010) list further virtues relevant for assessing the value of an explanation. These are the factual accuracy and the degree of integration into existing knowledge. Falsehood may be introduced into an explanation by, for instance, distorting certain values of variables through idealization. The integration of an explanation into a larger body of knowledge can make it better, since “the theoretical connections can expand the range of answers to different what if–questions” (2010, 213). I will restrict the discussion of explanatory power in epigenetics below to the criteria of precision, nonsensitivity, and cognitive salience. However, in principle, it can be expanded by considering the other criteria as well. 15. The idea of comparing explanatory agendas, including their standards, shows a certain similarity to Lakatos’s (1970, 155) attempt to define general rational principles, like content increase or the heuristic power to predict novel facts, according to which scientists chose research programs. Somewhat in line with Feyerabend’s (1975) rejection of Lakatos’s approach, the concept of explanatory power suggests “that there are no explicit [e.g., rational] general principles for goodness of explanations” (Ylikoski and Kuorikoski 2010, 206). Explanatory agendas can nevertheless be compared if we consider a field’s standards of explanation (e.g., to judge precision over sensitivity) as being learned, instanced, and changed via paradigmatic exemplar explanations (or models) of good explanation, which can be compared instead. An exemplar explanation in evolutionary biology might be Darwin’s explanation of differences in beak size and shapes in Galápagos finches. In molecular biology an influential exemplar is the explanation of gene regulation via the operon model of Jacob and Monod. 16. See also Woodward’s (2003, 295–302; 2006, 2010) and Mitchell’s (1997, 2000, 2002) concept of stability. 17. The add-on “and thus X can be selected” refers to the minor explanatory interest of molecular biologists (yet major interest of many epigeneticists) in putting their results into an evolutionary context. * 18. In (Eme ) the condition “under a certain combination of values of a set of

genetic and (other) environmental variables Zn” lists those non-epigenetic variables considered to be highly relevant for an epigenetic explanation. 19. The only mechanistic detail that is included in the explanans is that the vyA allele is not involved in bringing about the dependency relation under study. 20. The relatedness of the two explanations stems from the fact that both might share explanans variables; their explananda might overlap as well.

240 ◂ NOTES TO PAGES 188–193

21. It must be mentioned that many explanandum phenomena of epigenetics and evolutionary biology might be identical in the first place. Take, for instance, the question “Why has a certain character X evolved?” However, this rather phenomenological level of investigation is soon replaced by particular research agendas in epigenetics (seeking a “developmental” answer to the explanandum phenomenon) and in orthodox evolutionary biology (seeking an “adaptationist” explanans). This fission of evolutionary explananda is described below. 22. For evidence supporting an ultimate cause or proximate cause interpretation of the evolution of bird wing morphology, see the article by Vargas and Fal- lon (2005). 23. For a review of the debate on the thrifty gene hypothesis and other alternative explanations, see the work of Speakman and O’Rahilly (2012). 24. For this strategy to use a description of the non-occurrence of a phenomenon-to-be-explained as an explanandum’s contrast class in population genetics and ecological explanation, see the work of Marcel Weber (1999, 81–83). As Weber claims, often this contrast class is directly derived from the nonap- plicability of an ecological or evolutionary law, like the competition exclusion principle or the Hardy-Weinberg law, to a particular phenomenon. See also the article by Raerinne and Baedke (2015). 25. Actually, Prentice et al. (2008) develop two counterfactual community scenarios built upon anthropological and historical information: a hunter-gatherer scenario and a constant food supply scenario. In both scenarios no positive selection on obesity in humans occurs. 26. This contrastive explanatory strategy bears similarities to accounts of the “how-possible explanation” (Brandon 1990; Forber 2010) and to the so-called “viability explanation” (Wouters 1995), both of which are prominent in evolutionary biology and ecology. How-possible explanations outline how certain events might possibly occur, often in contrast to other events that could not have occurred. In addition, how-possible explanations usually offer a (e.g., evolutionary) narrative accounting for the event’s occurrence. Via- bility explanations describe in a contrastive manner why organisms have a particular trait, rather than another, by referring to a particular need of this organism. 27. Use of animal models of course broadens the original explanatory focus from the evolution of obesity in humans to, for example, the evolution of obesity in mammals. For the sake of the argument, let us assume in this case that

▸ 241 NOTES TO PAGES 193–208

extrapolation is legitimate and that it leads to a more general explanation, the explanation of obesity in humans being one instance. ** 28. Therefore, the explanandum of (Epc ) could also be described as {body fat

percentage a1, serum leptin level b1, and insulin tolerance level c1 in rats}

[body fat percentage a1+n, leptin level b1+n, and insulin tolerance level c1+n in rats]. 29. For example, evo-devoists are divided over whether their research is in line with a neo-Darwinian view and which concepts (e.g., evolvability or body plan) are the most crucial in their field (see Laubichler 2010; and Minelli 2010).

CONCLUSION 1. Although theoretical biology had not been established as an institutionally distinct biological field before the time of these meetings, the idea of investigating theoretical (especially conceptual) issues of biology can be traced back at least to Reinke’s Einleitung in die theoretische Biologie (1901). In fact, in the first half of the twentieth century, organicists in particular, including Joseph Needham, Joseph Henry Woodger, Joseph Edward Russell, John Scott Haldane, William Emerson Ritter, Paul Weiss, and Ludwig von Bertalanffy, drew no real distinction between theoretical biology and philosophy of biology (see Nicholson and Gawne 2015). On this early theoretical biology, see also the work of (other) members of the Theoretical Biology Club in Cam- bridge in the 1930s (Abir-Am 1987; Peterson 2016), as well as the monographs in Julius Schaxel’s (1919–31) book series Abhandlungen zur theoretischen Biologie. 2. Mayr’s (1961) paper “Cause and Effect in Biology” was even circulated before the first meeting and reprinted in the first volume of the proceedings (Wad- dington 1968). 3. This overview does not present an exhaustive catalog of issues to be addressed by an all-embracing philosophy of epigenetics. It should instead provide the reader insight into the wide spectrum of additional research topics in this field. 4. For general recent discussions on how model organisms represent theory, how they function as exemplary of higher taxons, and what their limitations are, see, for example, works by Keller (2000a), Bolker (2009), Ankeny and Leonelli (2011), and Leonelli and Ankeny (2013). In evo-devo a debate on

242 ◂ NOTES TO PAGES 211–214

epistemic and metaphysical issues related to the use of model organisms has been in full bloom (see, e.g., Minelli and Fusco 2004; Jenner and Wills 2007; Love 2009; Bolker 2014; Minelli and Baedke 2014; and Zuk et al. 2014). In contrast, philosophical challenges related to model organism selection in epigenetics have been widely neglected so far. 5. This view resembles some interesting similarities to process philosophy, from which central views in classical and contemporary epigenetics emerged (see chaps. 2 and 4). 6. Sometimes evo-devoists simply translate “epigenetic” as “developmental” (see, e.g., G. Wagner 2014, 27). 7. According to Müller and Wagner (1991), evo-devo qualitatively assesses phenotypic variation so that one can conceptualize an evolutionary novelty as an emerging developmental trait that is homologous neither to any trait in the ancestral species nor to a trait in the same species (or even to a trait in the same organism). 8. For a review of these positions in developmental systems theory, as well a distinction between this field and evo-devo, see the discussion by Griffiths and Gray (2005). 9. For studies on the historical development of evo-devo, see works by, for example, Amundson (2005), Laubichler and Maienschein (2007), and Müller (2008), and the essays in the volume edited by Love (2015). Philosophical analyses of evo-devo’s theoretical framework focus on the concepts of “evolvability” (Brigandt 2015b), “evolutionary novelty” (Brigandt and Love 2012), “modularity” (Winther 2005), “body plan” (Lewens 2009), and “mechanism” and “mechanistic explanation” in evo-devo (Calcott 2009; Brigandt 2015a). On the reconciliation of evo-devo concepts with other biological concepts, see essays in the volumes edited by Minelli and Fusco (2008) and by Laubich- ler and Maienschein (2009), and see also the article by Craig (2010). Addi- tionally, even the pedagogical dimension of evo-devo has been approached; see, for example, Hiatt et al.’s (2013) work on concepts and challenges for students learning evolutionary developmental biology.

▸ 243

REFERENCES

Abir-Am, Pnina G. 1987. “The Biotheoretical Gathering, Trans-Disciplinary Authority and the Incipient Legitimation of Molecular Biology in the 1930s: New Perspective on the Historical Sociology of Science.” History of Science 25:1–70. Aiken, Catherine E., and Susan E. Ozanne. 2014. “Transgenerational Develop- mental Programming.” Human Reproduction Update 20:63–75. Alberts, Bruce, Alexander Johnson, Julian Lewis, Martin Raff, Keith Roberts, and Peter Walter. 2002. Molecular Biology of the Cell. 4th ed. New York: Garland. Alcazar, Rosa M., Rueyling Lin, and Andrew Z. Fire. 2008. “Transmission Dynam- ics of Heritable Silencing Induced by Double-Stranded RNA in Caenorhabditis elegans.” Genetics 180:1275–88. Amundson, Ron. 2005. The Changing Role of the Embryo in Evolutionary Thought: Roots of Evo-Devo. Cambridge: Cambridge University Press. Andersen, Holly. 2012. “The Case for Regularity in Mechanistic Causal Explana- tion.” Synthese 189:415–32. Ankeny, Rachel A., and Sabina Leonelli. 2011. “What’s So Special about Model Organisms?” Studies in History and Philosophy of Science 42:313–23. REFERENCES

Anway, Matthew D., Andrea S. Cupp, Mehmet Uzumcu, and Michael K. Skin- ner. 2005. “Epigenetic Transgenerational Actions of Endocrine Disruptors and Male Fertility.” Science 308:1466–69. Arai, Junko A., and Larry A. Feig. 2011. “Long-Lasting and Transgenerational Effects of an Environmental Enrichment on Memory Formation.”Brain Research Bulletin 85:30–35. Arai, Junko A., Shaomin Li, Dean M. Hartley, and Larry A. Feig. 2009. “Transgen- erational Rescue of a Genetic Defect in Long-Term Potentiation and Memory Formation by Juvenile Enrichment.” Journal of Neuroscience 29:1496–502. Ariew, André. 2003. “Ernst Mayr’s ‘Ultimate/Proximate’ Distinction Reconsidered and Reconstructed.” Biology and Philosophy 18:553–65. Aristotle. 1984. Physics. In The Complete Works of Aristotle. Vol. 1. Translated by R. P. Hardie and R. K. Gaye. Edited by Jonathan Barnes. Princeton: Princeton University Press. Astbury, William T. 1941. “Protein and Virus Studies in Relation to the Problem of the Gene.” Proceedings of the Seventh International Genetical Congress of Genet- ics, 49–51. Cambridge: Cambridge University Press. Ayala, Francisco J. 1989. “Thermodynamics, Information, and Evolution: The Problem of Reductionism.” History and Philosophy of the Life Sciences 11:115–20. Backer, Philippe De, Danny De Waele, and Linda Van Speybroeck. 2010. “Ins and Outs of Systems Biology vis-à-vis Molecular Biology: Continuation or Clear Cut?” Acta Biotheoretica 58:15–49. Bacon, Francis. (1620) 1994. The Novum Organon. Edited by Peter Uhrbach and John Gibson. Chicago: Open Court. Baedke, Jan. 2012. “Causal Explanation beyond the Gene: Manipulation and Cau- sality in Epigenetics.” Theoria 27:153–74. Baedke, Jan. 2013. “The Epigenetic Landscape in the Course of Time: Conrad Hal Waddington’s Methodological Impact on the Life Sciences.” Studies in History and Philosophy of Biological and Biomedical Sciences 44:756–73. Baedke, Jan. 2014. “Challenges to Characterizing the Notion of Causation across Disciplinary Boundaries: Comment on Faye.” In New Directions in the Philos- ophy of Science, edited by M. C. Galavotti, M. Hartmann, Marcel Weber, W. Gonzalez, Dennis Dieks, and T. Uebel, 191–202. Dordrecht: Springer. Baedke, Jan. 2017a. “Expanding Views of Evolution and Causality (Workshop Report ‘Cause and Process in Evolution,’ Konrad Lorenz Institute for Evolu- tion and Cognition Research [KLI], Vienna, May 11th–14th, 2017).” Journal for General Philosophy of Science 48:591–94.

246 ◂ REFERENCES

Baedke, Jan. 2017b. “The New Biology of the Social: Shaping Humans’ Future, Science, and Public Health.” In Imagined Futures in Science, Technology and Society, edited by Gert Verschraegen, Frédéric Vandermoere, Luc Brackmans, and Barbara Segaert, 45–64. London: Routledge. Baedke, Jan, and Tobias Schöttler. 2017. “Visual Metaphors in the Sciences: The Case of Epigenetic Landscape Images.” Journal for General Philosophy of Sci- ence 48:173–95. Baetu, Tudor M. 2014. “Models and the Mosaic of Scientific Knowledge: The Case of Immunology.” Studies in History and Philosophy of Biological and Biomedical Sciences 45:49–56. Baker, John R. 1938. “The Evolution of Breeding Seasons.” In Evolution: Essays on Aspects of Evolutionary Biology, edited by Gavin R. De Beer, 161–77. Oxford: Clarendon Press. Baker, Monya. 2010. “Epigenome: Mapping in Motion.” Nature Methods 7:181–86. Balter, Michael. 2000. “Was Lamarck Just a Little Bit Right?” Science 288:38. Bapteste, Eric, and John Dupré. 2013. “Towards a Processual Microbial Ontology.” Biology and Philosophy 28:379–404. Barros, Sylvana P., and Steven Offenbacher. 2009. “Epigenetics: Connecting Envi- ronment and Genotype to Phenotype and Disease.” Journal of Dental Research 88:400–408. Barth, Teresa K., and Axel Imhof. 2010. “Fast Signals and Slow Marks: The Dynamics of Histone Modifications.”Trends in Biochemical Sciences 35:618–26. Bateson, Patrick. 2005. “The Return of the Whole Organism.”Journal of Biosci- ences 30:31–39. Bateson, Patrick. 2012. “The Impact of the Organism on Its Descendants.” Genetics Research International 2012:640612. doi:10.1155/2012/640612. Bateson, Patrick, and Peter Gluckman. 2011. Plasticity, Robustness, Development and Evolution. Cambridge: Cambridge University Press. Batterman, Robert W. 2002. The Devil in the Details: Asymptotic Reasoning in Explanation, Reduction, and Emergence. Oxford: Oxford University Press. Baumgartner, Michael. 2009. “Interventionist Causal Exclusion and Non-Reduc- tive Physicalism.” International Studies in the Philosophy of Science 23:161–78. Beatty, John. 1994. “The Proximate/Ultimate Distinction in the Multiple Careers of Ernst Mayr.” Biology and Philosophy 9:333–56. Beatty, John. 1995. “The Evolutionary Contingency Thesis.” InConcepts, Theories, and Rationality in the Biological Sciences, edited by Gereon Wolters and J. Len- nox, 83–97. Pittsburgh: University of Pittsburgh Press.

▸ 247 REFERENCES

Bechtel, William. 2006. Discovering Cell Mechanisms: The Creation of Modern Cell Biology. Cambridge: Cambridge University Press. Bechtel, William. 2008. Mental Mechanisms: Philosophical Perspectives on Cogni- tive Neuroscience. London: Routledge. Bechtel, William. 2011. “Mechanism and Biological Explanation.” Philosophy of Science 78:533–57. Bechtel, William. 2013a. “From Molecules to Networks: Adoption of Systems Approaches in Circadian Rhythm Research.” In New Challenges to Philosophy of Science, edited by Hanne Andersen, Dennis Dieks, Wenceslao J. Gonzalez, Thomas Uebel, and Gregory Wheeler, 211–23. Dordrecht: Springer. Bechtel, William. 2013b. “Understanding Biological Mechanisms: Using Illustra- tions from Circadian Rhythm Research.” In The Philosophy of Biology, edited by Kostas Kampourakis, 487–510. Dordrecht: Springer. Bechtel, William, and Adele Abrahamsen. 2005. “Explanation: A Mechanist Alter- n at i v e .” Studies in History and Philosophy of Biological and Biomedical Sciences 36:421–41. Bechtel, William, and Adele Abrahamsen. 2010. “Dynamic Mechanistic Expla- nation: Computational Modeling of Circadian Rhythms as an Exemplar for Cognitive Science.” Studies in History and Philosophy of Science 41:321–33. Bechtel, William, and Adele Abrahamsen. 2011. “Complex Biological Mecha- nisms: Cyclic, Oscillatory, and Autonomous.” In Philosophy of Complex Sys- tems, edited by Clifford Hooker, 257–85. New York: Elsevier. Bechtel, William, and Adele A. Abrahamsen. 2013. “Thinking Dynamically about Biological Mechanisms: Networks of Coupled Oscillators.” Foundations of Sci- ence 18:707–23. Bechtel, William, and Robert C. Richardson. 1993. Discovering Complexity: Decomposition and Localization as Strategies in Scientific Research. Princeton: Princeton University Press. Bell, Chris. 2013. “Epigenetics: How to Alter Your Genes.” Telegraph, 16 October. http://www.telegraph.co.uk/science/10369861/Epigenetics-How-to-alter-your- genes.html. Benkalfat, Nassira B., Hafida Merzouk, Samira Bouanane, Sid-Ahmed Merzouk, Jérôme Bellenger, Joseph Gresti, Christian Tessier, and Michel Narce. 2011. “Altered Adipose Tissue Metabolism in Offspring of Dietary Obese Rat Dams.” Clinical Science 121:19–28. Bernard, Claude. (1865) 1949. An Introduction to the Study of Experimental Medi- cine. New York: Schuman.

248 ◂ REFERENCES

Bernstein, Emily, and C. David Allis. 2005. “RNA Meets Chromatin.” Genes and Development 19:1635–55. Bhattacharya, Sudin, Qiang Zhang, and Melvin E. Andersen. 2011. “A Determin- istic Map of Waddington’s Epigenetic Landscape for Cell Fate Specification.” BMC Systems Biology 5:85. doi:10.1186/1752–0509–5-85. Bird, Adrian. 2007. “Perceptions of Epigenetics.” Nature 447:396–98. Blech, Jörg. 2010. Gene sind kein Schicksal [Genes are not destiny]. Frankfurt: Fischer. Bogen, Jim. 2005. “Regularities and Causality: Generalizations and Causal Expla- nations.” Studies in History and Philosophy of Biological and Biomedical Sci- ences 36:397–420. Bolker, Jessica A. 2009. “Exemplary and Surrogate Models: Two Modes of Repre- sentation in Biology.” Perspectives in Biology and Medicine 52:485–99. Bolker, Jessica A. 2014. “Model Species in Evo-Devo: A Philosophical Perspec- t i v e .” Evolution and Development 16:49–56. Bonasio, Roberto, Shengjiang Tu, and Danny Reinberg. 2010. “Molecular Signals of Epigenetic States.” Science 330:612–16. Bonduriansky, Russell, Angela J. Crean, and Troy Day. 2012. “The Implications of Nongenetic Inheritance for Evolution in Changing Environments.” Evolution- ary Applications 5:192–201. Boniolo, Giovanni. 2013. “Is an Account of Identity Necessary for Bioethics? What Post-Genomic Biomedicine Can Teach Us.” Studies in History and Philosophy of Biological and Biomedical Sciences 44:401–11. Boniolo, Giovanni, and Giuseppe Testa. 2012. “The Identity of Living Beings, Epi- genetics, and the Modesty of Philosophy.” Erkenntnis 76:279–98. Borelli, Giovanni Alfonso. 1680. De motu animalium. Rome: A. Bernabo. Borges, Jorge Luis. 1975. A Universal History of Iniquity. Translated by Norman T. di Giovanni. London: Penguin. Bossdorf, Oliver, Christina L. Richards, and Massimo Pigliucci. 2008. “Epigenetics for Ecologists.” Ecology Letters 11:106–15. Bossdorf, Oliver, and Yuanye Zhang. 2011. “A Truly Ecological Epigenetics Study.” Molecular Ecology 20:1572–74. Boudry, Maarten, and Massimo Pigliucci. 2013. “The Mismeasure of Machine: Synthetic Biology and the Trouble with Engineering Metaphors.” Studies in History and Philosophy of Biological and Biomedical Sciences 44:660–68. Boulter, Stephen. 2012. “Contrastive Explanations in Evolutionary Biology.” Ratio 25:425–41.

▸ 249 REFERENCES

Bowler, Peter J. 1983. Evolution: The History of an Idea. Berkeley: University of California Press. Brakefield, Paul M., and W. Anthony Frankino. 2009. “Polyphenisms in Lepidop- tera: Multidisciplinary Approaches to Studies of Evolution and Development.” In Phenotypic Plasticity in Insects: Mechanisms and Consequences, edited by Douglas W. Whitman and Taracad N. Ananthakrishnan, 121–51. Plymouth: Science Publishers. Brandon, Robert N. 1990. Adaptation and Environment. Princeton: Princeton University Press. Bredekamp, Horst. 2005. “Denkende Hände: Überlegungen zur Bildkunst der Naturwissenschaften” [Thinking hands: Reflections on the pictorial art of the natural sciences]. In Von der Wahrnehmung zur Erkenntnis—From Perception to Understanding, edited by Monika Lessl and Jürgen Mittelstraß, 109–32. Hei- delberg: Springer. Brigandt, Ingo. 2013. “Systems Biology and the Integration of Mechanistic Expla- nation and Mathematical Explanation.” Studies in History and Philosophy of Biological and Biomedical Sciences 44:477–92. Brigandt, Ingo. 2015a. “Evolutionary Developmental Biology and the Limits of Philosophical Accounts of Mechanistic Explanation.” In Explanation in Biol- ogy: An Enquiry into the Diversity of Explanatory Patterns in the Life Sciences, edited by Pierre-Alain Braillard and Christophe Malaterre, 135–73. Berlin: Springer. Brigandt, Ingo. 2015b. “From Developmental Constraint to Evolvability: How Concepts Figure in Explanation and Disciplinary Identity.” In Conceptual Change in Biology: Scientific and Philosophical Perspectives on Evolution and Development, edited by Alan Love, 305–25. Berlin: Springer. Brigandt, Ingo, and Alan C. Love. 2012. “Conceptualizing Evolutionary Novelty: Moving beyond Definitional Debates.” Journal of Experimental Zoology B 318:417–27. Broad, Charlie D. 1925. The Mind and Its Place in Nature. London: Routledge. Broadbent, Alex. 2011. “Inferring Causation in Epidemiology: Mechanisms, Black Boxes, and Contrasts.” In Causality in the Sciences, edited by Phyllis M. Illari, Federica Russo, and John Williamson, 45–69. Oxford: Oxford Univer- sity Press. Bronfman, Zohar Z., Simona Ginsburg, and Eva Jablonka. 2014. “Shaping the Learning Curve: Epigenetic Dynamics in Neural Plasticity.” Frontiers in Inte- grative Neuroscience 8:55. doi:10.3389/fnint.2014.00055.

250 ◂ REFERENCES

Brooks, Daniel S. 2017. “In Defense of Levels: Layer Cakes and Guilt by Associa- tion.” Biological Theory. doi:10.1007/s13752–017–0272–8. Brown, Rachael L. 2014. “What Evolvability Really Is.” British Journal for the Phi- losophy of Science 65:549–72. Brown, Rachael L. 2015. “Why Development Matters.” Biology and Philosophy 30:889–99. Buck, Christopher B., and Benes L. Trus. 2012. “The Papillomavirus Virion: A Machine Built to Hide Molecular Achilles’ Heels.” Advances in Experimental Medicine and Biology 726:403–22. Burggren, Warren W. 2015. “Dynamics of Epigenetic Phenomena: Intergenera- tional and Intragenerational Phenotype ‘Washout.’” Journal of Experimental Biology 218:80–87. Burian, Richard M. 2005. “Lillie’s Paradox: Or Some Hazards of Cellular Geog- raphy.” In The Epistemology of Development, Evolution, and Genetics: Selected Essays, 183–209. Cambridge: Cambridge University Press. Burkeman, Oliver. 2010. “Why Everything You’ve Been Told about Evolution Is Wron g .” The Guardian, 19 March. http://www.theguardian.com/science/2010/ mar/19/evolution-darwin-natural-selection-genes-wrong. Burkhardt, Richard W. 1995. The Spirit of System: Lamarck and Evolutionary Biol- ogy. Cambridge: Harvard University Press. Cabej, Nelson R. 2008. Epigenetic Principles of Evolution. Dumont: Albanet. Caianiello, Silvia. 2009. “Adaptive versus Epigenetic Landscape: A Visual Chapter in the History of Evolution and Development.” In Graphing Genes, Cells and Embryos: Cultures of Seeing 3D and Beyond, edited by Sabine Brauckmann, Christina Brandt, and Gerd Thieffry, 65–81. Berlin: Max Planck Institute for the History of Science. Calcott, Brett. 2009. “Lineage Explanations: Explaining How Biological Mecha- nisms Change.” British Journal for the Philosophy of Science 60:51–78. Calcott, Brett. 2013a. “Why How and Why Aren’t Enough: More Problems with Mayr’s Proximate-Ultimate Distinction.” Biology and Philosophy 28:767–80. Calcott, Brett. 2013b. “Why the Proximate-Ultimate Distinction Is Misleading, and Why It Matters for Understanding the Evolution of Cooperation.” In Coopera- tion and Its Evolution, edited by Kim Sterelny, Richard Joyce, Brett Calcott, and Ben Fraser, 249–64. Cambridge: MIT Press. Callebaut, Werner. 2010. “The Dialectics of Dis/Unity in the Evolutionary Synthe- sis and Its Extensions.” In Evolution: The Extended Synthesis, edited by Mas- simo Pigliucci and Gerd B. Muller, 443–81. Cambridge: MIT Press.

▸ 251 REFERENCES

Calvert, Jane, and Joan H. Fujimura. 2011. “Calculating Life? Duelling Discourses in Interdisciplinary Systems Biology.” Studies in History and Philosophy of Bio- logical and Biomedical Sciences 42:155–63. Campbell, Donald T. 1974. “‘Downward Causation’ in Hierarchically Organised Biological Systems.” In Studies in the Philosophy of Biology: Reduction and Related Problems, edited by Francisco J. Ayala and Theodosius G. Dobzhansky, 179–86. London: Macmillan. Carey, Nessa. 2011. The Epigenetics Revolution: How Modern Biology Is Rewriting Our Understanding of Genetics, Disease, and Inheritance. London: Icon. Carnap, Rudolf. 1928. Der logische Aufbau der Welt[The logical structure of the world]. Berlin: Schlachtensee. Cartwright, Nancy. 2001. “Modularity: It Can—and Generally Does—Fail.” In Stochastic Causality, edited by Domenico Costantini, Maria C. Galavotti, and Patrick Suppes, 65–84. Stanford: CSLI Publications. Cartwright, Nancy. 2002. “Against Modularity, the Causal Markov Condition, and Any Link between the Two: Comments on Hausman and Woodward.” British Journal for the Philosophy of Science 53:411–53. Carusi, Annamaria. 2012. “Making the Visual Visible in Philosophy of Science.” Spontaneous Generations 6:106–14. Casasa, Sofia, Daniel B. Schwab, and Armin P. Moczek. 2017. “Developmental Regulation and Evolution of Scaling: Novel Insights through the Study of Onthophagus Beetles.” Current Opinion in Insect Science 19:52–60. Cavalier-Smith, Thomas. 2004. “The Membranome and Membrane Heredity in Development and Evolution.” In Organelles, Genomes and Eukaryote Phylog- eny, edited by David S. Horner and Robert P. Hirt, 335–52. Boca Raton: CRC Press. Champagne, Frances A., Ian C. G. Weaver, Josie Diorio, Shakti Sharma, and Michael J. Meaney. 2003. “Natural Variations in Maternal Care Are Associ- ated with Estrogen Receptor Alpha Expression and Estrogen Sensitivity in the Medial Preoptic Area.” Endocrinology 144:4720–24. Chang, Hannah H., Martin Hemberg, Mauricio Barahona, Donald E. Ingber, and Sui Huang. 2008. “Transcriptome-Wide Noise Controls Lineage Choice in Mammalian Progenitor Cells.” Nature 453:544–47. Charlesworth, Deborah, Nicholas H. Barton, and Brian Charlesworth. 2017. “The Sources of Adaptive Variation.” Proceedings of the Royal Society B 284:20162864. doi:10.1098/rspb.2016.2864. Cheedipudi, Sirisha, Oriana Genolet, and Gergana D. Dobreva. 2014. “Epigenetic

252 ◂ REFERENCES

Inheritance of Cell Fates during Embryonic Development.” Epigenomics and Epigenetics 5:19. doi:10.3389/fgene.2014.00019. Chiu, Lynn, and Scott F. Gilbert. 2015. “The Birth of the Holobiont: Multi-Species Birthing through Mutual Scaffolding and Niche Construction.” Biosemiotics 8:191–210. Chong, Suyinn, Nicola Vickaryous, Alyson Ashe, Natasha Zamudio, Neil Young- son, Sarah Hemley, Tomas Stopka, Arthur Skoultchi, Jacqui Matthews, Hamish S. Scott, David de Kretser, Moira O’Bryan, Marnie Blewitt, and Emma Whitelaw. 2007. “Modifiers of Epigenetic Reprogramming Show Paternal Effects in the Mouse.” Nature Genetics 39:614–22. Church, Dawson. 2009. The Genie in Your Genes: Epigenetic Medicine and the New Biology of Intention. Santa Rosa: Elite Books. Clare, Daniel K., and Helen R. Saibil. 2013. “ATP-Driven Molecular Chaperone Machines.” Biopolymers 99:846–59. Collingwood, Robin G. 1940. An Essay on Metaphysics. Oxford: Clarendon Press. Cornfield, Jerome, William Haenszel, E. Cuyler Hammond, Abraham M. Lil- ienfeld, Michael B. Shimkin, and Ernst L. Wynder. 1959. “Smoking and Lung Cancer: Recent Evidence and a Discussion of Some Questions.” Journal of the National Cancer Institute 22:173–203. Cortijo, Sandra, René Wardenaar, Maria Colomé-Tatché, Arthur Gilly, Mathilde Etcheverry, Karine Labadie, Erwann Caillieux, Frédéric Hospital, Jean-Marc Aury, Patrick Wincker, François Roudier, Ritsert C. Jansen, Vincent Colot, and Frank Johannes. 2014. “Mapping the Epigenetic Basis of Complex Traits.” Sci- ence 343:1145–48. Costa, Fabricio F. 2008. “Non-Coding RNAs, Epigenetics and Complexity.” Gene 410:9–17. Coyne, Jerry. 2011a. “Epigenetics Again: Will It Cause a Revolution in Evolu- tion?” 26 September. http://whyevolutionistrue.wordpress.com/2011/09/26/ epigenetics-again-will-it-cause-a-revolution-in-evolution/. Coyne, Jerry. 2011b. “Is ‘Epigenetics’ a Revolution in Evolution?” Why Revolution Is True (blog), 21 August. http://whyevolutionistrue.wordpress.com/2011/08/21/ is-epigenetics-a-revolution-in-evolution/. Craig, Lindsay R. 2010. “The So-Called Extended Synthesis and Population Genetics.” Biological Theory 5:117–23. Craver, Carl F. 2002a. “Interlevel Experiments and Multilevel Mechanisms in the Neuroscience of Memory.” Philosophy of Science 69:S83–S97. Craver, Carl F. 2002b. “Structures of Scientific Theories.” InBlackwell Guide to

▸ 253 REFERENCES

the Philosophy of Science, edited by Peter Machamer and Michael Silberstein, 55–79. Oxford: Blackwell. Craver, Carl F. 2006. “When Mechanistic Models Explain.” Synthese 153:355–76. Craver, Carl F. 2007. Explaining the Brain: Mechanisms and the Mosaic Unity of Neuroscience. Oxford: Oxford University Press. Craver, Carl F. 2008. “Physical Law and Mechanistic Explanation in the Hodg- kin and Huxley Model of the Action Potential.” Philosophy of Science 75:1022–33. Craver, Carl F., and William Bechtel. 2006. “Mechanism.” In The Philosophy of Science: An Encyclopedia, edited by Sahotra Sarkar and Jessica Pfeifer, 469–78. New York: Routledge. Craver, Carl F., and William Bechtel. 2007. “Top-Down Causation without Top- Down Causes.” Biology and Philosophy 22:547–63. Craver, Carl F., and Lindley Darden. 2005. “Introduction.” Studies in History and Philosophy of Biological and Biomedical Sciences 36:233–44. Craver, Carl F., and Lindley Darden. 2013. In Search of Mechanisms: Discoveries across the Life Sciences. Chicago: University of Chicago Press. Crawford, C. 1998. “The Theory of Evolution in the Study of Human Behavior: An Introduction and Overview.” In Handbook of Evolutionary Psychology: Ideas, Issues, and Application, edited by Charles Crawford and Dennis L. Krebs, 3–41. Hillsdale: Erlbaum. Crews, David, Andrea C. Gore, Timothy S. Hsu, Nygerma L. Dangleben, Michael Spinetta, Timothy Schallert, Matthew D. Anway, and Michael K. Skinner. 2007. “Transgenerational Epigenetic Imprints on Mate Preference.” Proceedings of the National Academy of Sciences 104:5942–46. Crick, Francis. 1958. “On Protein Synthesis.” Symposium of the Society of Experi- mental Biology 12:138–63. Crick, Francis. 1966. Of Molecules and Men. Seattle: University of Washington Press. Crick, Francis. 1970. “Central Dogma of Molecular Biology.” Nature 227:561–63. Cummins, Robert C. 1975. “Functional Analysis.” Journal of Philosophy 72:741–64. Daly, Martin, and Margo Wilson. 1978. Sex, Evolution, and Behavior. Scituate: Duxbury Press. Danchin, Étienne, Anne Charmantier, Frances A. Champagne, Alex Mesoudi, Benoit Pujol, and Simon Blanchet. 2011. “Beyond DNA: Integrating Inclusive Inheritance into an Extended Theory of Evolution.” Nature Reviews Genetics 12:475–86.

254 ◂ REFERENCES

Danielli, James F., and Marie A. DiBerardino. 1979. Nuclear Transplantation. New York: Academic Press. Darden, Lindley. 2005. “Relations among Fields: Mendelian, Cytological and Molecular Mechanisms.” Studies in History and Philosophy of Biological and Biomedical Sciences 36:349–71. Darden, Lindley. 2006. Reasoning in Biological Discoveries: Essays on Mechanisms, Interfield Relations, and Anomaly Resolution. Cambridge: Cambridge Univer- sity Press. Darden, Lindley, and Nancy Maull. 1977. “Interfield Theories.”Philosophy of Sci- ence 44:43–64. Darden, Lindley, and James Tabery. 2010. “Molecular Biology.” In The Stanford Encyclopedia of Philosophy, Fall ed., edited by Edward N. Zalta. http://plato .stanford.edu/archives/fall2010/entries/molecular-biology/. Darwin, Charles. 1859. On the Origin of Species by Means of Natural Selection or, the Preservation of Favoured Races in the Struggle for Life. London: Murray. Darwin, Charles. 1868. The Variation of Plants and Animals under Domestication. London: Murray. Davidson, Eric H. 2006. The Regulatory Genome: Gene Regulatory Networks in Development and Evolution. Amsterdam: Elsevier. Davies, Paul C. W. 2012. “The Epigenome and Top-Down Causation.” Interface Focus 2:42–48. Dawkins, Richard. 1976. The Selfish Gene. Oxford: Oxford University Press. Dawkins, Richard. 1978. “Replicator Selection and the Extended Phenotype.” Zeit- schrift für Tierpsychologie 47:61–76. Dawkins, Richard. 1982. The Extended Phenotype. London: Freeman. Dawkins, Richard. 1986. The Blind Watchmaker. New York: Norton. Dawkins, Richard. 2004. “Extended Phenotype—But Not Too Extended: A Reply to Laland, Turner and Jablonka.” Biology and Philosophy 19:377–96. Day, Rachel L., Kevin N. Laland, and F. John Odling-Smee. 2003. “Rethinking Adaptation: The Niche-Construction Perspective.”Perspectives in Biology and Medicine 46:80–95. Day, Troy, and Russell Bonduriansky. 2011. “A Unified Approach to the Evolution- ary Consequences of Genetic and Nongenetic Inheritance.” American Natural- ist 178:E18–36. De Beer, Gavin. 1954. Embryos and Ancestors. 2nd ed. Oxford: Oxford University Press. Deichmann, Ute. 2016a. “Epigenetics: The Origins and Evolution of a Fashionable Topic.” Developmental Biology 416:249–54.

▸ 255 REFERENCES

Deichmann, Ute. 2016b. “Why Epigenetics Is Not a Vindication of Lamarckism— and Why That Matters.” Studies in History and Philosophy of Biological and Biomedical Sciences 57:80–82. Delcuve, Geneviève P., Dilshad H. Khan, and James R. Davie. 2012. “Roles of His- tone Deacetylases in Epigenetic Regulation: Emerging Paradigms from Studies with Inhibitors.” Clinical Epigenetics 4:5. doi:10.1186/1868–7083–4-5. Dennett, Daniel C. 1995. Darwin’s Dangerous Idea: Evolution and the Meanings of Life. New York: Simon and Schuster. Descartes, René. (1664) 1972. Treatise of Man. Edited by T. S. Hall. Cambridge: Harvard University Press. Descartes, René. (1667) 1998. “Description of the Human Body.” In The World and Other Writings, edited by Stephen Gaukroger, 170–205. Cambridge: Cam- bridge University Press. Dewsbury, Donald A. 1992. “Essay on Contemporary Issues in Ethology: On the Problems Studied in Ethology, Comparative Psychology, and Animal Behav- ior.” Ethology 92:89–107. Dewsbury, Donald A. 1999. “The Proximate and the Ultimate: Past, Present, and Future.” Behavioural Processes 46:189–99. Dias, Brian G., and Kerry J. Ressler. 2014. “Parental Olfactory Experience Influ- ences Behavior and Neural Structure in Subsequent Generations.” Nature Neu- roscience 17:89–96. Dickins, Thomas E., and Robert Barton. 2013. “Reciprocal Causation and the Proximate-Ultimate Distinction.” Biology and Philosophy 28:747–56. Dickins, Thomas E., and Benjamin J. A. Dickins. 2007. “Designed Calibration: Naturally Selected Flexibility, Not Non-Genetic Inheritance.” Behavioral and Brain Sciences 30:368–69. Dickins, Thomas E., and Benjamin J. A. Dickins. 2008. “Mother Nature’s Tolerant Ways: Why Non-Genetic Inheritance Has Nothing to Do with Evolution.” New Ideas in Psychology 26:41–54. Dickins, Thomas E., and Qazi Rahman. 2012. “The Extended Evolutionary Syn- thesis and the Role of Soft Inheritance in Evolution.”Proceedings of the Royal Society B 279:2913–21. DiFrisco, James. 2017. “Time Scales and Levels of Organization.” Erkenntnis 82:795–818. Do, Duy N., Anders B. Strathe, Tage Ostersen, Just Jensen, Thomas Mark, and Haja N. Kadarmideen. 2013. “Genome-Wide Association Study Reveals Genetic Architecture of Eating Behavior in Pigs and Its Implications for

256 ◂ REFERENCES

Human Obesity by Comparative Mapping.” PLoS ONE 8: 71509. doi:10.1371/ journal.pone.0071509. Dobzhansky, Theodosius G. 1951.Genetics and the Origin of Species. 3rd ed. New York: Columbia University Press. Doshi, Tanvi, Smita S. Mehta, Vikas Dighe, Nafisa Balasinor, and Geeta Vanage. 2011. “Hypermethylation of Estrogen Receptor Promoter Region in Adult Tes- tis of Rats Exposed Neonatally to Bisphenol A.” Toxicology 289:74–82. Dove, William F. 1993. “The Gene, the Polygene, and the Genome.” Genetics 134:999–1002. Dowe, Phil. 1995. “Causality and Conserved Quantities: A Reply to Salmon.” Phi- losophy of Science 62:321–33. Dowe, Phil. 2000. Physical Causation. Cambridge: Cambridge University Press. Doyle, Julie. 2007. “Picturing the Clima(c)tic: Greenpeace and the Representa- tional Politics of Climate Change Communication.” Science as Culture 16:129– 50. Dunn, Gregory A., and Tracy L. Bale. 2009. “Maternal High-Fat Diet Promotes Body Length Increases and Insulin Insensitivity in Second-Generation Mice.” Endocrinology 150:4999–5009. Dunn, Gregory A., and Tracy L. Bale. 2011. “Maternal High-Fat Diet Effects on Third-Generation Female Body Size via the Paternal Lineage.”Endocrinology 152:2228–36. Dupré, John. 2013. “I—Living Causes.” Aristotelian Society Supplementary Volume 87:19–37. Dupré, John. 2014. Processes of Life: Essays in the Philosophy of Biology. Reprint ed. Oxford: Oxford University Press. Dupré, John, and Stephan Guttinger. 2016. “Viruses as Living Processes.” Studies in History and Philosophy of Biological and Biomedical Sciences 59:109–16. Eccleston, Alex, Natalie DeWitt, Chris Gunter, Barbara Marte, and Deepa Nath. 2007. “Epigenetics.” Nature 447:395. Edmonds, Bruce. 1999a. “Modelling Bounded Rationality in Agent-Based Simu- lations Using the Evolution of Mental Models.” In Computational Techniques for Modelling Learning in Economics, edited by Thomas Brenner, 305–32. Dor- drecht: Kluwer. Edmonds, Bruce. 1999b. “What Is Complexity? The Philosophy of Complexity per se with Application to Some Examples in Evolution.” In The Evolution of Com- plexity, edited by Francis Heylighen, Johan Bollen, and Alexander Rieger, 1–18. Dordrecht: Kluwer.

▸ 257 REFERENCES

Egger, Gerda, Gangning Liang, Ana Aparicio, and Peter A. Jones. 2004. “Epi- genetics in Human Disease and Prospects for Epigenetic Therapy.” Nature 429:457–63. Ehlert, Tobias, Perikles Simon, and Dirk A. Moser. 2013. “Epigenetics in Sports.” Sports Medicine 43:93–110. Eichten, Steve R., Ruth A. Swanson-Wagner, James C. Schnable, Amanda J. Waters, Peter J. Hermanson, Sanzhen Liu, Cheng-Ting Yeh, Yi Jia, Karla Gen- dler, Michael Freeling, Patrick S. Schnable, Matthew W. Vaughn, and Nathan M. Springer. 2011. “Heritable Epigenetic Variation among Maize Inbreds.” PLoS Genetics 7:1002372. doi:10.1371/journal.pgen.1002372. Eichten, Steven R., Matthew W. Vaughn, Peter J. Hermanson, and Nathan M. Springer. 2013. “Variation in DNA Methylation Patterns Is More Common among Maize Inbreds than among Tissues.” Plant Genome 6:2. doi:10.3835/ plantgenome2012.06.0009. Eldredge, Nils, and Stephen J. Gould. 1972. “Punctuated Equilibria: An Alternative to Phyletic Gradualism.” In Models in Paleobiology, edited by Thomas J. M. Schopf, 82–115. San Francisco: Freeman. Emmeche, Claus. 1997. “Aspects of Complexity in Life and Science.” Philosophica 59:41–68. Emmeche, Claus, Simo Koppe, and Frederick Stjernfelt. 2000. “Levels, Emer- gence, and Three Versions of Downward Causation.” In Downward Causation: Minds, Bodies, and Matter, edited by Peter Anderson, Claus Emmeche, Niels Finnemann, and Peder Christiansen, 13–34. Aarhus: University of Aarhus Press. Engel, Katharina, Thomas Schreder, and Ralph Tollrian. 2014. “Morphological Defences of Invasive Daphnia lumholtzi Protect against Vertebrate and Inver- tebrate Predators.” Journal of Plankton Research 36:1140–45. Epigenetics. 2017. “Aims and Scope.” Accessed 14 August. http://www.tandfonline. com/action/journalInformation?show=aimsScope&journalCode=kepi20. Eronen, Markus I. 2013. “No Levels, No Problems: Downward Causation in Neu- roscience.” Philosophy of Science 80:1042–52. Fagan, Melinda B. 2012a. “The Joint Account of Mechanistic Explanation.”Philos - ophy of Science 79:448–72. Fagan, Melinda B. 2012b. “Waddington Redux: Models and Explanation in Stem Cell and Systems Biology.” Biology and Philosophy 27:179–213. Falk, Raphael. 2011. “Evolution as Progressing Complexity.” In Transformations of Lamarckism: From Subtle Fluids to Molecular Biology, edited by Snait Gissis and Eva Jablonka, 381–84. Cambridge: MIT Press.

258 ◂ REFERENCES

Fazekas, Peter, and Gergely Kertész. 2011. “Causation at Different Levels: Track- ing the Commitments of Mechanistic Explanations.” Biology and Philosophy 26:365–83. Feyerabend, Paul K. 1975. Against Method. London: New Left Books. Filipowicz, Witold, Lukasz Jaskiewicz, Fabrice A. Kolb, and Ramesh S. Pillai. 2005. “Post-Transcriptional Gene Silencing by siRNAs and miRNAs.” Current Opinion in Structural Biology 15:331–41. Fisher, Amanda G., and Matthias Merkenschlager. 2010. “Fresh Powder on Wad- dington’s Slopes.” EMBO Reports 11:490–92. Fisher, Ronald A. 1930. The Genetical Theory of Natural Selection. Oxford: Clar- endon Press. Forber, Patrick. 2005. “On the Explanatory Roles of Natural Selection.” Biology and Philosophy 20:329–42. Forber, Patrick. 2010. “Confirmation and Explaining How Possible.” Studies in History and Philosophy of Biological and Biomedical Sciences 41:32–40. Francis, Gregory. 2014. “Too Much Success for Recent Groundbreaking Epigenet- ic Experiments.” Genetics 198:449–51. Francis, Richard C. 1990. “Causes, Proximate and Ultimate.” Biology and Philos- ophy 5:401–15. Franklin, Allan, Anthony W. F. Edwards, Daniel J. Fairbanks, and Daniel L. Hartl. 2008. Ending the Mendel-Fisher Controversy. Pittsburgh: University of Pitts- burgh Press. Franklin, Tamara B., Holger Russig, Isabelle C. Weiss, Johannes Gräff, Natacha Linder, Aubin Michalon, Sandor Vizi, and Isabelle M. Mansuy. 2010. “Epige- netic Transmission of the Impact of Early Stress across Generations.” Biological Psychiatry 68:408–15. Freedman, David. 1997. “From Association to Causation via Regression.” Advances in Applied Mathematics 18:59–110. Furusawa, Chikara, and Kunihiko Kaneko. 2012. “A Dynamical-Systems View of Stem Cell Biology.” Science 338:215–17. Futuyma, Douglas J. 2011. “Expand or Revise? The Evolutionary Synthesis Today.” Quarterly Review of Biology 86:203–8. Galaty, David H. 1974. “The Philosophical Basis of Mid-Nineteenth Century German Reductionism.” Journal of the History of Medicine and Allied Sciences 29:295–316. Gardner, Andy. 2013. “Ultimate Explanations Concern the Adaptive Rationale for Organism Design.” Biology and Philosophy 28:787–91.

▸ 259 REFERENCES

Garfinkel, Alan. 1981.Forms of Explanation: Rethinking the Questions in Social Theory. New Haven: Yale University Press. Gel’fand, Israel M., and Mikhail L. Tsetlin. 1962. “Some Methods of Control for Complex Systems.” Russian Mathematical Surveys 17:95–117. Gel’fand, Israel M., and Mikhail L. Tsetlin. 1971. “Mathematical Modeling of Mechanisms of the Central Nervous System.” In Models of the Structural-Func- tional Organization of Certain Biological Systems, edited by Israel M. Gel’fand, Victor S. Gurfinkel, Sergei V. Fomin, and Mikhail L. Tsetlin, 1–22. Cambridge: MIT Press. Gel’fand, Israel M., and Mikhail L. Tsetlin. 1973. “Mathematical Simulation of the Principles of the Functioning of the Central Nervous System.” In Automaton Theory and Modeling of Biological Systems, edited by Mikhail L. Tsetlin, 131–53. New York: Academic Press. Geoghegan, Jemma L., and Hamish G. Spencer. 2012. “Population-Epigenetic Models of Selection.” Theoretical Population Biology 81:232–42. Giere, Ronald N. 1988. Explaining Science: A Cognitive Approach. Chicago: Uni- versity of Chicago Press. Giere, Ronald N. 1999. Science without Laws. Chicago: University of Chicago Press. Gilbert, Scott F. 1991. “Epigenetic Landscaping: Waddington’s Use of Cell Fate Bifurcation Diagrams.” Biology and Philosophy 6:135–54. Gilbert, Scott F. 2000. “Diachronic Biology Meets Evo-Devo: C. H. Wadding- ton’s Approach to Evolutionary Developmental Biology.” American Zoologist 40:729–37. Gilbert, Scott F. 2012. “Commentary: ‘The Epigenotype’ by C. H. Waddington.” International Journal of Epidemiology 41:20–23. Gilbert, Scott F. 2014. “A Holobiont Birth Narrative: The Epigenetic Transmis- sion of the Human Microbiome.” Frontiers in Genetics 5:282. doi:10.3389/ fgene.2014.00282. Gilbert, Scott F., and David Epel. 2009. Ecological Developmental Biology: Inte- grating Epigenetics, Medicine, and Evolution. Sunderland: Sinauer Associ- ates. Gilbert, Scott F., Jan Sapp, and Alfred I. Tauber. 2012. “A Symbiotic View of Life: We Have Never Been Individuals.” Quarterly Review of Biology 87:325–41. Gilbert, Walter. 1991. “Towards a Paradigm Shift in Biology.” Nature 349:99. Gissis, Snait, and Eva Jablonka, eds. 2011. Transformations of Lamarckism: From Subtle Fluids to Molecular Biology. Cambridge: MIT Press.

260 ◂ REFERENCES

Glennan, Stuart. 1996. “Mechanisms and the Nature of Causation.” Erkenntnis 44:49–71. Glennan, Stuart. 2002. “Rethinking Mechanistic Explanation.” Philosophy of Sci- ence 69:S342–53. Glennan, Stuart. 2005. “Modeling Mechanisms.” Studies in History and Philosophy of Biological and Biomedical Sciences 36:443–64. Glennan, Stuart. 2009. “Mechanisms.” In The Oxford Handbook of Causation, edited by Helen Beebee, Christopher Hitchcock, and Peter Menzies, 315–25. Oxford: Oxford University Press. Gliboff, Sander. 2010. “Did Paul Kammerer Discover Epigenetic Inheritance? No and Why Not.” Journal of Experimental Zoology B 314:616–24. Gluckman, Peter D., and Mark A. Hanson. 2004. “Living with the Past: Evolution, Development, and Patterns of Disease.” Science 305:1733–36. Gluckman, Peter D., Mark A. Hanson, and Tatjana Buklijas. 2011. “Maternal and Transgenerational Influence on Human Health.” InTransformations of Lamarckism: From Subtle Fluids to Molecular Biology, edited by Snait Gissis and Eva Jablonka, 237–50. Cambridge: MIT Press. Gluckman, Peter D., Mark A. Hanson, Cyrus Cooper, and Kent L. Thornburg. 2008. “Effect of In Utero and Early-Life Conditions on Adult Health and Dis- e a s e .” New England Journal of Medicine 359:61–73. Godfrey-Smith, Peter. 1996. Complexity and the Function of Mind in Nature. Cam- bridge: Cambridge University Press. Goldberg, Aaron D., C. David Allis, and Emily Bernstein. 2007. “Epigenetics: A Landscape Takes Shape.” Cell 128:635–38. Goldschmidt, Richard. 1940. The Material Basis of Evolution. New Haven: Yale University Press. Goodwin, Brian C. 1963. Temporal Organization in Cells. London: Academic Press. Goodwin, Brian C. 1965. “Oscillatory Behavior in Enzymatic Control Processes.” Advances in Enzyme Regulation 3:425–37. Goodwin, Brian C., and Peter T. Saunders. 1989. Theoretical Biology: Epigenetic and Evolutionary Order from Complex Systems. Edinburgh: Edinburgh Uni- versity Press. Gould, Stephen J., and Nils Eldredge. 1977. “Punctuated Equilibria: The Tempo and Mode of Evolution Reconsidered.” Paleobiology 3:115–51. Gould, Stephen J., and Richard C. Lewontin. 1979. “The Spandrels of San Marco and the Panglossian Paradigm: A Critique of the Adaptationist Programme.” Philosophical Transactions of the Royal Society B 205:581–98.

▸ 261 REFERENCES

Grant, Bob. 2010. “Should Evolutionary Theory Evolve?” Scientist 24, 1 January. http://www.the-scientist.com/article/print/56251/. Gravlee, Clarence C. 2009. “How Race Becomes Biology: Embodiment of Social Inequality.” American Journal of Physical Anthropology 139:47–57. Green, Sara, Melinda B. Fagan, and Johannes Jaeger. 2015. “Explanatory Integra- tion Challenges in Evolutionary Systems Biology.” Biological Theory 10:18–35. Griesemer, James. 1990. “Material Models in Biology.” PSA: Proceedings of the Biennial Meeting of the Philosophy of Science Association 1990 (2): 79–93. Griesemer, James 1991. “Must Scientific Diagrams Be Eliminable? The Case of Path Analysis.” Biology and Philosophy 6:155–80. Griesemer, James. 2002. “What Is ‘Epi’ about Epigenetics?” Annals of the New York Academy of Sciences 981:97–110. Griesemer, James. 2011a. “Heuristic Reductionism and the Relative Significance of Epigenetic Inheritance.” In Epigenetics: Linking Genotype and Phenotype in Development and Evolution, edited by Benedikt Hallgrímsson and Brian K. Hall, 14–40. Berkeley: University of California Press. Griesemer, James. 2011b. “Philosophy and Tinkering.” Biology and Philosophy 26:269–79. Griesemer, James. 2011c. “The Relative Significance of Epigenetic Inheritance in Evolution: Some Philosophical Considerations.” In Transformations of Lamarckism: From Subtle Fluids to Molecular Biology, edited by Snait Gissis and Eva Jablonka, 331–44. Cambridge: MIT Press. Griffiths, Paul E. 2011. “Our Plastic Nature.” InTransformations of Lamarck- ism: From Subtle Fluids to Molecular Biology, edited by Snait Gissis and Eva Jablonka, 319–30. Cambridge: MIT Press. Griffiths, Paul E., and Russell D. Gray. 1994. “Developmental Systems and Evolu- tionary Explanation.” Journal of Philosophy 91:277–304. Griffiths, Paul E., and Russell D. Gray. 2005. “Discussion: Three Ways to Misun- derstand Developmental Systems Theory.” Biology and Philosophy 20:417–25. Griffiths, Paul E., Arnaud Pocheville, Brett Calcott, Karola Stotz, Hyunju Kim, and Rob Knight. 2015. “Measuring Causal Specificity.” Philosophy of Science 82:529–55. Griffiths, Paul E., and Karola Stotz. 2006. “Genes in the Postgenomic Era.” Theo- retical Medicine and Bioethics 27:499–521. Griffiths, Paul, and Karola Stotz. 2013. Genetics and Philosophy: An Introduction. Cambridge: Cambridge University Press.

262 ◂ REFERENCES

Grishok, Alla. 2005. “RNAi Mechanisms in Caenorhabditis elegans.” FEBS Letters 579:5932–39. Gross, Charles G. 1998. “Claude Bernard and the Constancy of the Internal Envi- ronment.” Neuroscientist 4:380–85. Gross, Fridolin. 2015. “The Relevance of Irrelevance: Explanation in Systems Biol- ogy.” In Explanation in Biology: An Enquiry into the Diversity of Explanatory Patterns in the Life Sciences, edited by Pierre-Alain Braillard and Christophe Malaterre, 175–98. Berlin: Springer. Guerrero-Bosagna, Carlos. 2012. “Finalism in Darwinian and Lamarckian Evo- lution: Lessons from Epigenetics and Developmental Biology.” Evolutionary Biology 39:283–300. Guerrero-Bosagna, Carlos, Matthew Settles, Ben Lucker, and Michael K. Skin- ner. 2010. “Epigenetic Transgenerational Actions of Vinclozolin on Promoter Regions of the Sperm Epigenome.” PLoS ONE 5:13100. doi:10.1371/journal. pone.0013100. Gunawardena, Jeremy. 2010. “Models in Systems Biology: The Parameter Problem and the Meanings of Robustness.” In Elements of Computational Systems Biol- ogy, edited by Huma M. Lodhi and Stephen H. Muggleton, 19–47. Hoboken: Wiley. Gurdon, John. 2006. “From Nuclear Transfer to Nuclear Reprogramming: The Reversal of Cell Differentiation.” Annual Review of Cell and Developmental Biology 22:1–22. Haig, David. 2004. “The (Dual) Origin of Epigenetics.”Cold Spring Harbor Sym- posia on Quantitative Biology 69:67–70. Haig, David. 2007. “Weismann Rules! OK? Epigenetics and the Lamarckian Temp- tation.” Biology and Philosophy 22:415–28. Haig, David. 2011. “Lamarck Ascending!” Philosophy and Theory in Biology 3:e204. doi:http://dx.doi.org/10.3998/ptb.6959004.0003.004. Haig, David. 2012. “Commentary: The Epidemiology of Epigenetics.” Interna- tional Journal of Epidemiology 41:13–16. Haig, David. 2013. “Proximate and Ultimate Causes: How Come? And What For?” Biology and Philosophy 28:781–86. Hall, Brian K. 1992. “Waddington’s Legacy in Development and Evolution.” Amer- ican Zoologist 32:113–22. Hall, Brian K. 1998. “Epigenetics: Regulation Not Replication.” Journal of Evolu- tionary Biology 11:201–5.

▸ 263 REFERENCES

Hall, Brian K. 2000. “Evo-Devo or Devo-Evo—Does It Matter?” Evolution and Development 2:177–78. Hall, Brian K. 2012. “Lamarck, Lamarckism, Epigenetics and Epigenetic Inheri- tance.” Metascience 21:375–78. Hallgrímsson, Benedikt, Katherine Willmore, and Brian K. Hall. 2002. “Canal- ization, Developmental Stability, and Morphological Integration in Primate Limbs.” American Journal of Physical Anthropology 35:131–58. Hamburger, Viktor. 1980. “Embryology and the Modern Synthesis.” In The Evolutionary Synthesis: Perspectives on the Unification of Biology, edited by Ernst Mayr and William B. Provine, 97–112. Cambridge: Harvard Univer- sity Press. Hanson, Norwood R. 1962. “The Irrelevance of History of Science to the Philoso- phy of Science.” Journal of Philosophy 59:574–85. Haraway, Donna J. 1976. Crystals, Fabrics, and Fields: Metaphors That Shape Embryos. New Haven: Yale University Press. Harris, Kami D. M., Nicholas J. Bartlett, and Vett K. Lloyd. 2012. “Daphnia as an Emerging Epigenetic Model Organism.” Genetics Research International 2012:147892. doi:10.1155/2012/147892. Harrison, Ross G. 1937. “Embryology and Its Relations.” Science 85:369–74. Hartmann, Max. 1937. Philosophie der Naturwissenschaften. [Philosophy of the natural sciences]. Berlin: Springer. Hausman, Daniel M., and James Woodward. 1999. “Independence, Invariance and the Causal Markov Condition.” British Journal for the Philosophy of Science 50:521–83. He, Lin, and Gregory J. Hannon. 2004. “MicroRNAs: Small RNAs with a Big Role in Gene Regulation.” Nature Reviews Genetics 5:522–31. Heard, Edith, and Robert A. Martienssen. 2014. “Transgenerational Epigenetic Inheritance: Myths and Mechanisms.” Cell 157:95–109. Hedlund, Maria. 2012. “Epigenetic Responsibility.” Medicine Studies 3:171–83. Helmholtz, Hermann von. 1847. Ueber die Erhaltung der Kraft [On the conservation of force]. Berlin: G. Reimer. Hempel, Carl G. 1965. Aspects of Scientific Explanation. New York: Free Press. Hendrikse, Jesse L., Trish E. Parsons, and Benedikt Hallgrímsson. 2007. “Evolv- ability as the Proper Focus of Evolutionary Developmental Biology.” Evolution and Development 9:393–401. Hendry, Robin F. 2006. “Is There Downward Causation in Chemistry?” In Phi-

264 ◂ REFERENCES

losophy of Chemistry, edited by Davis Baird, Eric Scerri, and Lee McIntyre, 173–89. Dordrecht: Springer. Henning, Brian G., and Adam C. Scarfe, eds. 2013. Beyond Mechanism: Putting Life Back into Biology. Lanham: Rowman & Littlefield. Herman, Jacob J., Hamish G. Spencer, Kathleen Donohue, and Sonia E. Sultan. 2013. “How Stable ‘Should’ Epigenetic Modifications Be? Insights from Adap- tive Plasticity and Bet Hedging.” Evolution 68:632–43. Herrera, Carlos M., and Pilar Bazaga. 2011. “Untangling Individual Variation in Natural Populations: Ecological, Genetic and Epigenetic Correlates of Long- Term Inequality in Herbivory.” Molecular Ecology 20:1675–88. Hertwig, Oscar. 1900. Die Entwicklung der Biologie im 19. Jahrhundert [The development of biology in the 19th century]. Jena: G. Fisher. Hesse, Mary B. 1966. Models and Analogies in Science. Notre Dame: University of Notre Dame Press. Hiatt, Anna, Gregory K. Davis, Caleb Trujillo, Mark Terry, Donald P. French, Rebecca M. Price, and Kathryn E. Perez. 2013. “Getting to Evo-Devo: Concepts and Challenges for Students Learning Evolutionary Developmental Biology.” CBE Life Sciences Education 12:494–508. Hitchcock, Christopher. 1996. “The Role of Contrast in Causal and Explanatory Claims.” Synthese 107:395–419. Hitchcock, Christopher, and James Woodward. 2003. “Explanatory Generaliza- tions, Part II: Plumbing Explanatory Depth.” Noûs 37:181–99. Ho, Mae-Wan, and Peter T. Saunders. 1979. “Beyond Neo-Darwinism: An Epigen- etic Approach to Evolution.” Journal of Theoretical Biology 78:573–91. Hochedlinger, Konrad, and Kathrin Plath. 2009. “Epigenetic Reprogramming and Induced Pluripotency.” Development 136:509–23. Hochman, Adam. 2013. “The Phylogeny Fallacy and the Ontogeny Fallacy.” Biol- ogy and Philosophy 28:593–612. Hodgkin, Jonathan. 1998. “Seven Types of Pleiotropy.” International Journal of Developmental Biology 42:501–5. Holland, John H. 1995. Hidden Order: How Adaptation Builds Complexity. Read- ing: Addison-Wesley. Holland, John H. 1998. Emergence: From Chaos to Order. Oxford: Oxford Univer- sity Press. Holland, Paul W. 1986. “Statistics and Causal Inference.” Journal of the American Statistical Association 81:945–60.

▸ 265 REFERENCES

Holliday, Robin. 1994. “Epigenetics: An Overview.” Developmental Genetics 15:453–57. Holliday, Robin. 2002. “Epigenetics Comes of Age in the Twenty-First Century.” Journal of Genetics 81:1–4. Holliday, Robin. 2006. “Epigenetics: A Historical Overview.” Epigenetics 1:76–80. Holm, Sune, and Russell Powell. 2013. “Organism, Machine, Artifact: The Con- ceptual and Normative Challenges of Synthetic Biology.” Studies in History and Philosophy of Biological and Biomedical Sciences 44:627–31. Holmquist, Gerald P., and Terry Ashley. 2006. “Chromosome Organization and Chromatin Modification: Influence on Genome Function and Evolution.” Cytogenetic and Genome Research 114:96–125. Houchmandzadeh, Bahram, Eric Wieschaus, and Stanislas Leibler. 2002. “Estab- lishment of Developmental Precision and Proportions in the Early Drosophila Embryo.” Nature 415:798–802. Houri-Ze’evi, Leah, Yael Korem, Hila Sheftel, Lior Faigenbloom, Itai A. Toker, Yael Dagan, Lama Awad, Luba Degani, Uri Alon, and Oded Rechavi. 2016. “A Tunable Mechanism Determines the Duration of the Transgenerational Small RNA Inheritance in C. elegans.” Cell 165:88–99. Huang, Sui. 2009a. “Non-Genetic Heterogeneity of Cells in Development: More Than Just Noise.” Development 136:3853–62. Huang, Sui. 2009b. “Reprogramming Cell Fates: Reconciling Rarity with Robust- ness.” Bioessays 31:546–60. Huang, Sui. 2011. “Systems Biology of Stem Cells: Three Useful Perspectives to Help Overcome the Paradigm of Linear Pathways.” Philosophical Transactions of the Royal Society B 366:2247–59. Huang, Sui. 2012a. “The Molecular and Mathematical Basis of Waddington’s Epi- genetic Landscape: A Framework for Post-Darwinian Biology?” BioEssays 34:149–57, S1–5. Huang, Sui. 2012b. “Tumor Progression: Chance and Necessity in Darwinian and Lamarckian Somatic (Mutationless) Evolution.” Progress in Biophysics and Molecular Biology 110:69–86. Huang, Sui, Ingemar Ernberg, and Stuart Kauffman. 2009. “Cancer Attractors: A Systems View of Tumors from a Gene Network Dynamics and Developmental Perspective.” Seminars in Cell and Developmental Biology 20:869–76. Hull, David L. 1974. Philosophy of Biological Science. Englewood Cliffs: Pren- tice-Hall. Huneman, Philippe. 2010. “Assessing the Prospects for a Return of Organisms in Evolutionary Biology.” History and Philosophy of the Life Sciences 32:341–71.

266 ◂ REFERENCES

Huo, Fang-Jun, Yuan-Qiang Sun, Jing Su, Yu-Tao Yang, Cai-Xia Yin, and Jian- Bin Chao. 2010. “Chromene ‘Lock,’ Thiol ‘Key,’ and Mercury(II) Ion ‘Hand’: A Single Molecular Machine Recognition System.” Organic Letters 12:4756–59. Huxley, Julian S. 1916. “Bird-Watching and Biological Science: Some Observations on the Study of Courtship in Birds.” Auk 33:142–61. Illari, Phyllis M., and Jon Williamson. 2010. “Function and Organization: Com- paring the Mechanisms of Protein Synthesis and Natural Selection.” Studies in History and Philosophy of Biological and Biomedical Sciences 41:279–91. Illari, Phyllis M., and Jon Williamson. 2012. “What Is a Mechanism? Thinking about Mechanisms across the Sciences.” European Journal for Philosophy of Science 2:119–35. Jablonka, Eva. 2002. “Information: Its Interpretation, Its Inheritance, and Its Shar- ing.” Philosophy of Science 69:578–605. Jablonka, Eva. 2006. “Genes as Followers in Evolution—a Post-Synthesis Synthe- sis?” Biology and Philosophy 21:143–54. Jablonka, Eva. 2013. “Epigenetic Inheritance and Plasticity: The Responsive Germ- line.” Progress in Biophysics and Molecular Biology 111:99–107. Jablonka, Eva, Michael Lachmann, and Marion J. Lamb. 1992. “Evidence, Mecha- nisms and Models for the Inheritance of Acquired Characters.” Journal of The- oretical Biology 158:245–68. Jablonka, Eva, and Marion J. Lamb. 1989. “The Inheritance of Acquired Epigenetic Variations.” Journal of Theoretical Biology 139:69–83. Jablonka, Eva, and Marion J. Lamb. 1995. Epigenetic Inheritance and Evolution: The Lamarckian Dimension. Oxford: Oxford University Press. Jablonka, Eva, and Marion J. Lamb. 2002. “The Changing Concept of Epigenetics.” Annals of the New York Academy of Sciences 981:82–96. Jablonka, Eva, and Marion J. Lamb. 2005. Evolution in Four Dimensions: Genetic, Epigenetic, Behavioral, and Symbolic Variation in the History of Life. Cam- bridge: MIT Press. Jablonka, Eva, and Marion J. Lamb. 2007. “The Expanded Evolutionary Synthesis: A Response to Godfrey-Smith, Haig, and West-Eberhard.” Biology and Philos- ophy 22:453–72. Jablonka, Eva, and Marion J. Lamb. 2008. “Soft Inheritance: Challenging the Modern Synthesis.” Genetics and Molecular Biology 31:389–95. Jablonka, Eva, and Marion J. Lamb. 2010. “Transgeneration Epigenetic Inheri- tance.” In Evolution: The Extended Synthesis, edited by Massimo Pigliucci and Gerd B. Müller, 137–74. Cambridge: MIT Press.

▸ 267 REFERENCES

Jablonka, Eva, Marion J. Lamb, and Eytan Avital. 1998. “‘Lamarckian’ Mechanisms in Darwinian Evolution.” Trends in Ecology and Evolution 13:206–10. Jablonka, Eva, and Ehud Lamm. 2012. “Commentary: The Epigenotype; A Dynamic Network View of Development.” International Journal of Epidemi- ology 41:16–20. Jablonka, Eva, Marjori Matzke, Denis Thieffry, and Linda Van Speybroeck. 2002. “The Genome in Context: Biologists and Philosophers on Epigenetics.” Bioes- says 24:392–94. Jablonka, Eva, and Gal Raz. 2009. “Transgenerational Epigenetic Inheritance: Prevalence, Mechanisms, and Implications for the Study of Heredity and Evo- lution.” Quarterly Review of Biology 84:131–76. Jacob, François. 1974. The Logic of Life: A History of Heredity. New York: Pantheon. Jacob, François, and Jacques Monod. 1959. “Gènes de structure et gènes de régula- tion dans la biosynthèse des proteins” [Genes of structure and genes of regulation in the biosynthesis of proteins]. Comptes Rendus de l’Académie des Sciences de Paris 249:1282–84. Jacob, François, and Jacques Monod. 1961. “Genetic Regulatory Mechanisms in the Synthesis of Proteins.” Journal of Molecular Biology 3:318–56. Janzen, Fredric J., and Patrick C. Phillips. 2006. “Exploring the Evolution of Envi- ronmental Sex Determination, Especially in Reptiles.” Journal of Evolutionary Biology 19:1775–84. Jenner, Ronald A. 2006. “Unburdening Evo-Devo: Ancestral Attractions, Model Organisms, and Basal Baloney.” Development Genes and Evolution 216:385–94. Jenner, Ronald A., and Matthew A. Wills. 2007. “The Choice of Model Organisms in Evo-Devo.” Nature Reviews Genetics 8:311–14. Jinek, Martin, Krzysztof Chylinski, Ines Fonfara, Michael Hauer, Jennifer A. Doudna, and Emmanuelle Charpentier. 2012. “A Programmable Dual-RNA– Guided DNA Endonuclease in Adaptive Bacterial Immunity.” Science 337:816– 21. Johannes, Frank, Vincent Colot, and Ritsert C. Jansen. 2008. “Epigenome Dynam- ics: A Quantitative Genetics Perspective.” Nature Reviews Genetics 9:883–90. Johannes, Frank, Emmanuelle Porcher, Felipe K. Teixeira, Vera Saliba-Colom- bani, Matthieu Simon, Nicolas Agier, Agnès Bulski, Juliette Albuisson, Fabi- ana Heredia, Pascal Audigier, David Bouchez, Christine Dillmann, Philippe Guerche, Frédéric Hospital, and Vincent Colot. 2009. “Assessing the Impact of Transgenerational Epigenetic Variation on Complex Traits.” PLoS Genetics 5:1000530. doi:10.1371/journal.pgen.1000530.

268 ◂ REFERENCES

Johnson, Norman, and Adam Porter. 2001. “Toward a New Synthesis: Population Genetics and Evolutionary Developmental Biology.” Genetica 112–13:45–58. Johnston, Mark. 1992. “Constitution Is Not Identity.” Mind 101:89–106. Jones, Martin R. 2005. “Idealization and Abstraction: A Framework.” In Ideal- ization XII: Correcting the Model; Idealization and Abstraction in the Sciences, edited by Nancy Cartwright and Martin R. Jones, 173–217. Amsterdam: Rodopi. Joo, Ji-Hoon E., Roberta H. Andronikos, and Richard Saffery. 2011. “Metabolic Regulation of DNA Methylation in Mammals.” In Handbook of Epigenetics, edited by Trygve Tollefsbol, 281–93. San Diego: Academic Press. Just, Ernest Everett. 1939. The Biology of the Cell Surface. Philadelphia: Blakiston. Kahn, Fritz. 1926. Das Leben des Menschen [The life of man]. Vol. 3. Stuttgart: Franckh’sche Verlagshandlung. Kaiser, Marie I. 2011. “The Limits of Reductionism in the Life Sciences.” History and Philosophy of the Life Sciences 33:453–76. Kammerer, Paul. 1906. “Experimentelle Veränderung der Fortpflanzungstätigkeit bei Geburtshelferkröte (Alytes obstetricans) und Laubfrosch (Hyla arborea).” Archiv für Entwicklungsmechanik der Organismen 22:48–140. Kammerer, Paul. 1924. The Inheritance of Acquired Characteristics. New York: Boni and Liveright. Kappeler, Laurent, and Michael J. Meaney. 2010. “Epigenetics and Parental Effects.” BioEssays 32:818–27. Kauffman, Stuart A. 1969. “Metabolic Stability and Epigenesis in Randomly Con- structed Genetic Nets.” Journal of Theoretical Biology 22:437–67. Kauffman, Stuart A. 1971a. “Articulation of Parts Explanation in Biology and the Rational Search for Them.” PSA: Proceedings of the Biennial Meeting of the Phi- losophy of Science Association 1970:257–72. Kauffman, Stuart A. 1971b. “Differentiation of Malignant to Benign Cells.” Journal of Theoretical Biology 31:429–51. Kay, Lily E. 2000. Who Wrote the Book of Life? A History of the Genetic Code. Stan- ford: Stanford University Press. Kegel, Bernhard. 2009. Epigenetik: Wie Erfahrungen vererbt werden [Epigenetics: How experience is inherited]. Cologne: DuMont. Keijzer, Fred. 2015. “Moving and Sensing without Input and Output: Early Ner- vous Systems and the Origins of the Animal Sensorimotor Organization.” Biol- ogy and Philosophy 30:311–31. Keller, Evelyn F. 1995. Refiguring Life: Metaphors of Twentieth-Century Biology. New York: Columbia University Press.

▸ 269 REFERENCES

Keller, Evelyn F. 2000a. “Models of and Models For: Theory and Practice in Con- temporary Biology.” Philosophy of Science 67:86. Keller, Evelyn F. 2000b. The Century of the Gene. Cambridge: Harvard University Press. Keller, Evelyn F. 2002. Making Sense of Life: Explaining Biological Development with Models, Metaphors, and Machines. Cambridge: Harvard University Press. Kelly, William G., and Rodolfo Aramayo. 2007. “Meiotic Silencing and the Epi- genetics of Sex.” Chromosome Research: An International Journal on the Molec- ular, Supramolecular and Evolutionary Aspects of Chromosome Biology 15:633– 51. Kim, Jaegwon. 1999. “Making Sense of Emergence.” Philosophical Studies 95:3–36. Kim, Jaegwon. 2006. “Emergence: Core Ideas and Issues.” Synthese 151:547–59. Kirschner, Marc, and John Gerhart. 1998. “Evolvability.” Proceedings of the National Academy of Sciences 95:8420–27. Kirschner, Marc W., and John C. Gerhart. 2010. “Facilitated Variation.” In Evolu- tion: The Extended Synthesis, edited by Massimo Pigliucci and Gerd B. Müller, 253–80. Cambridge: MIT Press. Kitcher, Philip. 1989. “Explanatory Unification and the Causal Structure of the World.” In Scientific Explanation, edited by Philip Kitcher and Wesley C. Salmon, 410–505. Minneapolis: University of Minnesota Press. Klironomos, Filippos D., Johannes Berg, and Sinéad Collins. 2013. “How Epigen- etic Mutations Can Affect Genetic Evolution: Model and Mechanism.”BioEs - says 35:571–78. Kohler, Robert E. 1994. Lords of the Fly: Drosophila Genetics and the Experimental Life. Chicago: University of Chicago Press. Koonin, Eugene V., and Yuri I. Wolf. 2009. “Is Evolution Darwinian or/and Lamarckian?” Biology Direct 4:42. doi:10.1186/1745–6150–4-42. Krimsky, Sheldon, and Jeremy Gruber, eds. 2013. Genetic Explanations: Sense and Nonsense. Cambridge: Harvard University Press. Kuhlmann, Meinard, and Stuart Glennan. 2014. “On the Relation between Quan- tum Mechanical and Neo-Mechanistic Ontologies and Explanatory Strategies.” European Journal for Philosophy of Science 4:337–59. Kuhn, Thomas S. 1962.The Structure of Scientific Revolutions. Chicago: University of Chicago Press. Kulvicki, John. 2010. “Knowing with Images: Medium and Message.” Philosophy of Science 77:295–313.

270 ◂ REFERENCES

Kuorikoski, Jaakko. 2009. “Two Concepts of Mechanism: Componential Causal System and Abstract Form of Interaction.” International Studies in the Philoso- phy of Science 23:143–60. Kuorikoski, Jaakko. 2012. “Mechanisms, Modularity and Constitutive Explana- tion.” Erkenntnis 77:361–80. Kuorikoski, Jaakko, and Petri Ylikoski. 2013. “How Organization Explains.” In EPSA11 Perspectives and Foundational Problems in Philosophy of Science, edited by Vassilios Karakostas and Dennis Dieks, 69–80. Dordrecht: Springer. Kutschera, Ulrich, and Karl J. Niklas. 2004. “The Modern Theory of Biological Evolution: An Expanded Synthesis.” Naturwissenschaften 91:255–76. Kuzawa, Christopher W., and Elizabeth Sweet. 2009. “Epigenetics and the Embod- iment of Race: Developmental Origins of US Racial Disparities in Cardiovas- cular Health.” American Journal of Human Biology 21:2–15. Lack, David L. 1954. The Natural Regulation of Animal Numbers. Oxford: Oxford University Press. Laforsch, Christian, and Ralph Tollrian. 2004. “Embryological Aspects of Induc- ible Morphological Defenses in Daphnia.” Journal of Morphology 262:701–7. Lakatos, Imre. 1970. “Falsification and the Methodology of Scientific Research Programmes.” In Criticism and the Growth of Knowledge, edited by Imre Laka- tos and Alan Musgrave, 91–195. Cambridge: Cambridge University Press. Lakatos, Imre. 1971. “History of Science and Its Rational Reconstructions.” In PSA: Proceedings of the Biennial Meeting of the Philosophy of Science Association 1970:91–136. Lakoff, George, and Mark Johnson. 1980. Metaphors We Live By. Chicago: Univer- sity of Chicago Press. Lakoff, George, and Mark Johnson. 1999. Philosophy in the Flesh: The Embodied Mind and Its Challenge to Western Thought. New York: Basic Books. Laland, Kevin N., John Odling-Smee, Marcus W. Feldman, and Jeremy Kendal. 2009. “Conceptual Barriers to Progress within Evolutionary Biology.” Founda- tions of Science 14:195–216. Laland, Kevin N., John Odling-Smee, and Scott F. Gilbert. 2008. “EvoDevo and Niche Construction: Building Bridges.” Journal of Experimental Zoology B 310:549–66. Laland, Kevin N., John Odling-Smee, William Hoppitt, and Tobias Uller. 2013a. “More on How and Why: A Response to Commentaries.” Biology and Philos- ophy 28:793–810.

▸ 271 REFERENCES

Laland, Kevin N., John Odling-Smee, William Hoppitt, and Tobias Uller. 2013b. “More on How and Why: Cause and Effect in Biology Revisited.” Biology and Philosophy 28:719–45. Laland, Kevin N., and Kim Sterelny. 2006. “Perspective: Seven Reasons (Not) to Neglect Niche Construction.” Evolution 60:1751–62. Laland, Kevin N., Kim Sterelny, John Odling-Smee, William Hoppitt, and Tobias Uller. 2011. “Cause and Effect in Biology Revisited: Is Mayr’s Proximate-Ulti- mate Dichotomy Still Useful?” Science 334:1512–16. Laland, Kevin, Tobias Uller, Marc Feldman, Kim Sterelny, Gerd B. Müller, Armin Moczek, Eva Jablonka, and John Odling-Smee. 2014. “Does Evolutionary The- ory Need a Rethink? Yes, Urgently.” Nature News 514:161–64. Laland, Kevin N., Tobias Uller, Marcus W. Feldman, Kim Sterelny, Gerd B. Müller, Armin Moczek, Eva Jablonka, and John Odling-Smee. 2015. “The Extended Evolutionary Synthesis: Its Structure, Assumptions and Predictions.” Proceed- ings of the Royal Society B 282:1813. doi:10.1098/rspb.2015.1019. Lamarck, Jean-Baptiste de. 1809. Philosophie zoologique, ou exposition des considé- rations relatives à l’histoire naturelle des animaux. Paris: Dentu. Lamm, Ehud. 2013. “Epigenetic Mechanisms Underlie Genome Development.” International Journal of Developmental Science 7:87–91. Lamm, Ehud. 2014. “Inheritance Systems.” In The Stanford Encyclopedia of Phi- losophy, Winter ed., edited by Edward N. Zalta. http://plato.stanford.edu/ archives/spr2014/entries/inheritance-systems/. Landman, Otto E. 1991. “The Inheritance of Acquired Characteristics.”Annual Review of Genetics 25:1–20. Lange, Ulrike C., and Robert Schneider. 2010. “What an Epigenome Remembers.” BioEssays 32:659–68. Lango Allen, Hana, Karol Estrada, Guillaume Lettre, Sonja I. Berndt, Michael N. Weedon, Fernando Rivadeneira, Cristen J. Willer, Anne U. Jackson, Sailaja Vedantam, Soumya Raychaudhuri, Teresa Ferreira, Andrew R. Wood, Robert J. Weyant, Ayellet V. Segrè, Elizabeth K. Speliotes, Eleanor Wheeler, Nicole Soranzo, Ju-Hyun Park, Jian Yang, Daniel Gudbjartsson, et al. 2010. “Hundreds of Variants Clustered in Genomic Loci and Biological Pathways Affect Human Height.” Nature 467:832–38. Latzel, Vít, Eric Allan, Amanda Bortolini Silveira, Vincent Colot, Markus Fischer, and Oliver Bossdorf. 2013. “Epigenetic Diversity Increases the Productivity and Stability of Plant Populations.” Nature Communications 4:2875. doi:10.1038/ ncomms3875.

272 ◂ REFERENCES

Laubichler, Manfred D. 2005. “Das Forschungsprogramm der evolutionären Ent- wicklungsbiologie.” [The research program of evolutionary developmental biology]. In Philosophie der Biologie: Eine Einführung, edited by Ulrich Krohs and Georg Toepfer, 322–32. Frankfurt: Suhrkamp. Laubichler, Manfred D. 2007. “Evolutionary Developmental Biology.” In The Cambridge Companion to the Philosophy of Biology, edited by David Hull and Michael Ruse, 342–60. Cambridge: Cambridge University Press. Laubichler, Manfred D. 2010. “Evolutionary Developmental Biology Offers a Sig- nificant Challenge to the Neo-Darwinian Paradigm.” In Contemporary Debates in Philosophy of Biology, edited by Francisco J. Ayala and Robert Arp, 199–212. Malden: Wiley. Laubichler, Manfred D., and Jane Maienschein. 2007. From Embryology to Evo- Devo: A History of Developmental Evolution. Cambridge: MIT Press. Laubichler, Manfred D., and Jane Maienschein, eds. 2009. Form and Function in Developmental Evolution. Cambridge: Cambridge University Press. Ledón-Rettig, Cris C. 2013. “Ecological Epigenetics: An Introduction to the Sym- posium.” Integrative and Comparative Biology 53:307–18. Ledón-Rettig, Cris C., and David W. Pfennig. 2011. “Emerging Model Systems in Eco-Evo-Devo: The Environmentally Responsive Spadefoot Toad.” Evolution and Development 13:391–400. Ledón-Rettig, Cris C., Christina L. Richards, and Lynn B. Martin. 2013. “Epi- genetics for Behavioral Ecologists.” Behavioral Ecology 24:311–24. Lehman, Hugh. 1965. “Functional Explanation in Biology.” Philosophy of Science 32:1–20. Lem, Stanisław. 2013. Summa Technologiae. Minneapolis: University of Minnesota Press. Leonelli, Sabina. 2007. “What Is in a Model?” In Modeling Biology: Structures, Behavior, Evolution, edited by Manfred D. Laubichler and Gerd B. Müller, 15–36. Cambridge: MIT Press. Leonelli, Sabina. 2013. “Integrating Data to Acquire New Knowledge: Three Modes of Integration in Plant Science.” Studies in History and Philosophy of Biological and Biomedical Sciences 44:503–14. Leonelli, Sabina, and Rachel A. Ankeny. 2013. “What Makes a Model Organism?” Endeavour 37:209–12. Lepzelter, David, and Jin Wang. 2008. “Exact Probabilistic Solution of Spatial-De- pendent Stochastics and Associated Spatial Potential Landscape for the Bicoid Protein.” Physical Review E 77:041917. doi:10.1103/PhysRevE.77.041917.

▸ 273 REFERENCES

Leuridan, Bert. 2010. “Can Mechanisms Really Replace Laws of Nature?” Philos- ophy of Science 77:317–40. Leuridan, Bert. 2012. “Three Problems for the Mutual Manipulability Account of Constitutive Relevance in Mechanisms.” British Journal for the Philosophy of Science 63:399–427. Leuridan, Bert, and Anton Froeyman. 2012. “On Lawfulness in History and His- toriography.” History and Theory 51:172–92. Levins, Richard, and Richard C. Lewontin. 1985. The Dialectical Biologist. Cam- bridge: Harvard University Press. Levy, Arnon. 2013. “Three Kinds of New Mechanism.” Biology and Philosophy 28:99–114. Levy, Arnon. 2014. “What Was Hodgkin and Huxley’s Achievement?” British Jour- nal for the Philosophy of Science 65:469–92. Levy, Arnon, and William Bechtel. 2013. “Abstraction and the Organization of Mechanisms.” Philosophy of Science 80:241–61. doi:10.1086/670300. Lewens, Tim. 2009. “Evo-Devo and ‘Typological Thinking’: An Exculpation.” Journal of Experimental Zoology B 312:789–96. Lewis, David. 1973. Counterfactuals. Oxford: Blackwell Publishers. Lewis, David. 2000. “Causation as Influence.” Journal of Philosophy 97:182–97. Lewontin, Richard C. 1982. “Organism and Environment.” In Learning, Develop- ment, and Culture: Essays in Evolutionary Epistemology, edited by Henry C. Plotkin, 151–72. New York: Wiley. Li, Jiong, Jin Huang, Jian-Shuang Li, Hong Chen, Kun Huang, and Ling Zheng. 2012. “Accumulation of Endoplasmic Reticulum Stress and Lipogenesis in the Liver through Generational Effects of High Fat Diets.” Journal of Hepatology 56:900–907. Lickliter, Robert, and Thomas D. Berry. 1990. “The Phylogeny Fallacy: Develop- mental Psychology’s Misapplication of Evolutionary Theory.” Developmental Review 10:348–64. Lilienfeld, Abraham M., and David E. Lilienfeld. 1980. Foundations of Epidemiol- ogy. 2nd ed. Oxford: Oxford University Press. Lillie, Frank R. 1927. “The Gene and the Ontogenetic Process.” Science 66: 361–68. Lipton, Bruce H. 2005. The Biology of Belief: Unleashing the Power of Conscious- ness, Matter and Miracles. Santa Rosa: Mountain Love/Elite Books. Lipton, Peter. 1990. “Contrastive Explanation.” Royal Institute of Philosophy Sup- plement 27:247–66.

274 ◂ REFERENCES

Lipton, Peter. 1993. “Making a Difference.” Philosophica 51:39–54. Lister, Ryan, Mattia Pelizzola, Robert H. Dowen, R. David Hawkins, Gary Hon, Julian Tonti-Filippini, Joseph R. Nery, Leonard Lee, Zhen Ye, Que-Minh Ngo, Lee Edsall, Jessica Antosiewicz-Bourget, Ron Stewart, Victor Ruotti, A. Har- vey Millar, James A. Thomson, Bing Ren, and Joseph R. Ecker. 2009. “Human DNA Methylomes at Base Resolution Show Widespread Epigenomic Differ- ences.” Nature 462:315–22. Lloyd, Seth. 2001. “Measures of Complexity: A Nonexhaustive List.” IEEE Control Systems 21:7–8. Lloyd, Vett K., Jennifer A. Brisson, Kathleen A. Fitzpatrick, Lori A. McEach- ern, and Eveline C. Verhulst. 2012. “The Epigenetics of Emerging and Nonmodel Organisms.” Genetics Research International 2012:491204. doi:10.1155/2012/491204. Locke, John. (1689) 1979. An Essay Concerning Human Understanding. Edited by P. H. Nidditch. Oxford: Clarendon Press. Love, Alan C. 2003. “Evolutionary Morphology, Innovation, and the Synthe- sis of Evolutionary and Developmental Biology.” Biology and Philosophy 18: 309–45. Love, Alan C. 2009. “Marine Invertebrates, Model Organisms, and the Modern Synthesis: Epistemic Values, Evo-Devo, and Exclusion.” Theory in Biosciences 128:19–42. Love, Alan C. 2015. Conceptual Change in Biology: Scientific and Philosophical Per- spectives on Evolution and Development. Dordrecht: Springer. Love, Alan C., and Gary L. Lugar. 2013. “Dimensions of Integration in Interdisci- plinary Explanations of the Origin of Evolutionary Novelty.” Studies in History and Philosophy of Biological and Biomedical Sciences 44:537–50. Lucas, Robert. 1976. “Econometric Policy Evaluation: A Critique.” In The Phillips Curve and Labor Markets, edited by Karl Brunner and Allan H. Meltzer, 19–46. Amsterdam: North-Holland. Lux, Vanessa. 2013. “With Gottlieb beyond Gottlieb: The Role of Epigenetics in Psychobiological Development.” International Journal of Developmental Sci- ence 7:69–78. Lynch, Michael. 2007. “The Frailty of Adaptive Hypotheses for the Origins of Organismal Complexity.” Proceedings of the National Academy of Sciences 104:8597–604. MacArthur, Robert. 1955. “Fluctuations of Animal Populations and a Measure of Community Stability.” Ecology 36:533–36.

▸ 275 REFERENCES

Machamer, Peter, Lindley Darden, and Carl F. Craver. 2000. “Thinking about Mechanisms.” Philosophy of Science 67:1–25. Mackie, John L. 1974. The Cement of the Universe: A Study of Causation. Oxford: Oxford University Press. Manolio, Teri A., Lisa D. Brooks, and Francis S. Collins. 2008. “A HapMap Har- vest of Insights into the Genetics of Common Disease.” Journal of Clinical Investigation 118:1590–605. Marchionni, Caterina, and Jack Vromen. 2009. “The Ultimate/Proximate Dis- tinction in Recent Accounts of Human Cooperation.” Tijdschrift voor Filosofie 71:87–117. Martienssen, Robert A. 2010. “Small RNA Makes Its Move.” Science 328:834–35. Martínez, Maximiliano, and Maurizio Esposito. 2014. “Multilevel Causation and the Extended Synthesis.” Biological Theory 9:209–20. Mateiescu, Sebastian. 2012. “The Limits of Interventionism—Causality in the Social Sciences.” In Probabilities, Laws, and Structures, edited by Dennis Dieks, Wenceslao J. Gonzalez, Stephan Hartmann, Michel Stöltzner, and Marcel Weber, 141–56. Dordrecht: Springer. Mather, Kenneth. 1943. “Polygenic Balance in the Canalization of Development.” Nature 151:68–71. Matthewson, John, and Brett Calcott. 2011. “Mechanistic Models of Popula- tion-Level Phenomena.” Biology and Philosophy 26:737–56. Maynard Smith, John. 1966. The Theory of Evolution. 2nd ed. Middlesex: Penguin. Maynard Smith, John. 1982. Evolution and the Theory of Games. Cambridge: Cam- bridge University Press. Mayr, Ernst. 1961. “Cause and Effect in Biology.” Science 134:1501–6. Mayr, Ernst. 1980. “Prologue: Some Thoughts on the History of the Evolutionary Synthesis.” In The Evolutionary Synthesis: Perspectives on the Unification of Biol- ogy, edited by Ernst Mayr and William B. Provine, 1–48. Cambridge: Harvard University Press. Mayr, Ernst. 1984. “The Triumph of Evolutionary Synthesis.” Times Literary Sup- plement 2:1261–62. Mayr, Ernst. 2002. Foreword to Acquiring Genomes: A Theory of the Origins of Species, edited by Lynn Margulis and Dorion Sagan, xi–xiv. New York: Basic Books. McCurdy, Edward, ed. 1906. Leonardo da Vinci’s Note-Books. London: Duckworth & Co. McLaughlin, Brian P. 1997. “Emergence and Supervenience.” Intellectica 2:25–43. Mc Manus, Fabrizzio Guerrero. 2012. “Development and Mechanistic Explana-

276 ◂ REFERENCES

tion.” Studies in History and Philosophy of Biological and Biomedical Sciences 43:532–41. Meaney, Michael J. 2001. “Maternal Care, Gene Expression, and the Transmis- sion of Individual Differences in Stress Reactivity across Generations.” Annual Review of Neuroscience 24:1161–92. Meaney, Michael J., and Moshe Szyf. 2005. “Environmental Programming of Stress Responses through DNA Methylation: Life at the Interface between a Dynamic Environment and a Fixed Genome.” Dialogues in Clinical Neuroscience 7:103–23. Meloni, Maurizio. 2014a. “How Biology Became Social, and What It Means for Social Theory.” Sociological Review 62:593–614. Meloni, Maurizio. 2014b. “The Social Brain Meets the Reactive Genome: Neuro- science, Epigenetics and the New Social Biology.” Frontiers in Human Neuro- science 8:309. doi:10.3389/fnhum.2014.00309. Meloni, Maurizio. 2016. Political Biology: Science and Social Values in Human Heredity from Eugenics to Epigenetics. New York: Springer. Menzies, Peter. 2007. “Causation in Context.” In Causation, Physics, and the Con- stitution of Reality: Russell’s Republic Revisited, edited by Huw Price and Rich- ard Corry, 191–223. Oxford: Oxford University Press. Menzies, Peter. 2012. “The Causal Structure of Mechanisms.” Studies in History and Philosophy of Biological and Biomedical Sciences 43:796–805. Menzies, Peter, and Christian List. 2010. “The Causal Autonomy of the Special Sciences.” In Emergence in Mind, edited by Cynthia MacDonald and Graham MacDonald, 108–29. Oxford: Oxford University Press. Menzies, Peter, and Huw Price. 1993. “Causation as a Secondary Quality.” British Journal for the Philosophy of Science 44:187–203. Merlin, Francesca. 2010. “Evolutionary Chance Mutation: A Defense of the Modern Synthesis’ Consensus View.” Philosophy and Theory in Biology 2:103. doi:http://dx.doi.org/10.3998/ptb.6959004.0002.003. Mesoudi, Alex, Simon Blanchet, Anne Charmantier, Étienne Danchin, Laurel Fogarty, Eva Jablonka, Kevin N. Laland, Thomas J. H. Morgan, Gerd B. Müller, F. John Odling-Smee, and Benoît Pujol. 2013. “Is Non-Genetic Inheritance Just a Proximate Mechanism? A Corroboration of the Extended Evolutionary Syn- thesis.” Biological Theory 7:189–95. Mill, John Stuart. 1843. A System of Logic, Ratiocinative and Inductive. London: Parker. Minelli, Alessandro. 2009. Forms of Becoming: The Evolutionary Biology of Devel- opment. Princeton: Princeton University Press.

▸ 277 REFERENCES

Minelli, Alessandro. 2010. “Evolutionary Developmental Biology Does Not Offer a Significant Challenge to the Neo-Darwinian Paradigm.” In Contemporary Debates in Philosophy of Biology, edited by Francisco J. Ayala and Robert Arp, 213–26. Malden: Wiley. Minelli, Alessandro. 2015. “Biological Systematics in the Evo-Devo Era.” European Journal of Taxonomy 125:1–23. doi:10.5852/ejt.2015.125. Minelli, Alessandro, and Jan Baedke. 2014. “Model Organisms in Evo-Devo: Promises and Pitfalls of the Comparative Approach.” History and Philosophy of the Life Sciences 36:42–59. Minelli, Alessandro, and Giuseppe Fusco. 2004. “Evo-Devo Perspectives on Seg- mentation: Model Organisms, and Beyond.” Trends in Ecology and Evolution 19:423–29. Minelli, Alessandro, and Giuseppe Fusco, eds. 2008. Evolving Pathways: Key Themes in Evolutionary Developmental Biology. Cambridge: Cambridge Uni- versity Press. Minelli, Alessandro, and Giuseppe Fusco. 2010. “Developmental Plasticity and the Evolution of Animal Complex Life Cycles.” Philosophical Transactions of the Royal Society B 365:631–40. Mitchell, Sandra D. 1997. “Pragmatic Laws.” Philosophy of Science 64:479. Mitchell, Sandra D. 2000. “Dimensions of Scientific Law.” Philosophy of Science 67:242–65. Mitchell, Sandra D. 2002. “Ceteris Paribus—an Inadequate Representation for Biological Contingency.” Erkenntnis 57:329–50. Mitchell, Sandra D. 2003. Biological Complexity and Integrative Pluralism. Cam- bridge: Cambridge University Press. Mitchell, Sandra D. 2008. “Exporting Causal Knowledge in Evolutionary and Developmental Biology.” Philosophy of Science 75:697–706. Mitchell, Sandra D. 2009. Unsimple Truths: Science, Complexity, and Policy. Chi- cago: University of Chicago Press. Moczek, Armin P. 2007. “Developmental Capacitance, Genetic Accommodation, and Adaptive Evolution.” Evolution and Development 9:299–305. Moczek, Armin P. 2010. “Phenotypic Plasticity and Diversity in Insects.” Philo- sophical Transactions of the Royal Society B 365:593–603. Moczek, Armin P., and Douglas J. Emlen. 1999. “Proximate Determination of Male Horn Dimorphism in the Beetle Onthophagus taurus (Coleoptera: Scarabaei- dae).” Journal of Evolutionary Biology 12:27–37.

278 ◂ REFERENCES

Molnar, Attila, Charles W. Melnyk, Andrew Bassett, Thomas J. Hardcastle, Ruth Dunn, and David C. Baulcombe. 2010. “Small Silencing RNAs in Plants Are Mobile and Direct Epigenetic Modification in Recipient Cells.”Science 328:872–75. Monod, Jacques. 1977. Chance and Necessity: An Essay on the Natural Philosophy of Modern Biology. Glasgow: Collins. Morange, Michel. 2000a. A History of Molecular Biology. Translated by Matthew Cobb. Cambridge: Harvard University Press. Morange, Michel. 2000b. “The Developmental Gene Concept: History and Lim- its.” In The Concept of the Gene in Development and Evolution, edited by Peter Beurton, Raphael Falk, and Hans-Jörg Rheinberger, 193–215. Cambridge: Cam- bridge University Press. Morange, Michel. 2001. The Misunderstood Gene. Cambridge: Harvard University Press. Morgan, Mary S. 2003. “Experiments without Material Intervention: Model Experiments, Virtual Experiments, and Virtually Experiments.” In The Philos- ophy of Scientific Experimentation, edited by Hans Radder, 216–35. Pittsburgh: University of Pittsburgh Press. Morgan, Thomas Hunt. 1926. The Theory of the Gene. New Haven: Yale University Press. Morrison, Margaret, and Mary S. Morgan. 1999. “Models as Mediating Instru- ments.” In Models as Mediators: Perspectives on Natural and Social Science, edited by Mary S. Morgan and Margaret Morrison, 10–37. Cambridge: Cam- bridge University Press. Morse, Douglass H. 1980. Behavioral Mechanisms in Ecology. Cambridge: Harvard University Press. Moss, Lenny. 2001. “Deconstructing the Gene and Reconstructing Molecular Developmental Systems.” In Cycles of Contingency: Developmental Systems and Evolution, edited by Susan Oyama, Paul E. Griffiths, and Russell D. Gray, 85–98. Cambridge: MIT Press. Mossio, Matteo, and Alvaro Moreno. 2010. “Organisational Closure in Biological Organisms.” History and Philosophy of the Life Sciences 32:269–88. Mossio, Matteo, Cristian Saborido, and Alvaro Moreno. 2009. “An Organizational Account of Biological Functions.” British Journal for the Philosophy of Science 60:813–41. Müller, Gerd B. 2007. “Evo-Devo: Extending the Evolutionary Synthesis.” Nature Reviews Genetics 8:943–49.

▸ 279 REFERENCES

Müller, Gerd B. 2008. “Evo-Devo as a Discipline.” In Evolving Pathways: Key Themes in Evolutionary Developmental Biology, edited by Alessandro Minelli and Giuseppe Fusco, 307–32. Cambridge: Cambridge University Press. Müller, Gerd B. 2010. “Epigentic Innovation.” In Evolution: The Extended Synthesis, edited by Massimo Pigliucci and Gerd B. Muller, 307–32. Cambridge: MIT Press. Müller, Gerd B., and Stuart A. Newman. 2003. Origination of Organismal Form: Beyond the Gene in Developmental and Evolutionary Biology. Cambridge: MIT Press. Müller, Gerd B., and Massimo Pigliucci. 2010. “Extended Synthesis: Theory Expansion or Alternative.” Biological Theory 5:275–76. Müller, Gerd B., and Günter P. Wagner. 1991. “Novelty in Evolution: Restructuring the Concept.” Annual Review of Ecology and Systematics 22:229–56. Nagel, Ernest. 1949. “The Meaning of Reduction in the Natural Sciences.” InSci - ence and Civilization, edited by Robert C. Stauffer, 99–135. Madison: University of Wisconsin Press. Nanney, David L. 1958. “Epigenetic Control Systems.” Proceedings of the National Academy of Sciences 44:712–17. Nathan, Marco J. 2012. “The Varieties of Molecular Explanation.” Philosophy of Science 79:233–54. Nathan, Marco J. 2014. “Causation by Concentration.” British Journal for the Phi- losophy of Science 65:191–212. Nature Biotechnology. 2010. “Making a Mark” (editorial). 28:1010–31. Neander, Karen. 1995. “Pruning the Tree of Life.” British Journal for the Philosophy of Science 46:59–80. Needham, Joseph. 1929. The Skeptical Biologist. London: Chatto & Windus. Neel, James V. 1962. “Diabetes Mellitus: A ‘Thrifty’ Genotype Rendered Detri- mental by ‘Progress’?” American Journal of Human Genetics 14:353–62. Nelson, Vicki R., Jason D. Heaney, Paul J. Tesar, Nicholas O. Davidson, and Joseph H. Nadeau. 2012. “Transgenerational Epigenetic Effects of the Apobec1 Cyt- idine Deaminase Deficiency on Testicular Germ Cell Tumor Susceptibility and Embryonic Viability.” Proceedings of the National Academy of Sciences 109:E2766–73. Neuman, Yair. 2007. “The Rest Is Silence.” Review of The Music of Life: Biology beyond the Genome, by Denis Noble. Perspectives in Biology and Medicine 50:625–28. Newman, Stuart A. 2010. “Dynamical Patterning Modules.” In Evolution: The Extended Synthesis, edited by Massimo Pigliucci and Gerd B. Muller, 281–306. Cambridge: MIT Press.

280 ◂ REFERENCES

Ney, Alyssa. 2009. “Physical Causation and Difference-Making.” British Journal for the Philosophy of Science 60:737–64. Nicholson, Daniel J. 2012. “The Concept of Mechanism in Biology.” Studies in His- tory and Philosophy of Biological and Biomedical Sciences 43:152–63. Nicholson, Daniel J. 2013. “Organisms ≠ Machines.” Studies in History and Philos- ophy of Biological and Biomedical Sciences 44:669–78. Nicholson, Daniel J. 2014a. “The Machine Conception of the Organism in Devel- opment and Evolution: A Critical Analysis.” Studies in History and Philosophy of Biological and Biomedical Sciences 48:162–74. Nicholson, Daniel J. 2014b. “The Return of the Organism as a Fundamental Explanatory Concept in Biology.” Philosophy Compass 9:347–59. Nicholson, Daniel J. 2018. “Reconceptualizing the Organism: From Complex Machine to Flowing Stream.” In Everything Flows: Towards a Processual Philos- ophy of Biology, edited by Daniel J. Nicholson and John Dupré, 139–66. Oxford: Oxford University Press. Nicholson, Daniel J., and John Dupré, eds. 2018. Everything Flows: Towards a Pro- cessual Philosophy of Biology. Oxford: Oxford University Press. Nicholson, Daniel J., and Richard Gawne. 2014. “Rethinking Woodger’s Legacy in the Philosophy of Biology.” Journal of the History of Biology 47:243–92. Nicholson, Daniel J., and Richard Gawne. 2015. “Neither Logical Empiricism nor Vitalism, but Organicism: What the Philosophy of Biology Was.” History and Philosophy of the Life Sciences 37:345–81. Nicolis, Grégoire, and Ilya Prigogine. 1977. Self-Organization in Nonequilibrium Systems: From Dissipative Structures to Order through Fluctuations. New York: Wiley. Nicolosi, Guido, and Guido Ruivenkamp. 2012. “The Epigenetic Turn: Some Notes about the Epistemological Change of Perspective in Biosciences.” Medi- cine, Health Care, and Philosophy 15:309–19. Niewöhner, Jörg. 2011. “Epigenetics: Embedded Bodies and the Molecularisation of Biography and Milieu.” BioSocieties 6:279–98. Nijhout, H. Frederik. 2002. “The Nature of Robustness in Development.”BioEs - says 24:553–63. Nilsson-Ehle, Herman. 1914. “Vilka erfarenheter hava hittills vunnits rörande möjligheten av växters acklimatisering?” Kunglig Landtbruksakadamiens Handlingar och Tidskrift 53:537–72. Noble, Denis. 2008a. “Claude Bernard, the First Systems Biologist, and the Future of Physiology.” Experimental Physiology 93:16–26.

▸ 281 REFERENCES

Noble, Denis. 2008b. “Genes and Causation.” Philosophical Transactions of the Royal Society A 366:3001–15. Noble, Denis. 2012. “A Theory of Biological Relativity: No Privileged Level of Causation.” Interface Focus 2:55–64. Noble, Denis. 2015. “Evolution beyond Neo-Darwinism: A New Conceptual Framework.” Journal of Experimental Biology 218:7–13. Northcott, Robert. 2013. “Degree of Explanation.” Synthese 190:3087–105. Northrop, Robert B. 2010. Introduction to Complexity and Complex Systems. Boca Raton: CRC Press. Odling-Smee, John. 2010. “Niche Inheritance.” In Evolution: The Extended Syn- thesis, edited by Massimo Pigliucci and Gerd B. Muller, 175–207. Cambridge: MIT Press. Odling-Smee, John, Kevin N. Laland, and Marcus W. Feldman. 2003. Niche Con- struction: The Neglected Process in Evolution. Princeton: Princeton University Press. Olby, Robert C. 1990. “The Molecular Revolution in Biology.” In Companion to the History of Modern Science, edited by Robert C. Olby, Geoffrey N. Cantor, John R. R. Christie, and Jonathan S. Hodge, 503–20. London: Routledge. O’Malley, Maureen A., Ingo Brigandt, Alan C. Love, John W. Crawford, Jack A. Gilbert, Rob Knight, Sandra D. Mitchell, and Forest Rohwer. 2014. “Multilevel Research Strategies and Biological Systems.” Philosophy of Science 81:811–28. O’Malley, Maureen A., Orkun S. Soyer, and Mark L. Siegal. 2015. “A Philosophical Perspective on Evolutionary Systems Biology.” Biological Theory 10:6–17. Oppenheim, Paul, and Hilary Putnam. 1958. “Unity of Science as a Working Hypothesis.” In Concepts, Theories, and the Mind-Body Problem, edited by Herbert Feigl, Michael Scriven, and Grover Maxwell, 3–36. Minneapolis: Uni- versity of Minnesota Press. Otsuka, Jun. 2015. “Using Causal Models to Integrate Proximate and Ultimate Causation.” Biology and Philosophy 30:19–37. Oyama, Susan. 1985. The Ontogeny of Information: Developmental Systems and Evolution. Cambridge: Cambridge University Press. Oyama, Susan. 2000. “Causal Democracy and Causal Contributions in Develop- mental Systems Theory.” PSA: Proceedings of the Philosophy of Science Associ- ation 1998:S332–47. Oyama, Susan, Paul E. Griffiths, and Russell D. Gray, eds. 2001. Cycles of Contin- gency: Developmental Systems and Evolution. Cambridge: MIT Press.

282 ◂ REFERENCES

Pál, Csaba, and Laurence D. Hurst. 2004. “Epigenetic Inheritance and Evolution- ary Adaptation.” In Organelles, Genomes and Eukaryote Phylogeny: An Evolu- tionary Synthesis in the Age of Genomics, edited by Robert P. Hirt and David S. Horner, 353–70. London: Taylor & Francis. Parente, Lyana B., Marcia B. Aguila, and Carlos A. Mandarim-de-Lacerda. 2008. “Deleterious Effects of High-Fat Diet on Perinatal and Postweaning Periods in Adult Rat Offspring.” Clinical Nutrition 27:623–34. Pearl, Judea. 2000. Causality: Models, Reasoning, and Inference. New York: Cam- bridge University Press. Pearson, Helen. 2006. “Genetics: What Is a Gene?” Nature 441:398–401. Pennisi, Elizabeth. 2008. “Modernizing the Modern Synthesis.” Science 321:196–97. Pennisi, Elizabeth. 2009. “The Case of the Midwife Toad: Fraud or Epigenetics?” Science 325:1194–95. Pennisi, Elizabeth. 2013. “Evolution Heresy? Epigenetics Underlies Heritable Plant Traits.” Science 341:1055. Pentinat, Thais, Marta Ramon-Krauel, Judith Cebria, Ruben Diaz, and Josep C. Jimenez-Chillaron. 2010. “Transgenerational Inheritance of Glucose Intoler- ance in a Mouse Model of Neonatal Overnutrition.” Endocrinology 151:5617–23. Perini, Laura. 2005. “Explanation in Two Dimensions: Diagrams and Biological Explanation.” Biology and Philosophy 20:257–69. Perini, Laura. 2012. “Truth-Bearers or Truth-Makers?” Spontaneous Generations 6:142–47. Peter, Isabelle S., and Eric H. Davidson. 2011. “A Gene Regulatory Network Con- trolling the Embryonic Specification of Endoderm.” Nature 474:635–39. Peterson, Erik L. 2011. “The Excluded Philosophy of Evo-Devo? Revisiting C. H. Waddington’s Failed Attempt to Embed Alfred North Whitehead’s ‘Organicism’ in Evolutionary Biology.” History and Philosophy of the Life Sciences 33:301–20. Peterson, Erik L. 2014. “The Conquest of Vitalism or the Eclipse of Organicism? The 1930s Cambridge Organizer Project and the Social Network of Mid-Twen- tieth-Century Biology.” British Journal for the History of Science 47:281–304. Peterson, Erik L. 2016. The Life Organic: The Theoretical Biology Club and the Roots of Epigenetics. Pittsburgh: University of Pittsburgh Press. Piaget, Jean. 1978. Behavior and Evolution. New York: Pantheon. Pickersgill, Martyn, Jörg Niewöhner, Ruth Müller, Paul Martin, and Sarah Cun- ningham-Burley. 2013. “Mapping the New Molecular Landscape: Social Dimensions of Epigenetics.” New Genetics and Society 32:429–47.

▸ 283 REFERENCES

Piersma, Theunis, and Jan A. van Gils. 2011. The Flexible Phenotype: A Body-Centred Integration of Ecology, Physiology, and Behaviour. Oxford: Oxford University Press. Pigliucci, Massimo. 2001. Phenotypic Plasticity: Beyond Nature and Nurture. Balti- more: Johns Hopkins University Press. Pigliucci, Massimo. 2007. “Do We Need an Extended Evolutionary Synthesis?” Evolution 61:2743–49. Pigliucci, Massimo. 2008. “The Proper Role of Population Genetics in Modern Evolutionary Theory.” Biological Theory 3:316–24. Pigliucci, Massimo. 2009. “An Extended Synthesis for Evolutionary Biology.” Annals of the New York Academy of Sciences 1168:218–28. Pigliucci, Massimo. 2010. “Genotype–Phenotype Mapping and the End of the ‘Genes as Blueprint’ Metaphor.” Philosophical Transactions of the Royal Society B 365:557–66. Pigliucci, Massimo, and Maarten Boudry, eds. 2013. Philosophy of Pseudoscience: Reconsidering the Demarcation Problem. Chicago: University of Chicago Press. Pigliucci, Massimo, and Leonard Finkelman. 2014. “The Extended (Evolutionary) Synthesis Debate: Where Science Meets Philosophy.” BioScience 64:511–16. Pigliucci, Massimo, and Gerd B. Müller. 2010a. “Elements of an Extended Evolu- tionary Synthesis.” In Evolution—The Extended Synthesis, edited by Massimo Pigliucci and Gerd B. Muller, 3–17. Cambridge: MIT Press. Pigliucci, Massimo, and Gerd B. Muller, eds. 2010b. Evolution—The Extended Syn- thesis. Cambridge: MIT Press. Pigliucci, Massimo, Courtney J. Murren, and Carl D. Schlichting. 2006. “Pheno- typic Plasticity and Evolution by Genetic Assimilation.” Journal of Experimen- tal Biology 209:2362–67. Pittendrigh, Colin S. 1958. “Adaptation, Natural Selection, and Behavior.” In Behavior and Evolution, edited by Anne Roe and George. G. Simpson, 390–416. New Haven: Yale University Press. Plate, Ludwig. 1931. “Warum muß der Vererbungsforscher an der Annahme einer Vererbung erworbener Eigenschaften festhalten?” [Why must research of heredity hold on to the assumption of the inheritance of acquired characteristics?]. Zeitschrift für Induktive Abstammungs- und Vererbungslehre [Journal of inductive theory of descent and inheritance] 58:266–92. Pleiss, Jeffrey A., Gregg B. Whitworth, Megan Bergkessel, and Christine Guthrie. 2007. “Rapid, Transcript-Specific Changes in Splicing in Response to Environ- mental Stress.” Molecular Cell 27:928–37.

284 ◂ REFERENCES

Popper, Karl. 1935. Logik der Forschung. Vienna: Springer. Popper, Karl. 1963. Conjectures and Refutations: The Growth of Scientific Knowl- edge. London: Routledge. Por, Francis D. 2006. “The Actuality of Lamarck: Towards the Bicentenary of His Philosophie Zoologique.” Integrative Zoology 1:48–52. Powell, Alexander, and John Dupré. 2009. “From Molecules to Systems: The Importance of Looking Both Ways.” Studies in History and Philosophy of Bio- logical and Biomedical Sciences 40:54–64. Powell, Russell, and Nicholas Shea. 2014. “Homology across Inheritance Systems.” Biology and Philosophy 29:781–806. Prentice, Andrew M. 2005. “Starvation in Humans: Evolutionary Background and Contemporary Implications.” Mechanisms of Ageing and Development 126:976–81. Prentice, Andrew M., Branwen J. Hennig, and Anthony J. C. Fulford. 2008. “Evo- lutionary Origins of the Obesity Epidemic: Natural Selection of Thrifty Genes or Genetic Drift Following Predation Release?” International Journal of Obe- sity 32:1607–10. Probst, Aline V., Elaine Dunleavy, and Geneviève Almouzni. 2009. “Epigenetic Inheritance during the Cell Cycle.” Nature Reviews Molecular Cell Biology 10:192–206. Provine, William B. 1989. “Progress in Evolution and Meaning in Life.” In Evolu- tionary Progress, edited by Matthew H. Nitecki, 49–74. Chicago: University of Chicago Press. Psillos, Stathis. 2002. Causation and Explanation. Chesham: Acumen. Ptashne, Mark. 2007. “On the Use of the Word ‘Epigenetic.’” Current Biology 17:R233–36. Pugesek, Bruce H., Adrian Tomer, and Alexander von Eye. 2003. Structural Equa- tion Modeling: Applications in Ecological and Evolutionary Biology. Cambridge: Cambridge University Press. Pujadas, Elisabet, and Andrew P. Feinberg. 2012. “Regulated Noise in the Epigen- etic Landscape of Development and Disease.” Cell 148:1123–31. Purcell, Shaun M., Naomi R. Wray, Jennifer L. Stone, Peter M. Visscher, Michael C. O’Donovan, Patrick F. Sullivan, Pamela Sklar, Douglas M. Ruderfer, Andrew McQuillin, Derek W. Morris, Colm T. O’Dushlaine, Aiden Corvin, Peter A. Hol- mans, Michael C. O’Donovan, Stuart MacGregor, Hugh Gurling, Douglas H. R. Blackwood, Nick J. Craddock, Michael Gill, et al. 2009. “Common Polygenic Varia- tion Contributes to Risk of Schizophrenia and Bipolar Disorder.” Nature 460:748–52.

▸ 285 REFERENCES

Quine, Willard V. 1973. The Roots of Reference. La Salle: Open Court. Raatikainen, Panu. 2010. “Causation, Exclusion, and the Special Sciences.” Erkennt- nis 73:349–63. Raeker, Tim, Niss Ole Carstensen, and Bernd Hartke. 2012. “Simulating a Molecu- lar Machine in Action.” Journal of Physical Chemistry A 116:11241–48. Raerinne, Jani. 2011. “Causal and Mechanistic Explanations in Ecology.” Acta Bio- theoretica 59:251–71. Raerinne, Jani, and Jan Baedke. 2015. “Exclusions, Explanations, and Exceptions: On the Causal and Lawlike Status of the Competitive Exclusion Principle.” Phi- losophy and Theory in Biology 7:e602. doi:10.1007/s10441-010-9122-9. Raff, Rudolf A. 1996. The Shape of Life: Genes, Development, and the Evolution of Animal Form. Chicago: University of Chicago Press. Railton, Peter. 1978. “A Deductive-Nomological Model of Probabilistic Explana- tion.” Philosophy of Science 45:206–26. Rankinen, Tuomo, Aamir Zuberi, Yvon C. Chagnon, Stanley J. Weisnagel, George Argyropoulos, Brandon Walts, Louis Pérusse, and Claude Bouchard. 2006. “The Human Obesity Gene Map: The 2005 Update.” Obesity 14:529–644. Rassoulzadegan, Minoo. 2011. “An Evolutionary Role for RNA-Mediated Epigene- tic Variation?” In Transformations of Lamarckism: From Subtle Fluids to Molec- ular Biology, edited by Snait B. Gissis and Eva Jablonka, 227–35. Cambridge: MIT Press. Rassoulzadegan, Minoo, Valérie Grandjean, Pierre Gounon, Stéphane Vincent, Isabelle Gillot, and François Cuzin. 2006. “RNA-Mediated Non-Mendelian Inheritance of an Epigenetic Change in the Mouse.” Nature 441:469–74. Rechavi, Oded, Gregory Minevich, and Oliver Hobert. 2011. “Transgenerational Inheritance of an Acquired Small RNA-Based Antiviral Response in C. elegans.” Cell 147:1248–56. Reinders, Jon, Brande B. H. Wulff, Marie Mirouze, Arturo Marí-Ordóñez, Mélanie Dapp, Wilfried Rozhon, Etienne Bucher, Grégory Theiler, and Jerzy Paszkowski. 2009. “Compromised Stability of DNA Methylation and Transposon Immobili- zation in Mosaic Arabidopsis Epigenomes.” Genes and Development 23:939–50. Reinke, Johannes. 1901. Einleitung in die theoretische Biologie. [Introduction to theoretical biology]. Berlin: Paetel. Reutlinger, Alexander. 2012. “Getting Rid of Interventions.” Studies in History and Philosophy of Biological and Biomedical Sciences 43:787–95. Rheinberger, Hans-Jörg, and Staffan Müller-Wille.2012. Heredity. Chicago: Uni- versity of Chicago Press.

286 ◂ REFERENCES

Rheinberger, Hans-Jörg, Staffan Müller-Wille, and Robert Meunier. 2015. “Gene.” In The Stanford Encyclopedia of Philosophy, Spring ed., edited by Edward N. Zalta. http://plato.stanford.edu/archives/spr2015/entries/gene/. Ricci, Dante P., and Thomas J. Silhavy. 2012. “The Bam Machine: A Molecular Cooper.” Biochimica et Biophysica Acta (BBA)—Biomembranes 1818:1067–84. Richards, Christina L., Oliver Bossdorf, and Massimo Pigliucci. 2010a. “What Role Does Heritable Epigenetic Variation Play in Phenotypic Evolution?” Bio- Science 60:232–37. Richards, Christina L., Oliver Bossdorf, and Koen J. F. Verhoeven. 2010b. “Under- standing Natural Epigenetic Variation.” New Phytologist 187:562–64. Richards, Eric J. 2006. “Inherited Epigenetic Variation: Revisiting Soft Inheri - tance.” Nature Reviews Genetics 7:395–401. Richards, Eric J. 2008. “Population Epigenetics.” Current Opinion in Genetics and Development 18:221–26. Richards, Eric J. 2009. “Quantitative Epigenetics: DNA Sequence Variation Need Not Apply.” Genes and Development 23:1601–5. Risch, Neil, and Kathleen R. Merikangas. 1996. “The Future of Genetic Studies of Complex Human Diseases.” Science 273:1516–17. Riskin, Jessica. 2016. The Restless Clock: A History of the Centuries-Long Argument over What Makes Living Things Tick. Chicago: University of Chicago Press. Rivoire, Olivier, and Stanislas Leibler. 2014. “A Model for the Generation and Transmission of Variations in Evolution.” Proceedings of the National Academy of Sciences 111:E1940–49. Robertson, Alan. 1977. “Conrad Hal Waddington, 8 November 1905—26 Septem- ber 1975.” Biographical Memoirs of Fellows of the Royal Society 23:575–622. Rorty, Richard. 1989. Contingency, Irony, and Solidarity. Cambridge: Cambridge University Press. Rosen, Robert. 1991. Life Itself: A Comprehensive Inquiry into the Nature, Origin, and Fabrication of Life. New York: Columbia University Press. Rosen, Robert. 1999. Essays on Life Itself. New York: Columbia University Press. Rosenberg, Alexander. 1994. Instrumental Biology, or The Disunity of Science. Chi- cago: University of Chicago Press. Royle, J. Andrew, and Robert M. Dorazio. 2008. Hierarchical Modeling and Infer- ence in Ecology: The Analysis of Data from Populations, Metapopulations and Communities. San Diego: Academic Press. Ruden, Douglas M., Li Xiao, Mark D. Garfinkel, and Xiangyi Lu. 2005. “Hsp90 and Environmental Impacts on Epigenetic States: A Model for the Trans-

▸ 287 REFERENCES

Generational Effects of Diethylstibesterol on Uterine Development and Can- cer.” Human Molecular Genetics 14:R149–55. Russell, Bertrand. 1912/1913. “On the Notion of Cause.” Proceedings of the Aristo- telian Society 7:1–26. Russo, Federica. 2009. Causality and Causal Modelling in the Social Sciences: Mea- suring Variations. New York: Springer. Russo, Federica. 2011. “Correlational Data, Causal Hypotheses, and Validity.” Jour- nal for General Philosophy of Science 42:85–107. Russo, Federica. 2012. “On Empirical Generalisations.” In Probabilities, Laws, and Structures, edited by Dennis Dieks, Wenceslao J. Gonzalez, Stephen Hartmann, Michel Stöltzner, and Marcel Weber, 123–39. Dordrecht: Springer. Russo, Federica, and Jon Williamson. 2007. “Interpreting Causality in the Health Sciences.” International Studies in the Philosophy of Science 21:157–70. Russo, Federica, Guillaume Wunsch, and Michel Mouchart. 2011. “Inferring Cau- sality through Counterfactuals in Observational Studies: Some Epistemologi- cal Issues.” Bulletin of Sociological Methodology/Bulletin de Méthodologie Socio- logique 111:43–64. Russo, Vincenzo E. A., Robert A. Martienssen, and Arthur D. Riggs, eds. 1996. Epigenetic Mechanisms of Gene Regulation. Cold Spring Harbor: Cold Spring Harbor Laboratory Press. Rutishauser, Rolf, and Philip Moline. 2005. “Evo-Devo and the Search for Homol- ogy (‘Sameness’) in Biological Systems.” Theory in Biosciences 124:213–41. Ryan, Bryce C., and John G. Vandenbergh. 2002. “Intrauterine Position Effects.” Neuroscience and Biobehavioral Reviews 26:665–78. Salian, Smita, Tanvi Doshi, and Geeta Vanage. 2009. “Impairment in Protein Expression Profile of Testicular Steroid Receptor Coregulators in Male Rat Offspring Perinatally Exposed to Bisphenol A.” Life Sciences 85:11–18. Salmon, Wesley C. 1984. Scientific Explanation and the Causal Structure of the World. Princeton: Princeton University Press. Salmon, Wesley C. 1989. Four Decades of Scientific Explanation. Minneapolis: Uni- versity of Minnesota Press. Salmon, Wesley C. 1994. “Causality without Counterfactuals.” Philosophy of Sci- ence 61:297–312. Salmon, Wesley C. 1997. “Causality and Explanation: A Reply to Two Critiques.” Philosophy of Science 64:461–77. Sano, Hiroshi. 2010. “Inheritance of Acquired Traits in Plants.” Plant Signaling and Behavior 5:346–48.

288 ◂ REFERENCES

Sansom, Roger, and Robert N. Brandon. 2007. Integrating Evolution and Develop- ment: From Theory to Practice. Cambridge: MIT Press. Sarkar, Sahotra. 1992. “Models of Reduction and Categories of Reductionism.” Synthese 91:167–94. Sarkar, Sahotra. 1998. Genetics and Reductionism. Cambridge: Cambridge Uni- versity Press. Scarfe, Adam C. 2013. “Epigenetics, Soft Inheritance, Mechanistic Metaphysics, and Bioethics.” In Beyond Mechanism: Putting Life Back into Biology, edited by Brian G. Henning and Adam C. Scarfe, 369–409. Lanham: Rowman & Little- field. Schaffer, Jonathan. 2005. “Contrastive Causation.” Philosophical Review 114:327– 58. Schaffner, Kenneth F. 1967a. “Antireductionism and Molecular Biology: Though the Antireductionist Thesis Is Unwarranted, Research in Classical Biology May Well Be of Value.” Science 157:644–47. Schaffner, Kenneth F. 1967b. “Approaches to Reduction.”Philosophy of Science 34:137–47. Schaffner, Kenneth F. 1969. “The Watson-Crick Model and Reductionism.” British Journal for the Philosophy of Science 20:325–48. Schlichting, Carl, and Massimo Pigliucci. 1998. Phenotypic Evolution: A Reaction Norm Perspective. Sunderland: Sinauer. Schmalhausen, Ivan I. 1949. Factors of Evolution: The Theory of Stabilizing Selec- tion. Translated by Isadore Dordick. Philadelphia: Blakiston. Scholl, Raphael, and Massimo Pigliucci. 2014. “The Proximate-Ultimate Dis- tinction and Evolutionary Developmental Biology: Causal Irrelevance versus Explanatory Abstraction.” Biology and Philosophy 30:653–70. Schrey, Aaron W., Christina L. Richards, Victoria Meller, Vincent Sollars, and Doug- las M. Ruden. 2012. “The Role of Epigenetics in Evolution: The Extended Syn- thesis.” Genetics Research International 2012:286164. doi:10.1155/2012/286164. Schrödinger, Erwin. 1944. What Is Life? The Physical Aspect of the Living Cell. New York: Macmillan. Schulz, Laura, Tamar Kushnir, and Alison Gopnik. 2007. “Learning from Doing: Intervention and Causal Inference.” In Causal Learning: Psychology, Philos- ophy, and Computation, edited by Alison Gopnik and Laura Schulz, 67–85. Oxford: Oxford University Press. Schwander, Tanja, and Olof Leimar. 2011. “Genes as Leaders and Followers in Evo- lution.” Trends in Ecology and Evolution 26:143–51.

▸ 289 REFERENCES

Scott-Phillips, Thomas C., Thomas E. Dickins, and Stuart A. West. 2011. “Evolu- tionary Theory and the Ultimate-Proximate Distinction in the Human Behav- ioral Sciences.” Perspectives on Psychological Science 6:38–47. Shankar, Kartik, Amanda Harrell, Xiaoli Liu, Janet M. Gilchrist, Martin J. J. Ronis, and Thomas M. Badger. 2008. “Maternal Obesity at Conception Programs Obesity in the Offspring.” American Journal of Physiology—Regulatory, Inte- grative and Comparative Physiology 294:R528–38. Shea, Nicholas. 2007. “Representation in the Genome and in Other Inheritance Systems.” Biology and Philosophy 22:313–31. Shea, Nicholas. 2013. “Inherited Representations Are Read in Development.” Brit- ish Journal for the Philosophy of Science 64:1–31. Shea, Nicholas, Ido Pen, and Tobias Uller. 2011. “Three Epigenetic Information Channels and Their Different Roles in Evolution.”Journal of Evolutionary Biol- ogy 24:1178–87. Shi, Xinghua, and Taekjip Ha. 2011. “Seeing a Molecular Machine Self-Renew.” Proceedings of the National Academy of Sciences 108:3459–60. Shipley, Bill. 2000. Cause and Correlation in Biology: A User’s Guide to Path Anal- ysis, Structural Equations and Causal Inference. Cambridge: Cambridge Uni- versity Press. Shorter, James, and Susan Lindquist. 2005. “Prions as Adaptive Conduits of Mem- ory and Inheritance.” Nature Reviews Genetics 6:435–50. Simon, Herbert A. 1962. “The Architecture of Complexity.”Proceedings of the American Philosophical Society 106:467–82. Simpson, Stephen J., and Meir P. Pener. 2009. “Locust Phase Polyphenism: An Update.” Advances in Insect Physiology 36:1–272. Siomi, Haruhiko, and Mikiko C. Siomi. 2009. “On the Road to Reading the RNA-Interference Code.” Nature 457:396–404. Skinner, Michael K. 2008. “What Is an Epigenetic Transgenerational Phenotype? F3 or F2.” Reproductive Toxicology 25:2–6. Skinner, Michael K., Matthew D. Anway, Marina I. Savenkova, Andrea C. Gore, and David Crews. 2008. “Transgenerational Epigenetic Programming of the Brain Transcriptome and Anxiety Behavior.” PLoS ONE 3:3745. doi:10.1371/ journal.pone.0003745. Skinner, Michael K., Carlos Guerrero-Bosagna, M. Muksitul Haque, Eric Nilsson, Ramji Bhandari, and John R. McCarrey. 2013. “Environmentally Induced Trans- generational Epigenetic Reprogramming of Primordial Germ Cells and the Subsequent Germ Line.” PloS One 8:66318. doi:10.1371/journal.pone.0066318.

290 ◂ REFERENCES

Skinner, Michael K., Mohan Manikkam, and Carlos Guerrero-Bosagna. 2010. “Epigenetic Transgenerational Actions of Environmental Factors in Disease Etiology.” Trends in Endocrinology and Metabolism 21:214–22. Skinner, Michael K., Mohan Manikkam, and Carlos Guerrero-Bosagna. 2011. “Epigenetic Transgenerational Actions of Endocrine Disruptors.” Reproductive Toxicology 31:337–43. Skipper, Robert A., and Roberta L. Millstein. 2005. “Thinking about Evolutionary Mechanisms: Natural Selection.” Studies in History and Philosophy of Biological and Biomedical Sciences 36:327–47. Slack, Jonathan M. W. 2002. “Conrad Hal Waddington: The Last Renaissance Biologist?” Nature Reviews Genetics 3:889–95. Smart, John J. C. 1959. “Can Biology Be an Exact Science?” Synthese 11:359–68. Smits, Wiep K., Oscar P. Kuipers, and Jan-Willem Veening. 2006. “Phenotypic Variation in Bacteria: The Role of Feedback Regulation.” Nature Reviews Microbiology 4:259–71. Sober, Elliott. 1975. Simplicity. Oxford: Clarendon Press. Sober, Elliott. 1984. The Nature of Selection: Evolutionary Theory in Philosophical Focus. Chicago: University of Chicago Press. Sober, Elliott. 1986. “Explanatory Presupposition.” Australasian Journal of Philos- ophy 64:143–49. Sober, Elliott. 1988. “Apportioning Causal Responsibility.” Journal of Philosophy 85:303–18. Sober, Elliott. 1997. “Two Outbreaks of Lawlessness in Recent Philosophy of Biol- ogy.” PSA: Proceedings of the Biennial Meeting of the Philosophy of Science Asso- ciation 1996:S458–67. Sollars, Vincent, Xiangyi Lu, Li Xiao, Xiaoyan Wang, Mark D. Garfinkel, and Doug- las M. Ruden. 2003. “Evidence for an Epigenetic Mechanism by Which Hsp90 Acts as a Capacitor for Morphological Evolution.” Nature Genetics 33:70–74. Speakman, John R., and Stephen O’Rahilly. 2012. “Fat: An Evolving Issue.” Disease Models and Mechanisms 5:569–73. Spector, Tim. 2012. Identically Different: Why You Can Change Your Genes. Lon- don: Hachette. Srinivasan, Dayalan G., and Jennifer A. Brisson. 2012. “Aphids: A Model for Polyphenism and Epigenetics.” Genetics Research International 2012:431531. doi:10.1155/2012/431531. Stearns, Frank W. 2010. “One Hundred Years of Pleiotropy: A Retrospective.” Genetics 186:767–73.

▸ 291 REFERENCES

Steel, Daniel. 2008. Across the Boundaries: Extrapolation in Biology and Social Sci- ence. Oxford: Oxford University Press. Stegmann, Ulrich E. 2010. “What Can Natural Selection Explain?” Studies in His- tory and Philosophy of Biological and Biomedical Sciences 41:61–66. Sterelny, Kim. 1992. “Evolutionary Explanations of Human Behaviour.” Austral- asian Journal of Philosophy 70:156–73. Sterelny, Kim. 2001. “Niche Construction, Developmental Systems, and the Extended Replicator.” In Cycles of Contingency: Developmental Systems and Evolution, edited by Susan Oyama, Paul E. Griffiths, and Russell D. Gray, 333– 49. Cambridge: MIT Press. Stotz, Karola. 2008. “The Ingredients for a Postgenomic Synthesis of Nature and Nurture.” Philosophical Psychology 21:359–81. Stotz, Karola, and Paul E. Griffiths. 2015. “Dissecting Developmental Biology.” Stud- ies in History and Philosophy of Biological and Biomedical Sciences 53:134–38. Stotz, Karola, and Paul E. Griffiths. 2016. “Epigenetics: Ambiguities and Implica- tions.” History and Philosophy of the Life Sciences 38:22. Strevens, Michael. 2008. Depth: An Account of Scientific Explanation. Cambridge: Harvard University Press. Strohman, Richard C. 1995. “Linear Genetics, Non-Linear Epigenetics: Comple- mentary Approaches to Understanding Complex Diseases.” Integrative Physio- logical and Behavioral Science 30:273–82. Sturm, Richard. A. 2009. “Molecular Genetics of Human Pigmentation Diversity.” Human Molecular Genetics 18:R9–17. Sultan, Sonia E. 2015. Organism and Environment: Ecological Development, Niche Construction, and Adaptation. New York: Oxford University Press. Suzuki, Miho M., and Adrian Bird. 2008. “DNA Methylation Landscapes: Provoc- ative Insights from Epigenomics.” Nature Reviews Genetics 9:465–76. Svardal, Hannes. 2010. “Can Epigenetics Solve the Case of the Midwife Toad? A Comment on Vargas.” Journal of Experimental Zoology B 314:625–28. Sword, Gregory A. 2002. “A Role for Phenotypic Plasticity in the Evolution of Aposematism.” Proceedings of the Royal Society B 269:1639–44. Szyf, Moshe. 2014. “Lamarck Revisited: Epigenetic Inheritance of Ancestral Odor Fear Conditioning.” Nature Neuroscience 17:2–4. Tabery, James. 2004. “Synthesizing Activities and Interactions in the Concept of a Mechanism.” Philosophy of Science 71:1–15. Tabery, James. 2009. “Difference Mechanisms: Explaining Variation with Mecha- nisms.” Biology and Philosophy 24:645–64.

292 ◂ REFERENCES

Takahashi, Kazutoshi, and Shinya Yamanaka. 2006. “Induction of Pluripotent Stem Cells from Mouse Embryonic and Adult Fibroblast Cultures by Defined Factors.” Cell 126:663–76. Tal, Omri, Eva Kisdi, and Eva Jablonka. 2010. “Epigenetic Contribution to Covar- iance between Relatives.” Genetics 184:1037–50. Tauber, Alfred I. 2011. “Epigenetics and the ‘New Biology’: Enlisting in the Assault on Reductionism.” In Transformations of Lamarckism: From Subtle Fluids to Molecular Biology, edited by Snait B. Gissis and Eva Jablonka, 385–88. Cam- bridge: MIT Press. Tebbich, Sabine, Andrea S. Griffin, Markus F. Peschl, and Kim Sterelny. 2016. “From Mechanisms to Function: An Integrated Framework of Animal Inno- vation.” Philosophical Transactions of the Royal Society B 371:195. doi:10.1098/ rstb.2015.0195. Thakore, Pratiksha I., Joshua B. Black, Isaac B. Hilton, and Charles A. Gersbach. 2016. “Editing the Epigenome: Technologies for Programmable Transcription and Epigenetic Modulation.” Nature Methods 13:127–37. Thierry, Bernard. 2005. “Integrating Proximate and Ultimate Causation: Just One More Go!” Current Science 89:1180–83. Thom, René. 1969. “Topological Models in Biology.” Topology 8:313–35. Thom, René. 1975.Structural Stability and Morphogenesis: An Outline of a Gen- eral Theory of Models. Translated by David H. Fowler. Foreword by Conrad H. Waddington. Reading: W. A. Benjamin. Tinbergen, Nikolaas. 1951. The Study of Instinct. Oxford: Clarendon Press. Tinbergen, Nikolaas. 1963. “On Aims and Methods of Ethology.” Zeitschrift für Tierpsychologie 20:410–33. Tollefsbol, Trygve. 2011. “Epigenetics: The New Science of Genetics.” In Handbook of Epigenetics, edited by Trygve Tollefsbol, 1–6. San Diego: Academic Press. University of Groningen. 2014. “Revolutionary New View on Heritability in Plants.” Accessed 14 August 2017. http://www.Rug.Nl/News-and-Events/News/ Archief2014/Nieuwsberichten/0207epigenetica?Lang=en. Valena, Sophie, and Armin P. Moczek. 2012. “Epigenetic Mechanisms Underlying Developmental Plasticity in Horned Beetles.” Genetics Research International 2012:576303. doi:10.1155/2012/576303. van Fraasen, Bas C. 1980. The Scientific Image. Oxford: Oxford University Press. Van Speybroeck, Linda. 2002. “From Epigenesis to Epigenetics.” Annals of the New York Academy of Sciences 981:61–81. Van Speybroeck, Linda, Gertrudis Van de Vijver, and Dani De Waele. 2007.

▸ 293 REFERENCES

“Epi-Geneticization.” In The Influence of Genetics on Contemporary Thinking, edited by Anne Fagot-Largeault, Shahid Rahman, and Juan Manuel Torres, 115–33. Dordrecht: Springer. Vargas, Alexander O. 2009. “Did Paul Kammerer Discover Epigenetic Inheri- tance? A Modern Look at the Controversial Midwife Toad Experiments.” Jour- nal of Experimental Zoology B 312:667–78. Vargas, Alexander O., and John F. Fallon. 2005. “Birds Have Dinosaur Wings: The Molecular Evidence.” Journal of Experimental Zoology B 304B:86–90. Vargas, Alexander O., Quirin Krabichler, and Carlos Guerrero-Bosagna. 2016. “An Epigenetic Perspective on the Midwife Toad Experiments of Paul Kam- merer (1880–1926).” Journal of Experimental Zoology B 328:179–92. Vastenhouw, Nadine L., Karin Brunschwig, Kristy L. Okihara, Fritz Muller, Mar- cel Tijsterman, and Ronald H. A. Plasterk. 2006. “Gene Expression: Long- Term Gene Silencing by RNAi.” Nature 442:882. Venter, J. Craig. 2014. “A Conversation with J. Craig Venter, PhD.” Industrial Bio- technology 10:7–10. Villota-Salazar, Nubia A., Artemio Mendoza-Mendoza, and Juan M. González-Pri- eto. 2016. “Epigenetics: From the Past to the Present.” Frontiers in Life Science 9:347–70. von Neumann, John. 1966. Theory of Self-Reproducing Automata. Edited by Arthur W. Burks. Urbana: University of Illinois Press. von Wright, Georg Henrik. 1971. Explanation and Understanding. Ithaca: Cornell University Press. Waddington, Conrad H. 1940. Organisers and Genes. Cambridge: Cambridge University Press. Waddington, Conrad H. 1942a. “Canalization of Development and the Inheri- tance of Acquired Characters.” Nature 150:563–65. Waddington, Conrad H. 1942b. “The Epigenotype.” Endeavour 1:18–20. Waddington, Conrad H. 1943. “Polygenes and Oligogenes.” Nature 151:394. Waddington, Conrad H. 1952a. “Selection of the Genetic Basis for an Acquired Character.” Nature 169:625–26. Waddington, Conrad H. 1952b. The Epigenetics of Birds. Cambridge: Cambridge University Press. Waddington, Conrad H. 1953. “Genetic Assimilation of an Acquired Character.” Evolution 7:118–26. Waddington, Conrad H. 1956. “Genetic Assimilation of the Bithorax Phenotype.” Evolution 10:1–13.

294 ◂ REFERENCES

Waddington, Conrad H. 1957. The Strategy of the Genes: A Discussion of Some Aspects of Theoretical Biology. London: Allen & Unwin. Waddington, Conrad H., ed. 1968. Towards a Theoretical Biology. Vol. 1, Prolegom- ena. Edinburgh: Edinburgh University Press. Waddington, Conrad H. 1969a. “Biological Organization and Physical Systems.” In Biology and the Physical Sciences, edited by Samuel Devons, 172–85. New York: Columbia University Press. Waddington, Conrad H., ed. 1969b. Towards a Theoretical Biology. Vol. 2, Sketches. Edinburgh: Edinburgh University Press. Waddington, Conrad H. 1970a. Behind Appearance: A Study of the Relations between Painting and the Natural Sciences in This Century. Cambridge: MIT Press. Waddington, Conrad H., ed. 1970b. Towards a Theoretical Biology. Vol. 3, Drafts. Edinburgh: Edinburgh University Press. Waddington, Conrad H., ed. 1972. Towards a Theoretical Biology. Vol. 4, Essays. Edinburgh: Edinburgh University Press. Waddington, Conrad H. 1974. “A Catastrophe Theory of Evolution.” Annals of the New York Academy of Sciences 231:32–41. Waddington, Conrad H. 1977a. “Stabilisation in Systems: Chreods and Epigenetic Landscapes.” Futures 9:139–46. Waddington, Conrad H. 1977b. Tools for Thought: How to Understand and Apply the Latest Scientific Techniques of Problem Solving. London: Jonathan Cape. Waddington, Conrad H. 1977c. “Whitehead and Modern Science.” In Mind in Nature: Essays on the Interface of Science and Philosophy, edited by John B. Cobb and David R. Griffin, 143–46. Washington, DC: University Press of America. Waggoner, Miranda R., and Tobias Uller. 2015. “Epigenetic Determinism in Sci- ence and Society.” New Genetics and Society 34:177–95. Wagner, Andreas. 1996. “Can Nonlinear Epigenetic Interactions Obscure Causal Relations between Genotype and Phenotype?” Nonlinearity 9:607–29. Wagner, Günter P. 2014. Homology, Genes, and Evolutionary Innovation. Prince- ton: Princeton University Press. Wagner, Günter P., and Jianzhi Zhang. 2011. “The Pleiotropic Structure of the Genotype-Phenotype Map: The Evolvability of Complex Organisms.”Nature Reviews Genetics 12:204–13. Walsh, Denis M. 2015. Organisms, Agency, and Evolution. Cambridge: Cambridge University Press.

▸ 295 REFERENCES

Walsh, J. Bruce. 1996. “The Emperor’s New Genes.” Evolution 50:2115–18. Wang, Jin, Kun Zhang, Li Xu, and Erkang Wang. 2011. “Quantifying the Wad- dington Landscape and Biological Paths for Development and Differentiation.” Proceedings of the National Academy of Sciences 108:8257–62. Wang, Yan, Huijie Liu, and Zhongsheng Sun. 2017. “Lamarck Rises from His Grave: Parental Environment-Induced Epigenetic Inheritance in Model Organisms and Humans.” Biological Reviews. doi:10.1111/brv.12322. Warren, Ian A., J. Cristobal Vera, Annika Johns, Robert Zinna, James H. Marden, Douglas J. Emlen, Ian Dworkin, and Laura C. Lavine. 2014. “Insights into the Development and Evolution of Exaggerated Traits Using De Novo Tran- scriptomes of Two Species of Horned Scarab Beetles.” PLoS ONE 9:88364. doi:10.1371/journal.pone.0088364. Waterland, Robert A., Michael Travisano, K. G. Tahiliani, M. T. Rached, and S. Mirza. 2008. “Methyl Donor Supplementation Prevents Transgenerational Amplification of Obesity.” International Journal of Obesity 32:1373–79. Waters, C. Kenneth. 2007. “Causes That Make a Difference.”Journal of Philosophy 104:551–79. Watson, James D., and Francis H. C. Crick. 1953. “Molecular Structure of Nucleic Acids: A Structure for Deoxyribose Nucleic Acid.” Nature 171:737–38. Weaver, Ian C. G., Michael J. Meaney, and Moshe Szyf. 2006. “Maternal Care Effects on the Hippocampal Transcriptome and Anxiety-Mediated Behaviors in the Offspring That Are Reversible in Adulthood.” Proceedings of the National Academy of Sciences 103:3480–85. Weaver, Warren. 1938. “The Natural Sciences.” InThe Rockefeller Foundation, Annual Report 1938, 203–25. New York: Rockefeller Foundation. Weber, Kenneth, Robert Eisman, Lisa Morey, April Patty, Joshua Sparks, Michele Tausek, and Zhao-Bang Zeng. 1999. “An Analysis of Polygenes Affecting Wing Shape on Chromosome 3 in Drosophila melanogaster.” Genetics 153:773–86. Weber, Marcel. 1999. “The Aim and Structure of Ecological Theory.”Philosophy of Science 66:71–93. Weber, Marcel. 2008. “Causes without Mechanisms: Experimental Regularities, Physical Laws, and Neuroscientific Explanation.”Philosophy of Science 75:995– 1007. Wei, Yanchang, Cai-Rong Yang, Yan-Ping Wei, Zhen-Ao Zhao, Yi Hou, Heide Schatten, and Qing-Yuan Sun. 2014. “Paternally Induced Transgenerational Inheritance of Susceptibility to Diabetes in Mammals.” Proceedings of the National Academy of Sciences 111:1873–78.

296 ◂ REFERENCES

Weismann, August. 1892. Das Keimplasma: Eine Theorie der Vererbung. [The germ plasm: A theory of inheritance]. Jena: Fischer. Welberg, Leonie. 2014. “Epigenetics: A Lingering Smell?” Nature Reviews Neuro- science 15:1. Welch, John J. 2017. “What’s Wrong with Evolutionary Biology?” Biology and Phi- losophy 32:263–79. Wells, Jonathan C. K. 2010. “Maternal Capital and the Metabolic Ghetto: An Evo- lutionary Perspective on the Transgenerational Basis of Health Inequalities.” American Journal of Human Biology 22:1–17. Weslake, Brad. 2010. “Explanatory Depth.” Philosophy of Science 77:273–94. West-Eberhard, Mary Jane. 2003. Developmental Plasticity and Evolution. Oxford: Oxford University Press. West-Eberhard, Mary Jane. 2005. “Developmental Plasticity and the Origin of Species Differences.” Proceedings of the National Academy of Sciences 102:6543– 49. West-Eberhard, Mary Jane. 2007. “Dancing with DNA and Flirting with the Ghost of Lamarck.” Biology and Philosophy 22:439–51. White, Christy L., Megan N. Purpera, and Christopher D. Morrison. 2009. “Maternal Obesity Is Necessary for Programming Effect of High-Fat Diet on Offspring.” American Journal of Physiology—Regulatory, Integrative and Com- parative Physiology 296:R1464–72. Whitehead, Alfred N. 1929. Process and Reality. New York: Macmillan. Whitfield, John. 2008. “Biological Theory: Postmodern Evolution?”Nature News 455:281–84. Wigler, Michael. H. 1981. “The Inheritance of Methylation Patterns in Vertebrates.” Cell 24:285–86. Williams, George C. 1966. Adaptation and Natural Selection: A Critique of Some Current Evolutionary Thought. Princeton: Princeton University Press. Williams, George C. 1992. Natural Selection: Domains, Levels, and Challenges. Oxford: Oxford University Press. Williamson, Jon. 2009. “Probabilistic Theories.” InThe Oxford Handbook of Causation, edited by Helen Beebee, Christopher Hitchcock, and Peter Men- zies, 185–212. Oxford: Oxford University Press. Wilson, Edward O. 2000. Sociobiology: The New Synthesis. Cambridge: Belknap Press of Harvard University Press. Wimsatt, William C. 1972. “Teleology and the Logical Structure of Function State- ments.” Studies in History and Philosophy of Science 3:1–80.

▸ 297 REFERENCES

Wimsatt, William C. 1974. “Complexity and Organization.” PSA: Proceedings of the Biennial Meeting of the Philosophy of Science Association 1972:67–86. Wimsatt, William C. 1976. “Reductive Explanation: A Functional Account.” PSA: Proceedings of the Biennial Meeting of the Philosophy of Science Association 1974:671–710. Wimsatt, William C. 1980. “Reductionistic Research Strategies and Their Biases in the Units of Selection Controversy.” In Scientific Discovery: Case Studies, edited by Thomas Nickles, 213–59. Dordrecht: Springer. Wimsatt, William C. 1986. “Forms of Aggregativity.” In Human Nature and Natu- ral Knowledge, edited by Alan Donagan, Anthony N. Perovich Jr., and Michael V. Wedin, 259–91. Boston: Reidel. Wimsatt, William C. 1994. “The Ontology of Complex Systems: Levels of Orga- nization, Perspectives, and Causal Thickets.”Canadian Journal of Philosophy 20:207–74. Wimsatt, William C. 2007. Re-Engineering Philosophy for Limited Beings: Piecewise Approximations to Reality. Cambridge: Harvard University Press. Winther, Rasmus G. 2005. “Evolutionary Developmental Biology Meets Levels of Selection: Modular Integration or Competition, or Both?” In Modularity: Understanding the Development and Evolution of Natural Complex Systems, edited by Werner Callebaut and Diego Rasskin-Gutman, 61–97. Cambridge: MIT Press. Winther, Rasmus G. 2015. “Evo-Devo as a Trading Zone.” In Conceptual Change in Biology: Scientific and Philosophical Perspectives on Evolution and Develop- ment, edited by Alan C. Love, 459–82. Berlin: Springer. Wittgenstein, Ludwig. 1958. Philosophische Untersuchungen. Frankfurt: Suhrkamp, 1958. Woese, Carl R. 2004. “A New Biology for a New Century.” Microbiology and Molecular Biology Reviews 68:173–86. Woodward, James. 1984. “A Theory of Singular Causal Explanation.” Erkenntnis 21:231–62. Woodward, James. 1997. “Explanation, Invariance, and Intervention.” Philosophy of Science 64:41. Woodward, James. 2001. “Law and Explanation in Biology: Invariance Is the Kind of Stability That Matters.” Philosophy of Science 68:1–20. Woodward, James. 2002. “What Is a Mechanism? A Counterfactual Account.” PSA: Proceedings of the Biennial Meeting of the Philosophy of Science Associa- tion 2000:S366–77.

298 ◂ REFERENCES

Woodward, James. 2003. Making Things Happen: A Theory of Causal Explanation. Oxford: Oxford University Press. Woodward, James. 2004. “Counterfactuals and Causal Explanation.” International Studies in the Philosophy of Science 18:41–72. Woodward, James. 2006. “Sensitive and Insensitive Causation.” Philosophical Review 115:1–50. Woodward, James. 2007a. “Causation with a Human Face.” In Causation, Physics, and the Constitution of Reality: Russell’s Republic Revisited, edited by Huw Price and Richard Corry, 66–105. Oxford: Oxford University Press. Woodward, James. 2007b. “Interventionist Theories of Causation in Psychologi- cal Perspective.” In Causal Learning: Psychology, Philosophy, and Computation, edited by Alison Gopnik and Laura Schulz, 19–37. Oxford: Oxford University Press. Woodward, James. 2008. “Mental Causation and Neural Mechanisms.” In Being Reduced: New Essays on Reduction, Explanation, and Causation, edited by Jakob Hohwy and Jesper Kallestrup, 218–26. Oxford: Oxford University Press. Woodward, James. 2009. “Agency and Interventionist Theories.” InThe Oxford Handbook of Causation, edited by Helen Beebee, Christopher Hitchcock, and Peter Menzies, 234–62. Oxford: Oxford University Press. Woodward, James. 2010. “Causation in Biology: Stability, Specificity, and the Choice of Levels of Explanation.” Biology and Philosophy 25:287–318. Woodward, James. 2011. “Mechanisms Revisited.” Synthese 183:409–27. Woodward, James. 2013. “II—Mechanistic Explanation: Its Scope and Limits.” Aristotelian Society Supplementary Volume 87:39–65. Woodward, James, and Christopher Hitchcock. 2003. “Explanatory Generaliza- tions, Part I: A Counterfactual Account.” Noûs 37:1–24. Wouters, Arno. 1995. “Viability Explanation.” Biology and Philosophy 10:435–57. Wray, Gregory A., Hopi E. Hoekstra, Douglas J. Futuyma, Richard E. Lenski, Trudy F. C. Mackay, Dolph Schluter, and Joan E. Strassmann. 2014. “Does Evo- lutionary Theory Need a Rethink? No, All Is Well.” Nature News 514:161–64. Wright, Sewall. 1922. “Coefficients of Inbreeding and Relationship.” American Naturalist 56:330–38. Wu, C. Ting, and James R. Morris. 2001. “Genes, Genetics, and Epigenetics: A Correspondence.” Science 293:1103–5. Yamanaka, Shinya. 2009. “Elite and Stochastic Models for Induced Pluripotent Stem Cell Generation.” Nature 460:49–52. Ylikoski, Petri. 2007. “The Idea of Contrastive Explanandum.” In Rethinking

▸ 299 REFERENCES

Explanation, edited by Johannes Persson and Petri Ylikoski. Dordrecht: Springer. Ylikoski, Petri. 2013. “Causal and Constitutive Explanation Compared.” Erkennt- nis 78:277–97. Ylikoski, Petri, and Jaakko Kuorikoski. 2010. “Dissecting Explanatory Power.” Philosophical Studies 148:201–19. Young, Emma. 2008. “Rewriting Darwin: The New Non-Genetic Inheri- tance.” New Scientist 199, 9 July. https://www.newscientist.com/article/ mg19926641–500-rewriting-darwin-the-new-non-genetic-inheritance/. Zambrano, Elena, Claudia, J. Bautista, M. Deás, Paola M. Martínez-Samayoa, Margarita González-Zamorano, Hector Ledesma, J. Morales, Fernando Lar- rea, and Peter. W. Nathanielsz. 2006. “A Low Maternal Protein Diet during Pregnancy and Lactation Has Sex- and Window of Exposure-Specific Effects on Offspring Growth and Food Intake, Glucose Metabolism and Serum Leptin in the Rat.” Journal of Physiology 571:221–30. Zhang, Yuan-Ye, Markus Fischer, Vincent Colot, and Oliver Bossdorf. 2013. “Epi- genetic Variation Creates Potential for Evolution of Plant Phenotypic Plas- ticity.” New Phytologist 197:314–22. Zhou, Qiao, and Douglas A. Melton. 2008. “Extreme Makeover: Converting One Cell into Another.” Cell Stem Cell 3:382–88. Zordan, Rebecca E., David J. Galgoczy, and Alexander D. Johnson. 2006. “Epige- netic Properties of White-Opaque Switching in Candida albicans Are Based on a Self-Sustaining Transcriptional Feedback Loop.” Proceedings of the National Academy of Sciences 103:12807–12. Zuk, Marlene, Francisco Garcia-Gonzalez, Marie E. Herberstein, and Leigh W. Simmons. 2014. “Model Systems, Taxonomic Bias, and Sexual Selection: Beyond Drosophila.” Annual Review of Entomology 59:321–38.

300 ◂ INDEX

Abrahamsen, Adele, 143–44 Caenorhabditis elegans, 34–35, 45–46, agency, 37, 58, 92, 173, 207 100, 118, 120, 207, 223 allele, 28, 34, 56, 109, 111, 184. See also canalization, 20–25, 38, 71, 98, 158, gene 218n6 aphids, 118, 207 cancer, 1, 26, 80, 104, 198 Aristotle, 2, 14, 55, 223n17 Calcott, Brett, 59–60, 62, 65–66 artificiality, 45–46, 54, 222n9 catastrophe theory, 72, 225n34 causal explanation, 8, 10, 15, 45, 52, 87, bacteria, 46, 100–101, 138 97, 113, 134, 179, 198; and change-re- Bazaga, Pilar, 48, 50–51, 103–4, 222n13 latedness, 164; in classical epi- Bechtel, William, 131–32, 141–44, 153, genetics, 98; compared to mech- 156–57, 234n16, 234n19, 236n32 anistic explanation, 91, 130–31, Bernard, Claude, 12, 96–97, 230n18 204; and counterfactuals, 187; and Bhattacharya, Sudin, 160, 237n37 difference-making, 118–19, 127, 159; body plan, 37, 212, 243n9 interventionist account of, 91–96, Boniolo, Giovanni, 67–68, 210, 224n29 104–5, 110, 128; and manipulation, Borges, Jorge Luis, 17, 39, 166, 205 101–2; and “mechanistic detail,” 107–8 (see also “causes without INDEX

mechanisms”); nonmechanistic, tion of, 15–16, 20; dynamic, 9, 15, 109, 127, 130, 169, 186; and proxi- 22, 39; epigenetic, 9–10, 19, 25, 27, mate-ultimate distinction, 60, 62, 36–40, 67, 82, 166, 203–5, 208, 215; 66 of genotype-phenotype relation- causation, 8, 11, 45, 47, 61, 92, 117; ship, 22, 81, 111, 117, 119; multilevel, conserved quantity theory of, 131; representation problem of, 105; and counterfactuals, 93, 102, 16–17, 39, 166, 204; science, 202; 105; downward, 106, 157; and structural, 9, 15, 22, 25, 39; Wad- epigenetics, 10, 102, 108, 124, 128; dington on, 202 on higher levels of organization, composition, 56, 144. See also decom- 132; interlevel, 156; and invariance, position 108–9, 127; mechanistic theory of, constitutive explanation, 10, 61, 91, 10, 108, 122, 126, 166, 204; mental, 134, 156, 159. See also mechanistic 106; multifactorial, 22, 184; recip- explanation rocal, 57–58. See also proximate-ul- counterfactual, 14, 52, 98, 102–3, timate distinction 107, 191; account of causation, “causes without mechanisms,” 10, 93; explanation, 101; notion of 127–28, 166, 178, 199, 204 invariance, 93, 104–5; population, cell differentiation, 18, 61, 78, 80, 86, 51, 53, 104; questions, 51, 53, 94, 97 136, 143, 150, 154; (bi)stable, 160–66; (see also what-if-things-had-been- changes in, 15, 162; field of, 26, 79, different-questions); situations, 46, 159; processes, 68–69, 74; Wad- 94, 175–78, 180, 187, 195–96 dington’s epigenetic landscape Craig, Lindsay, 43, 221n3 model of, 71–72. See also repro- Craver, Carl, 136, 161; on interlevel gramming, cell relations, 156–59; on interven- cell division, 9, 24, 40 tionist explanations, 107, 134, 155, central dogma (of molecular biology), 229n16; on the Hodgkin-Huxley 30, 80, 82, 107, 136, 183, 220n20 model, 127; and machine meta- chromatin, 34–35, 68, 130; marking, phors, 142, 234nn17–18 25, 130; modification, 26, 32, 69, Crick, Francis, 12, 30, 107, 130, 138, 80–81 146, 201 cluster concept, 7, 88, 203. See also CRISPR, 154, 236n29 Wittgenstein Cummins, Robert, 133, 142, 149 complexity, 7, 9, 15, 18, 20, 31, 45, 69, 112, 122, 154, 162, 214; biological, 1, Darden, Lindley, 125, 142, 234nn17–18 9, 26, 36, 39, 91, 203, 206; Borges Darwin, Charles, 30, 42, 186, 240n15 problem of, 17, 39, 166, 205; defini- Davidson, Eric, 154, 226n40

302 ◂ INDEX

Davies, Paul W. C., 147, 150, 235nn22– 64, 115, 183, 208. See also canaliza- 23 tion; plasticity Dawkins, Richard, 18, 110; on developmental biology, 37, 44, 72, 110, Lamarck, 5; on replicator-vehicle 157; exclusion of from evolutionary concept, 31, 83; and selfish genes, theory, 29, 56 22, 207 Developmental Systems Theory decomposition 133, 143–44, 148, (DST), 114, 172, 183, 213–14 150; functional, 235n22–23; and “development first view” of evolution, localization, 131, 146–47, 150, 153; 28, 36, 38, 43 structural, 235n22–23 Dickins, Thomas, 58, 64, 66, 189, Descartes, René, 137, 144 224n27 determination, 22, 32, 210 difference-making, 95, 97, 114, 127–28, development, 24, 31, 54, 60, 65, 80, 110, 130, 155, 175, 230n17; actual, 91, 115, 139–40, 155, 169, 173; causal analysis 117–20, 124–26, 128; interlevel, of, 3, 19, 202; changes during, 3, 159; and (non-)change-relating 26, 98, 118; complexity of, 19; of relations, 94, 165, 237n37; potential, CRISPR, 154; embryonic, 18, 154, 91, 117, 127; and robust systems, 160; environmental influence on, 162–63; Waters on, 110–12 4, 21, 25, 27, 33, 102, 111, 158, 190; disease, 2, 18, 181, 186–87. See also epigenetic explanation of, 189, 193, cancer; obesity; schizophrenia 214; epigenetic regulation of, 26, DNA, 1, 14, 31, 69, 80, 82–83, 107, 110, 44, 115, 128, 150; of epigenetics, 23, 114, 119, 122, 181, 209; code, 140; 27, 85, 119, 174, 207, 209, 215; and and definition of gene, 230n19; evolution, 23, 58, 192, 199; of exper- epistemic status of, 115, 118; level, iments, 186; and gene-regulatory 5, 18, 221n6; and machines, 138–40; networks, 154, 226n40; and genes, methylation (see methylation, 19–20, 83, 85, 158, 172, 214; germ DNA); no influence of epigenetic cell, 32; and heredity, 5, 8–10, 30, variation on, 220n8, 220n20; onto- 39–40, 42, 55, 102, 107, 109, 115, 119, logical status of, 113; replication, 126; human, 193; plastic, 4, 9, 149, 130; and RNA, 129–30, 132, 135; 203; representation of with sequence, 3–4, 18, 24, 26, 46, 48, 54, epigenetic landscape model, 69, 69, 83, 100, 103, 112–13, 118, 146–47, 71–72, 76, 78–79; role of in evolu- 222n10; structure of, 15, 130; Wat- tion, 2, 22, 28, 36–37, 56, 58, 62–64, son and Crick on, 130 83; phases of, 61, 148; process of Drosophila (fruit fly), 20, 35, 98, 110, development, 151; stability of, 159, 207 161, 218n6, 237n35; trait, 20, 56, 61, Dupré, John, 151, 154

▸ 303 INDEX

embryology, 3, 19, 37. See also develop- 221n5; systems, 45, 80; transgener- mental biology ational, 32, 64, 84–85, 87, 182. See environment, 15, 18, 20, 28, 47, 53, also chromatin, marking; RNA 63, 65, 101, 120, 146, 176; (extra) interference (RNAi) cellular, 4, 18; early life, 32–33, 187, epigenetic landscape (EL), 10, 42; in 193; enriched, 33, 100; inducing, stem cell biology, 76–79, 86; 120, 122; internal, 96; natural, 53; Waddington’s, 68–69, 71–74, 79, normal, 34; and organism, 16, 21, 87, 151. See also canalization; gene- 29, 38, 46, 51, 57, 63, 104, 119, 151, regulatory network (GRN); 181, 208, 211; (extra)organismic, 4, Waddington 18, 183; persistent, 84 epigenetic turn, 4–7, 11, 119, 203 epigenetic explanation, 10, 125, epigenetics, concept of, 7, 24, 81, 205. 168–69, 203, 205, 224n26; cluster See also Jablonka; Waddington concept of, 6, 227n1; and evolu- epigenetics, ecological/evolutionary, tionary phenomena, 65, 67, 88, 8, 55, 85, 91, 93, 188; integrate with 178, 189; explanatory power of, 12, lab experiments, 45, 48, 88, 52, 169, 171, 173–74, 176, 178, 185, 187, 54, 168; methodologies in, 47–48, 189, 195–96, 199–200, 206; and 104–5, 222n10 extended evolutionary synthe- epigenetics, molecular, 8, 10, 84, sis, 11, 200; higher-level, 92, 180, 88–89, 93, 105; (causal) expla- 232n29; integration of into evo- nation in, 45, 47, 108–9, 127–28, lutionary theory, 59, 169, 174, 186, 179–80, 181; integrate with stem 200; and model organisms, 208; cell epigenetics, 85; Jablonka’s molecular, 9, 42, 88, 178, 183; role of view of, 80; epigenetics, stem cell, genes in, 120, 182–83, 213, 230n19; 10, 68, 88–89, 158, 160; integrate role of organism in, 207; structure with molecular epigenetics, 85–86; of, 38–39, 44, 166 Huang’s view of, 83–84; and Wad- epigenetic inheritance, 9, 26, 51, 86, 91, dington’s epigenetic landscape 98, 101, 174, 208, 211; in C. elegans, model, 78–79 118; channels of, 32; definition of, epigenome, 3, 25, 146, 154, 220n20; 24, 220n17; examples of, 32–35; Human Epigenome Project, explanation of, 64, 79, 87, 101, 181; 2–3; problem to locate, 147, 150, in humans, 120; and Kammerer, 235n22–23 209; and Lamarckian inheritance, epigenotype, 3, 69, 82 27, 35, 38, 80, 84, 226n41; mecha- evolution, 27, 28, 29, 37, 38, 43, 66, 82, nisms of, 130; origins of, 58; role of 138–39, 188–89, 196; and develop- in evolution, 47, 57, 83; stability of, ment, 23, 58, 80, 169, 172, 192, 199,

304 ◂ INDEX

213–14; developmental systems the- opmental explanation, 59, 61–65, orists’ view of, 214; Dobzhansky 88; evolutionary, 6, 11, 39, 58, 60, on, 56; gene-centered view of, 4–5, 63, 169–70, 173, 178–79, 187–89, 31, 207; of gene networks, 83; and 196, 200, 206, 239n5; explanato- heredity, 187, 197, 214; Lamarck’s riness of, 66, 169, 171, 239n11; of view of evolution, 29, 36; mecha- extended evolutionary synthesis, nistic view of, 140; of obesity, 193, 11, 197; functional, 14, 133, 142, 149; 241n27; role of development in, 2, gene-centered, 84; hybrid, 62, 148; 22, 28–29, 36, 39, 56–57, 58, 62–64, interventionist (account of), 90–91, 83, 199, 220n17; Waddington on, 87. 95–96, 105, 107–10, 127, 229n14; See also “development first view” likeliness of, 171, 239n6; lineage, of evolution; theory of evolution 60, 224n23; and mechanistic detail, evolutionary developmental biology 107–8, 179; molecular, 146, 155, 169, (evo-devo), 67, 208, 242n4, 243n7; 178–81, 183, 187, 200, 204, 229n13; compared to epigenetics and/or neo-Darwinian, 42, 172; neo- niche construction theory, 37–38, Lamarckian, 172; of obesity, 190– 57, 197, 199, 211–12, 214, 243n6; 94; organism-centered, 207; part- explanation in, 60, 227n1; and whole, 153; proximate cause (see extended evolutionary synthesis, proximate-ultimate distinction); 43, 65, 242n29; as trading zone, 88; pseudoscientific, 169; riskiness ecological and evolutionary devel- of, 173–74, 239n11; of robustness, opmental biology (eco-evo-devo), 159, 163–64, 237n35, 237n37; role of 118, 197. See also evolutionary genes in, 119–25, 128, 204; statis- novelty; evolvability tical, 45; theory of, 8, 204–5, 212; evolutionary novelty, 29, 37, 58, 211–12, ultimate cause (see proximate- 243n7 ultimate distinction); Waters’s evolvability, 37, 43, 221n22 view of, 113. See also causal expla- explanation, 7–8, 13, 44, 59, 86, 106, nation; constitutive explanation; 202–3, 213; attractiveness of, epigenetic explanation; mechanis- 172–73; biological, 1, 6, 11, 45, 90, tic explanation 106, 114–15, 118, 158, 165, 171, 177, explanatory power, 11, 94, 177–78, 190, 179, 200, 215; causal power of, 172; 206; compared to explanatori- concept of scientific, 90; contras- ness, 169–70, 206; and contrastive tive, 94–95, 174–87, 199, 217n3, explanation, 174–75, 189; definition 239nn12–15, 241n26; covering law of, 66, 169–70, 172, 174, 239n11, account of, 14, 132, 95–96, 131–32, 240nn14–15; of epigenetic expla- 238n2; counterfactual, 101; devel- nations, 187, 189, 195–96, 198; and

▸ 305 INDEX

extended evolutionary synthesis, gene-centrism, 43, 68, 84, 88, 197, 208 198–200, 206; and pluralism, 171, gene expression, 76, 79, 84, 113, 112; 200, 206; and precision, 175–76, and gene-regulatory networks, 78, 177, 179, 184–85; and sensitivity, 178, 83, 86, 152, 161–62, 218n4; heritable 180; and theoretical integration, changes in, 3, 24, 34; and histone 66, 198–99 modification, 25; patterns, 25–26, extended evolutionary synthesis 34; variation in, 20, 24–25, 33, 85 (EES), 9, 43, 169, 219n13; challenges gene regulation, 17, 27, 80–81, 136, 174, of, 42–43, 63, 65–66, 179, 197–200, 210 221n3; compared to modern syn- gene-regulatory network (GRN), 76, thesis, 65, 85, 221n1; and Lamarck, 78, 84, 154, 160–62 9; and philosophy, 215; pluralistic genetic assimilation, 27–29, 32, 36, 38, framework of, 11, 66, 171, 206; and 98, 219n11 proximate-ultimate distinction, 58, genetics, 19, 55, 80, 112, 136, 208; com- 64; role of development in, 36 pared to epigenetics, 2, 6, 100, 110, 113, 126; developmental, 12, 20, 23, Fagan, Melinda Bonnie, 79, 86, 237n39 79, 81, 158; early twentieth-century, Feyerabend, Paul, 173, 199, 240n15 12, 19; molecular, 15, 81, 125, 180, fitness, 22, 28, 63 199–200; population, 28, 43, 56, 65, 202, 221n2; Waddington’s views Gel’fand, Israel, 12, 73, 78, 225n34 on, 3, 12, 71, 79, 87 gene, 84, 109, 146, 153, 183; “above,” 2; genome, 14, 54, 61, 83–84, 101, 154–55, and characters, 111; compared to 157; compared to epigenome, the “epigene,” 146; definition of, 18, 147; hidden variability in, 20, 28; 230n18; and difference-making, 91, human, 18, 78; Human Genome 110, 117–18, 125; environment, 18, Project, 18–19, 27, 183; level of, 4, 218n4; for eye color in Drosophila, 229n13; and plasticity, 21; and syn- 110; gfp, 46, 100; hda-4, 46, 100, thetic biology, 139; viral, 34; -wide 122; rde-1, 120, 124; level of, 107; association studies, 181, 184, 199 and m-RNA, 130; mutual inhibi- genotype, 27, 82, 84, 107, 109, 191; as tion of, 161; as replicator, 83; role “blueprint,” 22; changes in, 28, 58, of in twentieth-century biology, 4, 60; and canalization, 20, 71; and epi- 217n1; selfish, 18, 22, 207. See also genotype, 3, 69, 82; frequencies, 28, DNA; gene expression; gene-regu- 58, 60; and phenotypes, 84, 111, 114, latory network (GRN) 117, 119, 125, 208; and plasticity, 149 gene-centered, 84, 86; view of evolu- genotype-phenotype map, 20, 22–23, tion, 4–5 81–82, 117

306 ◂ INDEX germ line, 24, 100; epigenetic inheri- 4, 14; obesity in, 181, 190–92, 195, tance through, 32–36, 38, 84, 100, 241n25 102, 220n17; segregation of, 30–31, Human Genome Project, 18–19, 27, 183 57, 220n20 Huxley, Andrew Fielding, 127, 229n16 Gilbert, Scott, 211, 214 Glennan, Stuart, 132, 141, 155 identity, biological, 67, 210–11, 224n29 Goodwin, Brian, 17, 152, 158, 201, 209 induction, 65, 76, 120; environmental, Griesemer, James, 30, 173–74, 206, 214 32, 190, 223n16; somatic, 32, 220n20 Gross, Fridolin, 160, 237n36 information, 10, 30, 51–53, 60, 65, Guerrero-Bosagna, Carlos, 189–90, 78, 91, 110, 113–15, 148; biological, 195–96 25, 208; causal, 54; about difference-making, 122, 124, 127, 162–63, Haig, David, 172–73 231n28, 237n37; environmental, 177, Helmholtz, Hermann Von, 106, 137, 222; epigenetic, 24, 44, 64, 80, 83, 229n11 109, 147; and explanatory power, Hempel, Carl Gustav, 95, 164, 228n5, 175–76, 179, 184–87, 195; genetic, 1, 238nn1–2, 238n4 4, 25, 140, 147, 154, 208, 214, 235n22; heredity, 30, 37–38, 173–74, 187, 197, irrelevant, 177; and laws, 164–65; 208, 214; broader notion of, 5; about mechanisms, 107–8, 128, 131, Darwin’s view of, 186; and devel- 169, 180, 187, 204; and models, 16, opment, 8, 9–10, 23, 39–40, 42, 105; transfer from DNA to protein, 102, 107, 183, 203; and epigenetics, 129–30, 132 2, 26–27, 40, 44, 80, 98, 102, 107, inheritance of acquired character- 110, 126; and natural selection, 4; istics, 29, 173, 219n14; and epi- neo-Darwinian view of, 29, 98; genetics, 5, 27, 35, 38, 42, 80, 203. nongenetic cell, 24, 26, 209; Weis- See also soft inheritance mannian model of, 30–31, 220n15 inheritance systems, 27, 45, 64, 80, 208 Herrera, Carlos, 48, 50–51, 103–4, integration, theoretical, 54, 179; of 222n13 epigenetics, 6, 9, 42–43, 59, 67, 85, histone modification, 25, 69, 115, 117, 174, 205, 209; and the extended 122, 124 evolutionary synthesis, 66 Hodgkin, Jonathan, 127, 229n16 interventions, 93, 100, 102, 155, 177, Holliday, Robin, 24, 27 180, 185, 229n10, 236n33; in biology, homology, 208, 212 45, 97–98, 100–102, 104, 113, 120, Huang, Sui, 78, 80–87, 149, 226n41 122; and counterfactuals, 93–94, humans, 3, 25, 33, 111, 137, 210; and 104, 195, 238n39; definition of, genetic diseases, 18; genome of, 92–93, 228n2: interlevel, 156–58;

▸ 307 INDEX

and invariance, 10, 92–94, 101–2, laws, 86–87, 230n17; biological, 14, 91, 108–9, 122, 127, 229n16; non-an- 95–96; coextistence, 163–64; and thropocentric, 92; and probabilis- invariance, 95; of nature, 14, 94; tic causation, 15; and specificity, succession, 163–64 113; statistical, 105 Lem, Stanislaw, 138–39 levels or organization, 54, 58, 87, 149, Jablonka, Eva, 20, 53, 80, 82, 84–85, 158, 162, 184; boundaries of, 143, 118, 130, 197, 214, 223n16, 226n39; 152; higher, 9, 14, 39, 132, 152; lower, on Human Genome Project, 18; on 10, 108, 131, 138, 152, 204; relations Lamarckian dimension of epigene- between, 60–62, 64, 107, 133–34, tic inheritance, 4, 27, 29 136, 143, 145–46, 148, 156. See also Jacob, François, 129, 132,134, 136, time scale 229n13, 240n15; and operon model, Leuridan, Bert, 155, 157 12, 17, 72; on molecular machines, Lewontin, Richard, 201–2 138, 233n10 Lillie, Frank Rattray, 19–20, 69 Johannes, Frank, 54–56, 86 mathematical models, 8, 10, 76, 86–87, Kauffman, Stuart, 18, 201 144, 202 Kisdi, Eva, 53, 226n16 Maynard Smith, John, 57, 201 Kuorikoski, Jaakko, 155–56, 170, Mayr, Ernst, 201, 207; on soft inheri- 177–78, 180, 240n14 tance, 27; on proximate-ultimate distinction, 14, 55–56, 59–60, 62, Lakatos, Imre, 208, 240n15 64–66, 84, 88, 187, 189, 196, 202, Laland, Kevin, 57–58, 63, 65–66, 223nn17–18 223n19 mechanism, 15, 45, 54, 58–59, 139, 155, Lamarck, Jean-Baptiste, 5, 9, 29, 36, 157, 165, 176–77, 183, 202, 209–11; 39, 42, 219n14 cell differentiation, 78, 149, 159, Lamarckian inheritance. See inheri- 161; concept of, 8, 10–11, 108, 129, tance of acquired characteristics; 130–32, 137, 140, 142–45, 148, 152–53, soft inheritance 155–56, 163, 232n1, 232n5; develop- Lamarckism, (neo-), 9, 36, 39, 68, 88; mental, 60, 148, 199, 207; ecologi- compared to modern synthesis, cal, 45; and epigenetic inheritance, 172–73; and Dawkins, 5; and epi- 3, 11, 25, 27, 69, 109, 120, 130; and genetics, 37; Huang on, 226n41 gene regulation, 24, 27, 150, 173–74, Lamb, Marion, 20, 84, 118, 214; on epi- 184; genetic, 120, 125, 129–30, genetics, 4, 27, 29, 197; on Human 179–81, 185, 187; as a machine, 130, Genome Project, 18 138, 140–46, 148, 150, 154, 204–5,

308 ◂ INDEX

233n10, 234nn17–19; properties of, itance in, 33–34, 36, 45–47, 109, 60; robust, 157–64, 166, 236n33, 120, 175, 184, 186; Weismann and 238n39; role of in explanation, experiment on, 30 107–9, 125–28, 131–37, 141, 147, 149, Mill, John Stuart, 13, 106, 164 165–66, 169, 189, 229n15, 230n18, Mitchell, Sandra, 163, 227n46 232n2, 233n7, 236n31. See also mitosis. See cell division mechanistic explanation model organisms, 8, 48, 53–54, 67, “mechanisms without causes,” 11, 207–8, 217n3, 242n4 165–66, 199, 205 models, 8, 11, 15, 44, 46, 50, 54, 88, 130, mechanistic explanation, 8, 10, 107, 137, 139, 153, 174, 182, 193, 202, 205; 132, 137, 153, 155, 199; and causal causal, 8, 20, 51–53, 98, 105; and com- explanation, 110, 131, 134, 185, 204, plexity, 16; epigenetic landscape, 74, 233n7; and covering law account 76–77, 79, 226n36–38; of gene-reg- of explanation, 131; and decom- ulatory networks, 17, 77, 84, 86–87; position, 133; demachinized, 204; mathematical, 8, 10, 76–77, 86, 144, and developmental explanation, 202; mechanistic, 8, 132, 144, 162–63, 62, 148; examples of, 136; and 202; statistical, 44–45, 48; visual, 8, interlevel relations, 156, 236n31; 79; population, 64, 22n10 interventionist view of, 91, 134; of modern synthesis, 30, 36–38, 43, 68, living systems, 15; and machines, 137, 211; compared to extended 140–45; of robustness, 159, 164–65, evolutionary synthesis, 58, 65, 237n39 (see also “mechanisms 67, 172, 200, 221n1; exclusion of without causes”). See also constitu- development from, 22–23, 28–29; tive explanation; new mechanistic integration of epigenetics into, 9, philosophy 85, 174 Mendel, Gregor Johann, 24, 28, 36, 84, modularity, 133, 162, 232n3, 236n33 125, 136, 222n8 molecular biology, 19, 24, 44, 52, 54, Mendelian genetics, 28, 84, 125, 136 82, 90–91, 98, 119, 128, 139, 155, metaphysics, 150–51, 155, 210 199, 227n1; birth and growth of, methylation, DNA, 26, 82, 117–18, 147, 12, 18, 71, 129, 130, 138, 150, 233n10; 184; and epigenetic inheritance, explanation in, 10, 80, 92, 106, 110, 32–34, 54, 68, 109, 115, 186; molec- 114, 130, 136, 143, 145, 153–54, 166, ular mechanisms of, 24, 150; regu- 168, 174, 177, 179–80, 183, 187, 204, lation of gene activity through, 25, 235n27; and Waddington, 71 147, 173; variation in patterns of, 32, Morange, Michel, 18, 214, 228n1 222n10 Morgan, Thomas Hunt, 12, 19, 23, mice, 149, 175, 177; epigenetic inher- 110–12, 115–17

▸ 309 INDEX

Monod, Jacques, 12, 129, 132, 134, 136, organisms, 14–16, 26, 46, 54, 60–61, 229n13; on molecular machines, 139; 64, 96, 101, 149, 157, 170, 173, 193, and operon model, 17, 72, 240n15 196; concept of, 151, 207, 218n2; Müller, Gerd B., 197, 214, 221n3, 243n7 Dawkins on, 22, 207; and envi- mutation, 34, 64, 82, 84, 184, 219n11; ronment, 4, 46, 58, 100; Lamarck directed, 220n20; epimutation, 82, on, 29; as machines, 140, 150, 207; 182; paramutation, 47, 109; rates, multicellular, 24, 69, 149, 211; 44, 182 Needham on, 16; and niche con- mutual manipulability account of struction, 37–38, 57; and plasticity, constitutive relevance, 157–58. See 28; and reproduction, 56; trans- also Craver mission between, 9, 15, 24, 27, 32, 40; Waddington’s view of, 22–23, natural selection, 4, 22, 56, 60, 170–71, 151, 158. See also model organisms 173, 191–92, 208, 233n8; blind, 140; of epigenetic variants, 55; and paradigm, 88, 119 genetic assimilation, 28; and niche parthood. See composition construction, 57; as an ultimate phenotypes, 21, 51, 117, 181, 190, 211; cause, 58, 61, 63–65, 189, 223n17 cell, 24, 154; environmentally Needham, Joseph, 16, 71, 242n1 induced, 32, 51, 220n17; and epi- Neel, James, 190–91 genetic regulation, 25, 28, 32, 152; neo-Darwinism, 41, 68, 82, 88. See and evolution, 58; genotypes devel- also Modern Synthesis oping into, 3; range of potential, neo-Lamarckism. See Lamarckism 21, 149; Waddington’s view of, 23. new mechanistic philosophy, 131, 133, See also genotype-phenotype map; 140, 150, 153 plasticity Ney, Alyssa, 230n17, 231n20 philosophy of biology, 14–15, 39, 136, niche construction (theory), 43, 57, 62, 170–71, 201, 212–13; philosophy for 65, 67, 187, 207, 211n23; compared biology, 206, 213; and theoretical to epigenetics, 37–38, 211, 214; and biology, 242n1 developmental systems theory, philosophy of epigenetics, 11, 12, 202, 213–14; and earthworms, 63 208, 210–15, 242n3 Nicholson, Daniel, 138, 140–41, 234n14 philosophy of science, 14, 132, 202, nutrition, 25, 100, 193 204, 206, 209, 213 physics, 9, 87, 138, 229n10 obesity, 1, 125, 181, 183–86, 190–95 Pigliucci, Massimo, 36, 197, 221n3, organicism, 71, 209, 235n24 224n28 organicists, 202, 242n1 plants, 48, 50–51, 53, 55, 104, 222n12

310 ◂ INDEX plasticity, 1, 22, 24, 37–39, 45, 52, 62, RNA, 18, 45, 47, 113, 124, 130, 230n19, 117, 173, 219n11; definition of, 21, 236n29; double-stranded (dsRNA), 218n6, 220n17; and environmental 100–101; messenger (mRNA), 118, change, 51; multigenerational, 36, 129–33, 135, 136, 221n6, 231n26; 22n17; role of and evolution, 57–58, small (microRNA), 25, 32, 34–35, 149, 207, 211; synaptic, 33, 136 45, 109, 218n4, 221n6 (see also RNA polygeny, 111, 116–17 interference) polyphenism, 22, 117 RNA interference (RNAi), 45, 100–101, popular science, 35, 210 120, 221n6, 223n15, 231n26 Popper, Karl Raimund, 169, 174, robustness, developmental. See canal- 239n11 ization Prentice, Andrew, 190–91, 193, 241n25 Russo, Federica, 102–4, 228n7, 229n15 proteins, 25, 130, 134–36, 218n4, 226n37, 230n19 Salmon, Wesley, 132, 177, 232n5 proximate cause. See proximate-ulti- schizophrenia, 1, 111 mate distinction selfish genes, 18, 22, 207 proximate-ultimate distinction, 6, 44, self-sustaining feedback loops, 25, 130 55, 148, 205, 223n19; and explan- Simon, Herbert, 133, 142 atory power, 179, 189–90, 192, Sober, Elliott, 114, 238n3 194–200; history of, 56–57, 223n18, soft inheritance, 5, 27, 30–31, 35. See 223n20; Mayr’s view of, 55–56, 60, also inheritance of acquired char- 88, 187, 196, 202; and reciprocal acteristics causation, 57–58; rejection of, 9, 42, Speybroeck, Linda Van, 84, 147, 214 57–63, 65–67, 84, 88, 168, 178, 197 stem cell, 8, 80, 86, 209; epigenetics, 10, 68–69, 78–79, 83–86, 88–89, Rahman, Quazi, 64, 189, 224n27 158–60; modeling differentiation Rassoulzadegan, Minoo, 45, 47, 109, of, 74, 76–79, 266n36, 266n38; and 120 pluripotency, 61, 74, 159; and repro- rats, 32–34, 118, 193, 194, 207, 242n28 gramming, 68, 76, 159; therapy, 1, reductionism, 153, 230–31, 235 198 replicator-vehicle distinction, 31, 57, 83 stress, 25, 32, 34, 118, 141, 196, 222n12 reproduction, 15, 56, 100, 120, 122, 140 structural equation modeling (SEM), reprogramming, cell, 26, 68, 74, 85, 159, 48, 52, 102–5, 222n11 198; and epigenetic landscape mod- systems biology, 80, 154, 199, 224n30, els, 42, 74, 76, 78; history of, 76, 209 227n43, 237n36; compare to epi- Richardson, Robert, 131–32, 141, 143, genetics, 7, 166, 197, 211; and stem 234n16 cell research, 68, 79

▸ 311 INDEX

Tal, Omri, 53, 226n16 Waddington, Conrad Hal, 12, 37, 39, temperature, 20, 25, 118 84, 98, 218n5; on canalization, Testa, Giuseppe, 67–68, 210, 224n29 20–21, 71, 98, 158, and epigenetic theoretical biology, 12, 201–2, 209, landscape model, 10, 42, 68–69, 242n1 71–74, 76, 78–79, 82, 86–87, 151, Theoretical Biology Club, 71, 209, 225n34, 226nn36–37; and epi- 242n1 genetics, 2–3, 9, 19–20, 22–23, 40, theory of evolution, 5, 31, 36, 43, 46, 68–69, 79, 81–82, 85, 89, 152, 202–6, 66, 80, 186. See also extended evo- 212; on genetic assimilation, 27–29, lutionary synthesis (EES); modern 98; on integrating development synthesis; neo-Darwinism and evolution, 23, 28, 87; and Thom, René, 12, 18, 201, 209, 225n34 molecular biology, 3; recogni- time scale, 144, 152 tion of work of, 68, 76, 209; and Tinbergen, Nikolaas, 56–57 theoretical biology, 12, 201–2; and “Towards a Theory of Biology” (Wad- Theoretical Biology Club, 71, 209; dington), 12, 201 on time scales, 152; and “Towards transcription, 15, 217, 236n29 a Theoretical Biology,” 12, 201, 214; Tsetlin, Mikhail L’vovich, 73, 78, and Whitehead’s process philoso- 225n34 phy, 71, 151, 209, 225n33 Waters, Kenneth, 46, 96, 110, 112–18, ultimate cause. See proximate-ulti- 231n20 mate distinction Watson, James, 12, 130, 146 Weber, Marcel, 127, 241n24 Vastenhouw, Nadine, 46, 99–101, 118, Weismann, August, 30–31, 36, 57, 120, 120, 124–25 197, 220n15, 220n20 variation, 28, 54, 58, 64, 65, 103–4, 114, Weismannism, molecular, 31, 197, 118, 219n11; Darwin’s explanation 220n15, 220n20 of, 30; epigenetic, (heritable), 5, 26, West-Eberhard, Mary Jane, 28, 212 27, 32–34, 44, 46–48, 50–51, 54–55, what-if-things-had-been-different 80, 82, 86, 100–101, 103–4, 120, 212, question (w-questions), 97–98, 220n20, 222n10, 223n15; in gene 101–2, 104; and explanatory power, expression, 25, 83; genetic, 44–45, 175–76, 178, 185, 194; Woodward 48, 71, 83–84, 101, 103, 118, 219n6, on, 94 228n7; hidden/cryptic, 28; origin Whitehead, Alfred North, 71, 82, 151, of variation, 173, 188; phenotypic, 209, 225n33 20, 28, 36, 117–18, 221n22, 243n7 Wittgenstein, Ludwig, 7, 88, 203, vitalism, 142, 202 227n1

312 ◂ INDEX

Woodger, Joseph Henry, 71, 209, 242n1 Woodward, James, 14, 91, 93, 96, 100, 107, 110, 155–57, 170, 175, 228n5; and change-relatedness, 163–65; on intervention, 45, 92, 101, 108, 228n2, 229n16; on invariance, 92–93, 103–5, 108, 134, 229n16

Ylikoski, Petri, 61–62, 148, 155, 170, 177–78, 180, 240n14

Zeeman, Christopher, 201, 225n34

▸ 313