Dædalus coming up in Dædalus: Dædalus

on happiness Martin E. P. Seligman, Richard A. Easterlin, Martha C. Nussbaum, Journal of the American Academy of Arts & Sciences Anna Wierzbicka, Bernard Reginster, Robert H. Frank, Julia E. Annas, Roger Shattuck, Darrin M. McMahon, and Winter 2004 Winter 2004: on on progress Joseph Stiglitz, John Gray, Charles Larmore, Randall Kennedy, on learning What we do & don’t know 5 Sakiko Fukuda-Parr, Jagdish Bhagwati, Richard A. Shweder, and others The of learning: a short history 13 Finding our inner scientist 21 on human , , Vernon Smith, Richard Wrangham, Daniel John Povinelli Of apes & men 29 Patrick Bateson, Thomas Sowell, Jonathan Haidt, and Donald Patricia Smith Brown Churchland How do neurons know? 42 Learning through others 51 on race Kenneth Prewitt, Orlando Patterson, George Fredrickson, Ian On the origin of concepts 59 Hacking, Jennifer Hochschild, Glenn Loury, David Hollinger, Clark Glymour The automation of discovery69 Victoria Hattam, and others

poetry David Ferry October 78 on imperialism Niall Ferguson, Kenneth Pomeranz, Tzvetan Todorov, Anthony Pagden, Jack Snyder, Akira Iriye, Molly Greene, William Easterly, ½ction Robin Blackburn, Henk Wesseling, and others Chuck Wachtel The Annunciation 79

notes Jennifer Hochschild on the social science wars 91 plus poetry by Richard Wilbur, Franz Wright, Rachel Hadas, Gerald Early on literature & childhood 95 W. S. Merwin, Charles Wright, Richard Howard &c.; ½ction by Richard Stern on a writer’s endgame 98 Erin McGraw, Roxana Robinson, Victor LaValle, Sigrid Nunez &c.; and notes by Michael Wood, S. George H. Philander, Linda letter on Cambridge’s ½rst African Ph.D. 103 Hutcheon, Charles Altieri, Donald Green, Shelley Taylor, Robert Nagel, Philip L. Quinn, Michael Kremer, Hue-Tam Ho Tai, Peter Hohendahl, Morris E. Fine, Richard Sobel, Stephen Hubbell &c.

U.S. $13 www.amacad.org

Inside front cover: Megan, a fourteen-year-old chimpanzee, uses a simple tool to reach for an apple. “Surely chimpanzees form con- cepts about concrete things. . . . But what about more abstract concepts? Concepts like ghosts, gravity, and God?” See Daniel John Povinelli on Behind the ape’s appearance: escaping anthropocentrism in the study of other , pages 29–41. Photograph courtesy of the Cognitive Evolution Group, University of Louisiana at Lafayette. James Miller, Editor of Dædalus

Phyllis S. Bendell, Managing Editor and Director of Publications

Janet Foxman, Assistant Editor

Contributing Editors: Robert S. Boynton, D. Graham Burnett

Board of editors

Steven Marcus, Editor of the Academy

Russell Banks, Fiction Adviser

Rosanna Warren, Poetry Adviser

Joyce Appleby (u.s. history, ucla), Stanley Hoffmann (government, Harvard), Donald Kennedy (environmental science, Stanford), Martha C. Nussbaum (law and , Chicago), Neil J. Smelser (sociology, Berkeley), Steven Weinberg (physics, University of Texas–Austin); ex of½cio: Patricia Meyer Spacks (President of the Academy), Leslie Cohen Berlowitz (Executive Of½cer)

Editorial advisers

Daniel Bell (sociology, Harvard), Michael Boudin (law, u.s. Court of Appeals), Wendy Doniger (religion, Chicago), Howard Gardner (education, Harvard), Clifford Geertz (anthropology, Institute for Advanced Study), Carol Gluck (Asian history, Columbia), Stephen Greenblatt (English, Harvard), Thomas Laqueur (European history, Berkeley), Alan Lightman (English and physics, mit), Steven Pinker (psychology, Harvard), Diane Ravitch (education, nyu), Richard Shweder (human development, Chicago), Frank Wilczek (physics, mit)

Dædalus is designed by Alvin Eisenman Dædalus

Journal of the American Academy of Arts & Sciences

The labyrinth designed by Daedalus for King Minos of Crete, on a silver tetradrachma from Cnossos, Crete, c. 350–300 b.c. (35 mm, Cabinet des Médailles, Bibliothèque National, Paris). “Such was the work, so intricate the place, / That scarce the workman all its turns cou’d trace; / And Daedalus was puzzled how to ½nd / The secret ways of what himself design’d.”–Ovid, Metamorphoses, Book 8

Dædalus was founded in 1955 and established as a quarterly in 1958. The journal’s namesake was renowned in ancient Greece as an inventor, scientist, and unriddler of riddles. Its emblem, a maze seen from above, symbolizes the aspiration of its founders to “lift each of us above his cell in the labyrinth of learning in order that he may see the entire structure as if from above, where each separate part loses its comfortable separateness.” The American Academy of Arts & Sciences, like its journal, brings together distinguished individuals from every ½eld of human endeavor. It was chartered in 1780 as a forum “to cultivate every art and science which may tend to advance the interest, honour, dignity, and happiness of a free, independent, and virtuous people.” Now in its third century, the Academy, with its more than four thousand elected members, continues to provide intellectual leadership to meet the critical challenges facing our world. Dædalus Winter 2004 Subscription rates: Electronic only for non- Issued as Volume 133, Number 1 member individuals–$36; institutions–$72. Canadians add 7% gst. Print and electronic © 2004 by the American Academy for nonmember individuals–$40; institu- of Arts & Sciences tions–$80. Canadians add 7% gst. Outside October the and Canada add $20 for © 2004 by David Ferry postage and handling. Prices subject to change The Annunciation without notice. © 2004 by Chuck Wachtel Institutional subscriptions are on a volume- Editorial of½ces: Dædalus, Norton’s Woods, year basis. All other subscriptions begin with 136 Irving Street, Cambridge ma 02138. the next available issue. Phone: 617 491 2600. Fax: 617 576 5088. Email: [email protected]. Single issues: current issues–$13; back issues for individuals–$13; back issues for institu- Library of Congress Catalog No. 12-30299 tions–$26. Outside the United States and isbn 0-87724-042-6 Canada add $5 per issue for postage and han- dling. Prices subject to change without notice. Dædalus publishes by invitation only, and as- sumes no responsibility for unsolicited manu- Claims for missing issues will be honored free scripts. The views expressed are those of the of charge if made within three months of the author of each article, and not necessarily of publication date of the issue. Claims may be the American Academy of Arts & Sciences. submitted to [email protected]. Mem- issn bers of the American Academy please direct all Dædalus ( 0011-5266) is published quar- questions and claims to [email protected]. terly (winter, spring, summer, fall) by The mit Press, Cambridge ma 02142, for the American Newsstand distribution by Ingram Periodicals Academy of Arts & Sciences. An electronic full Inc., 1240 Heil Quaker Blvd., La Vergne tn text version of Dædalus is available from The 37086. Phone: 800 627 6247. mit Press. Subscription and address changes Advertising and mailing list inquiries may be should be addressed to mit Press Journals, Five addressed to Marketing Department, mit Cambridge Center, Cambridge ma 02142-1407. Press Journals, Five Cambridge Center, Cam- Phone: 617 253 2889. Fax: 617 577 1545. bridge ma 02142-1407. Phone: 617 253 2866. Email: [email protected]. Fax: 617 258 5028. Email: journals-info@ Printed in the United States of America by mit.edu. Cadmus Professional Communications, Permission to photocopy articles for internal Science Press Division, 300 West Chestnut or personal use is granted by the copyright Street, Ephrata pa 17522. owner for users registered with the Copyright Postmaster: Send address changes to Clearance Center (ccc) Transactional Report- Dædalus, Five Cambridge Center, Cambridge ing Service, provided that the per copy fee ma 02142-1407. Periodicals postage paid at of $10 per article is paid directly to the ccc, Boston ma and at additional mailing of½ces. 222 Rosewood Drive, Danvers ma 01923. The Fee Code for users of the Transactional Report- The typeface is Cycles, designed by Sumner ing Service is 0011-5266/04. Address all other Stone at the Stone Type Foundry of Guinda inquiries to the Subsidiary Rights Manager, ca. Each size of Cycles has been separately mit Press Journals, Five Cambridge Center, designed in the tradition of metal types. Cambridge ma 02142. Phone: 617 253 2864. Fax: 617 258 5028. Email: journals-rights@ mit.edu. Howard Gardner

What we do & don’t know about learning

Suppose that we were commissioned or so, and how can these revolutionary to create a museum of learning. I don’t insights inform the education of future mean a stuffy, hands-off collection of generations? Finally, what puzzles old manuscripts or ½lms, but rather a remain? state-of-the-art exploratorium that In all probability our museum’s ½rst displayed the full spectrum of learning displays would show humans learning: types and, in vivid form, everything that infants crawling, walking, talking; tod- is known about learning.1 Suppose, fur- dlers engaged in rough-and-tumble or ther, that we had a budget for consul- imaginative play; youngsters (or old- tants and were able to hire the seven sters) at school, learning their 3Rs and experts whose essays are collected here. going on to master the disciplines and Presumably we would have in a perhaps engage in interdisciplinary number of guiding questions, among work. Casting our net more widely, we them: What examples of learning might exhibit a child learning to play a should we include? How should we musical instrument, an apprentice work- conceptualize this enterprise? What ing alongside a master builder, a medical progress has taken place in our under- student attending rounds, a recruit in standing of learning in the last century the military, a physically injured person recovering the ability to walk, a victim of Howard Gardner, John H. and Elisabeth A. a stroke learning to talk or reason again. Hobbs Professor in and Education at Casting our net more widely still, to the Harvard Graduate , is encompass the full range of learning best known in educational circles for his theory of among animals, our museum might in- multiple , a critique of the notion that clude exhibits of sea slugs that learn to there exists but a single human that move in certain directions while avoid- can be assessed by standard psychometric instru- ing others; ½sh that ‘imprint’ on certain ments. His most recent books are “Changing physical forms and trail after these privi- Minds” (2004), “Making Good” (2004), and leged forms throughout development; “Good Work” (with Mihaly Csikszentmihaly and 1 This supposition is not idle. At Harvard William Damon, 2001). Gardner has been a Project Zero, a research group concerned with Fellow of the American Academy since 1995. learning with which I have been af½liated since 1967, we are planning to construct such a muse- © 2004 by the American Academy of Arts um–at ½rst virtually, ultimately in bricks and & Sciences mortar.

Dædalus Winter 2004 5 Howard that learn their way around goal that is intended by another entity, Gardner natural and man-made mazes; pigeons which we may designate a teacher or on learning that can play ping-pong, trace missiles, trainer. Such a formulation allows us to and recognize human beings in photo- distinguish learning experiences from graphs; chimpanzees that can use sticks the sheer accumulation of mud on a tire, to wipe off termites or to hide treasures on the one hand, and from a computer from their fellow chimps, and that seem program that may accomplish amazing able to learn some language-like sys- feats, but always in precisely the same tems. way, on the other. We will also need to create displays From studies of preliterate cultures that show the organic structure of the and naive children, we can with some and explain how learning takes con½dence delineate the major folk the- place at microscopic levels: in regions ories of learning: Human beings learn by of the brain (like the hippocampus), observing others who are more knowl- in neural column networks (like those edgeable and by imitating, implicitly or involved in recognizing lines of different explicitly, what they do; asking ques- orientation), and even in single nerve tions of and listening to what more cells that form synaptic connections to knowledgeable individuals say; practic- other nerve cells and have those connec- ing a skill and noting its improvement; tions bolstered or weakened as a result receiving clear rewards or punishments of experience. that signal which course of behavior And our museum will also have to ex- should be pursued and which should be hibit machines that learn and think, so eschewed. In literate cultures, the theo- to speak: smart machines that can play ries of learning expand to include the (and improve in) chess, that can under- reading of texts and the taking of class- stand much of natural language, interact es; and there are of course more idiosyn- with human beings, and engage in scien- cratic theories that call attention to the ti½c problem solving. learning potential embodied in dreams, drugs, and deities. It is interesting to As the range of exhibits and displays note the extent to which more formal suggests, our imaginary museum, like theories of psychology and the group of consultants it has retained, venture beyond these ‘commonsense’ reflects a wide range of theories about views of the learning process. learning and about the appropriate level Before there was a formal psychology at which to analyze and understand the or of learning, it was philoso- phenomenon. phers who addressed issues of knowl- Still, as we plan our museum, it is edge–, learning, , important to step back and to provide and the like. With Greek and Roman both a de½nition and a little history. As thinkers as background, the philoso- a rough and ready approximation, learn- phers of the Renaissance and the En- ing occurs under the following condi- lightenment carved out positions that tion: An organism or entity represents a continue to serve as points of orienta- certain amount of information or data at tion today. Descartes took a strong men- time X; at time Y it can represent new or talistic position, arguing that the mind additional or transformed information operated according to its own principles that brings it closer to a goal–either a and that it came stocked with innate goal of the entity’s own choosing or a ideas. We see echoes of this perspective

6 Dædalus Winter 2004 in the writings of and tion. To be sure, were crucial What we Jerry Fodor, self-identi½ed nativists. for lower organisms. But it was the ver- do & don’t know about The British empiricists, led by Locke, tebrates–and especially mammals–ca- learning took an opposing perspective, according pable of problem solving and planning to which the mind was initially a blank who emerged as victors in the struggle slate; experience etched ideas onto the for survival. As soon as the implications slate and these ideas become associated of Darwin’s writings became clear, his with one another. Twentieth-century be- way of thinking came to dominate both haviorists like the Russian physiologist the theories and the empirical work of , the American scientists interested in learning. B. F. Skinner, and the ‘learning theorists’ The ½rst generation of modern schol- portrayed here by Jerome Bruner sub- ars of learning did not shrink from at- scribed to this empiricist point of view. tending to the more complex forms of In the eighteenth century, two new reasoning in human beings and other perspectives on learning took hold. Im- primates. But beginning in the early manuel Kant described the basic episte- twentieth century, the territory of learn- mological categories–time, space, num- ing was largely ceded to those research- ber, –that human beings neces- ers who stressed the continuity of learn- sarily imposed on their sensations and ing across the animal kingdom; avoided . Individuals did learn from issues of language, , and experience, but that experience was higher-order ratiocination; and strove to necessarily apprehended in temporal, explain any intellectual achievement in spatial, and causal ways. The Kantian the most parsimonious and reductionist problematic had a great effect on the fashion. Interestingly, this was true not research program of twentieth-century only for those experimentalists who Swiss developmental psychologist Jean worked primarily with rats and pigeons Piaget, who sought to describe the devel- (the two most common ‘model organ- opment in infants and young children isms’) but also for those, like Edward of these categories of experience. Piaget Lee Thorndike, who studied the acquisi- was also influenced by the writing of tion of skills in school-age children. For his countryman Jean-Jacques Rousseau, the ½rst half of the twentieth century, who discerned genius in the mind of this approach to learning held sway. the child and believed that And indeed, it might still hold sway should be allowed to unfold within, today had it not been for the develop- rather than be imposed didactically ment of high-speed computers and the upon, the child. complex programs that have permitted Studies of learning were influenced these electronic entities to compute enormously by the rise of evolutionary and solve various kinds of human-scale thinking, chiefly emanating from the in- problems. Once it became clear that sights of Charles Darwin and, to a lesser computers could mimic human thought extent, other British scholars like Alfred processes and–in the view of many– Wallace, Thomas Huxley, and Herbert bootstrap themselves over time to a Spencer. These writers all stressed the higher level of performance, scientists continuities between human beings and could no longer withhold such intellec- other animals, and the importance of tual competences from human beings. mental capacities that allowed individ- Thus was born the , ual organisms to survive until reproduc- an important intellectual movement

Dædalus Winter 2004 7 Howard among whose forefathers were the com- panzees. While both have documented Gardner on puter scientists Herbert Simon and Mar- the impressive capacities of chimpan- learning vin Minsky, the linguist Noam Chomsky, zees, they elect in their essays here to and the George Miller and focus on the fault line between chim- Jerome Bruner. I view the cognitive revo- panzees and children. Povinelli claims lution as a contemporary interdiscipli- that chimpanzees are incapable of ab- nary effort to provide scienti½c answers stract thought; that all of their achieve- to long-standing epistemological ques- ment is the result of of con- tions, such as our present consideration crete objects and events. Tomasello doc- –the nature of learning, why it is possi- uments that chimpanzees have only the ble, how it occurs. most meager capacities to imitate mod- In a broad , all of the consultants els, to infer the motives of others, and to to our museum–the contributors to this transmit any kind of cultural knowledge. issue of Dædalus, and the majority of The chimpanzee emerges in their ac- current workers on issues of learning– counts as an organism that is incredibly are offspring of this intellectual revolu- skilled at making use of the information tion of ½fty years ago. They recognize at hand, but that is unable either to con- the relationship between the long- ceptualize what is not present or to make standing philosophical agenda sketched use of the incidental knowledge attained above, on the one hand, and discoveries by other individuals in its group. Here in psychology, linguistics, anthropology, lies the huge fault line that separates , cognitive science, and chimpanzees from human children, who other relevant disciplines, on the other. from early on can engage in pretend play, And they believe that progress is being imitate elders, and rapidly assimilate the made in understanding the nature of knowledge that earlier generations have various kinds of learning, though they accumulated. may differ on how best to describe that Not represented in this collection but learning and the nature of that progress. worthy of note is the recent claim by pri- matologist Marc Hauser and his col- Now that I’ve surveyed the historical leagues Noam Chomsky and Tecumseh context to our current understanding of Fitch that they have identi½ed a crucial learning, it is timely to suggest the major capacity that is absent in nonhuman pri- dimensions against which to evaluate mates: the capacity for recursion.2 Bo- the speci½c contributions of our consul- nono apes are able to master language- tants–as well as those of some other like strings of symbols, provided that the consultants who might have been re- syntax of the string does not depend on tained. the capacity to embed one unit within Two dimensions seem particularly another. For instance, an ape may appre- useful: the learning of species or entities ciate the logic in the “Mom- to which these consultants compare hu- my sleeps,” but could be completely sty- man learning; and the type and extent of mied by an expression like “Baby said entailed in their efforts to that Mommy sleeps,” let alone “Daddy explain all manner of learning. Continu- said that Baby said that Mommy sleeps.” ing in the tradition laid out by Darwin 2 Marc Hauser, Noam Chomsky, and Tecum- and his successors, Daniel Povinelli and seh Fitch, “The Faculty of Language: What Is Michael Tomasello ½nd it productive to It, Who Has It, and How Did It Evolve?” Science delineate the nature of learning in chim- 298 (2002): 1569–1579.

8 Dædalus Winter 2004 In her essay here Alison Gopnik focus- crucial missing, something that calls for What we do & don’t es on the characteristics of young chil- an explanation at the level of the human know about dren, but she brings to bear an entirely organism? learning different comparison group. Like several As I read her essay, Churchland stress- other contemporary developmental psy- es the importance of understanding the chologists, Gopnik ½nds it useful to nervous system and chides those philos- think of the child in comparison to the ophers who do place the same value on working scientist–a worker rather like it. But she does not feel the need to dis- herself. In making this analogy, Gopnik pense with a psychological level of ex- revisits a theme introduced decades ago planation. For most of his essay, Gly- by (as well as themes ½rst mour embraces a tougher-minded re- articulated by the philosopher of science ductionism: we should stop trying to Thomas Kuhn in his discussion of para- solve problems that are too compli- digmatic scienti½c revolutions). How- cated for us and instead turn them over ever, Gopnik goes well beyond Piaget, to those ever smarter computers. But in who tended to emphasize the limitations the end, Glymour acknowledges that the in childrens’ thinking and the disconti- problems solved by computers are ones nuities between child tinkerer and that human beings have formulated scientist. She argues that, like scientists, and that, at both ends of the process, very young children are capable of put- we need human judgment after all. ting forth theories, carrying out experi- ments, observing the of As one who surveyed the cognitive sci- others, and discerning statistical pat- ences in some detail twenty years ago,3 terns. In an accompanying essay, Su- I am very impressed by the knowledge san Carey characterizes the conceptual that has accumulated in the past few de- growth from child to scientist as a boot- cades. Thanks to theoretical and empiri- strapping operation; only through such cal researchers like those represented self-constructing operations can chil- here, we know a great deal more about dren proceed, for example, from early the nature of early learning and under- intuitions about quantity to a full- standing in human beings, and can point blown sense of number. with far greater precision to the ways in Possibly because they are philoso- which humans differ from their closest phers, and Clark biological relatives. Our accumulating Glymour are less concerned with the knowledge of the nervous system is speci½cs of experiments involving in- even more impressive, and the bridges fants or chimpanzees. Churchland dis- between cognitive science and neuro- cusses learning at the level of individual science are sturdier. The accomplish- nerve cells, while Glymour describes the ments of computers are also striking; powerful operations that can be carried and while some of these accomplish- out by high-speed computers. ments are achieved by methods quite From my vantage point, the key issue remote from those used by Homo sapi- their juxtaposed essays raise is the ques- ens, we are beginning to have software tion of reductionism. To put it sharply, and hardware that in important respects can human learning and thinking be learn in ways that resemble our own adequately reduced to the operations of 3 Howard Gardner, The Mind’s New Science: A neurons, on the one hand, or to chips of History of the Cognitive Revolution (: silicon, on the other? Or is something Basic Books, 1985).

Dædalus Winter 2004 9 Howard learning. In fact I have heard that good Albert Einstein’s coming up with the Gardner on chess masters now study computer theory of relativity; Marcel Proust’s pen- learning games and learn new strategies from ning À la recherche de temps perdu; Andrew those inanimate models. Ultimately, of Wiles’s solution of Fermat’s theorem; or course, debates similar to those between Mahatma Gandhi’s creation of peaceful the rationalists and the empiricists in the nonviolence; or, indeed, Ramon y Cajal’s eighteenth century, and between the studies of the nervous system, or John learning theorists and the Gestaltists in von Neumann’s formulation of the na- the ½rst half of the twentieth century, ture of computer programming? Even are still being waged. Yet, as Jerome the high-speed computers that can Bruner indicates in his essay, the current handle far more variables than ordinary debates are conducted in much more so- mortals can do not illuminate the nature phisticated ways. of original artistic, scienti½c, or political But as one who seeks to build a muse- thought. um of learning, I am struck by the limi- tations, reflected in the essays collected Moreover, the contributions of cog- here, of the still evolving cognitive sci- nitive science to schooling–the chief ences. Let me mention the principal institution devoted to learning–remain ones. modest. As one who spends his life at a First of all, these disciplines almost school of education and has devoted always deem the scienti½c mind as the much time to school reform, I could not proper model of human thought. The help but be struck by the virtual absence claim that all cognition, learning, devel- here of any reference to schools, formal opment, and intelligence are best rep- teaching, the 3Rs, and the scholarly dis- resented by those of the scientist was ciplines–in short, the realms that most Piaget’s great contribution, and also his individuals think of nowadays when weakness. And we see his sentiments in they think of learning. Part of the ex- most of the essays here. But human planation for this is undoubtedly that thinking and learning is achieved as schools are very complex institutions well by artists, musicians, politicians, and the processes of learning that businesspeople, inventors, religious are supposed to take place there over leaders, and dreamers–we must under- months or years are dif½cult to capture stand their forms of learning and the in scienti½c research. ways in which they may differ from the Still, I think that more can be said cognition of the theoretical physicist or about how our current understanding of the benchtop chemist. learning might influence education, as Second, the instances of learning that well as the obstacles that make such ap- are most frequently examined typically plications dif½cult. In my own work I take place over brief periods of time: have recently focused on two lines of nanoseconds in the case of computers, research. The ½rst outlines the various milliseconds in the case of nerve cells, misconceptions that readily arise in ear- an hour or less in the case of most exper- ly childhood and that considerably com- imental trials. Yet the forms of human plexify the mastery of the disciplines. thought that are most valued often re- It turns out that young children readily quire the investment of months or even embrace creationist accounts of the ori- years. What of the learning involved in gin of the species, a phenomenon that

10 Dædalus Winter 2004 makes dif½cult the learning of evolu- that respond principally to political What we do & don’t tionary theory; embrace Aristotelian ac- pressures and institutional impera- know about counts of the behavior of physical mat- tives.5 This, in short, is the reason that learning ter, which render the mastery of New- politicians in America talk incessantly tonian physics problematic; and readily about test scores and international com- embrace one-dimensional accounts of parisons, and rarely if ever mention historical and political events, thus mak- what has been learned about learning. ing it dif½cult to appreciate complex and Independent schools have somewhat multicausal accounts. Recognition of greater latitude in what they prescribe the early emergence and dogged persis- but they are by no means immune from tence of these misconceptions is an es- these social pressures. Only in home- sential component of effective pedagogy. schooling or individual tutoring can the Teachers must directly confront these student readily bene½t from our growing misconceptions and give students ample understanding of learning. And only opportunities to air their understand- when ways are found to bridge the gap ings and misunderstandings, to discover between the knowledge being accumu- where they are inadequate and where lated by scholars and the typical opera- they require revisions.4 tions of schools in the nation-state will I have also collected evidence against the pipeline between research and prac- the contention that intelligence is a sin- tice be opened. gle, all-encompassing human capacity. I favor an alternative account: that all Finally, I believe that there is also a human beings possess a range of intel- conceptual gap that needs to be ad- ligences and that we differ from one dressed–both by our consultants and another in our intellectual pro½les. by our hypothetical museum of learning. Throughout most of history, educators That gap concerns the fact that human have ignored this possibility and have beings are social, cultural, and historical taught subjects in one way–thereby creatures as much as we are neurologi- inevitably favoring students who are cal, psychological, and computational strong in linguistic and logical ways of creatures. We evolved to do many things thinking. It is possible to reverse this well–but we did not evolve to create uniform approach, and so reach more calculus or write the U.S. Constitution students, by presenting materials in a or compose classical music or invent multitude of ways and giving students airplanes and the pill. Nor could anyone options in how they may convey their have anticipated, even ½fty years ago, own understandings. the civil rights revolution, or the femi- Alas, even if we had exquisitely de- nist revolution, or the fall of commu- tailed and powerful theories of learning, nism, or the proliferation of nuclear this would be no guarantor that they weapons, or the rise of the World Wide would be adopted widely in schools. As Web in an increasingly globalized civi- David Olson argues in his recent book, lization. Yet somehow individuals grow- schools are bureaucratic organizations ing up in the early twenty-½rst century must be able to master these bodies of 4 See Howard Gardner, The Unschooled Mind (New York: Basic Books, 1991) and The Disci- 5 David Olson, Psychological Theory and Educa- plined Mind (New York: Penguin Putnam tional Reform (New York: Cambridge University Books, 2000). Press, 2003).

Dædalus Winter 2004 11 Howard knowledge and cope with these events Gardner on of historical signi½cance. Little in our learning sciences of learning addresses issues of this scale; our cultural, historical, and literary studies do not make much con- tact with our scienti½c approaches; an interdisciplinary span across these broad disciplinary terrains still eludes us. In short, there will be a great deal of interest on display in our hypothetical museum of learning. But there will also be a large number of empty rooms–for there is still so much that we have to learn about learning.

12 Dædalus Winter 2004 Jerome Bruner

A short history of psychological theories of learning

Learning remains an elusive topic, de- form-wise? How do we learn the lay of spite the endless research lavished on it. the land? How do we learn to concen- And what we mean by it, of course, is trate our attention? shaped by how we choose to study it. And then there are questions about Concentrate on how children master differences in how learning occurs. Do their native language and you arrive at all species learn in the same way and do a very different conception of learning the bright and the dull go about it in like than had you researched how under- manner? And what about external in- graduates memorize nonsense syllables. ducements, rewards, and punishments? Does learning to ½nger a Bach cello so- Are all learning situations comparable? nata tap the same learning processes as I used to give the star performers of learning to trace your way through a the experiments I’d just completed to ½nger maze? Is all learning alike, re- my young daughter. These rats seemed ducible to a common set of principles? to develop a more open curiosity under Two learning tasks are said to be alike her magnanimous care. What, indeed, if mastering one makes mastering the does domestication do to an animal’s other easier–the so-called transfer cri- approach to learning? Were those in- terion. But what is transferred? Is it re- sights achieved by Wolfgang Koehler’s sponses? Rules? Or do we simply learn pampered chimpanzees–their ½guring how to learn, as when with enough out how to rake in an out-of-reach ba- practice we become exam-wise or tax- nana by putting two sticks together, for instance–simply the result of the Jerome Bruner, a Fellow of the American Acade- leisurely tutelage they received on that 1 my since 1954, is University Professor at New York German island of Tenerife? It used University, where he teaches principally in the to be said, only half jokingly, that Yale School of Law. With George Miller he founded -response-reinforcement learn- Harvard’s Center for Cognitive Studies in the ing theory was different from more cog- early 1960s. He has published widely, his work nitive California theory because Clark principally focusing on the interaction of mind Hull in New Haven taught his graduate and culture. His latest book is “Making Stories: students that rats “should get on with Law, Literature, Life” (2002). 1 Wolfgang Koehler, The Mentality of Apes © 2004 by the American Academy of Arts (New York: Harcourt Brace, 1926). This was originally published in German in 1917. & Sciences

Dædalus Winter 2004 13 Jerome it,” while Edward Tolman counseled his goes, psychology is forever subject to Bruner on at Berkeley that rats need time enough to physics envy.) The atomism of learning learning pause at the choice points in a maze. theory embodies the notion that learn- And ½nally, do we learn for learning’s ing consists of the association of ideas, sake, or must we be extrinsically moti- , sensations, whatever; at its vated to do so? Assuming the latter, the heart is the conception of the associative Yerkes-Dodson law tells us that too bond, the linkage that co-occurrence or much or too little reduces spatial proximity produces between two learning. I checked that out once myself sensations or ideas. While association- and got a surprise. I found that very hun- ism is of ancient provenance, it had gry and just moderately hungry rats more recent philosophical adherents learned to ½nd their way through a suc- as well–not only Aristotle, but Locke, cession of pairs of doors. The correct Berkeley, Hume, and pére et ½ls Mills. path through was marked redundantly Indeed, by the mid-nineteenth century, in two ways: follow a left-right-left-right philosopher-psychologist Johann Frie- path, or just choose the darker door at drich Herbart had proclaimed the asso- each choice point. The hungry rats ciative bond as the keystone of the new learned only one of the two cues; the psychology. moderately hungry rats learned both. This paradigm found further, if indi- The less hungry rats had a more open rect, support in the newly burgeoning curiosity–like my daughter’s pets. brain physiology of those times. As the nineteenth century entered its last quar- Given all this, it is natural enough that ter, the older phrenology of the days of scientists would want somehow to sim- Gall and Spurzheim was reformulated in plify what we mean by ‘studying learn- terms of newly discovered cortically lo- ing.’ And, of course, the standard way of calized ‘centers’ in the cerebral cortex, doing that is to agree on some paradigm each dedicated to a particular function. that would make it possible to compare Perhaps the most compelling localiza- results. That is exactly what happened at tion study was the one conducted in 1870 the very start of learning research. But, by the German physiologists Fritsch and as often happens, rival paradigms came Hitzig. In their study, electrical stimula- into existence and, alas, this research tion of different spots in the medial- soon became a war of would-be para- lateral cortex produced particular, quite digms. Indeed, the learning theory wars ½nite motor responses: stimulating one that resulted came to dominate the psy- spot produced flexion of a monkey’s chological research scene from the latter forearm, another would turn his eyes nineteenth century until a decade after upward, still another would turn them World War II, with various ‘schools’ downward.2 If the brain were organized devising clever experiments to demon- in this localized punctate way, psycholo- strate how well their paradigm worked, gists asked, why not the mind as well? or how poorly rival ones fared. One needs to remember that the pre- There were two competing paradigms vailing philosophical view among those from the start, each with its variants. The principal one, a child of its times, 2 The classic article was Gustav Fritsch and was molecular associationism, a meta- Eduard Hitzig, “Ueber die elektrische Errig- phoric extension of the atomism of barkheit des Grosshirns,” Archiv der Anatomie nineteenth-century physics. (As the quip und Physiologie (1870): 300–332.

14 Dædalus Winter 2004 scholars was psychophysical parallelism, the beginning or end–are easily repro- Psychological 3 theories of which held that mind and brain move ducible. learning along parallel tracks. But the associative bond, even be- Their critics, however, championed tween nonsense syllables, soon came to another model–that of molar con½gura- seem mentalistic, too fragile to suit the tionism. This paradigm took as its major scienti½c taste of the times. So by the premise that mind and brain alike oper- turn of the century it was replaced by ate as integral systems controlling the Pavlov’s more scienti½cally solid ‘condi- functioning of component parts. Like its tioned reflex.’ Pavlov’s paradigm physi- rival, it too rested its case on brain phys- calized associationism, turning its con- iology, for there was already plenty of tent into something more measurable evidence that overall cortical processes while preserving its associative form in- controlled localized centers–the neural tact. All his paradigm required was link- ‘mass action’ holism represented by the ing and relinking stimuli and responses: renowned Pierre Flourens. a salivary reflex, once produced by food, The brain’s mass action was analogous was now evoked by a bell signaling the to the phenomenology of everyday life– coming of food. Pavlov’s Nobel Prize in that ordinary experience transcends its physiology seemed to clinch the triumph bits and pieces. The ‘urban scene,’ after of physicalism. But Pavlov himself was all, is more than just a collection of taxis, not altogether pleased, as we’ll see later. buildings, pedestrians; its properties as a Now turn to con½gurationism, which whole shape the elements that make it had no shortage of psychologists to sup- up. was, of course, port it, dubious as many were of associa- the most direct expression of this view, tionism’s abstractness and its remote- and it had much to say about how learn- ness from ordinary experience. Con½g- ing was a matter of overall organization urationism had the support of brain re- rather than of local associative linkages. search as well, with the holistic neurolo- gy of the indomitable Flourens still very Consider now the rise of the associa- much in vogue. Also in those ½n de siècle tionist paradigm. That closing quarter times there was a rising tide of interest of the nineteenth century was a time of in how language and culture shaped many new studies of learning–mostly mind, with ½gures like Emile Durkheim concerned with the memorization of and Max Weber in the neighboring dis- lists of words or pairs of words to be as- cipline of sociology urging that culture– sociated. But it was the nonsense syllable not just individual encounters with the principally that gave associative bonding world of physical nature–also forms its scienti½c flavor. Hermann Ebbing- mind. haus used nonsense syllables in order to 3 Ebbinghaus’s 1885 classic is available in Eng- rule out past experience and ‘’ lish only in brief, but representative excerpts in explanations of learning. Ebbing- may be found in Wayne Dennis, Readings in haus’s 1885 Ueber das Gedaechtnis is a te- the (New York: Appleton- dious account of learning lists of non- Century-Crofts, 1948), 304–313. Interestingly sense syllables (with Ebbinghaus him- enough, Ebbinghaus’s original monograph was published in its entirety in English translation self as the subject of most of the experi- in 1913 by Teachers College, Columbia Univer- ments). His ½ndings–for example, that sity–very much in keeping with the then dom- nonsense syllables in the middle of the inant emphasis on in American list are more slowly learned than ones at education. It has long been out of print.

Dædalus Winter 2004 15 Jerome Gestalt theory was the prime exemplar parent movement, not the light points Bruner on of the con½gurationist trend in those moving. The whole, then, is indeed dif- learning early years, though it hit its full stride ferent from a sum of its parts. only after World War I. Its credo was Now as it happens, Pavlov himself that all systems–physical, biological, came to advocate a kind of linguistic and mental–have the intrinsic character con½gurationism. How does the condi- of controlling the local elements that tioned response square with an ordered compose them. Field theory in physics phenomenon such as language? Does was its model, and its proclaimed maxim language change how stimuli are inter- was “The whole is greater than the sum preted, how a conditioned stimulus is of its parts,” which the Gestaltists pro- substituted for an unconditioned one in ceeded to con½rm with a steady stream the case of human beings? Troubled by of clever studies on human perception. such issues in his later years, Pavlov pro- The Koehler chimpanzee studies on Ten- posed a Second Signal System whose erife were intended to make the same stimuli were not raw physical inputs, but point where learning was concerned: language imbedded in codes and cate- There was no way in which those chim- gories. Thus linguistic synonymy influ- panzees could turn a pair of sticks into a enced stimulus substitution in ordinary reaching tool by the simple ‘association’ conditioning. of elements. It took an act of insight to Some say that Pavlov was driven to do so, a way of con½guring the whole sit- his new views by communist ideologues uation. with prematurely Gramscian leanings, Koehler had a deep in the ubiq- but in fact his Second Signal System was uitousness of con½gurationism in all of quite in keeping with the European tra- nature. He launched one of his ½rst ma- dition of human studies, Geisteswissen- jor attacks on associationism by arguing schaft, rather than with Naturwissenschaft the insuf½ciency of atomism, in a book –a well-revered tradition among the bearing the forbidding, if telltale, title Russian intelligentsia. Still, structural- Ueber die physische Gestalten im ruhe und im ism was virtually the hallmark of the stationaren Zustanden (On physical con- lively Russian literary and linguistic ½gurations at rest and in stationary scene of Pavlov’s day, and the Second states). If atomism was insuf½cient even Signal System was certainly, to some de- in physics, Koehler asked, how could it gree, a response to that scene. I recall fly- serve as a paradigm for psychology?4 ing to Moscow from Paris in the 1960s He applied a phenomenon in visual per- with the celebrated Russian emigré lin- ception to make an analogy that would guist Roman Jakobson. He laughed when drive home his point: When two nearby I told him about Pavlov’s later turn and points of light are briefly flashed one af- about the accusation that he had knuck- ter the other, the eye perceives pure ap- led under to the nomenklatura. “No, no, Jerry, communist ideologues weren’t 4 For Koehler’s philosophical allegiances, see needed, just being Russian was enough. Mary Henle, ed., The Selected Papers of Wolfgang And being a Russian intellectual besides! Koehler (New York: Liveright, 1971). Perhaps Not even Pavlov could live with the idea the best and most accessible account of Gestalt that language makes no difference, that psychology’s empirical accomplishments (mostly before Hitler’s rise to power) is Kurt people learn like dogs!” Koffka, Principles of Gestalt Psychology (New Small wonder that cultural theorists York: Harcourt Brace, 1935). like Vygotsky and Luria took over after

16 Dædalus Winter 2004 Pavlov and that many of the post- all distinguished, self-professed Psychological theories of Pavlovian young studied Gestalt psy- stimulus-response learning theorists. learning chology at the Institute of Psychology in Their forte was the well-designed ani- in the years after.5 mal : maze running, discrim- ination learning, operant conditioning The climax of the rivalry between asso- à la Skinner box, and the like–mostly ciationism and con½gurationism came with rats as subjects, but sometimes pi- in America in the years before World geons, and occasionally monkeys. Un- War I. Nourished by the imposing Ed- dergraduates were used as well, but ward Lee Thorndike of Teachers Col- again, mostly in rote learning experi- lege, , the associa- ments–in what was referred to in my tive paradigm had flourished in the graduate student days at Harvard as United States. Thorndike had been a ‘dustbowl .’ It was in these postdoctoral student at one of the major days that Pavlov’s dog became a meta- centers of associationism in Germany. phor for American know-nothing anti- On his return to America (and Teachers intellectualism. College) he popularized practice and The burden of the behaviorists’ ½nd- repetition as the routes to pro½cient ings, taken collectively, was that repeti- school learning: practice and repeat as tion of a task, with suitable reinforce- you would were you memorizing non- ment for completing each trial, im- sense syllables.6 proved performance. There were sub- But the associationist research pro- tleties, to be sure–like the deleterious gram soon changed in America under effects of massing trials rather than the influence of Pavlov. J. B. Watson, the spacing them, creating interference by founder of American , pop- setting positive and negative reinforce- ularized Pavlov and gave his ½ndings an ment in a conflicting relationship, and American twist, by stressing how all the like. But the overall outcome of the learning occurred through stimulus and work, where ordinary everyday learning response. I sometimes wonder whether was concerned, was, I believe, much as it was Watson’s oversimpli½cations that I’ve stated it. I’ll return to this matter eventually drove American association- later. ist learning theorists to their zealous rig- But, as in Europe earlier, a contrarian or in exploring Pavlov’s ideas. It was the con½gurationism soon came into being. energy and determination of their re- Partly it was influenced by Gestalt theo- search that made America for half a cen- rists, now in America and sparking the tury the home of later Pavlovianism, a opposition, but it had American roots as half century dominated by the likes well, nourished particularly by Edward of Walter Hunter, Clark Hull, Edward Tolman, who was sympathetic to the Guthrie, B. F. Skinner, Kenneth Spence– work of Koehler and was a close friend of , a latter-day leader in the 5 , Thought and Language (Cam- Berlin Gestalt group. Tolman’s brother bridge, Mass.: mit Press, 1962); Aleksandr Romanovich Luria, The Role of Speech in the Reg- Richard, moreover, was a distinguished ulation of Normal and Abnormal Behavior (New nuclear physicist and shielded him well York: Liveright, 1961). from old-fashioned atomistic notions– and, indeed, from physicalistic tempta- 6 Edward L. Thorndike’s classic is the three- tions. Tolman, from the start, was a cog- volume , which appeared in 1913–1914. nitivist.

Dædalus Winter 2004 17 Jerome Tolman’s ½rst major book appeared maps are not mirrors of the happen- Bruner on in 1932 and it quickly gained adherents stance of our encounters with the world, learning among the discontented, and there were but a record of our strivings and what plenty of them. His students–notably has proved relevant to their outcome. , but many others as well– His views in this sense were basically also joined the battle against associa- pragmatist, perhaps because of his years tionism. By World War II, there was vir- of exposure as a psychology graduate tually open conflict in America between student to Harvard’s pragmatist philoso- con½gurationist and associationist learn- phers, particularly C. I. Lewis, whom he ing theorists–the ½rst holding that greatly admired. Following Tolman’s learning is principally a task of organiz- lead, David Krech went to the extent of ing knowledge from the top down, the proposing that learning is hypothesis second insisting that it is accreting it driven, not just passive registration. from the bottom up. The con½guration- Even rats, Krech tried to show, generate ists, though still a minority, had been of- hypotheses.8 ½cially well received on the American It’s revealing to compare Tolman with scene when they fled Hitler’s Europe. the leading, perhaps most radical associ- Koehler was invited to deliver the Wil- ationist behaviorist of the same period, liam James Lectures at Harvard, and B. F. Skinner. Skinner was surely as com- Kurt Lewin became virtually a cult ½g- pelling in defense of operant condition- ure in . The displaced ing as Tolman was of cognitive map the- members of the old Gestalt group were ory. His central concept was the operant soon well placed in leading American response–an act not initially under the universities. They made commonsense direct control of some particular feature phenomenology seem commonsense of the immediate environment. An ex- rather than arcane, an achievement ample of an operant response is provid- given the hold of behaviorist American ed by a starting pigeon in a Skinner box psychology. Learning began to be under- whose pecking of the button on the stood as grasping things in context, not box’s wall either produces or fails to pro- in bits. duce a reinforcement (a grain of seed, Take Edward Tolman’s research as an say). Any reinforcement increases the example. He taught that learning is like likelihood of the operant response mapmaking and that to learn is to organ- occurring again, the level of likelihood ize things in the light of their utility for depending upon whether the reinforce- achieving ends. In “Cognitive Maps in ment always follows the response or Rats and Men,” his still renowned Re- does so only sometimes, and whether it search Lecture to the Berkeley faculty in does so regularly (periodically) or irreg- 1947, Tolman claimed that trial and error is not so much acting out habits to dis- ry, 1932). His Berkeley lecture was later elab- cover which are effective, but rather a orated in his International Congress address, looking back and forth to get the lay of “Cognitive Maps in Rats and Men,” Psychologi- the land in order to construct a solution. cal Review 55 (4) (1948): 189–208. That is why he urged his graduate stu- 8 Krech’s bold study (written under his origi- dents not to rush their rats through the nal name) is I. Krechevsky, “‘Hypothesis’ ver- maze.7 He believed that our cognitive sus ‘Chance’ in the Presolution Period in Dis- crimination Learning,” University of California 7 Tolman’s most influential book was Purposive Publications in Psychology 6 (1932): 27–44. Behavior in Animals and Men (New York: Centu-

18 Dædalus Winter 2004 ularly (aperiodically). Partial aperiodic sic sense died around 1960–though Psychological theories of reinforcement, for example, evokes a there are still Skinnerians who stalwart- learning rather more persistent response than ly continue to publish operant ½ndings, one might expect, though Skinner would mostly for each other. I know of no more scoff at interpreting such persistence as Tolmanians or Hullians.11 hope springing eternal. Learning, in It was the cognitive revolution that Skinner’s austere terms, is under the sole brought down learning theory or, per- control of schedules of reinforcement: haps, focused attention elsewhere. After reinforcement can only be positive; pun- 1960, say, stimulus-response learning ishment does not affect learning. And theory seemed quaintly stunted, that is about it. As Skinner would some- hemmed in by its own self-denial. As for times say, a bit ironically, learning more molar, cognitive learning theories, scarcely needs a theory.9 many of their ideas were restated and Not all behavioral associationists, to absorbed into general cognitive theories be sure, shared Skinner’s disdain for the- such as Newell and Simon’s on problem ory. Clark Hull at Yale, indeed, elaborat- solving, or Bruner, Goodnow, and Aus- ed his theory into a highly re½ned set of tin’s on thinking, or Miller, Galanter, axioms about what constitutes positive and Pribram’s on planning.12 By the lat- and negative reinforcement, what makes ter 1960s, learning was being translated a conditioned stimulus generalize along into the concepts of information pro- a certain gradient, how organisms an- cessing, with no compulsion to elevate ticipate reinforcers, and the like–all in one kind of learning over another in rather exquisite and specialized detail. terms of its ‘basic’ properties. Certainly, His ½rst books–the 1943 Principles of Be- the old wars were over. And so, interest- havior and the more triumphally titled ingly, were the old labs and their 1952 A Behavior System–bristle with ta- ubiquitous mazes. bles and idealized learning curves and As I reflect on the transition period, I with abstract formulae for relating those think that it was the study of language ½ndings to his central axioms–perhaps and particularly of a prophetic effort to devise a mathemati- that precipitated learning theory’s de- cal model of learning, the preoccupation cline. Language use and its acquisition of computational psychologists a gener- are too out of reach of piecemeal S-R ation later.10 learning: efforts to bring them into the fold soon become absurd, and linguists The conflicts between Hull and Skin- have mostly dismissed them as such. ner, and between both of them and Tol- man, were the last battles of the learning 11 The most detailed and authoritative volume theory wars. Learning theory in the clas- on the classic learning theories is Ernest R. Hil- gard, Theories of Learning, 2d ed. (New York: Appleton-Century-Crofts, 1956). 9 B. F. Skinner, “Are Theories of Learning Necessary?” 57 (1950): 12 Alan Newell and Herbert A. Simon, Human 193–216. Problem Solving (Englewood, N.J.: Prentice Hall, 1972); Jerome Bruner, Jacqueline Goodnow, 10 Clark L. Hull, Principles of Behavior (New and George A. Austin, A Study of Thinking (New York: Appleton-Century-Crofts, 1942) and A York: Wiley, 1956); George A. Miller, Eugene Behavior System: An Introduction to Behavior The- Galanter, and Karl Pribram, Plans and the Struc- ory Concerning the Individual Organism (New ture of Behavior (New York: Holt, Rinehart, and Haven, Conn.: Press, 1952). Winston, 1960).

Dædalus Winter 2004 19 Jerome The contemporary linguistic assault on sounds than of foreign ones. And so it Bruner on associationist learning theory began does: French babies babble in French, learning with Noam Chomsky’s gloves-off criti- Spanish in Spanish, etc. With such cal review of Skinner’s Verbal Behavior.13 experiments, one tests in context, not in But the mentalist, problem-solving em- a maze, and knows without extrapola- phasis it introduced has now expanded tion whether the experiment has any beyond language as such. One now asks bearing on real learning by real people in whether cultural codes are learned in real life. some language-like way. Neither psy- Shall we conclude, then, that three- cholinguistics nor cultural psychologists quarters of a century of warfare between think of learning in the old-fashioned associationist and con½gurational learn- learning-theory way. ing theories taught us little or nothing I think it would be fair to say that, about the real nature of learning? That under this new dispensation, more has would be a mistake. been learned during the last three de- Both Pavlov’s dogs and Koehler’s cades about language acquisition than chimpanzees did, in fact, learn, though in any prior century–more, indeed, than in different ways and in different cir- in all of them combined. And it’s well to cumstances. And we have ample reason remember that the flood of research that to suspect that neither of their approach- made this possible was precipitated by es can be reduced to the other. In the the linguist Chomsky, not by a learning next turn of things perhaps we will ½g- theorist. ure out how to put them together. But of The turn to language, moreover, has one thing at least I am quite convinced. shifted learning-related research away You cannot strip learning of its content, from many of the older, arti½cial experi- nor study it in a ‘neutral’ context. It is mental paradigms–mazes, paired-asso- always situated, always related to some ciate word lists, nonsense syllables, and ongoing enterprise. Perhaps there is no the rest. Let me give an example: the such thing as ‘learning in general’–and prediction that children must be so early perhaps that is what we should learn tuned to the structure of their native lan- from Pavlov’s dogs, Koehler’s chimps, guage that they pick up its phonemic dis- and the disputes over learning that they tinctions in parental talk even before once symbolized. they learn to understand or talk the language proper. It is a prediction that grows out of linguistic and developmen- tal theory. And you can test it in context directly–by seeing whether childrens’ prelinguistic babbling has a higher fre- quency of native-language phoneme

13 B. F. Skinner, Verbal Behavior (Cambridge, Mass.: Press, 1947). Chom- sky’s structuralist-mentalist views were ½rst published in his Syntactic Structures (The Hague: Mouton, 1957). His frontal attack on Skinner’s Verbal Behavior (in Language 35 [1959]: 26–34) only two years later came as a rather unexpect- ed, though scarcely a conceptual, surprise.

20 Dædalus Winter 2004 Alison Gopnik

Finding our inner scientist

In 1946, the philosopher of science Karl described was the problem of causal in- Popper had a fateful meeting with the duction: How is it possible for us to cor- philosopher of language Ludwig Witt- rectly infer the causal structure of the genstein at the Cambridge Philosophy world from our limited and fragmentary Club. In a talk to the Club, with Wittgen- experience? Popper claimed that this stein in the audience, Popper described problem would one day be solved, and several “philosophical problems”– he turned out to be right. Surprisingly, important, dif½cult questions that he at least part of the solution to the prob- thought would one day be answered. lem comes from a source about as far Here Popper was issuing a direct chal- removed from the chilly Cambridge lenge to Wittgenstein, who had argued seminar room of ½fty years ago as pos- that philosophy could only analyze lin- sible–it comes from babies and young guistic puzzles–not solve any real prob- children. lems. The past thirty years have been a gold- The visit has become most famous for en age for the study of cognitive devel- the subsequent controversy among eye- opment. We’ve learned more about what witnesses over whether or not Wittgen- babies and young children know, and stein’s response to this challenge was to when they know it, than we did in the angrily brandish a ½replace poker at preceding two thousand years. And this Popper. new science has completely overturned But there is a more interesting aspect traditional ideas about what children are to the story. One of the problems Popper like. The conventional wisdom, from Locke Alison Gopnik, professor of psychology at the to Freud and Piaget, had been that ba- University of California at Berkeley, conducts bies and young children are irrational, research on the ways that children come to under- egocentric, pre-causal, and solipsistic, stand the world around them. She has written governed by sensation rather than rea- many articles and monographs and is the co- son, and rather than intention. author (with Andrew N. Meltzoff and Patricia K. In contrast, the last thirty years of re- Kuhl) of “The Scientist in the Crib: What Early search have taught us that even the Learning Tells Us About the Mind” (1999). youngest infants–literally newborns –already know a great deal about a wide range of subjects. Moreover, we have © 2004 by the American Academy of Arts & Sciences been able to chart consistent changes

Dædalus Winter 2004 21 Alison in children’s knowledge of the world as inside another box. Then she gave the Gopnik they grow older. Those changes suggest children the boxes. The children cheer- on learning that even the youngest babies are solving fully opened the ‘happy’ box but kept Popper’s problem: somehow they accu- the ‘disgusted’ box shut.2 In another rately learn about the causal structure of experiment, infants seemed to predict the world from their experience. that a hand that had reached toward an object would continue to reach toward it Consider how children come to under- even when it was placed at a new loca- stand one particularly important aspect tion–just as their own hands would. of the world–the fact that other people (They did not, however, make this same have , desires, and beliefs and prediction about a stick that had made that those mental states cause their be- contact with an object.)3 havior. All of us know that other people By two, children seem to understand have minds in spite of the fact that we that their own desires may differ from only see the movements of their physical the desires of others. And by two and a bodies. This raises another ancient phil- half, they extend this understanding to osophical question: How do we come to perception. In one study, the experi- know other minds? menter demonstrated disgust toward a In the last ½fteen years, a great deal of food that the baby liked (gold½sh crack- empirical research has begun to illumi- ers) and happiness toward a food that nate the intuitive psychology of even the the baby did not like (raw broccoli), and youngest human beings. Infants seem to then asked the baby to “give [her] be born believing that people are special some.” Fourteen-month-olds always and that there are links between their gave her the crackers, but eighteen- own internal feelings and the internal month-olds gave her the broccoli.4 In feelings of others. For example, new- another experiment, thirty-month-old borns can imitate facial expressions: children could accurately predict that when an experimenter sticks his tongue someone on one side of an opaque out at the baby, the baby will stick out screen would see a toy placed there, but her own tongue; when he opens his someone on the other side of the screen mouth, she will open hers; and so on. would not.5 In order to do this, newborns must be able to link their own internal kinesthet- 2 Betty M. Repacholi, “Infants’ Use of Atten- tional Cues to Identify the Referent of Another ic sensations, the way their mouth feels Person’s Emotional Expression,” Developmental from the inside, to the facial gestures of Psychology 34 (5) (September 1998): 1017–1025. another person–that pink thing moving back and forth in the oval in front of 3 Amanda L. Woodward, Jessica A. Som- them.1 merville, and Jose J. Guajardo, “How Infants Make Sense of Intentional Action,” in Bertram By a year, babies seem to understand F. Malle and Louis J. Moses, eds., Intentions and that mental states can be caused by ex- Intentionality: Foundations of Social Cognition ternal objects. For example, fourteen- (Cambridge, Mass.: mit Press, 2001), 149–169. month-olds saw an experimenter make a disgusted face as she looked inside one 4 Betty M. Repacholi and Alison Gopnik, “Ear- ly Reasoning About Desires: Evidence from 14- box, and a happy face when she looked and 18-Month-Olds,” 1 Andrew N. Meltzoff and Wolfgang Prinz, 33 (1) (January 1997): 12–21. eds., The Imitative Mind: Development, Evolution, and Brain Bases (New York: Cambridge Univer- 5 John H. Flavell, Barbara A. Everett, and sity Press, 2002). Karen Croft, “Young Children’s Knowledge

22 Dædalus Winter 2004 By four, children can understand that of the causal structure of mental life– Finding beliefs, as well as desires and percep- misunderstandings that lead to surpris- our inner scientist tions, may differ, and that beliefs may be ingly inaccurate but consistent predic- false. For example, you can show chil- tions and explanations. As they get old- dren this age a candy box that, much to er, the misconceptions fade away and their surprise, turns out to be full of pen- their causal knowledge becomes more cils. Three-year-olds will say that they extensive and precise. always thought that there were pencils Evidence seems to play an important in the box, and that everyone else will role in these developments. For example, think that there are pencils inside, too. younger siblings from large families, But four-year-olds understand that they who have a lot of experience with a vari- and others may falsely believe that there ety of other minds, develop this under- are candies in the box.6 standing more quickly than solitary only By six, children start to understand children.8 We can also show that giving that beliefs may be the result of interpre- young children relevant evidence can tation, and that different people may actually accelerate their developing interpret the world differently. When understanding of the mind. For exam- you give ½ve-year-olds a small glimpse of ple, we can, shades of Popper, set out a picture–a triangular fragment that to show children who do not yet under- might imply a sailboat, or a witch’s hat, stand false beliefs that their predictions or many other things–they don’t under- about another person’s actions can be stand at ½rst that people might interpret systematically falsi½ed; we can show this fragment in different ways. But by them that someone who sees the closed six or so they get this right.7 box will, in fact, say there are candies in- At each point in development children side of it. A month later, children who know some quite abstract and sophisti- saw evidence that they were wrong were cated things about how the mind works, more likely to understand how false be- knowledge that leads them to surprising- liefs really work than children who did ly accurate and wide-ranging predictions not.9 and explanations. They seem to under- stand something about how events in We can tell very similar stories about the world cause different mental states, children’s developing causal knowledge and about the way these mental states of everyday physical phenomena, like in turn cause particular human actions. gravity and movement, and everyday Yet they fail to understand other aspects biological phenomena, like illness and growth. These patterns of development about Visual Perception: Further Evidence for have led many of us to draw an analogy the Level 1–Level 2 Distinction,” Developmen- between children’s learning and the his- tal Psychology 17 (1) (January 1981): 99–110. torical development of scienti½c theo- 6 Josef Perner, Susan R. Leekam, and Heinz Wimmer, “Three-Year-Olds’ Dif½culty with 8 Jennifer M. Jenkins and Janet Wilde Asting- False Belief: The Case for a Conceptual De½- ton, “Cognitive Factors and Family Structure cit,” British Journal of Developmental Psychology Associated with Theory of Mind Development 5 (2) (June 1987): 125–137. in Young Children,” Developmental Psychology 32 (1) (January 1996): 70–83. 7 Marjorie Taylor, “Conceptual Perspective Taking: Children’s Ability to Distinguish 9 V. Slaughter and Alison Gopnik, “Conceptual What They Know from What They See,” Coherence in the Child’s Theory of Mind,” Child Development 59 (3) (June 1988): 703–718. Child Development 67 (6) (1996): 2967–2989.

Dædalus Winter 2004 23 Alison ries, an analogy I’ve called the theory same unfortunate boat, convinced that Gopnik on theory. Like scientists, children seem the scientists and children they study learning to develop a succession of related intu- are getting to the , perhaps even itive causal theories of the world, theo- suspecting that they may be using some ries that they expand, elaborate, modify, of the same learning mechanisms to and revise in the light of new evidence. get there, but unable to determine how. There is only one problem with the So both groups have mostly ended up theory theory, and it harks back to Pop- waving their hands and talking vaguely per’s talk at Cambridge. We have had al- about paradigm shifts and construc- most no idea how scientists learn about tivism. the world; when we ‘theory theorists’ Ten years ago I would have said that turned to philosophers of science to this sad state of affairs was irremediable, ½nd out about scienti½c learning mecha- at least for the immediate future. Our nisms, we got the runaround. Philoso- generation of scientists would have to phers knew that insofar as a theory was labor over the details of the empirical a deductive system, you could say some- natural history of learning and leave it to thing about how one part of the theory the next generation to develop precise should follow from another; and they and convincing explanations of learning. knew something, though much less, But, rather remarkably, age has made about how evidence could con½rm or me more optimistic. Though we are still falsify a hypothesis that had been gener- very far from having the whole story, I ated by a theory (this, of course, was think there is a new line of work that is where Popper made his contribution). actually on the right track. We are begin- But they knew almost nothing about ning to understand not only what babies what has been called the logic of dis- (and scientists) know when–but also covery–the way that experience itself how they learn it and why they get it might lead to the generation of new right. theories or hypotheses. And notoriously, they knew even less about what psychol- The general structure of the explana- ogists call conceptual changes (and what tion comes from an entirely different the rest of the world, ad nauseam, calls part of cognitive science: the study of paradigm shifts), in which the very vo- vision.10 Indeed, the study of vision has cabulary of a theory seems to change in been the most striking, though unher- the light of new evidence. Some philoso- alded, success story in cognitive sci- phers said that to answer questions ence–a case of real rather than just-so about discovery and conceptual change . Although we you would have to go talk to psycholo- don’t typically think of vision as a kind gists. Others, even more discouragingly, of learning, there is a sense in which the said the questions were simply unan- two processes are quite similar. The swerable. And if there were no accurate visual system takes a pattern of retinal learning mechanisms that underlaid sci- input and generates accurate representa- ence, if Wittgenstein was right that the tions of three-dimensional objects mov- problems of induction, discovery, and ing through space. It has to solve what conceptual change were not solvable, has been called the inverse problem: the then the whole enterprise of science was three-dimensional world produces cer- in doubt. 10 Stephen E. Palmer, Vision Science: Photons to So philosophers of science and devel- Phenomenology (Cambridge, Mass.: mit Press, opmental psychologists have been in the 1999). 24 Dædalus Winter 2004 tain patterns at the retina and the brain ing train rather than two slightly differ- Finding our inner has to work backward to accurately re- ent two-dimensional photographs. scientist create the world from that information. But the consequences of those We have a remarkably good understand- assumptions are deductive. It is not ing of the computations, and even the always true that retinal images are gen- neurological mechanisms, that are in- erated by light reflecting off the same volved in this process. three-dimensional object onto two sepa- The visual system solves the inverse rate retinas. But if it is true, then we can problem by making certain very abstract say, as a geometrical fact, that only cer- and general assumptions about how the tain kinds of images will result. In fact, three-dimensional world creates pat- of course, in real life, without the de- terns on the retina. And we can explain monic View-Master to confuse things, the way the system works by describing the assumptions of the visual system will it in terms of these assumptions, and in almost always be correct. That’s why the terms of knowledge, rules, and infer- designers of computer vision systems ences–just as we can explain how my build those assumptions into their pro- computer works in this way. For exam- grams, and presumably that’s why evo- ple, the visual system seems to assume lution built those assumptions into the that the images at the retina of each eye design of the visual cortex. are projections of the same three-dimen- In learning, as in vision, our sional objects in the world, and that the may be performing computations that discrepancies between them are the re- we can’t perform consciously. We see a sult of geometry and optics. We can three-dimensional world or know about show mathematically that, given these a causal one, without having to bother assumptions, only some three-dimen- about the implicit computations that let sional con½gurations of objects, and not us generate that world from the data. In others, will be compatible with a partic- vision science, we ½gure out which com- ular set of retinal patterns. This enables putations the brain performs by giving us to also say mathematically whether a people particular patterns of retinal data visual system (human, animal, or robot- and recording what they see. In the same ic) generates the right representations of way, we can give babies and young chil- the spatial world from a particular pat- dren patterns of statistical data and re- tern of data. In fact, the human visual cord what they learn. system seems to be about as good at get- When trained scientists do statistics, ting the right representations as it could we make certain very general assump- possibly be. tions about what the underlying causal The assumptions that allow these in- structure of the world is like, and how ferences to take place are themselves that structure leads to particular pat- contingent and sometimes may be vio- terns of data. The data we consider are lated. For example, the View-Master patterns of dependence and indepen- toys and 3-D glasses of my youth and dence among variables. Just looking at a their modern virtual reality equivalents single dependency between two vari- arti½cially create retinal images that ables may not tell us a great deal about normally would be generated by three- causal structure, just as looking at a dimensional objects, and the visual sys- small piece of a picture won’t tell us tem gets it wrong as a result. We see a much about a spatial scene. But by look- three-dimensional Taj Mahal or oncom- ing at the entire pattern of dependence

Dædalus Winter 2004 25 Alison and independence among several types can also use patterns of conditional Gopnik on of variables, we can zero in on the right probability to make genuinely causal learning causal structure, and eliminate incorrect inferences. To do this, we show children hypotheses. Sometimes we can even use a machine called the blicket detector. these patterns to add to the vocabulary The machine is a square box that lights of the theory. For instance, if we ½nd up and plays music when particular otherwise unexplained dependencies blocks are placed on top of it. The blocks between two variables, we may decide are all different from one another, so the that there is a hidden unobserved vari- job for children is to identify which able that influences them both. Recently, blocks are blickets, that is, which blocks philosophers of science, computer scien- will cause the machine to light up. We tists, and statisticians working with can present the children with quite com- what is called the Bayes net formalism plex patterns of contingency between have begun to provide a precise mathe- the activation of the detector and vari- matical account of these kinds of infer- ous combinations of blocks. We can ask ences (see Clark Glymour’s essay in this them which blocks are blickets, and we issue). can ask them to activate the machine or It turns out that even very young get it to stop. And their answers are babies, as young as eight months old, almost always correct. They make the are sensitive to patterns of dependency. right inferences about the causal powers We can play babies strings of syllables in of the blocks. They make the sort of sta- various probabilistic combinations with tistical inferences a scientist would make particular patterns of dependency–for and, according to the Bayes net formal- example, ‘ba’ may usually precede ‘da,’ ism, should make. In similar experi- but rarely precede ‘ga.’ The babies can ments, we can even show that children use these patterns of probabilities to postulate unobserved variables to deal infer which combinations of syllables with otherwise inexplicable patterns of are likely to occur together, and they data.12 can also detect similar statistical pat- terns among musical tones or aspects of In order to make inferences about the a visual scene. Babies also seem able to causal structure of the world and causal map those probabilities onto representa- relations among variables, the scientist tions of the external world. They don’t, performs experiments. The scientist in- for example, just notice that certain syl- tentionally intervenes on a variable in lables tend to go together; they assume the world, forcing it to have a particular that these regularities occur because value and then observing what happens these combinations of syllables consti- to the values of other variables. Again tute words in the language they hear Bayes nets provide a precise mathemati- around them. In the example above, cal account of such inferences. they would assume that ‘bada’ is more In a similar way, even the youngest likely to be a word than ‘baga.’11 babies are particularly sensitive to the We have shown that, at least by the consequences of their interventions on time they are two and a half, children 12 Alison Gopnik, Clark Glymour, David Sobel, 11 Richard N. Aslin, Jenny R. Saffran, and Elis- Laura Schulz, Tamar Kushnir, and David sa L. Newport, “Computation of Conditional Danks, “A Theory of Causal Learning in Chil- Probability Statistics by 8-Month-Old Infants,” dren: Causal Maps and Bayes-Nets,” Psychologi- Psychological Science 9 (4) (July 1998): 321–324. cal Review 111 (1) (2004): 1–30.

26 Dædalus Winter 2004 the world. For example, with a ribbon set of experimental interventions with Finding our inner we can attach a mobile to a three- gear and switch to determine which gear scientist month-old baby’s leg; the baby will re- moves the other. gard her influence over the mobile with Of course these observations will not fascination, systematically exploring the surprise anyone who has spent much contingencies between various limb time with infants or young children, movements and the movements of the who are perpetually ‘getting into things.’ mobile.13 By the time they are a year old, In this sense, we may think of toddlers babies will systematically vary the kinds as causal learning machines. They are of actions they perform on objects, as small human versions of the Mars rovers they simultaneously observe the conse- that roam about getting into things on quences of those actions. And they may the red planet–except that children are watch the further consequences of the also mission control, interpreting the action ‘downstream’ and use that infor- data they collect. mation to design new actions. Give a Somewhere between statistical obser- one-year-old a set of blocks and you can vation and active experimentation, sci- see her trying different combinations, entists and babies alike learn from the placements, and angles, and gauging interventions of others. Scientists read which of these will produce stable tow- journals, go to talks, hold lab meetings, ers and which will end in equally satisfy- and visit other labs–and all those con- ing crashes. ferences surely have some function be- We have shown that by the time chil- yond assortative mating. We scientists dren are four they will intervene in the make the assumption that the interven- world in a way that lets them uncover tions of others are like our own inter- causal structure. My student Laura ventions, and that we can learn similar Schulz’s gear toy tests show how chil- things from both sources. dren learn about causal structure. This By at least nine months, human in- toy, like the blicket detector, presents fants seem to make the same assump- children with a new causal relation that tion. For example, in one study babies they must infer from evidence about see an experimenter enter the room and contingencies. It is a square box with touch the top of his head to a box that two gears on top and a switch on the then lights up. A day later, babies return side. When you flip the switch the gears to the room, see the box, and then im- turn simultaneously. If you remove gear mediately touch their heads against the A and then flip the switch, B turns by top of it.14 itself; if you remove gear B and flip the We have shown that by four, children switch, A doesn’t turn. With both of can use information about the interven- these pieces of evidence you can con- tions of others appropriately to make clude that B is making A move. We tell new causal inferences. Consider the gear the children that one of the gears makes toy experiment described above. Chil- the other one move, and then leave them dren will also solve this task if they sim- alone with the toy and a hidden camera. ply see an adult perform the right experi- The children swiftly produce the right ments on the toy. They not only learn 13 Carolyn Rovee-Collier and Rachel Barr, “Infant Learning and Memory,” in Gavin 14 Andrew N. Meltzoff, “Infant Imitation and Bremner and Alan Fogel, eds., Blackwell Hand- Memory: Nine-Month-Olds in Immediate and book of Infant Development (Malden, Mass.: Deferred Tests,” Child Development 59 (1) Blackwell Publishers, 2001), 139–168. (1988): 217–225.

Dædalus Winter 2004 27 Alison about the causal consequences of adult For example, children tend to assume Gopnik on actions, but also about the causal rela- that the samples of evidence they collect learning tions among the objects upon which are representative of the data. Similarly, perform those actions. they seem to assume that their own ac- Indeed, the three techniques of causal tions and the actions of others have all inference that I have described–analyz- the formal characteristics of an ideal ing statistics, performing experiments, experimental intervention. The self- and watching the experiments of oth- conscious methodological canons of ers–may give both scientists and chil- formal science–the courses on statistics dren their extraordinary learning pow- and experimental design–are intended ers. Elements of the ½rst two techniques to make these assumptions explicit rath- are probably in place even in nonhuman er than implicit and so ensure that they animals. In classical conditioning, ani- are correct in particular cases. For chil- mals calculate dependencies among par- dren, however, the assumptions may be ticularly important events, like shock close enough to the truth most of the and food. In operant conditioning, ani- time, and the evidence may be suf½cient- mals calculate the consequences of their ly rich, so that they mostly get things actions. This is not surprising given the right anyway. importance of causal knowledge for sur- If we want children, and lay adults, to vival. understand and appreciate science, we However, as Mike Tomasello and Dan- may need to make more connections ny Povinelli point out in this issue, there between their intuitive and implicit is much less clear evidence of the third causal inference methods and the self- type of learning–learning from the ac- conscious and explicit use of these meth- tions of others–in other animals. And ods in science. We may need, literally, a there is no evidence that other animals sort of scienti½c consciousness-raising. combine all three types and assume that Popper’s quarrel with Wittgenstein re- they provide information about the flected a larger argument between the causal structure of the external world. By view that science and philosophy tell us contrast, human children, at least by age new things about the world, and the three or four, do seem to put these types view that all they do is reflect social ar- of information together in this way. This rangements and linguistic conventions. ability may, in fact, be one of the crucial If we could put children in touch with abilities that give human beings their their inner scientists, we might be able unique intellectual capacities. It allows to bridge the divide between everyday them to learn far more about the world knowledge and the apparently intimi- around them than other animals, and to dating and elite apparatus of formal use that knowledge to change the world. science. We might be able to convince them that there is a deep link between My guess is that many of the mis- the realism of everyday life and scienti½c takes that children and adults make in realism. And if we were able to do that, learning don’t happen because they then we might win Popper’s argument make the wrong deductions from as- for him–without having to resort to sumptions and evidence, but rather pokers. because they make assumptions that are unwarranted under the particular cir- cumstances.

28 Dædalus Winter 2004 Daniel John Povinelli

Behind the ape’s appearance: escaping anthropocentrism in the study of other minds

Look at Megan. Not just at her distinc- dedicated his life to studying these tively chimpanzee features–her accen- remarkable animals–entertain the pos- tuated brow ridge, her prognathic face, sibility that their minds are, in profound her coarse black hair–but at the totality respects, radically different from our of her being: her darting eyes, her slow, own? How can I challenge the received studied movements, the gestures she wisdom of Darwin–con½rmed by my makes as her companion, Jadine, passes own initial impressions–that the mental nearby. Can there be any doubt that be- life of a chimpanzee is best compared to hind certain obvious differences in her that of a human child? appearance resides a mind nearly identi- Actually, it’s easy: I have learned to cal to our own? Indeed, is it even possi- have more respect for them than that. ble to spend an afternoon with her and I have come to see that we distort their not come to this conclusion? Upon re- true nature by conceiving of their minds flection, you will probably acknowledge as smaller, duller, less talkative versions that her mind is not identical to ours. of our own. Casting aside these insidious “But surely it’s not qualitatively differ- assumptions has been dif½cult, but it has ent, either,” you will still insist. “I mean, allowed me to see more clearly that the it’s obvious from watching her that we human mind is not the gold standard share the same kind of mind.” against which other minds must be Faced with the overwhelming similari- judged. For me it has also illuminated ty in the spontaneous, everyday behavior the possibility of creating a science that of humans and chimpanzees, how can is less contaminated by our deeply an- someone like me–someone who has thropocentric intuitions about the na- ture of other minds. Daniel John Povinelli is Louisiana Board of Re- gents Endowed Professor of Science at the Univer- The best available estimates suggest sity of Louisiana at Lafayette, and director of the that humans and chimpanzees originat- Cognitive Evolution Group and the Center for ed from a common ancestor about ½ve Child Studies. His latest book is “Folk Physics for or six million years ago.1 This is reflected Apes: The Chimpanzee’s Theory of How the 1 Galina V. Galzko and Masatoshi Nei, “Esti- World Works” (2000). mation of Divergence Times for Major Lineages of Primate Species,” Molecular Biology and Evolu- © 2004 by the American Academy of Arts tion 20 (2003): 424–434. & Sciences

Dædalus Winter 2004 29 Daniel John in estimates of our genetic similarity: we that our minds were 75 percent chim- Povinelli share, on average, about 98.6 percent of panzee? on learning our total nucleotide sequence in com- No, such coarse genetic comparisons mon. This statistic seems impressive. will hardly suf½ce to help us understand After all, such biological af½nity would the complex similarities and differences appear to be the ½nal nail in the cof½n that exist between the mental lives of of the notion that there could be any humans and chimpanzees. However, radical mental differences between in a climate where certain highly visible them and us: if chimpanzees and experts have radically anthropomor- humans share 98.6 percent of their phized chimpanzees,3 such statistics are genetic material, then doesn’t it follow heralded as establishing once and for all that there ought to be an extraordinarily that chimpanzees are, at the very least, high degree of mental similarity as well? mentally equivalent to two- or three- This idea has been paraded so frequently year-old human children, and should through the introductory paragraphs of therefore be granted human rights.4 both scholarly journal articles and the A few obvious biological facts may be popular press alike that it has come to worth noting here. To begin, it was the constitute a melody of sorts; an anthem human lineage, not the chimpanzee one, that if not sung raises doubts as to one’s that underwent radical changes after our allegiance to the cause of defending the respective geneologies began to diverge chimpanzee’s dignity. from their common ancestor. Since this But what does this 98.6 percent statis- split, humans have resculpted their bod- tic really mean? It should be of immedi- ies from head to toe–quite literally, in ate interest that it is almost invariably fact; as our lineage became bipedal, misreported. We do not share 98.6 per- the pelvis, the knee, and the foot were cent of our genes in common with chim- all drastically reshaped, with modi½ca- panzees; we share 98.6 percent of our tions in the hand (including new mus- nucleotide sequence. A single nucle- cles) soon following. To top it all off, otide difference in a string of four hun- we ultimately tripled the size of our dred may code for a different allele. brain, with disproportionate increases Furthermore, as the geneticist Jonathan probably occurring in the seat of higher Marks has pointed out in lucid detail, cognitive function, the prefrontal cortex. the 98.6 percent statistic has so little Oh yes, and at some point during all of grounding in the average mind that con- this (no one knows exactly when), natu- fronts it, as to render it essentially mean- ral language–perhaps the most notice- ingless.2 We might, after all, share 50 percent of our nucleotide sequences in 3 For examples, see Sue Savage-Rumbaugh, common with bananas and broccoli. Kanzi: The Ape at the Brink of the Human Mind But what on earth does it mean to say (New York: John Wiley & Sons, 1994); Jane Goodall, Through a Window (Boston: Houghton that we are 50 percent the same as a Mifflin, 1990); Roger Fouts, Next of Kin (New vegetable? I don’t know about you, but York: William Morrow and Co., 1997). I doubt my mind is 50 percent identical to that of the garden pea. And so what 4 Steven M. Wise, Rattling the Cage: Toward Le- would it mean, exactly, if we discovered gal Rights for Animals (Cambridge, Mass.: Per- seus Books, 2000); Paola Cavalieri and Peter 2 Jonathan Marks, What It Means to Be 98% Singer, eds., The Great Ape Project: Equality Be- Chimpanzee (Berkeley: University of California yond Humanity (New York: St. Martin’s Press, Press, 2002). 1993).

30 Dædalus Winter 2004 able of human adaptations–emerged as brains–of humans and chimpanzees. Behind well. So, exactly how similar are the brains the ape’s appearance In contrast, chimpanzees have proba- of humans and chimpanzees? After all, bly changed relatively little from the if we knew that, couldn’t we directly ad- common ancestor they shared with us dress the question of their mental sim- about ½ve million years ago. Indeed, ilarity? Well, it would be a start, any- of all of the members of the great ape/ how. Unfortunately, comparisons of the human group who shared a common brains of humans and apes have tradi- ancestor about ½fteen million years ago, tionally been limited to gross considera- none, indeed, has diverged as much as tions such as size and surface features humans. A simple thought experiment (such as lobes and sulcus patterns). may help to put this point into perspec- Remarkably, the details of the internal tive: line up all of the species in ques- organization of human and great ape tion–gorillas, orangutans, chimpan- brain systems and structures have been zees, bonobos, humans–and one of largely ignored, in part because it’s so them immediately stands out. Guess dif½cult to study these brains, but also which one? because most have fre- In fact, the more we compare humans quently assumed that despite great dif- and chimpanzees, the more the differ- ferences in size, all mammalian brains ences are becoming apparent. Even ge- are organized pretty much the same. neticists are starting to catch up with Fortunately, even this is beginning the reality of these differences. New re- to change. For example, Todd Preuss, search has shown that rough similarity working at the University of Louisiana, in our nucleotide sequences obscures recently made a startling discovery the fact that the same genes may have while comparing the brains of humans dramatically different activity levels in and chimpanzees. Turning his attention the two species. So even where humans away from the frontal lobes, his previous and chimpanzees share genes in com- area of research, Preuss decided to take mon, it turns out that there are what can a look at the primary visual cortex (V1), only be described as major differences in the area of the cerebral cortex that is the gene expression–that is, whether, when, ½rst way station into the processing of and for how long genes are actually visual information. The organization of working to produce the proteins for this area of the brain has been assumed which they code.5 This is the real stuff to be nearly identical across primates. of genetic comparison, and it casts our But there, in the middle of V1, Preuss crude genetic similarity to the garden and his colleagues uncovered a distinc- pea in a wholly different light. tively human specialization–a kind of What makes these differences in gene neural architecture not found even in expression signi½cant is that they ulti- chimpanzees.6 Preuss speculates that mately manifest themselves as differ- this specialization involves modi½ca- ences in the bodies–including the tions of the pathways related to spatial vision and motion processing. But, re- 5 Wolfgang Enard et al., “Intra- and Inter- speci½c Variation in Primate Gene Expression gardless of what it is for, it suggests that Patterns,” Science 296 (2002): 341–343; Mario Cáceres et al., “Elevated Gene Expression Lev- 6 Todd M. Preuss et al., “Distinctive Compart- els Distinguish Human from Non-Human Pri- mental Organization of Human Primary Visual mate Brains,” Proceedings of the National Acade- Cortex,” Proceedings of the National Academy of my of Sciences 100 (2003): 13030–13035. Sciences 96 (1999): 11601–11606.

Dædalus Winter 2004 31 Daniel John we need to rethink brain evolution in a opposed to, “How does thinking differ Povinelli on way that’s consistent with neo-Darwin- across species?” (the latter being a learning ian theory: similarity and difference decidedly more evolutionarily minded among species as comfortable bedfel- question). lows; a state of affairs accomplished by Assuming that chimpanzees and other weaving in new systems and structures species have mental states (a point I take alongside the old. “If we ½nd such differ- for granted), it seems to me that a more ences in the middle of the primary visual productive question to ask is, “What are cortex,” Preuss recently remarked to me, their mental states about?” Or, put an- “just imagine what we’re going to ½nd other way, “What kinds of concepts do when we start looking elsewhere.” they have at their disposal?” It would Some may be surprised (or even stand to reason that the mental states of afraid) to learn of such differences be- chimpanzees, ½rst and foremost, must tween humans and our nearest living rel- be concerned with the things most rele- atives. After several decades of being fed vant to their natural ecology–remem- a diet heavy on exaggerated claims of the bering the location of fruit trees, keeping degree of mental continuity between hu- an eye out for predators, and keeping mans and apes, many scientists and lay- track of the alpha male, for instance. persons alike now ½nd it dif½cult to con- And so surely chimpanzees form con- front the existence of radical differences. cepts about concrete things–things like But then, in retrospect, how viable was trees, facial expressions, threat vocaliza- the idea of seamless mental continuity in tions, leopards, and the like. But what the ½rst place? After all, it tended to por- about more abstract concepts? Con- tray chimpanzees as watered-down hu- cepts like ghosts, gravity, and God? mans, not-quite-½nished children. De- Admittedly, to use the term ‘concept’ spite the fact that aspects of this notion as loosely as I have will require the indul- can be traced straight to Darwin, it is an gence of certain scholars. But perhaps evolutionarily dubious proposition, to some progress can be made by noting say the least. that every concept is at least somewhat abstract if it extends beyond a particular If there are substantial differences be- example. For instance, if one has a no- tween the mental abilities of humans tion of an apple that is not limited to a and chimpanzees, in what areas are they single instance of that apple, then one likely to exist? Over the past couple of has made a generalization, and thus a thousand years, many potential rubicons kind of abstraction. Given that it has separating human and animal thinking been known for decades or more that have been proposed. Some of these have chimpanzees and many other species been particularly unhelpful, such as the form such abstractions,8 this cannot be radical behaviorists’ forgettable proposi- a de½ning feature of human thinking. tion that animals don’t ‘think’ at all (of course, these behaviorists were even 8 Suzette L. Astley and Edward A. Wasserman, skeptical about the existence of human “Object Concepts: Behavioral Research with Animals and Young Children,” in William T. thought!). And, unfortunately, in the O’Donohue, ed., Learning and Behavior Therapy popular imagination the question still (Boston: Allyn and Bacon, 1997), 440–463; appears to be, “Can animals think?”7 as Tom R. Zentall, “The Case for a Cognitive Approach to Animal Learning and Behavior,” 7 Eugene Linden, “Can Animals Think?” Time, Behavioural Processes 54 (2001): 65–78. 22 March 1993.

32 Dædalus Winter 2004 At the risk of oversimpli½cation, let about observable things; but, at the Behind the ape’s me instead propose a distinction be- same time, are radically different, in that appearance tween concepts that refer to objects humans form additional concepts about and events that can be directly observed inherently unobservable things.9 (that is, things that can be detected by Now, I realize that most people would the unaided ), versus hypothetical not be surprised if it were established entities and processes (things that are beyond doubt that chimpanzees lack a classically unobservable). Thus, I wish concept of God. But what about other, to separately consider all concepts that seemingly more prosaic concepts that refer to theoretical things: all the things infest our way of thinking about the that are not directly registered by the world? Consider the way in which we senses, but are merely posited to exist think about the social realm. In interact- on the basis of things we can observe. ing with each other (and with animals, Such concepts permeate our common- for that matter), we use a dual system sense way of thinking: we explain phys- of representation: we understand other ical events on the basis of things like beings both as part of the observable ‘forces’ (supernatural or otherwise) world (they engage in particular move- that we have never actually witnessed, ments of their hands and feet, and and account for the behavior of other their lips form particular contortions as humans on the basis of mental states sounds emerge from their mouths), and we have never seen (e.g., their beliefs, as entities with mental properties–un- desires, and emotions). These concepts observable attributes like emotions, in- serve as the bedrock for some of our tentions, desires, and beliefs. most fundamental explanations for why The proposal is that, in contrast to the world works the way it does. humans, chimpanzees rely strictly upon Meanwhile, we can directly contrast observable features of others to forge these sorts of concepts with ones that their social concepts. If correct, it would are derived from things that can be di- mean that chimpanzees do not realize rectly observed: apples and oranges, that there is more to others than their trees, flashes of lightning, facial expres- movements, facial expressions, and hab- sions–even the raising of a hand or the its of behavior. They would not under- sound of a train whistle blowing in the stand that other beings are repositories distance. Concepts about these things of private, internal experience. They share at least one property in common: would not appreciate that in addition they are all derived from the world of to things that go on in the observable macroscopic entities with which the pri- world, there are forever hidden things mary senses directly interact. Without that go on in the private life of the mind. additional justi½cation, I am therefore It would mean that chimpanzees do not asserting a distinction between con- reason about what others think, believe, cepts that refer to observable objects and feel–precisely because they do not and events, and ones that refer to strict- form such concepts in the ½rst place. ly hypothetical ones. 9 This discussion extends several previous o, here’s a proposal: the mental lives descriptions of this hypothesis, for example, S my article with Jesse Bering and Steve Giam- of humans and chimpanzees are similar, brone, “Toward a Science of Other Minds: in that both species form innumerable Escaping the Argument by Analogy,” Cognitive (and in many cases, identical) concepts Science 24 (2000): 509–541.

Dædalus Winter 2004 33 Daniel John Before we get too much further, let of ‘seeing’ (unobservable feature). So, Povinelli on me be honest: I recognize that this pro- the proposal isn’t that chimpanzees use learning posal has troubling implications. For one system and humans use another; one thing, if chimpanzees do not reason both species are purported to rely upon about unobservable entities, then we concepts about the observable proper- would frequently need distinctly differ- ties of others. Instead, the proposal is ent explanations for human and chim- that chimpanzees don’t form additional panzee behavior–even in situations concepts about the unobservable proper- where the behavior looks almost identi- ties of other beings (or the world in gen- cal. Mind you, we would not need com- eral, for that matter). pletely different explanations, just ones So, at face value, the proposal I have that are distinctive enough to capture made is worrying. In interpreting what the proposed difference. Nonetheless, would appear to be the exact same be- each time we witnessed a chimpanzee haviors in humans and chimpanzees in engage in a complex social behavior that different ways, I seem to be applying a resembles our own, we would have to double standard. believe that, unlike us, the chimpanzee But is this implication really problem- has only one conceptual system for en- atic, or does it just seem problematic coding and reasoning about what is hap- because it runs counter to some of our pening: a system that invokes concepts most deeply engrained–but fundamen- derived from observable features of the tally flawed–ways of thinking? world. Thus, when chimpanzees deceive each other (which they do regularly), Assume, for a moment, that you have they would never be trying to manipu- traveled back in time to a point when late what others believe, nor what oth- there were no chimpanzees on this plan- ers can see or hear, for constructs like et–and no humans, either. Imagine fur- ‘believing,’ ‘seeing,’ and ‘hearing’ are ther that you have come face to face with already deeply psychological. No, in members of the last common ancestor deciding what to do, the chimpanzee of humans and chimpanzees. Let’s stipu- would be thinking and reasoning solely late that these organisms are intelligent, about the abstracted statistical regulari- thinking creatures who deftly attend to ties that exist among certain events and and learn about the regularities that un- the behaviors, postures, and head move- fold in the world around them. But let ments (for example) of others–what we us also stipulate that they do not reason have called ‘behavioral abstractions.’10 about unobservable things; they have I should note that humans, too, rely no ideas about the ‘mind,’ no notion of heavily upon behavioral abstractions in ‘causation.’ their day-to-day interactions. We must As you return to your time machine be doing so: otherwise upon what basis and speed forward, you will observe new could we attribute additional, psycho- lineages spring to life from this common logical states to others? First, we recog- ancestor. Numerous ape-like species will nize the turn of the head and the direc- emerge, then disappear. As you approach tion of the eyes (observable features), the present day, you will even witness then we ascribe the internal experience the evolutionary birth of modern orang- 10 Daniel J. Povinelli and Jennifer Vonk, utans, chimpanzees, and gorillas. But “Chimpanzee Minds: Suspiciously Human?” amid all of this your attention will be Trends in Cognitive Science 7 (2003): 157–160. drawn to one particular offshoot of this

34 Dædalus Winter 2004 process, a peculiar genealogy that buds Now, is it really troubling to invoke a Behind the ape’s off numerous descendent species. This different explanation for what on the appearance particular lineage has evolved an eye- surface seem to be identical units of catching trick: it habitually stands up- behavior in humans and chimpanzees? right; it walks bipedally. And some of If the scenario I have outlined above is its descendants build upon this trick, correct, then the answer must be, no. capitalizing upon the new opportunities After all, for any given ability that hu- it offers. For reasons that we may never mans and chimpanzees share in com- fully know, tool use and manufacture in- mon, the two species would share a crease exponentially, language emerges, common set of psychological structures, brain size triples, and, as more time which, at the same time, humans would passes, human material and social cul- augment by relying upon a system or ture begins to accrete upon the shoul- systems unique to our species. The resid- ders of the lineage’s last surviving mem- ual effect of this would manifest itself in ber: Homo sapiens sapiens. Now, imagine numerous ways: some subtle (such as that as part of this process, this lineage tightly constrained changes in the de- evolved new conceptual structures (in- tails of things to which our visual sys- timately connected to the evolution of tems attend), others more profound language) that allow them to reason (such as the creation of cultural artifacts about things that cannot be observed: like the issue of Dædalus in which you mental states, physical forces, spiritual are now reading these words). deities. I have stipulated all of this so we can So much for theory. What about the confront the following question: If evo- empirical evidence; does it support the lution proceeded in this quite plausible proposal I have just offered? Although it manner, then how would we expect the will not surprise you to learn that I think spontaneous, everyday behavior of hu- it does, I have not always been of this mans to compare to that of chimpan- opinion; I used to believe that any dif- zees? The answer, I think, is that things ferences between humans and chim- would look pretty much the way they do panzees would have to be trivial. But now. After all, humans would not have the results of over two hundred studies abandoned the important, ancestral psy- that we have conducted during the past chological structures for keeping track ½fteen years have slowly changed my of other individuals within their groups, mind. Combined with ½ndings from nor jettisoned their systems for noticing other laboratories, this evidence has that something very different happens forced me to seriously confront the pos- when Joe turns his head toward so-and- sibility that chimpanzees do not reason so, just depending on whether or not his about inherently unobservable phenom- hair is standing on end. No, in evolving a ena. new psychological system for reasoning Let me briefly illustrate this evidence about hypothetical, internal mental with three simple examples: one from states, humans would not have (indeed, the social domain, one from the domain could not have!) abandoned the ancient of physics, and one from the domain of systems for reasoning about observable numerical reasoning. behavior. The new system by de½nition First, what does the experimental evi- would depend upon the presence of old- dence suggest about whether chimpan- er ones. zees reason about mental states? Al-

Dædalus Winter 2004 35 Daniel John though the opinions of experts differ would approach a familiar playmate or Povinelli on (and have swung back and forth over the caretaker to request a food treat using learning past several years), I believe that at pres- their species-typical begging gesture. ent there is no direct evidence that chim- Simple enough. But on the crucial test panzees conceive of mental states, and trials, the chimpanzees were confronted considerable evidence that they do not. with two individuals, only one of whom As an example, consider the well-studied could see them. For example, in one question of whether chimpanzees rea- condition, one caretaker had a blindfold son about the internal, visual experi- covering her mouth, whereas the other ences of others, that is, of whether had a blindfold covering her eyes. The they know anything about ‘seeing.’ question was to whom would the chim- To begin, no one doubts that chim- panzee gesture. panzees respond to, reason about, and Not surprisingly, in our trials with hu- form concepts related to the movements man children, even two-year-olds ges- of the head, face, and eyes of others; tured to whoever had the blindfold over these are aspects of behavior that can be her mouth (versus the eyes), probably readily witnessed.11 But what about the because they could represent her inner, idea that another being ‘sees’ things, psychological state (“She can see me!”). that others are loci of unobservable, In striking contrast, our chimpanzees visual experiences? did nothing of the kind. Indeed, in nu- Over the past ten years we have con- merous studies, our chimpanzees gave ducted dozens of studies of juvenile, ad- virtually no indication that they could olescent, and adult chimpanzees to ex- understand ‘seeing’ as an internal expe- plore this question. Perhaps the most rience of others. straightforward of these studies in- With enough trials of any given condi- volved examining how chimpanzees tion the chimpanzees were able to learn understand circumstances under which to select whoever was able to see them; others obviously can or cannot see after enough trials of not being handed a them.12 In these studies, chimpanzees banana when gesturing to someone with were exposed to a routine in which they a bucket over her head, the chimpanzees ½gured out to gesture to the other per- 11 See Daniel J. Povinelli and Timothy J. Eddy, son. Did this mean that they had ½nally “Chimpanzees: Joint Visual Attention,” Psycho- discerned what we were asking them? In logical Science 7 (1996): 129–135; Shoji Itakura, “An Exploratory Study of Gaze-Monitoring in numerous transfer tests in which we pit- Nonhuman Primates,” Japanese Psychological Re- ted the idea that the chimpanzees were search 38 (1996): 174–180; Michael Tomasello, learning about the observable cues (i.e., Brian Hare, and Josep Call, “Five Primate Spe- frontal posture, presence of the face or cies Follow the Visual Gaze of Conspeci½cs,” eyes) against the possibility that on the Animal Behaviour 58 (1998): 769–777. basis of such cues they were reasoning 12 Our laboratory’s empirical research of about who could ‘see’ them, the chim- chimpanzees’ understanding of ‘seeing’ has panzees consistently insisted (through been summarized in my article, “The Minds their behavior) that they were reasoning of Humans and Apes are Different Outcomes about observable features, not internal of an Evolutionary Experiment,” in Susan M. mental states, to guide their choices. Fitzpatrick and John T. Bruer, eds., Carving Our Destiny: Scienti½c Research Faces a New Millenni- In addition to what they learned in um (Washington, D.C.: National Academy of these tests, it also became apparent that Sciences and John Henry Press, 2001), 1–40. chimpanzees come pre-prepared, as it

36 Dædalus Winter 2004 were, to make sense of certain postures. Of course, some have challenged this Behind the ape’s For instance, in our tests they immedi- conclusion, arguing that we need to turn appearance ately knew what to do when confronted up the microscope and develop more with someone facing them versus some- tests that will allow chimpanzees to ex- one facing away, and this ½nding has press their less well-developed under- been replicated in several other labora- standing of such concepts.14 So, for ex- tories.13 “But if they make that distinc- ample, researchers at Emory University tion,” you wonder, “then why do they recently conducted tests in which a perform so differently on the other dominant and a subordinate chimpan- tests? Is it just because they’re con- zee were allowed to ½ght over food that fused? How are we to make sense of was positioned in an enclosure between such a puzzling pattern of ½ndings?” them.15 On the critical trials, two pieces Actually, these results are not puzzling of food were positioned equidistant at all if the ability to reason about men- from the animals. The catch was that tal states evolved in the manner that I one piece of food was placed behind an suggested earlier–that is, if humans opaque barrier so that only the subordi- wove a system for reasoning about men- nate could see it. The researchers report tal states into an existing system for rea- that when the subordinate was released soning about behavior. After all, if the into the enclosure, he or she tended to idea is correct, then chimpanzees may head for the food that was hidden from well be born predisposed to attend to the dominant’s view, suggesting, per- certain postures and behaviors related haps, that the subordinate was modeling to ‘seeing’–even though they know the visual experience of his or her domi- nothing at all about such mental states nant rival. per se–precisely because overt features But do such tests really help?16 Do of behavior are the tell-tale indicators of they reveal some weaker understanding the future behavior of others. But when such features are carefully teased apart 14 Michael Tomasello et al., “Chimpanzees Un- derstand Psychological States–The Question to probe for the presence of a mentalistic Is Which Ones and to What Extent,” Trends in construal of others, the chimpanzees Cognitive Science 7 (2003): 153–156, esp. 156. stare back blankly: this is not part of their biological endowment. Thus, if the 15 Brian Hare et al., “Chimpanzees Know evolutionary framework I have sketched What Conspeci½cs Do and Do Not See,” Ani- is correct, neither the chimpanzees nor mal Behaviour 59 (2000): 771–785; see also M. Rosalyn Karin-D’Arcy and Daniel J. Povinelli, the results are ‘confused’; that epithet “Do Chimpanzees Know What Each Other may fall squarely upon the shoulders of See? A Closer Look,” International Journal of we human experimenters and theorists 15 (2002): 21–54. who are so blinded by our own way of understanding the world that we are not 16 In a recent analysis of the diagnostic poten- tial of these and other tests, Jennifer Vonk and readily open to the chimpanzee’s way of I (see footnote 10) argued that the logic of cur- viewing things. rent tests with chimpanzees (and other ani- mals) cannot, in principle, provide evidence 13 For example, see Autumn B. Hostetter et al., that uniquely supports the notion that they are “Differential Use of Vocal and Gestural Com- reasoning about mental states (as opposed to munication by Chimpanzees (Pan troglodytes) in behavior alone), and we advocated a new par- Response to the Attentional Status of a Human adigm of tests that may have such diagnostic (Homo sapiens),” Journal of Comparative Psycholo- power. An alternative point of view is provided gy 115 (2001): 337–343. in the companion piece by Tomasello and col-

Dædalus Winter 2004 37 Daniel John of mental states in chimpanzees? These Table 1 Povinelli Theoretical causal constructs and their observ- on are precisely the situations in which learning chimpanzees will be evolutionarily able ‘ambassadors’ primed to use their abilities to form Theoretical Paired observable concepts about the actions of others to concept ‘ambassador’ guide their social behavior. So, for exam- ple, they can simply know to avoid food gravity downward object that is out in the open when a dominant trajectories animal is about to be released. “But transfer of force motion-contact-motion still,” the skeptic within you asks, sequences “that’s pretty smart, isn’t it? The chim- strength propensity for panzees would have to be paying atten- deformation tion to who’s behind the door, and what shape perceptual form that other individual is going to do when the door opens, right?” physical connection degree of contact Fair enough. But that, in the end, is weight muscle/tendon stretch the point: chimpanzees can be intelli- sensations gent, thinking creatures even if they do not possess a system for reasoning about psychological states like ‘seeing.’ A second example of the operation of If it turns out that this is a uniquely hu- what may be a uniquely human capacity man system, this should not detract to reason about unobservables comes from our sense of the evolved intelli- from comparisons of humans’ and gence of apes. By way of analogy, the chimpanzees’ commonsense under- fact that bats echolocate but humans standing of physics. Humans–even don’t, hardly constitutes an intellectual very young children–seem disposed to or evolutionary crisis. assume that there’s more to the physi- In the ½nal analysis, the best theory cal world than what meets the eye. For will be the one that explains both data example, when one ball collides with sets: the fact that chimpanzees reason another, stationary one, and the second about all the observable features of oth- speeds away, even quite young children ers that are associated with ‘seeing’– are insistent that the ½rst one caused and yet at the same time exhibit a strik- the second to move away. Indeed, as ing lack of knowledge when those fea- Michotte’s classic experiments revealed, tures are juxtaposed in a manner that this seems to be an automatic mental they have never witnessed before (i.e., process in adult humans.17 But what is blindfolds over eyes versus over the it, exactly, that humans believe causes the mouth). I submit that, at least for the movement of the second ball? As Hume time being, the evolutionary hypothesis noted long ago, they do not merely rec- I have described best meets this criteri- ognize that the objects touched; that’s on. just a re-description of the observed events.18 No, the ½rst one is seen as hav-

17 Albert Michotte, The Perception of Causality leagues. However, I believe that this view dra- (New York: Basic Books, 1963). matically underestimates the representational power of a psychological system that forms 18 , Treatise of Human Nature, vols. concepts solely about the observable aspects 1–2, ed. A. D. Lindsay (London: Dent, 1739; of behavior. 1911).

38 Dædalus Winter 2004 ing transmitted something to the second question was whether, when confronted Behind the ape’s object, some kind of ‘force.’ But where is with two new options, they would select appearance this force? Can it be seen? No, it is a the- the one involving genuine physical con- oretical thing. nection as opposed to mere ‘contact.’ In an initial ½ve-year study of ‘chim- Consistent with our ½ndings in other panzee physics,’ we focused our apes’ tests, they did not. Instead, ‘perceptual attention on simple tool-using prob- contact’ seemed to be their operating lems.19 Given their natural expertise concept. The observable property of with tools, our goal was to teach them contact (of any type) was generally suf- how to solve simple problems–tasks in- ½cient for them to think that a tool could volving pulling, pushing, poking, etc.– move another object. and then to use carefully designed trans- fer tests to assess their understanding of Finally, consider the chimpanzee’s nu- why the tool objects produced the effects merical understanding. Over the past they did. In this way, we attempted to decade or so, it has become apparent determine if they reason about things that many species share what Stanislas like gravity, transfer of force, weight, Dehaene has called a ‘number sense’– and physical connection, or merely form the ability to distinguish between larger concepts about spatio-temporal regular- and smaller quantities, even when the ities. To do so, we contrasted such con- quantities being compared occupy iden- cepts with their perceptual ‘ambassa- tical volumes.20 dors’ (see table 1), much in the same way In an attempt to explore the question that we had contrasted the unobservable of numerical reasoning in animals, sev- psychological state of ‘seeing’ against eral research laboratories have trained the observable behavioral regularities apes to match a speci½c quantity of that co-vary with ‘seeing.’ items (say, three jelly beans) with the To pick just one example: we explored appropriate Arabic numeral.21 That they in detail the chimpanzee’s understand- can accomplish this should not be the ing of physical connection–of the idea least bit surprising: humans and chim- that two objects are bound together panzees (and many other species) share through some unobservable interaction the ability to visually individuate ob- such as the force transmitted by the jects. After extensive training, further- mass of one object resting on another, more, the most apt of these pupils have or the frictional forces of one object gone on to exhibit some understanding against another; or conversely, the idea of ordinality (the idea that 5 represents a that simply because two objects are physically touching does not mean there 20 , The Number Sense (Ox- is any real form of ‘connection.’ We pre- ford: , 1997). sented our chimpanzees with numerous 21 For this discussion, I rely heavily on the problems, but consider one test in which detailed results from Ai, a twenty-½ve-year-old we ½rst taught them to use a simple tool chimpanzee whose numerical abilities have to hook a ring in order to drag a platform been studied since she was ½ve by a team led by with a food treat on it toward them. Al- Tetsuro Matsuzawa in Kyoto, Japan. See Dora though they learned to do so, our real Biro and Tetsuro Matsuzawa, “Chimpanzee Numerical Competence: Cardinal and Ordinal Skills,” in Tetsuro Matsuzawa, ed., Primate Ori- 19 Daniel J. Povinelli, Folk Physics for Apes (Ox- gins of Human Cognition and Behavior (Tokyo: ford: Oxford University Press, 2000). Springer, 2001), 199–225.

Dædalus Winter 2004 39 Daniel John larger quantity than 4, for example). So, when confronted with a scrambled ar- Povinelli on isn’t this evidence that chimpanzees ray of the numerals 1 to 9, she has the learning have a solid grasp of the notion of the remarkable ability to select them in number? ascending order. Let us scratch the surface a bit, to look But what does it mean that under the at these ½ndings from the perspective I right training regime we can guide a have been advocating. First, do these chimpanzee like Ai into a performance chimpanzees possess a dual understand- that looks, in many but not all respects, ing of numbers–both as associates of like human counting? One possibility real object sets and as inherently theo- is that a basic number sense–a system retical things–such that every succes- grounded to individual macroscopic sive number in the system is exactly ‘1’ objects–is widespread among animals, more than the previous number? The and that apes (and other animals) can training data even from Ai, the most use this ability (in concert with their mathematically educated of all chim- other cognitive skills) to ½gure out panzees, suggests that they do not. For ways to cope with the ‘rules’ that example, each time the next numeral humans establish in their tests. In con- was added into her training set, it took trast, the human system for counting (as her just as long to learn its association well as other mathematical ideas) could with the appropriate number of objects be seen as building upon these older sys- as it took with the previous numeral. In tems by reifying numbers as things in other words, there appeared to be little their own right–theoretical things. This evidence that Ai understood the symbols may seem like a subtle and unimportant as anything other than associates of the distinction for some tasks, but it may be object sets. Furthermore, even her dedi- one that leaves the ape mysti½ed when cated mentors suggest that she was not facing questions that treat numbers ‘counting’ at all: with quantities of up as things in their own right. to three or four objects, she performed As a striking example of the distinc- like humans, using an automatic process tion I have been trying to draw, consider (‘subitizing’) to make her judgments; zero, surely one of the purest examples but with larger quantities, instead of that exists of an inherently unobservable counting, it appears as if she was simply entity. If I am right, then zero ought to estimating ‘larger’ or ‘smaller.’ be virtually undetectable by the chim- What about ordinality? When ½rst panzee’s cognitive system. And indeed, tested for her understanding of the rela- the data seem to bear this out.22 For all tive ordering of numbers, Ai exhibited of her training, even Ai does not appear no evidence that this was part of her to have learned to understand zero in conceptual structure. That is, when pre- this sense. True, she (and other animals) sented with pairs of numbers, 1 versus 8, have quickly learned to pick the numeral for example, she did not seem to have 0 in response to the absence of objects any notion that the value of 1 is smaller (something easily explained by associa- than the value of 8–even though she had tive learning processes). But tests of or- been correctly matching these numerals to object sets for years! Of course, after 22 Dora Biro and Tetsuro Matsuzawa, “Use of Numerical Symbols by the Chimpanzee (Pan extended training, Ai did eventually ex- troglodytes): Cardinal, Ordinals, and the Intro- hibit evidence of this ability, and now, duction of Zero,” Animal Cognition 4 (2001): after more than ½fteen years of training, 193–199.

40 Dædalus Winter 2004 dinality involving zero (choosing wheth- And yet I cannot help but suspect that Behind the ape’s er 0 is greater or lesser than 6, for ex- many of you will react to what I have appearance ample) have consistently revealed what said with a feeling of dismay–perhaps I believe might be best described as the loss; a sense that if the possibility I have virtual absence of the concept. Although sketched here turns out to be correct, this training has gradually forced her then our world will be an even lonelier ‘understanding’ of zero into a position place than it was before. But for the further and further down the ‘number time being, at least, I ask you to stay this line,’ even to this day, after thousands of thought. After all, would it really be so trials, Ai still reliably confuses 0 with 1 disappointing if our ½rst, uncontaminat- (and in some tasks, with 2 or 3 as well). ed glimpse into the mind of another However one wishes to interpret such species revealed a world strikingly dif- ½ndings, they are certainly not consis- ferent from our own; or all that surpris- tent with an understanding of the very ing if the price of admission into that essence of zero-ness.23 world were that we check some of our most familiar ways of thinking at the Our work together is done. To the door? No, to me, the idea that there may best of my ability I have laid out the case be profound psychological differences for believing that chimpanzees can be between humans and chimpanzees no bright, alert, intelligent, fully cognitive longer seems unsettling. On the con- creatures, and yet still have minds of trary, it’s the sort of possibility that has, their own. From this perspective, it on at least some occasions, emboldened may be our species that is the peculiar our species to reach out and discover one–unsatis½ed in merely knowing new worlds with open minds and hearts. what things happen, but continually driven to explain why they happen, as well. Armed with a natural language that makes referring to abstract things easy, we continually pry behind appearances, probing ever deeper into the causal structure of things. Indeed, some tests we have conducted suggest that chim- panzees may not seek ‘explanations’ at all.24

23 One might retort that the numeral 0 ap- peared quite late in human history. But here’s a thought experiment. Return to our imaginary time machine (see above) and travel back to those civilizations that predate the invention of the numeral 0. How dif½cult would it be to teach those adult humans the position occupied by the symbol for zero?

24 Daniel J. Povinelli and Sarah Dunphy-Lelii, “Do Chimpanzees Seek Explanations? Prelimi- nary Comparative Investigations,” Canadian Journal of Comparative Psychology 55 (2001): 187–195.

Dædalus Winter 2004 41 Patricia Smith Churchland

How do neurons know?

My knowing anything depends on my ride a bicycle, I have to try over and neurons–the cells of my brain.1 More over; by contrast, learning to avoid eat- precisely, what I know depends on the ing oysters if they made me vomit the speci½c con½guration of connections last time just happens. Knowing how to among my trillion neurons, on the neu- change a tire depends on cultural arti- rochemical interactions between con- facts, but knowing how to clap does not. nected neurons, and on the response And neurons are at the bottom of it all. portfolio of different neuron types. All How did it come to pass that we know this is what makes me me. anything? The range of things I know is as di- Early in the history of living things, verse as the range of stuff at a yard sale. evolution stumbled upon the advantages Some is knowledge how, some knowl- accruing to animals whose nervous sys- edge that, some a bit of both, and some tems could make predictions based upon not exactly either. Some is fleeting, some past correlations. Unlike plants, who enduring. Some I can articulate, such as have to take what comes, animals are the instructions for changing a tire, movers, and having a brain that can some, such as how I construct a logical learn confers a competitive advantage in argument, I cannot. ½nding food, mates, and shelter and in Some learning is conscious, some not. avoiding dangers. Nervous systems earn To learn some things, such as how to their keep in the service of prediction, and, to that end, map the me-relevant Patricia Smith Churchland is uc President’s Pro- parts of the world–its spatial relations, fessor of Philosophy and chair of the philosophy social relations, dangers, and so on. And, department at the University of California, San of course, brains map their worlds in Diego, and adjunct professor at the Salk Institute. varying degrees of complexity, and rela- She is past president of the American Philosophi- tive to the needs, equipment, and life- 2 cal Association and the Society for Philosophy style of the organisms they inhabit. and Psychology. Her latest books are “Brain- 1 Portions of this paper are drawn from my Wise: Studies in ” (2002) and book Brain-Wise: Studies in Neurophilosophy “On the Contrary: Critical Essays, 1987–1997” (Cambridge, Mass.: mit Press, 2002). (with , 1998). 2 See Patricia Smith Churchland and Paul M. Churchland, “Neural Worlds and Real Worlds,” © 2004 by the American Academy of Arts Nature Reviews Neuroscience 3 (11) (November & Sciences 2002): 903–907.

42 Dædalus Winter 2004 Thus humans, dogs, and frogs will repre- latches its mouth onto a nipple, and be- How do neurons sent the same pond quite differently. The gins to suck. A kitten thrown into the air know? human, for example, may be interested rights itself and lands on its feet. A hu- in the pond’s water source, the potability man neonate will imitate a facial expres- of the water, or the potential for irriga- sion, such as an outstuck tongue. But tion. The dog may be interested in a cool other knowledge, such as how to weave swim and a good drink, and the frog, in a or make ½re, is obviously learned post- good place to lay eggs, ½nd flies, bask in natally. the sun, or hide. Such contrasts have seemed to imply Boiled down to essentials, the main that everything we know is either caused problems for the neuroscience of knowl- by genes or caused by experience, where edge are these: How do structural ar- these categories are construed as exclu- rangements in neural tissue embody sive and exhaustive. But recent discover- knowledge (the problem of representa- ies in molecular biology, neuroembryol- tions)? How, as a result of the animal’s ogy, and neurobiology have demolished experience, do neurons undergo changes this sharp distinction between nature in their structural features such that and nurture. One such discovery is that these changes constitute knowing some- normal development, right from the ear- thing new (the problem of learning)? liest stages, relies on both genes and epi- How is the genome organized so that the genetic conditions. For example, a fe- nervous system it builds is able to learn male (xx) fetus developing in a uterine what it needs to learn? environment that is unusually high in The spectacular progress, during the androgens may be born with male-look- last three or four decades, in genetics, ing genitalia and may have a masculin- psychology, , neuroem- ized area in the hypothalamus, a sexually bryology, and neurobiology has given dimorphic brain region. In mice, the the problems of how brains represent gender of adjacent siblings on the pla- and learn and get built an entirely new cental fetus line in the uterus will affect look. In the process, many revered para- such things as the male/female ratio of a digms have taken a pounding. From the given mouse’s subsequent offspring, and ashes of the old verities is arising a very even the longevity of those offspring. different framework for thinking about On the other hand, paradigmatic in- ourselves and how our brains make stances of long-term learning, such as sense of the world. memorizing a route through a forest, re- ly on genes to produce changes in cells Historically, philosophers have debat- that embody that learning. If you experi- ed how much of what we know is based ence a new kind of sensorimotor event on , and how much on experi- during the day–say, for example, you ence. At one extreme, the rationalists ar- learn to cast a ½shing line–and your gued that essentially all knowledge was brain rehearses that event during your innate. At the other, radical empiricists, deep cycle, then the gene zif-268 impressed by infant modi½ability and will be up-regulated. Improvement in by the impact of culture, argued that all casting the next day will depend on the knowledge was acquired. resulting gene products and their role in Knowledge displayed at birth is obvi- neuronal function. ously likely to be innate. A normal neo- Indeed, ½ve important and related dis- nate rat scrambles to the warmest place, coveries have made it increasingly clear

Dædalus Winter 2004 43 Patricia just how interrelated ‘nature’ and ‘nur- have only about thirty thousand genes, Smith Churchland ture’ are, and, consequently, how inade- and we differ from mice in only about on quate the old distinction is.3 three hundred of those;4 meanwhile, learning First, what genes do is code for pro- we share about 99.7 percent of our genes teins. Strictly speaking, there is no gene with chimpanzees. Our brains and those for a sucking reflex, let alone for female of other primates have the same organi- coyness or Scottish thriftiness or cogni- zation, the same gross structures in zance of the concept of zero. A gene is roughly the same proportions, the same simply a sequence of base pairs contain- neuron types, and, so far as we know, ing the information that allows rna to much the same developmental schedule string together a sequence of amino and patterns of connectivity. acids to constitute a protein. (This gene Fourth, given the high degree of con- is said to be ‘expressed’ when it is tran- servation, whence the diversity of multi- scribed into rna products, some of cellular organisms? Molecular biologists which, in turn, are translated into pro- have discovered that some genes regu- teins.) late the expression of other genes, and Second, natural selection cannot di- are themselves regulated by yet other rectly select particular wiring to support genes, in an intricate, interactive, and a particular domain of knowledge. Blind systematic organization. But genes (via luck aside, what determines whether the rna) make proteins, so the expression animal survives is its behavior; its equip- of one gene by another may be affected ment, neural and otherwise, underpins via sensitivity to protein products. Addi- that behavior. Representational prowess tionally, proteins, both within cells and in a nervous system can be selected for, in the extracellular space, may interact albeit indirectly, only if the representa- with each other to yield further contin- tional package informing the behavior gencies that can ½gure in an unfolding was what gave the animal the competi- regulatory cascade. Small differences in tive edge. Hence representational so- regulatory genes can have large and far- phistication and its wiring infrastructure reaching effects, owing to the intricate can be selected for only via the behavior hierarchy of regulatory linkages between they upgrade. them. The emergence of complex, inter- Third, there is a truly stunning degree active cause-effect pro½les for gene ex- of conservation in structures and devel- pression begets very fancy regulatory opmental organization across all verte- cascades that can beget very fancy or- brate animals, and a very high degree of ganisms–us, for example. conservation in basic cellular functions Fifth, various aspects of the develop- across phyla, from worms to spiders to ment of an organism from fertilized egg humans. All nervous systems use essen- to up-and-running critter depend on tially the same neurochemicals, and where and when cells are born. Neurons their neurons work in essentially the originate from the daughter cells of the same way, the variations being vastly last division of pre-neuron cells. Wheth- outweighed by the similarities. Humans er such a daughter cell becomes a glial (supporting) cell or a neuron, and which 3 In this discussion, I am greatly indebted to type of some hundred types of neurons Barbara Finlay, Richard Darlington, and Nich- olas Nicastro, “Developmental Structure in 4 See John Gerhart and Marc Kirschner, Cells, Brain Evolution,” Behavioral and Brain Sciences Embryos, and Evolution (Oxford: Blackwell, 24 (2) (April 2001): 263–278. 1997).

44 Dædalus Winter 2004 the cell becomes, depends on its epige- learning to read braille.6 So long as the How do neurons netic circumstances. Moreover, the blindfold remains ½rmly in place to pre- know? manner in which neurons from one area, vent any light from falling on the retina, such as the thalamus, connect to cells in performance of braille reading steadily the cortex depends very much on epige- improves. The blindfold is essential, for netic circumstances, e.g., on the sponta- normal visual stimuli that activate the neous activity, and later, the experience- visual cortex in the normal way impede driven activity, of the thalamic and corti- acquisition of the tactile skill. For exam- cal neurons. This is not to say that there ple, if after ½ve days the blindfold is re- are no causally signi½cant differences moved, even briefly while the subject between, for instance, the neonatal suck- watches a television program before ing reflex and knowing how to make a going to sleep, his braille performance ½re. Differences, obviously, there are. under blindfold the next day falls from The essential point is that the differ- its previous level. If the visual cortex can ences do not sort themselves into the be recruited in the processing of nonvi- archaic ‘nature’ versus ‘nurture’ bins. sual signals, what sense can we make of Genes and extragenetic factors collabo- the notion of the dedicated vision mod- rate in a complex interdependency.5 ule, and of the dedicated-modules hy- pothesis more generally? Recent discoveries in What is clear is that the nature versus point in this same direction. Hitherto, it nurture dichotomy is more of a liability was assumed that brain centers–mod- than an asset in framing the inquiry into ules dedicated to a speci½c task–were the origin of plasticity in human brains. wired up at birth. The idea was that we Its inadequacy is rather like the inade- were able to see because dedicated ‘visu- quacy of ‘good versus evil’ as a frame- al modules’ in the cortex were wired for work for understanding the complexity vision; we could feel because dedicated of political life in human societies. It is modules in the cortex were wired for not that there is nothing to it. But it is touch, and so on. like using a grub hoe to remove a splin- The truth turns out to be much more ter. puzzling. For example, the visual cortex of a An appealing idea is that if you learn blind subject is recruited during the something, such as how to tie a trucker’s reading of braille, a distinctly nonvisual, knot, then that information will be tactile skill–whether the subject has ac- stored in one particular location in the quired or congenital blindness. It turns brain, along with related knowledge– out, moreover, that stimulating the say, between reef knots and half-hitches. subject’s visual cortex with a magnet- That is, after all, a good method for stor- induced current will temporarily impede ing tools and paper ½les–in a particular his braille performance. Even more re- drawer at a particular location. But this markably, activity in the visual cortex is not the brain’s way, as occurs even in normal seeing subjects ½rst demonstrated in the 1920s. who are blindfolded for a few days while 6 See Alvaro Pascual-Leone et al., “Study and Modulation of Human Cortical Excitability 5 See also Steven Quartz and Terrence J. Sej- with Transcranial Magnetic Stimulation,” Jour- nowski, Liars, Lovers, and Heroes (New York: nal of Clinical 15 (1998): 333– William Morrow, 2002). 343.

Dædalus Winter 2004 45 Patricia Lashley reasoned that if a rat learned Neuroscientists are trying to under- Smith Churchland something, such as a route through a stand the structure of learning by using a on certain maze, and if that information variety of research strategies. One strat- learning was stored in a single, punctate location, egy consists of tracking down experi- then you should be able to extract it by ence-dependent changes at the level of lesioning the rat’s brain in the right the neuron to ½nd out what precisely place. Lashley trained twenty rats on his changes, when, and why. Another strate- maze. Next he removed a different area gy involves learning on a larger scale: of cortex from each animal, and allowed what happens in behavior and in partic- the rats time to recover. He then retested ular brain subsystems when there are le- each one to see which lesion removed sions, or during development, or when knowledge of the maze. Lashley discov- the subject performs a memory task ered that a rat’s knowledge could not while in a scanner, or, in the case of ex- be localized to any single region; it ap- perimental animals, when certain genes peared that all of the rats were some- are knocked out? At this level of inquiry, what impaired and yet somewhat com- psychology, neuroscience, and molecu- petent–although more extensive tissue lar biology closely interact. removal produced more serious memory Network-level research aims to strad- de½cit. dle the gap between the systems and the As improved experimental protocols neuronal levels. One challenge is to un- later showed, Lashley’s non-localization derstand how distinct local changes in conclusion was essentially correct. many different neurons yield a coherent There is no such thing as a dedicated global, system-level change and a task- memory organ in the brain; information suitable modi½cation of behavior. How is not stored on the ½ling cabinet model do diverse and far-flung changes in the at all, but distributed across neurons. brain underlie an improved golf swing A general understanding of what it or a better knowledge of quantum means for information to be distributed mechanics? over neurons in a network has emerged What kinds of experience-dependent from computer models. The basic idea modi½cations occur in the brain? From is that arti½cial neurons in a network, one day to the next, the neurons that col- by virtue of their connections to other lectively make me what I am undergo arti½cial neurons and of the variable many structural changes: new branches strengths of those connections, can pro- can sprout, existing branches can ex- duce a pattern that represents something tend, and new receptor sites for neuro- –such as a male face or a female face, or chemical signals can come into being. the face of Churchill. The connection On the other hand, pruning could de- strengths vary as the arti½cial network crease branches, and therewith decrease goes through a training phase, during the number of synaptic connections be- which it gets feedback about the adequa- tween neurons. Or the synapses on re- cy of its representations given its input. maining branches could be shut down But many details of how actual neural altogether. Or the whole cell might die, nets–as opposed to computer-simulated taking with it all the synapses it formerly ones–store and distribute information supported. Or, ½nally, in certain special have not yet been pinned down, and so regions, a whole new neuron might be computer models and neural experi- born and begin to establish synaptic ments are coevolving. connections in its region.

46 Dædalus Winter 2004 And that is not all. Repeated high rates more likely to spike. This joint input ac- How do neurons of synaptic ½ring (spiking) will deplete tivity creates a larger postsynaptic effect, know? the vesicles available triggering a cascade of events inside the for release, thus constituting a kind of neuron that strengthens the synapse. memory on the order of two to three This general arrangement allows for dis- seconds. The constituents of particular tinct but associated world events (e.g., neurons, the number of vesicles released blue flower and plenty of nectar) to be per spike, and the number of transmitter modeled by associated neuronal events. molecules contained in each vesicle, can change. And yet, somehow, my skills re- The nervous system enables animals to main much the same, and my autobio- make predictions.7 Unlike plants, ani- graphical memories remain intact, even mals can use past correlations between though my brain is never exactly the classes of events (e.g., between red cher- same from day to day, or even from min- ries and a satisfying taste) to judge the ute to minute. probability of future correlations. A cen- No ‘bandleader’ neurons exist to en- tral part of learning thus involves com- sure that diverse changes within neu- puting which speci½c properties predict rons and across neuronal populations the presence of which desirable effects. are properly orchestrated and collective- We correlate variable rewards with a fea- ly reflect the lessons of experience. Nev- ture to some degree of probability, so ertheless, several general assumptions good predictions will reflect both the guide research. For convenience, the expected value of the reward and the broad range of neuronal modi½ability probability of the reward’s occurring; can be condensed by referring simply to this is the expected utility. Humans and the modi½cation of synapses. The deci- bees alike, in the normal course of the sion to modify synapses can be made business of life, compute expected utili- either globally (broadcast widely) or ty, and some neuronal details are begin- locally (targeting speci½c synapses). If ning to emerge to explain how our made globally, then the signal for change brains do this. will be permissive, in effect saying, “You To the casual observer, bees seem to may change yourself now”–but not dic- visit flowers for nectar on a willy-nilly tating exactly where or by how much or basis. Closer , however, re- in what direction. If local, the decision veals that they forage methodically. Not will likely conform to a rule such as this: only do bees tend to remember which If distinct but simultaneous input signals individual flowers they have already vis- cause the receiving neuron to respond ited, but in a ½eld of mixed flowers with with a spike, then strengthen the con- varying amounts of nectar they also nection between the input neurons and learn to optimize their foraging strategy, the output neurons. On its own, a signal so that they get the most nectar for the from one presynaptic (sending) neuron least effort. is unlikely to cause the postsynaptic (re- Suppose you stock a small ½eld with ceiving) neuron to spike. But if two dis- two sets of plastic flowers–yellow and tinct presynaptic neurons–perhaps one blue–each with wells in the center into from the auditory system and one from which precise amounts of sucrose have the somatosensory system–connect to the same postsynaptic neuron at the 7 John Morgan Allman, Evolving Brains (New same time, then the receiving neuron is York: Scienti½c American Library, 1999).

Dædalus Winter 2004 47 Patricia been deposited.8 These flowers are ran- and ‘the goodies received this time.’9 Smith Churchland domly distributed around the enclosed Vum’s output is the release of a neuro- on ½eld and then baited with measured vol- modulator that targets a variety of cells, learning umes of ‘nectar’: all blue flowers have including those responsible for action two milliliters; one-third of the yellow selection. If that neuromodulator also flowers have six milliliters, two-thirds acts on the synapses connecting the sen- have none. This sucrose distribution sory neurons to vum, then the synapses ensures that the mean value of visiting will get stronger, depending on whether a population of blue flowers is the same the vum calculates ‘worse than expect- as that of visiting the yellow flowers, ed’ (less neuromodulator) or ‘better though the yellow flowers are more than expected’ (more neuromodulator). uncertain than the blues. Assuming that the Montague-Dayan After an initial random sampling of model is correct, then a surprisingly sim- the flowers, the bees quickly fall into a ple circuit, operating according to a fair- pattern of going to the blue flowers 85 ly simple weight-modi½cation algo- percent of the time. You can change rithm, underlies the bee’s adaptability their foraging pattern by raising the to foraging conditions. mean value of the yellow flowers–for Dependency relations between phe- example, by baiting one-third of them nomena can be very complex. In much with ten milliliters. The behavior of the of life, dependencies are conditional and bees displays a kind of trade-off between probabilistic: If I put a fresh worm on the reliability of the source type and the the hook, and if it is early afternoon, nectar volume of the source type, with then very probably I will catch a trout here. the bees showing a mild preference for As we learn more about the complexities reliability. What is interesting is this: of the world, we ‘upgrade’ our represen- depending on the reward pro½le taken tations of dependency relations;10 we in a sample of visits, the bees revise their learn, for example, that trout are more strategy. The bees appear to be calculat- likely to be caught when the water is ing expected utility. How do bees–mere cool, that shadowy pools are more bees–do this? promising ½sh havens than sunny pools, In the bee brain there is a neuron, and that talking to the worm, entreating though itself neither sensory nor motor, the trout, or wearing a ‘lucky’ hat makes that responds positively to reward. This no difference. Part of what we call intel- neuron, called VUMmx1 (‘vum’ for ligence in humans and other animals is short), projects very diffusely in the bee the capacity to acquire an increasingly brain, reaching both sensory and motor complex understanding of dependency regions, as it mediates reinforcement relations. This allows us to distinguish learning. Using an arti½cial neural net- work, Read Montague and Peter Dayan 9 See Read Montague and Peter Dayan, “Neu- robiological Modeling,” in William Bechtel, discovered that the activity of vum rep- George Graham, and D. A. Balota, eds., A Com- resents prediction error–that is, the dif- panion to Cognitive Science (Malden, Mass.: ference between ‘the goodies expected’ Blackwell, 1998).

10 Clark N. Glymour, The Mind’s Arrows (Cam- 8 This experiment was done by Leslie Real, bridge, Mass.: mit Press, 2001). See also Alison “Animal Choice Behavior and the Evolution of Gopnik, Andrew N. Meltzoff, and Patricia K. Cognitive Architecture,” Science (1991): 980– Kuhl, The Scientist in the Crib (New York: Wil- 986. liam Morrow & Co., 1999).

48 Dædalus Winter 2004 fortuitous correlations that are not gen- These discoveries open the door to How do neurons uinely predictive in the long run (e.g., understanding the neural organization know? breaking a tooth on Friday the thir- underlying prediction. They begin to teenth) from causal correlations that are forge the explanatory bridge between (e.g., breaking a tooth and chewing hard experience-dependent changes in single candy). This means that we can replace neurons and experience-dependent superstitious hypotheses with those that guidance of behavior. And they have be- pass empirical muster. gun to expose the neurobiology of addic- Like the bee, humans and other ani- tion. A complementary line of research, mals have a reward system that mediates meanwhile, is untangling the mecha- learning about how the world works. nisms for predicting what is nasty. Al- There are neurons in the mammalian though aversive learning depends upon a brain that, like vum, respond to re- different set of structures and networks ward.11 They shift their responsiveness than does reinforcement learning, here to a stimulus that predicts reward, or too the critical modi½cations happen indicates error if the reward is not forth- at the level of individual neurons, and coming. These neurons project from a these local modi½cations are coordinat- brainstem structure (the ventral tegmen- ed across neuronal populations and inte- tal area, or ‘vta’) to the frontal cortex, grated across time. and release dopamine onto the postsy- Within other areas of learning re- naptic neurons. The dopamine, only search, comparable explanatory threads one of the neurochemicals involved are beginning to tie together the many in the reward system, modulates the levels of nervous system organization. excitability of the target neurons to the This research has deepened our under- , thus setting up the standing of working memory (holding conditions for local learning of speci½c information at the ready during the ab- associations. sence of relevant stimuli) spatial learn- Reinforcing a behavior by increasing ing, autobiographical memory, motor pleasure and decreasing anxiety and skills, and logical inference. Granting works very ef½ciently. Nevertheless, the extraordinary research accomplish- such a system can be hijacked by plant- ments in the neuroscience of knowledge, derived molecules whose behavior mim- nevertheless it is vital to realize that ics the brain’s own reward system neu- these are still very early days for neuro- rochemicals. Changes in reward system science. Many surprises–and even a rev- pathways occur after administration of olution or two–are undoubtedly in cocaine, nicotine, or opiates, all of which store. bind to receptor sites on neurons and are similar to the brain’s own peptides. The Together, neuroscience, psychology, precise role in brain function of the large embryology, and molecular biology are number of brain peptides is one of neu- teaching us about ourselves as knowers– roscience’s continuing conundrums.12 about what it is to know, learn, remem- ber, and forget. But not all philosophers 11 See Paul W. Glimcher, Decisions, Uncertainty, mit embrace these developments as prog- and the Brain (Cambridge, Mass.: Press, ress.13 Some believe that what we call 2003). 13 I take it as a sign of the backwardness of aca- 12 I am grateful to Roger Guillemain for dis- demic philosophy that one of its most esteemed cussing this point with me. living practitioners, Jerry Fodor, is widely sup-

Dædalus Winter 2004 49 Patricia external reality is naught but an idea cre- their relative success in predicting and Smith Churchland ated in a nonphysical mind, a mind that explaining. on can be understood only through intro- But does all of this mean that there is a learning spection and reflection. To these philos- kind of fatal circularity in neuroscience ophers, developments in cognitive neu- –that the brain necessarily uses itself to roscience seem, at best, irrelevant. study itself? Not if you think about it. The element of truth in these philoso- The brain I study is seldom my own, but phers’ approach is their hunch that the that of other animals or humans, and I mind is not just a passive canvas on can reliably generalize to my own case. which reality paints. Indeed, we know Neuroepistemology involves many that brains are continually organizing, brains–correcting each other, testing structuring, extracting, and creating. As each other, and building models that can a central part of their predictive func- be rated as better or worse in character- tions, nervous systems are rigged to izing the neural world. make a coherent story of whatever input Is there anything left for the philoso- they get. ‘Coherencing,’ as I call it, pher to do? For the neurophilosopher, at sometimes entails seeing a fragment as least, questions abound: about the inte- a whole, or a contour where none exists; gration of distinct memory systems, the sometimes it involves predicting the im- nature of representation, the nature of minent perception of an object as yet reasoning and rationality, how informa- unperceived. As a result of learning, tion is used to make decisions, what brains come to recognize a stimulus as nervous systems interpret as informa- indicating the onset of meningitis in a tion, and so on. These are questions with child, or an eclipse of the Sun by the deep roots reaching back to the ancient Earth’s shadow. Such knowledge de- Greeks, with ramifying branches ex- pends upon stacks upon stacks of neural tending throughout the history and phi- networks. There is no apprehending the losophy of Western thought. They are nature of reality except via brains, and questions where experiment and theo- via the theories and artifacts that brains retical insight must jointly conspire, devise and interpret. where creativity in experimental design From this it does not follow, however, and creativity in theoretical speculation that reality is only a mind-created idea. It must egg each other on to unforeseen means, rather, that our brains have to discoveries.14 keep plugging along, trying to devise hypotheses that more accurately map 14 Many thanks to Ed McAmis and Paul the causal structure of reality. We build Churchland for their ideas and revisions. the next generation of theories upon the scaffolding–or the ruins–of the last. How do we know whether our hypothe- ses are increasingly adequate? Only by

ported for the following conviction: “If you want to know about the mind, study the mind –not the brain, and certainly not the genes” (Times Literary Supplement, 16 May 2003, 1–2). If philosophy is to have a future, it will have to do better than that.

50 Dædalus Winter 2004 Michael Tomasello

Learning through others

Learning is a biological adaptation. signi½cant degree processes of cogni- The majority of organisms on Earth tion. But the modern view that learning learn little or nothing during their in- assumes diverse forms in different spe- dividual lifetimes. On the other hand, cies and behavioral domains, and oper- many mammals are born in a highly ates in concert with cognitive processes immature state and so they must indi- that may be speci½c to particular species vidually learn things crucial for their or domains, has for the most part sus- survival. In order to ½nd food reliably, pended that search. youngsters of foraging species must For social species such as humans and learn the spatial layouts of their local other mammals, an especially important environments. In order to distinguish form of learning is social learning. Ob- friends from enemies, youngsters of serving the activities of others and learn- social species must learn to recognize ing about the world from or through the individuals who make up their them enables individuals to acquire in- social groups. formation with less effort and risk than For several decades, behaviorists at- if they were forced to learn on their own. tempted to ½nd the laws of learning that For instance, many species of rats learn applied equally to all species, for any and which foods to eat and which to avoid all tasks, and that did not involve to any by observing what other rats eat and then seeing what happens to them sub- Michael Tomasello is codirector of the Max sequently–clearly a safer strategy than 1 Planck Institute for Evolutionary Anthropology always trying out new foods for oneself. and honorary professor of psychology at the Despite an overall similarity in the University of Leipzig and at Manchester Univer- function of learning in the lives of differ- sity. His publications include “The Cultural Ori- ent species of mammals, social learning, gins of Human Cognition” (1999), “The New like individual learning, comes in many Psychology of Language: Cognitive and Func- different forms. In our empirical work tional Approaches to Language Structure” (1998), over the past ½fteen years, we have in- and “First Verbs: A of Early Gram- vestigated forms of social learning that matical Development” (1992). 1 Bennett G. Galef, Jr., “Social Influences in Food Choices of Norway Rats and Mate Choic- © 2004 by the American Academy of Art es of Japanese Quail,” International Journal of & Sciences Comparative Psychology 14 (1–2) (2001): 1–24.

Dædalus Winter 2004 51 Michael human beings share with other primate learning, most importantly because arti- Tomasello species, as well as forms that are unique- facts and practices–exempli½ed proto- on learning ly human. The unique forms mostly typically by the use of tools and linguis- derive, we believe, from some social- tic symbols–invariably point beyond cognitive processes that only humans themselves to the phenomena for which possess.2 they have been designed. To learn the In brief, because human beings per- conventional use of a tool or a symbol, ceive the behavior of others in intention- an individual must therefore come to al terms–that is, because they perceive understand why, toward what outside a person ‘cleaning the table’ or ‘opening end, another individual is using it.4 the drawer,’ rather than simply moving Chimpanzees, humans’ nearest pri- her limbs in a particular way–they learn mate relatives, do not learn from one from the behavior of others in unique another in this same way. In 1996, I re- ways. We have called this process ‘cul- viewed all of the experimental studies tural learning’ to distinguish it from of chimpanzee tool use, and I concluded processes of social learning in general, that chimpanzees are very good at learn- and also to highlight the crucial role of ing from others about the dynamic af- culture in the acquisition of many hu- fordances of objects, but are not skillful man skills. My colleagues and I have dis- at learning from others new behavioral tinguished three kinds of cultural learn- strategies or intentional activities per ing: imitative learning, instructed learn- se.5 For example, if a mother rolls over ing, and collaborative learning.3 The a log and eats the insects underneath, ability of individuals to imagine them- her child will very likely follow suit. selves in the ‘mental shoes’ of other peo- From her mother’s act the child has ple, to understand conspeci½cs as beings learned that there are insects under like themselves who have intentional this particular log–but she did not and mental lives like themselves, enables learn from her mother how to roll these types of cultural learning. Most of over a log or how to eat insects; she our empirical work has focused on only could have learned these on her own. one type of cultural learning–imitation Thus the youngster would have learned in children before about two years of the same thing if the wind, rather than age. So that will be my focus here. her mother, had exposed the ants under the log. This is an instance of ‘emulation The recognition of others as intentional learning,’ which concerns changes of beings like oneself is crucial in human state in the environment rather than a conspeci½c’s intentional activity or 2 Michael Tomasello, Ann C. Kruger, and behavioral strategy. Hillary H. Ratner, “Cultural Learning,” Be- In some circumstances, emulation havioral and Brain Sciences 16 (1993): 495–552; learning is a more adaptive strategy Michael Tomasello, “Do Apes Ape?” in Cecilia M. Heyes and Bennett G. Galef, Jr., eds., Social than learning by imitation. For example, Learning in Animals: The Roots of Culture (San Kathy Nagell, Kelly Olguin, and I pre- Diego, Calif.: Academic Press, 1996); Michael sented chimpanzees and two-year-old Tomasello, The Cultural Origins of Human Cogni- human children with a rake-like tool and tion (Cambridge, Mass.: Harvard University Press, 1999). 4 Tomasello, The Cultural Origins of Human Cognition. 3 Tomasello, Kruger, and Ratner, “Cultural Learning.” 5 Tomasello, “Do Apes Ape?”

52 Dædalus Winter 2004 an out-of-reach object.6 The tool could during the demonstration, perceiving Learning the actions of the demonstrator just, in through be used in either of two ways leading to others the same end result of obtaining the ob- effect, as other physical motions. The ject. Within each species, one group of intentional states of the demonstrator, subjects observed a demonstrator em- and thus her behavioral methods as dis- ploy a relatively inef½cient method of tinct entities, are simply not a part of tool use, while another group observed their experience. a more ef½cient method of tool use. The A similar story may be told about the result: human children in general copied gestural communication of chimpan- the method of the assigned demonstra- zees. In a series of studies, we explored tor (imitative learning), while chimpan- whether youngsters acquire their gestur- zees used the same methods to obtain al signals by imitative learning or by a the object no matter which demonstra- process of ontogenetic ritualization.7 In tion they observed (emulation learning). ontogenetic ritualization, two organisms The interesting point is that many chil- devise a communicatory signal through dren insisted on reproducing adult be- repeated instances of a social interac- havior even if it seemed inef½cient– tion. For example, an infant may initiate leading to a less successful performance nursing by going directly for the moth- than that of the chimpanzees. Imitation er’s nipple, perhaps grabbing and mov- is thus not a ‘higher’ or ‘more intelli- ing her mother’s arm in the process. So gent’ learning strategy than emulation; in some future encounter the mother it is simply a more culturally mediated might sense, and respond to, her in- strategy–which, in some circumstances fant’s hunger at the ½rst touch of her and for some behaviors, has some ad- arm, leading the infant to abbreviate vantages. her signal for hunger even further the Chimpanzees are very creative in using next time. This is presumably analo- tools, and intelligent about understand- gous to the way that most human in- ing changes in the environment brought fants learn the ‘arms over head’ gesture about by the tool use of others. But they to request that adults pick them up– do not seem to understand the instru- ½rst as a direct attempt to crawl up the mental behavior of conspeci½cs in the same way as humans do. Humans per- 7 Michael Tomasello, Barbara George, Ann ceive the demonstrator’s apparent inten- C. Kruger, Jeff Farrar, and Andrea Evans, “The tion as centrally important, and they un- Development of Gestural Communication in Young Chimpanzees,” Journal of Human Evolu- derstand this goal as something separate tion” 14 (1985): 175–186; Michael Tomasello, from the various behavioral means that Deborah Gust, and Thomas Frost, “A Longi- may be used to accomplish it. In the ab- tudinal Investigation of Gestural Communi- sence of this ability to understand goal cation in Young Chimpanzees,” Primates 30 and behavioral means as separable in the (1989): 35–50; Michael Tomasello, Josep Call, Katherine Nagell, Raquel Olguin, and Malinda actions of others, chimpanzees focus on Carpenter, “The Learning and Use of Gestural the changes of state (including changes Signals by Young Chimpanzees: A Trans- in the spatial position) of the objects generational Study,” Primates 37 (1994): 137– 154; Michael Tomasello, Josep Call, Jennifer 6 Katherine Nagell, Raquel Olguin, and Mi- Warren, Thomas Frost, Malinda Carpenter, chael Tomasello, “Processes of Social Learning and Katherine Nagell, “The Ontogeny of Chim- in the Tool Use of Chimpanzees (Pan troglo- panzee Gestural Signals: A Comparison Across dytes) and Human Children (Homo sapiens),” Groups and Generations,” Evolution of Commu- Journal of Comparative Psychology 107 (1993): nication 1 (1997): 223–253. 174–186.

Dædalus Winter 2004 53 Michael adult’s body, and then, as the adult explanation for emulation learning in Tomasello on anticipates the baby’s desire and picks the case of tool use. Like emulation learning her up, as an abbreviated, ritualized ver- learning, ontogenetic ritualization does sion of this crawling activity performed not require individuals to analyze the for communicative purposes only.8 behavior of others in terms of ends and All available evidence suggests that means in the same way as does imitative ontogenetic ritualization, not imitative learning. Imitatively learning an arm learning, is responsible for chimpanzees’ touch as a solicitation for nursing would acquisition of communicative gestures. require that an infant observe another Individual chimpanzees use a number infant using an arm touch and under- of idiosyncratic signals that must have stand that other infant’s goal. Ritualiz- been individually invented and ritual- ing the arm touch, on the other hand, ized–a ½nding that longitudinal analy- only requires the infant to anticipate the ses have con½rmed.9 Signi½cantly, cap- future behavior of a conspeci½c in a con- tive youngsters raised in peer groups text in which the infant already has the that have no opportunity to observe goal of nursing. Ontogenetic ritualiza- older conspeci½cs frequently use many tion is thus, like emulation learning, a of the same gestures that are common very useful learning process that is im- among other chimpanzee youngsters. In portant in all social species–but it is not an experimental study, colleagues and I a learning process by which individuals removed an individual from the group attempt to reproduce the intentional and taught her two different arbitrary activities or behavioral strategies of oth- gestures she could use to obtain desired ers; it is not cultural learning the way food from a human.10 When she re- humans practice it. turned to her group and used these sig- nals to obtain food from a human, not Human beings begin to learn through even one chimpanzee reproduced either imitation at around the ½rst birthday. of the new gestures–even though all of But it takes clever experimentation to the other individuals observed the ges- distinguish the unique features of this turer and were highly motivated for the form of learning from those of another. food. For example, if an adult takes the top off Chimpanzee youngsters thus acquire of a pen and a child then does the same, the majority, if not the totality, of their there are many possible explanations, gestures by individually ritualizing them including emulation and mimicking with one another. The explanation for (copying movements without knowing this learning process is analogous to the what they are for). Researchers have therefore devised ingenious techniques 8 Andrew Lock, “The Emergence of Language,” for analyzing the different components in Andrew Lock, ed., Action, Gesture, and Sym- of what the child perceives, understands, bol: The Emergence of Language (New York: Aca- and reproduces in a demonstrated act. demic Press, 1978). For example, according to the tech- 9 See Jane Goodall, The Chimpanzees of Gombe: nique Andy Meltzoff devised, fourteen- Patterns of Behavior (Cambridge, Mass.: Har- month-old infants saw an adult illumi- vard University Press, 1986). nate a box by bending down and touch- ing her head to the top of it.11 Although 10 Tomasello, Call, Warren, Frost, Carpenter, and Nagell, “The Ontogeny of Chimpanzee 11 Andrew Meltzoff, “Infant Imitation After a Gestural Signals.” One-Week Delay: Long-Term Memory for

54 Dædalus Winter 2004 infants could more easily have solved actions on objects.13 For each object, the Learning through this task by emulation (e.g., by touching pair of actions was followed by a striking others the box with their hand), they instead result–the sudden illumination of col- chose to use the same means as the ored lights, for example. In the key ex- adult, unusual as it was. These infants periment, one of the demonstrator’s could have been mimicking the adult’s paired actions was marked verbally as unusual action without understanding intentional (“There!”) while the other the goal of turning on the light. But if was marked verbally as accidental they had been copying this action with (“Woops!”), but otherwise the actions the same goal in mind, their behavior looked very similar. Instead of mimick- would have been an instance of imitative ing both of the actions they observed, learning. even the youngest infants reproduced In order to determine which of these the action marked as intentional signi½- two mechanisms was at work, Malinda cantly more often than the one marked Carpenter, Nagell, and I tested nine- to as accidental. ½fteen-month-old infants on a modi½ed Another study demonstrated that in- version of this task: we delayed the illu- fants were able to imagine the goal to- mination of the light after the infants’ ward which the adult was acting, even reproduction of the action, and noted though they never actually saw any con- whether they looked in anticipation crete results. In a 1995 experiment, to the light.12 We found that infants Meltzoff showed eighteen-month-olds twelve-months and older looked to the an adult either successfully completing light in anticipation before it came on. a task (pulling apart two halves of a If the light did not come on, they often dumbbell) or trying but failing to do so repeated their action or looked quizzi- (because the adult’s hands slipped off cally to the people in the room. This sug- the ends of the dumbbell). Infants were gests that they were adopting the adult’s able to complete the task whether or not means in order to achieve the same goal they had seen an adult successfully com- of turning on the light. Infants thus were plete it. Yet these eighteen-month-olds not just mimicking the adult’s action, were not able to achieve the same result but were engaging in imitative learning when they watched a machine either of a novel means to achieve a perceived successfully completing the same task end. or trying, or failing, to do so. Francesca In another experiment, infants were Bellagamba and I replicated the basic shown identical actions that produced ½ndings of this study with twelve- and identical results, but with different eighteen-month-old infants, but we expressed intentions. Carpenter, Na- found that twelve-month-olds could meera Akhtar, and I showed fourteen- not reproduce the adult’s intended to eighteen-month-olds a series of two action when they only saw her trying unsuccessfully to perform it.14 Novel Acts and Multiple Stimuli,” Developmen- tal Psychology 24 (1988): 470–476. 13 Malinda Carpenter, Nameera Akhtar, and Michael Tomasello, “Sixteen-Month-Old In- 12 Malinda Carpenter, Katherine Nagell, and fants Differentially Imitate Intentional and Michael Tomasello, “Social Cognition, Joint Accidental Actions,” Infant Behavior and Develop- Attention, and Communicative Competence ment 21 (1998): 315–330. from 9 to 15 Months of Age,” Monographs of the Society for Research in Child Development 63 14 Francesca Bellagamba and Michael Toma- (4) (1998). sello, “Reenacting Intended Acts: Comparing

Dædalus Winter 2004 55 Michael Other studies have manipulated the own choice of a behavioral means in Tomasello on social learning context in an effort to similar or different circumstances. learning influence what behavior children repro- A series of studies of older children duce and so gain insight into what they extends these ½ndings. For example, interpret as intentional action. Using Bekkering and his colleagues showed Meltzoff’s study as a starting point, three- to six-year-old children an exper- George Gergely, Harold Bekkering, and imenter touching a table in one of two Ilday Király showed fourteen-month- locations.16 In one condition there were olds an adult touching her head to the dots on the table in those locations, and top of a box to turn on a light.15 In their in another condition there were no dots. study, half of the infants saw the adult In the no-dot condition, children usual- turn on the light while her hands were ly matched the adult’s behavior exactly, occupied (she was holding a blanket even copying her crossed or straight around her shoulders), and half saw arm positions–presumably because her turn it on while her hands were there was no other apparent goal to free. Infants who saw the hands-free her actions than these arm movements. demonstration touched the box with In the dot condition, however, children their heads signi½cantly more often touched the same locations as the exper- than infants who saw the hands- imenter, but often did not match her ex- occupied demonstration. Infants thus act arm positions. This is presumably used the context of the situation to in- because when there were dots they in- terpret the adult’s behavior, appearing terpreted the adult’s goal as ‘touching to assume that if the adult’s hands were the dots,’ whereas when there were no free and she still chose to use her head, dots the only possible goal seemed to then there must be a good reason for this be ‘moving one’s arms like this.’ Bek- choice. Meanwhile, the infants who saw kering and his colleagues concluded that the other demonstration apparently in- young children’s imitation is guided by terpreted the use of her head as neces- their understanding of adults’ goals and sary given her circumstances (and so of the hierarchy of those goals, and that as an inessential part of her intention), children imitate what they perceive the and thus did not reproduce this action. adults’ main goal to be. These infants’ interpretation of the Subsequent studies have con½rmed adult’s goal thus differed across condi- that children use context to interpret tions: in the hands-occupied condition adults’ actions, and that this influences her apparent goal was ‘turn on the light’; what they learn. In one study, an adult in the hands-free condition it was ‘turn demonstrated to ½ve groups of children on the light with your head.’ By fourteen how to pull out a pin and open a box.17 months, infants thus evidence a very deep understanding of intentional ac- 16 Harold Bekkering, Andrew W. Wohlschlaeg- er, and Merideth Gattis, “Imitation of Ges- tion, of how it relates to the surrounding tures in Children is Goal-directed,” Quarterly context, and of what this means for their Journal of 53A (1) (2000): 153–164. 12- and 18-Month-Olds,” Infant Behavior and Development 22 (1999): 277–282. 17 Malinda Carpenter, Josep Call, and Michael Tomasello, “Understanding Others’ Prior 15 György Gergely, Harold Bekkering, and Intentions Enables 2-Year-Olds to Imitatively Ildikó Király, “Rational Imitation in Preverbal Learn a Complex Task,” Child Development 73 Infants,” Nature 415 (2002): 755. (2002): 1431–1442.

56 Dædalus Winter 2004 What differed among the groups was might blow a raspberry along her own Learning through what the children experienced just prior arm, in exactly the same place her moth- others to this demonstration. One group of er did, or alternatively, she might blow a children received information about raspberry along her mother’s arm–in- the adult’s goal ahead of time; another terpreting the behavior in this case group received none; the three other reciprocally as ‘blowing on the partner’s groups received varying amounts of arm.’ I have called this ‘role reversal imi- information about the adult’s goal. tation.’ In an ongoing study, my col- In the demonstration, the adult either leagues and I have found that eighteen- tugged unsuccessfully on the door of the month-olds are more likely to employ box, or showed the box already open, or this reciprocal interpretation than are visited and opened three different boxes twelve-month-olds. At both ages, chil- before demonstrating how to open the dren are more likely to reciprocate in the test box. Thus all of the children in all of situation where the adult, for example, these conditions saw a full demonstra- pats her own head (and the child pats his tion of how to open the box, but only the own), than in the case where the two children in the three prior-information partners act on one another. conditions could know what the adult A similar process occurs in the learn- was about to do before she began this ing of language, since learning to use lin- demonstration. Two- and two-and-a- guistic symbols is also reciprocal. Thus, half-year-old children were signi½cant- when an adult uses a linguistic symbol ly better at opening the box themselves in a communicative act, she directs the when they knew the adult’s goal ahead child to attend to something. Conse- of time. In this study, then, children quently, to learn to use a symbol as the interpreted the exact same behavior adult does, the child must learn to direct differently depending on whether they the adult’s attention as the adult had knew the adult’s goal ahead of time– directed the child’s.18 with no concurrent cues in adult emo- Interestingly, my colleagues and I have tional expression or the like. In other recently offered evidence that some- words, in the control conditions the thing similar goes on in children’s early children were not able to provide an in- symbolic play. Before two years of age, tentional description of ‘what the adult by watching adults children imitatively is doing,’ whereas in the prior informa- learn symbolic behaviors with objects, in tion conditions they were able to under- much the same way that they learn stand the behavior as the intentional instrumental actions with artifacts. As action ‘trying to open the box.’ they grow older, they look to the adult more often, and in some cases smile There are some kinds of actions that more often, when producing the sym- children observe and attempt to imitate bolic behaviors. This is evidence that that have a special structure because children of this age are reproducing a they involve people having goals toward special kind of intentionality–a kind one another reciprocally. For example, a of mutually reciprocal intentionality in mother might blow a raspberry along which for the moment the child and the her child’s arm; if the child wants to im- adult agree, for example, to treat a pencil itate this behavior, she is faced with a as if it were a horse. choice that depends on how she inter- 18 Tomasello, The Cultural Origins of Human prets her mother’s action. Thus, she Cognition.

Dædalus Winter 2004 57 Michael Given the general ability to learn a meek a different character themselves, Tomasello on symbol through role reversal imitation, children performed the action that the learning it is still the case that in learning particu- adult had marked as intentional. lar words on particular occasions chil- Like the studies of actions on objects, dren often need to read the adults inten- these word learning studies provide evi- tions in order to connect the word ap- dence that at a very early age children propriately to its intended referent. come to understand intentional action. Several language acquisition studies And human learning is what it is–name- show that children as young as eighteen ly, cultural learning–because human months can combine all of the types of beings, even when quite young, are intention reading we have discussed able to understand the intentional and above while imitatively learning nov- mental states of other human beings. el words. For example, in a study of Through this understanding, cultural twenty-four-month-olds, an adult an- processes take human cognition in some nounced her (prior) intention to ½nd a directions not possible in other species– target object by saying, “Let’s go ½nd the and make human cognition an essential- toma.”19 She searched through several ly collective enterprise. buckets, extracting and rejecting with a scowl the novel objects inside. She then extracted another novel object with an excited expression and stopped search- ing. In a later comprehension test, when asked to go get the toma themselves, children chose the object the adult had identi½ed as ful½lling her intention. This experiment used a modi½ed procedure to show that twenty-four-month-old children could identify the intended ref- erent even when the adult was unable to open the container with the target object inside–that is, when she had an unful- ½lled intention. Another study investi- gated children’s use of their understand- ing of intentional versus accidental ac- tions when learning novel words. In a study of twenty-four-month-olds, the adult announced her (prior) intention to perform a target action by saying, “I’m going to meek Big Bird!” She then per- formed, in counterbalanced order, one accidental action, which she verbalized by saying “Woops!” and one intentional action, which she indicated by saying “There!” Later, when they were asked to 19 Michael Tomasello and Michelle Barton, “Learning Words in Non-ostensive Contexts,” Developmental Psychology 30 (1994): 639–650.

58 Dædalus Winter 2004 Susan Carey

Bootstrapping & the origin of concepts

All animals learn. But only human ysis and our capacity to reason about beings create scienti½c theories, mathe- the mental states of others. Each of matics, literature, moral systems, and these factors doubtless contributes complex technology. And only humans to our prodigious ability to learn. have the capacity to acquire such cultur- But in my view another factor is even ally constructed knowledge in the nor- more important: our uniquely human mal course of immersion in the adult ability to ‘bootstrap.’ Many psycholo- world. gists, historians, and philosophers of There are many reasons for the differ- science have appealed to the metaphor ences between the minds of humans of bootstrapping in order to explain and other animals. We have bigger learning of a particularly dif½cult sort– brains, and hence more powerful infor- those cases in which the endpoint of the mation processors; sometimes differ- process transcends in some qualitative ences in the power of a processor can way the starting point. The choice of create what look like qualitative differ- metaphor may seem puzzling–it is self- ences in kind. And of course human evidently impossible to pull oneself up beings also have language–the main by one’s own bootstrap. After all, the medium for the cultural transmission process I describe below is not impos- of acquired knowledge. Comparative sible, but I keep the term because of studies of humans and other primates its historical credentials and because it suggest that we differ from them as well seeks to explain cases of learning that in our substantive cognitive abilities– many have argued are impossible. for example, our capacity for causal anal- Sometimes learning requires the cre- ation of new representational resources Susan Carey, professor of psychology at Harvard that are more powerful than those pres- University, has played a leading role in transform- ent at the outset. Early in the cultural ing our understanding of . history of mathematics, for instance, A Fellow of the American Academy since 2001, the concept of the number included only she is the author of numerous articles and essays positive integers: with subsequent de- and the book “Conceptual Change in Childhood” velopment the concept came to encom- (1985). pass zero, rational numbers (fractions), negative numbers, irrational numbers © 2004 by the American Academy of Arts like pi, and so on. & Sciences

Dædalus Winter 2004 59 Susan Bootstrapping is the process that un- in solving it, I will examine how children Carey derlies the creation of such new con- acquire one speci½c set of concepts: the on learning cepts, and thus it is part of the answer to positive integers–i.e., concepts such as the question: What is the origin of con- one, two, three, nine, eighteen, etc. cepts? Individual concepts are the units of Before they acquire language, infants thought. They are constituents of larger form several different types of represen- mental structures–of beliefs that are tation with numerical content, at least formed out of them and of systems of two of which they share with other ver- representation such as intuitive theories. tebrate animals. Concepts are individuated on the basis One, described by Stanislas Dehaene of two kinds of considerations: their ref- in his delightful book The Number Sense, erence to different entities in the world uses mental symbols that are neural and their role in distinct mental systems magnitudes linearly related to the num- of inferential relations. ber of individuals in a set. Because the How do human beings acquire con- symbols get bigger as the represented cepts? Logic dictates three parts to any entity gets bigger, they are called analog explanation of the origin of concepts. magnitudes. Figure 1 gives an external First, we must specify the innate repre- analog magnitude representation of sentations that provide the building number, where the symbol is a line, and blocks of the target concepts of interest. length is the magnitude linearly related Second, we must describe how the target to number. Mental computations using concepts differ from these innate repre- these symbols include comparison, to sentations–that is, we must describe de- establish numerical difference or equal- velopmental change. And third, we must ity, and also addition and subtraction. characterize the learning mechanisms Mental analog magnitudes represent that enable the construction of new con- many dimensions of experience–for cepts out of the prior representations. example, brightness, loudness, and tem- Claims about all three parts of the ex- poral duration. In each case as the physi- planation of the origin of concepts are cal magnitudes get bigger, it becomes highly controversial. Many believe that increasingly harder to discriminate be- innate representations are either percep- tween pairs of values that are separated tual or sensory, while others (including by the same absolute difference. You can myself ) hold that humans and other ani- see in ½gure 1 that it is harder to tell that mals are endowed with some innate rep- the symbol for seven is different from resentations with rich conceptual con- (and smaller than) that for eight than it tent. Some researchers also debate the is to tell that the symbol for two is dif- existence, even the possibility, of quali- ferent from (and smaller than) that for tative changes to the child’s initial repre- three. Analog magnitude representa- sentations. One argument for the impos- tions follow Weber’s law, according to sibility of such radical changes in the which the discriminability of two values course of development is the putative is a function of their ratio. lack of learning mechanisms that could You can con½rm for yourself that you explain them. This is the gap that my have an analog magnitude system of rep- appeal to bootstrapping is meant to ½ll. resentation of number that conforms to To make clear both what the problem Weber’s law. Tap out as fast as you can is, and what role bootstrapping may play without counting (you can prevent your-

60 Dædalus Winter 2004 Figure 1 dots. After habituation they were pre- Bootstrapping Analog magnitude models sented with new displays containing ei- & the origin Number represented by a quantity linearly related ther the same number of dots to which of concepts to the cardinal value of the set they had been habituated or the other one: –– number. Xu and Spelke found that the infants recovered interest to the new two: –––– number, and so concluded that they are capable of representing number. Xu and three: –––––– Spelke also found evidence for Weber’s law: infants could discriminate eight seven: –––––––––––––– from sixteen and sixteen from thirty- two, but not eight from twelve or six- eight: ––––––––––––––– teen from twenty-four.1 Infants and animals can form analog self from counting by thinking ‘the’ with magnitude representations of fairly large each tap) the following numbers of taps: sets, but these representations are only 4, 15, 7, and 28. If you carried this out approximate. Analog magnitude repre- several times, you’d ½nd the mean num- sentations of number fall short of the ber of taps to be 4, 15, 7, and 28, with the representational power of integers; in range of variation very tight around 4 this system one cannot represent exactly (usually 4, occasionally 3 or 5) and very ½fteen, or ½fteen as opposed to fourteen. great around 28 (from 14 to 40 taps, for Nonetheless, analog magnitude repre- example). Discriminability is a function sentations clearly have numerical con- of the absolute numerical value, as dic- tent: they refer to numerical values, tated by Weber’s law. Since you were not and number-relevant computations counting, some other numerical repre- are de½ned over them. sentation must have been guiding your tapping performance–presumably ana- A second system of representations log magnitudes, as your adherence to with numerical content works very Weber’s law, again, would seem to indi- differently. Infants and nonhuman pri- cate. mates have the capacity to form sym- Space precludes my reviewing the ele- bols for individuals and to create men- gant evidence for analog magnitude rep- tal models of ongoing events in which resentations of number in animals and each individual is represented by a single human infants, but let me give just one symbol. Figure 2 shows how, in this sys- example. Fei Xu and tem, sets of one, two, or three boxes showed infants arrays of dots, one dot might be represented. The ½gure repre- array at a time, until the infants got sents three different possibilities for bored with looking at them. All other the format and content of the symbols. variables that could have been con- 1 For an overview of the evidence for analog founded with number (total array size, magnitude representations of number in both total volume of dots, density of dots, nonhuman animals and human adults, see and so on) were controlled in these stud- Stanislas Dehaene, The Number Sense (Oxford: Oxford University Press, 1997). For evidence ies, such that the only possible basis for in human infants, see Fei Xu and Elizabeth S. the infants’ discrimination was numeric. Spelke, “Large Number Discrimination in 6- Seven-month-old infants were habituat- Month-Old Infants,” Cognition 74 (2000): ed either to arrays of eight or sixteen B1–B11.

Dædalus Winter 2004 61 Susan There is one symbol for each box, so Figure 2 Carey Parallel individuation models on number is implicitly represented; the learning symbols in the model stand in one-one correspondence with the objects in the Number Image Abstract Speci½c of boxes world. To give you a feel for the evidence that infants indeed employ such models, dis- 1 box  obj box tinct from the analog magnitude repre- sentations sketched above, consider the 2 boxes  obj obj box box following experiment from my laborato-  ry. Ten- to fourteen-month-old infants 3 boxes obj obj obj box box box are shown a box into which they can reach to retrieve objects, but into which Note: one symbol for each individual; no symbols for they cannot see. If you show infants integers. three objects being placed, one at a time numerical comparison is a function or all at once, into this box, and then of the ratios of the numbers being com- allow them to reach in to retrieve them pared, and that the representations can one at a time, they show by their pattern handle sets of objects at least as big as of reaching that they expect to ½nd ex- thirty-two. But in this reaching task, actly three objects there. If the infant infants succeed at ratios of 2:1 and 3:2, has a mental representation of a set of but fail at 4:2 and even 4:1; as soon as two objects (e.g., object, object) that are the set exceeds three, infants cannot hidden from view, and the infant sees a hold a model of distinct items in their new object being added to the set, the short-term memory.2 infant creates a mental representation of a set of three (object, object, object). In sum, human infants (and other pri- Further, computations of one-one corre- mates) are endowed with at least two spondence carried out over these models distinct systems of representation with allow the child to establish numerical numerical content. Both take sets of in- equivalence and number order (e.g., dividuals as input. One creates a summa- Have I got all the objects out of the box ry analog representation that is a linear or are there more?) function of the number of individuals So far, this is just another demon- in the set. This process is noisy, and the stration that infants represent number. noise is itself a linear function of the set However, an exploration of the limits size, with the consequence that the rep- on infants’ performance of this task im- resentations are merely approximate. plicates a different system of representa- For several reasons, this system is too tion from the analog magnitude system weak to represent the positive integers. sketched above. For one, there is likely an upper bound to Performance breaks down at four the set sizes that can be represented by objects. If the infants see four objects analog magnitudes. More importantly, being placed into the box and are al- lowed to retrieve two of them, or even 2 For evidence of the set-size limits on in- just one of them, they do not reach per- fants’ representations of small numbers of objects, see Lisa Feigenson and Susan Carey, sistently for the remaining objects. Re- “Tracking Individuals Via Object Files: Evi- member that in the analog magnitude dence from Infants’ Manual Search,” Develop- system of representation, success at mental Science 6 (2003): 568–584.

62 Dædalus Winter 2004 animals and infants cannot discriminate symbols in the list refer to cardinal val- Bootstrapping adjacent integer values once the sets ues exactly one apart: 5 is 4 plus 1, 6 is 5 & the origin of concepts contain more than three or four individ- plus 1, and so on. uals; that is, they cannot represent exact- I have argued so far that the count-list ly ½fteen or twenty-½ve or forty-nine, or representation of number transcends any other large exact integer. Finally, the representational power of both of analog magnitude representations the representational systems with nu- obscure one of the foundational rela- merical content that are available to pre- tions among successive integers–that verbal infants, for these precursors lack each one is exactly one more than the the capacity to represent integers. If this one before. It is this relation, called the is so, it should be dif½cult for children to successor relation, that underlies how come to understand the numerical func- counting algorithms work and provides tion of counting. the mathematical foundation of integer And so, indeed, it is dif½cult for chil- concepts. Since discriminabilty of ana- dren to learn how counting represents log magnitudes is a function of the ratio number, and details about the partial between them, the relation between two understanding they achieve along the and three is not experienced as the same way constrain our theories of the learn- as that between twenty-four and twenty- ing process. In the United States (and ½ve; indeed, the latter two values cannot every other place where early counting really even be discriminated within this has been studied, including Western system of representation. Europe, Russia, China, and Japan) The second system–one symbol for children learn to recite the count list each individual–falls even shorter as a as young two-year-olds, and at this age representation of integers. There are no can even engage in the routine of count- symbols for number in this system at all; ing–touching objects in a set one by one the symbols in ½gure 2 each represent an as they recite the list. But it takes anoth- individual object, unlike those in ½gure er year and a half before they work out 1, which represent an approximate cardi- how counting represents number, and nal value. Furthermore, what can be rep- in every culture yet studied, children go resented in this system is limited in through similar stages in working out number to sets of one, two, and three. the meanings of the number words in the count list. The count list (‘one, two, three . . .’) is First, children learn what ‘one’ means a system of representation that has the and take all other words in the list to power to represent the positive integers, contrast with ‘one,’ meaning ‘more than so long as it contains a generative sys- one’ or ‘some.’ The behaviors that dem- tem for creating an in½nite list. When onstrate this are quite striking. If you deployed in counting, it provides a rep- present young two-year-olds with a pile resentation of exact integer values based of pennies and ask them to give you one on the successor function. That is, when penny, they comply. If you ask for two applied in order, in one-one correspon- pennies or three pennies or ½ve pennies, dence with the individuals in a set, the they grab a bunch, always more than ordinal position of the last number word one, and hand them over. They do in the count provides a representation not create a larger set for ‘½ve’ than for of the cardinal value of the set–of how ‘two.’ You might suppose that the plural many individuals it contains. Successive in ‘pennies’ is doing the work here, but

Dædalus Winter 2004 63 Susan the same phenomenon is observed in and then to ½x the set, counters invari- Carey on China and Japan, even though Chinese ably adjust the set in the right direction, learning and Japanese do not have a singular- taking an object away if the set is too plural distinction, and also in the large or adding one if it is too small. United States when the contrast is One-, two-, and three-knowers, in con- between ‘one ½sh,’ ‘two ½sh,’ and trast, almost always add more to the ‘½ve ½sh.’ set–even if they had counted to ½ve or Let us call children at this stage of six or seven when they were checking working out the meanings of number whether it had four–con½rming that words ‘one-knowers.’ Many other tasks they really do not understand how provide additional evidence that one- counting determines the meaning of knowers truly know only the meaning number words. of the word ‘one’ among all the words in These data suggest that the partial their count list. For example, if you ask a meanings of number words seem to be one-knower to tell you what’s on a series organized initially by the semantics of of cards that contain one, two, or three quanti½ers–the singular-plural distinc- ½sh (up to eight ½sh), they say ‘a ½sh’ or tion and the meanings of words like ‘one ½sh’ for the card with one, and ‘two ‘some’ and ‘a.’ If this is right, then we ½sh’ or ‘two ½shes’ or ‘two ½shies’ for might expect that children learning lan- all of the other cards. This again indi- guages with quanti½er systems that cates a single cut between the meaning mark numerical contrasts differently of ‘one,’ which they grasp, and words for from English would entertain different the number of individuals in larger sets, hypotheses concerning the partial mean- which they do not. ings of number words. They might break After having been one-knowers for into the system differently. And indeed about six to nine months, children learn they do. what ‘two’ means. At this point they can Consider ½rst classi½er languages correctly give you two objects if you ask such as Chinese and Japanese that do for ‘two,’ but they still just grab a bunch not mark the distinction between singu- (always greater than two), if you ask for lar and plural in nouns, verbs, or adjec- ‘three,’ ‘four,’ ‘½ve,’ or ‘six.’ After some tives. Two independent studies have months as two-knowers, they become found that although children in China three-knowers, and some months later and Japan learn the count list as young induce how counting works. as English-speaking children do, they The performance of children who have become one-knowers several months worked out how counting works is quali- later and are relatively delayed at each tatively different from that of the one-, stage of the process. Conversely, Russian two-, and three-knowers in a variety of has a complex plural system in which the ways that reflect the conceptual under- morphological markers for sets of two, standing of counting. three, and four differ from those for ½ve To give just one example, in the task in through ten. Two independent studies which children are asked to give the ex- have shown that even Russian one- and perimenter a certain number of items, two-knowers distinguish between the say four, one-, two-, and three-knowers meanings of the number words ‘two,’ usually give the wrong number, and the ‘three,’ and ‘four,’ on the one hand, and young counters also sometimes make an ‘½ve,’ ‘six,’ ‘seven,’ and ‘eight,’ on the error. When asked to check by counting other. Unlike the one- and two-knowers

64 Dædalus Winter 2004 described above, Russian children in the have the power to represent natural Bootstrapping early stages of working out how count- numbers either. & the origin of concepts ing works grab smaller sets when asked to give the experimenter ‘two,’ ‘three,’ The problem of the origin of the posi- or ‘four’ than when asked to give the ex- tive integers arises at two different time perimenter ‘½ve’ or more, and use larger scales–historical and ontogenetic. At numbers for larger sets in the what’s-on- the dawn of modern anthropology, this-card task.3 when colonial of½cers went out into These phenomena concerning young the French and English colonial worlds, children’s partial understanding of the they discovered many systems of explicit meanings of number words support number representation that fell short of three interrelated conclusions. First, a full representation of natural number. that it is so dif½cult for children to learn They described languages that marked what ‘two’ means, let alone what ‘½ve’ number on nouns, adjectives, and verbs, and ‘eight’ mean, lends support to the and which had quanti½ers like the Eng- claims that preverbal number represen- lish ‘one,’ ‘two,’ ‘many,’ ‘some,’ ‘each,’ tations are not representations of inte- ‘every,’ and ‘more,’ but which had no gers, at least not in the format of an inte- count list. In this vein, the psychologist ger list. Young children–for a full six to Peter Gordon has described the language nine months before they work out what of the Piraha, an isolated Amazonian ‘two’ means, and a full year and a half people. He has shown that in addition before they work out how the count to linguistic quanti½ers meaning ‘one,’ list represents integers–know how to ‘two,’ and ‘many,’ the Piraha also have count, know what ‘one’ means, and access to the nonverbal systems de- know that ‘two,’ ‘three,’ ‘four,’ ‘½ve,’ scribed above (parallel individuation ‘six,’ ‘seven,’ and ‘eight’ represent num- of small sets and analog magnitude rep- bers larger than ‘one.’ Second, coming resentations of large numerosities). Gor- to understand how the count list repre- don con½rms that they have no repre- sents numbers reflects a qualitative sentations of large exact numerical val- change in the child’s representational ues. capacities; I would argue that it does Anthropologists and archeologists nothing less than create a representa- have described intermediate systems of tion of the positive integers where none integer representation, short of integer was available before. Finally, a third pos- lists, and these intermediate systems sible developmental source of natural provide evidence for a process of cultur- number representations, in addition to al construction over generations and the preverbal systems described above, centuries of historical time.4 Here I con- may be the representations of numbers centrate on ontogenetic time. How do within natural language quanti½er se- three-year-olds do it? How do they cre- mantics. Of course, natural language ate a representational system with more quanti½ers, other than the number 4 I would recommend the linguist James Hur- words in the count list itself, do not ford’s review of this literature to any reader in- terested in this process. James Hurford, Lan- 3 For a characterization of the early stages of guage and Number (Oxford: Basil Blackwell, counting in English, see , “Chil- 1987). For work on the Piraha, see Peter Gor- dren’s Acquisition of the Number Words and don, “The Role of Language in Numerical Cog- the Counting System,” 24 nition: Evidence from Amazonia” (under re- (2) (1992): 220–257. view).

Dædalus Winter 2004 65 Susan power than any on which it is built?5 As described above, the child learns Carey on In answering this question, I would the meanings of the ½rst number words learning appeal to bootstrapping processes. as natural language quanti½ers. Children Bootstrapping processes make essen- learn the meaning of ‘one’ just as they tial use of the human capacity for creat- learn the meaning of the singular deter- ing and using external symbols such as miner ‘a’ (indeed, in many languages, words and icons. Bootstrapping capital- such as French, they are the same lexi- izes on our ability to learn sets of sym- cal item). bols and the relations among them di- Some months later, ‘two’ is learned, rectly, independently of any meaning just as dual markers are in languages that assigned to them in terms of anteced- have singular/dual/plural morphology. ently interpreted mental representa- Languages with dual markers have a dif- tions. These external symbols then ferent plural af½x for sets of two than the serve as placeholders, to be ½lled in af½x for sets greater than two. It is as if with richer and richer meanings. The English nouns were declined ‘box’ (sin- processes that ½ll the placeholders gular), ‘boxesh’ (dual), ‘boxeesh’ (plu- create mappings between previously ral). In this system, the suf½x ‘esh’ separate systems of representation, would apply just when the set referred to drawing on the human capacity for contained exactly two items. By hypoth- analogical reasoning and inductive in- esis, children would learn the meaning ference. The power of the resulting sys- of the word ‘two’ just as they would tem of concepts derives from the com- learn the morphological marker ‘esh’– bination and integration of previously if English plural markers worked that distinct representational systems. way. By extension, some months later, Let’s see how this might work in the ‘three’ is learned just as trial markers present case. We must allow the child are in the rare languages that have sin- one more prenumerical capacity–that gular/dual/trial/plural morphology. of representing serial order. This is no In the early stages of being a one-, problem–young children learn a vari- two-, or three-knower, the child repre- ety of meaningless ordered lists, such sents other number words as quanti½ers, as ‘eeny, meeny, miney, mo.’ meaning ‘many,’ where ‘many’ is more We seek to explain how the child than any known number word. As I will learns the meanings of the number argue below, it is likely that the nonver- words–what ‘two’ means, what ‘seven’ bal number representations that support means–and how the child learns how the meanings of the known words is the the list itself represents number–that system of parallel individuation (½gure the cardinal value of a set enumerated 2), with natural language quanti½cation by counting is determined by the order articulation in terms of notions like ‘set’ on the list, and that successive numbers and ‘individual.’ on the list are related by the arithmetic Meanwhile, the child has learned the successor relation. count list, which initially has no seman- tic content other than its order. The 5 In a forthcoming book I argue that the same child knows one must recite ‘one, two, bootstrapping process underlying this marvel- ous feat in childhood also accounts for the de- three, four, ½ve,’ not ‘two, three, one, velopment in historical time, but that argument ½ve, four,’ just as one must say ‘a, b, c, d, is beyond the scope of this brief paper. e,’ not ‘c, a, e, d, b.’

66 Dædalus Winter 2004 The stage is now set for a series of would be empirically correct, at least for Bootstrapping mappings between representations. some children. We do know that chil- & the origin of concepts Children may here make a wild analo- dren come to integrate their integer list gy–that between the order of a particu- with analog magnitudes, such that ‘½ve’ lar quantity within an ordered list, and comes to mean both ‘one more than that between this quantity’s order in a four, which is one more than three. . .’ series of sets related by additional indi- and ‘––––––––––,’ the analog magnitude viduals. These are two quite different symbol for the cardinality of a set of ½ve bases of ordering–but if the child recog- individuals. This integration is undoubt- nizes this analogy, she is in the position edly very important; bootstrapping pro- to make the crucial induction: For any vides richer representations precisely word on the list whose quanti½cational through integration of previously dis- meaning is known, the next word on the tinct systems of representation. list refers to a set with another individ- As important as the integration of the ual added. Since the quanti½er for single integer list representation with analog individuals is ‘one,’ this is the equivalent magnitude representations may be, to the following induction: If number there is good reason to believe that this word X refers to a set with cardinal value integration is not part of the bootstrap- n, the next number word in the list re- ping process through which the concept fers to a set with cardinal value n + 1. of positive integers is ½rst understood. This bootstrapping story provides dif- Research suggests that it is not until after ferent answers for how the child learns children have worked out how the count the meaning of the word ‘two’ than for list represents number–in fact some six how she learns the meaning of ‘½ve.’ months later–that they know which According to the proposal, the child analog magnitudes correspond to which ascertains the meaning of ‘two’ from numbers above ½ve in their count list. the resources that underlie natural lan- That ½nding–along with the fact that guage quanti½ers, and from the system the precise meanings of number words of parallel individuation, whereas she are learned in the order ‘one,’ then ‘two,’ comes to know the meaning of ‘½ve’ then ‘three,’ followed by the induction through the bootstrapping process– of how the count list works–leads me to i.e., that ‘½ve’ means ‘one more than favor the bootstrapping proposal above. four, which is one more than three . . .’ I doubt that anybody would deny that –by integrating representations of natu- language helps us occupy the distinctive ral language quanti½ers with the exter- cognitive niche that we human beings nal serial ordered count list. enjoy. It is obvious that culturally con- structed knowledge is encoded in lan- I began by sketching two systems of guage and can then be passed on to new preverbal representation with numerical generations through verbal communica- content: the analog magnitude system tion–you can tell your children some- and the system of parallel individuation. thing, saving them from having to dis- You may have noticed that the analog cover it themselves. Still, this account magnitude system played no role in my misses the equally obvious point that bootstrapping story. It would be quite children are often unable to understand possible to imagine a role for this system what we tell them, because they lack the in a slightly different bootstrapping pro- concepts that underlie our words. The posal, and it may be that such a proposal problem then becomes accounting for

Dædalus Winter 2004 67 Susan how they acquire the relevant concepts Carey on they need to understand what we are learning telling them. I have argued that bootstrapping mechanisms provide part of the solution to this problem. In thinking about how bootstrapping might work, we are led to a fuller appreciation of the role of language in supporting the cultural transmission of knowledge. We cannot just teach our children to count and ex- pect that they will then know what ‘two’ or ‘½ve’ means. Learning such words, even without fully understanding them, creates a new structure, a structure that can then be ½lled in by mapping rela- tions between these novel words and other, familiar concepts. And so eventu- ally our children do know what ‘½ve’ means: through the medium of language and the bootstrapping process sketched here they have acquired a new concept.

68 Dædalus Winter 2004 Clark Glymour

The automation of discovery

Scienti½c revolutions are sometimes infrared spectrum), a computer iden- quiet. Despite a lack of public fanfare, ti½ed minerals in rocks from a Califor- there is mounting evidence that we are nia desert lake as accurately as had a in the midst of such a revolution–pre- team of human experts at the site who mised on the automation of scienti½c had access both to the spectra and to discovery made possible by modern the actual rocks; computers and new methods of acquir- • In Antarctica, a robot traversing a ½eld ing data. of ice and stones picked out the rare Consider, for example, the following meteorites from among the many developments: rocks; • Using data from the 1970s, about eight • Scientists at the Swedish Institute for years ago a team of data analysts work- Space Physics realized that an instru- ing in Holland predicted that low-level ment aboard a satellite was malfunc- lead exposure is more dangerous to tioning and they recalibrated it from children’s cognitive development than Earth; had previously been thought–a pre- diction con½rmed by recent reanalyses • An economist working for the World of later observations; Food Organization found that current foreign aid practices have no impact on • Using measurements of reflected solar extreme poverty; energy (technically, the visible-near • Climate researchers traced the global increase in vegetation and its causes Clark Glymour is Alumni University Professor of over the last twenty years; Philosophy at Carnegie Mellon University, and Senior Research Scientist at the Institute for Hu- • A team of biologists and computer sci- man and Machine Cognition and John Pace Emi- entists reported determinations of the nent Scholar at the University of West Florida. genes in yeast whose function is regu- He is the author or coauthor of numerous articles lated by any of a hundred regulator and books, including “The Mind’s Arrows” genes; (2001), “Computation, Causation and Discov- • A kidney transplant surgeon measured ery” (1999), and “Thinking Things Through” the behavior of rat genes that had been (1995). aboard the space shuttle;

• A biologist reported a determination of © 2004 by the American Academy of Arts (possibly) all of the human genes in & Sciences Dædalus Winter 2004 69 Clark cells lining the blood vessels that re- of the scienti½c community.2 We can see Glymour on spond to changes in liquid flow across the revolution at work by looking more learning the cells.1 closely at three of the examples I men- All of these developments–and they tioned above. are simply more or less random exam- ples I happen to know–reflect a new Lead was long a component of paint, way of learning about the world. Thanks and the Mobil Oil Company introduced to innovations in computer software, in tetraethyl lead into gasoline in the 1930s. laboratory techniques, and in observa- From these and other sources, low-level tional technology, scientists today can lead exposure became common in the measure things on a scale inconceivable United States and elsewhere. Large only a few years ago. New laboratory doses of lead and other heavy metals and computational methods allow evalu- were known to disrupt mental faculties, ation of vast numbers of hypotheses in but the effects of low-level exposure order to identify those few that have a were unknown. Besides, low-level expo- reasonable chance of being true, and sure was hard to measure: low-level lead simple oversights of human judgment concentrations fluctuate in blood and do can be corrected by computer. The not indicate how much lead the body has change is from the textbook scienti½c absorbed over time. paradigm in which one or a very few hy- In the 1970s, Herbert Needleman potheses are entertained and tested by a found an ingenious way to measure cu- very few experiments, to a framework in mulative lead exposure using the lead which algorithms take in data and use it concentration in children’s baby teeth. to search over many hypotheses, as ex- He also measured the children’s iq perimental procedures simultaneously scores and many family and social vari- establish not one but many relation- ables that might conceivably be relevant ships. While there are consequences to the children’s cognitive abilities. Re- even for small collections of data, the viewing the data by analysis of variance, automation of scienti½c inquiry is a standard statistical technique intro- chiefly driven by novel abilities to duced early in the twentieth century, acquire, store, and access previously Needleman concluded that lead expo- inconceivable amounts of data, far too sure has a small but robust effect–it much for humans to survey by hand lowers children’s iq scores. and eye. Methodology has moved in Since a lot of money was at stake, crit- consequence; in a growing number of icism naturally followed, and in 1983 a ½elds, automated search and data selec- scienti½c review panel formed by Ron- tion methods have become indispensa- ald Reagan’s Environmental Protection ble. Agency asked Needleman to reanalyze This may not seem revolutionary, but the data with stepwise regression, a it has all of the earmarks of scienti½c more modern statistical technique. The revolution that Thomas Kuhn empha- idea behind this technique is very simple sized years ago: novel results, novel even if the mathematics is not: Suppose kinds of theory, novel problems, intense any of several measured variables might and often irrational hostility from parts 2 Thomas S. Kuhn, The Structure of Scienti½c 1 References can be found at . Press, 1996).

70 Dædalus Winter 2004 influence iq scores. But start with the interesting theorem that implied that for The Needleman’s data, with the assumptions automation assumption that none of the variables of discovery influence iq. Change that assumption about measurement error, the true effect by entertaining as a causal factor which- of lead exposure on cognitive ability ever variable is most highly correlated could be positive or negative or zero. with iq score, then keep adding causal The elimination of lead from gasoline, it factors by a mathematical measure that seemed, had been based on a statistical takes account of the correlation already mistake.3 The story doesn’t end here, explained by previously considered fac- however. But before continuing, a di- tors. Stop when additional variables gression into the statistics of causality don’t explain anything more. (This pro- is necessary. cedure can also be run in reverse, start- ing with the assumption that all of the In the early 1980s, several statisticians measured variables influence iq scores, developed a network representation of and then throwing out the least explana- probability relations that formalized and tory factors, one by one.) Needleman generalized ideas that had been used for had measured about forty variables that a long while in biology, social science, might account for variations in his sub- and elsewhere. According to their repre- jects’ iq scores, and stepwise regression sentation, suppose we have data for a eliminated all but six of them. Lead ex- number of variables, each of which takes posure remained among the causal fac- a de½nite value in each individual object tors, and using a standard method (in- or case (the variables might be height, deed, the oldest method in statistics, weight, ratio of Democrats to Republi- originating with Legendre’s essay on cans, whatever; the individual objects, comets in 1808) to estimate the depen- or cases, could be people, rats, cells, state dence of iq score on lead exposure, Nee- governments, whatever). Represent each dleman again found a small negative variable as a node and draw arrows from effect. some nodes to other nodes, e.g., C  B  Many years after the con½rmation of A. This particular diagram represents the Needleman’s results had helped to elimi- claim that the information that the val- nate lead from gasoline, two economists, ues of A and B together provide about Stephen Klepper and Mark Kamlet, re- the value of C is the same as the informa- analyzed Needleman’s data–with a dif- tion that the value of B provides all by ference. Reasonably, they assumed that itself. And, symmetrically, the informa- the measured values Needleman report- tion that values of C and B provide about ed were not perfectly accurate: iq scores A is the same as the information that the did not perfectly measure cognitive abil- value of B alone provides. ity; lead concentrations in teeth did not perfectly measure lead exposure; and so 3 Herbert Needleman et al., “De½cits in Psy- chologic and Classroom Performance of Chil- on. Each of Needleman’s six remaining dren with Elevated Dontine Lead Levels,” New variables perhaps influenced cognitive England Journal of Medicine 300 (1979): 389; ability, but the true values of those vari- Needleman et al., “Lead and iq Scores: A Re- ables were not recorded in his data. The analysis,” Science 227 (1985): 701–704; Stephen data consisted of measurements pro- Klepper et al., “Regressor Diagnostics for the Errors-in-Variables Model–An Application to duced by the true value of each variable the Health Effects of Pollution,” Journal of Envi- for each child, and also by unknown ronmental Economics and Management 24 (1993): measurement errors. Klepper proved an 190–211.

Dædalus Winter 2004 71 Clark In other words, you can use the dia- also hold in a related way if the arrows Glymour on gram described above when, for predict- represent causal relations, and, relying learning ing C, if you know the value of B, then on the Markov condition, they gave a the value of A doesn’t tell you anything general characterization of the relation more about the probabilities of the val- between network structure, probabili- ues of C. In more technical terms, C is ties, and causal claims. independent of A conditional on B. (C The idea is easiest to see for interrup- would also be independent of A condi- tions of a simple causal chain. For in- tional on B if the structure were C  B  stance, if pushing the doorbell button A or C  B  A, but not if it were C  B  A causes the bell to ring, which in turn or B  C  A, etc.) The general version of causes the house parrot to say “hello,” this connection between networks and then if you intervene to keep the bell probabilities, known as the Markov con- from ringing, pushing or not pushing dition, was introduced explicitly by stat- the doorbell button will not change the isticians around 1980, though it was used probability that the parrot says “hello.” implicitly in many subjects long before After your intervention, the state of the that time, and almost formalized by button and the state of the parrot will be the philosopher Hans Reichenbach in independent of each other; neither will the 1950s. Without clearly formulating provide information about the other. But the general idea, biologists, psycholo- if you do not intervene to disconnect the gists, sociologists, and even biblical his- bell, pushing the button will be inde- torians had used such diagrams to repre- pendent of the parrot’s speech condi- sent causal hypotheses and the probabil- tional on the state of the bell, ringing ity relationships of their variables.4 In or not ringing; if you know whether the 1980s a group at ucla, led by Judea the bell is ringing, the parrot’s speech Pearl, developed a fast algorithm for won’t give you any more information computing any conditional indepen- as to whether someone is at the door. In dence relations implied by the Markov many cases, the independence relations condition when applied to such a dia- produced by interventions in a system gram, now called a Bayes net. parallel the conditional independence In the early 1990s a group of philoso- relations implied by the network repre- phers and statisticians at Carnegie Mel- sentation of the causal structure of the lon noted that many of the information system. restrictions, or conditional indepen- These connections between causation, dence facts, represented in a network probability, and network representations suggested that with appropriate assump- 4 See William Farmer, The Synoptic Problem tions and background knowledge, some- (Macon, Ga.: Mercer University Press, 1981) thing about the causal structure can be and, for a more recent discussion, Donald learned from observation, and the out- Akenson, Surpassing Wonder: The Invention of the Bible and the Talmuds (Montreal: McGill- comes of some ideal interventions can Queen’s University Press, 1998). The statistical be predicted. If C and A are independent work is reported and illustrated in Harry Kiiveri conditional on B, and no other indepen- and Terry Speed, “Structural Analysis of Multi- dence relation holds, then C and A are variate Data: A Review,” in Samuel Leinhardt, causally connected only through B, ed., Sociological Methodology (San Francisco: Jossey-Bass, 1982) and in , Probabi- which functions either as a common listic Reasoning in Intelligent Systems (San Mateo, cause or an intermediary. Inferences Calif.: Morgan Kaufmann Publishers, 1988). like this readily combine with other

72 Dædalus Winter 2004 information one might have–for exam- grams ‘converge’ to giving correct infor- The automation ple, if the same probability relations mation about the causal structure be- of discovery hold and B occurs before A and C, then hind the data.6 the causal structure should be A  B  C.5 The old shibboleth that correlation does Back to lead. In collaboration with not imply causation is true for any pair Dutch statisticians, Richard Scheines, of variables considered in isolation, but, one of the Carnegie Mellon researchers, when combined with otherwise routine applied a program implementing these assumptions, is not necessarily true for new search techniques to Needleman’s sets of correlations among several vari- data. ables. The problem is to say in a mathe- What the program found was simple matically precise and useful way just but astonishing: three of the six predic- what causal information can be extract- tion variables that had remained after ed from such dependencies, and under Needleman’s stepwise regression had what assumptions. no correlation with iq scores–a fact The class of alternative networks that that had somehow eluded Needleman, might conceivably describe the causal his collaborators, his critics, and, in- relations among a set of variables, before deed, the stepwise regression procedure data is collected, is astronomical even alike. Of the initial variables possibly for small numbers of variables, and with correlated with iq that Needleman had larger numbers of variables remains ½rst considered, only lead and two other huge even if some of the variables are factors now remained. But, with the ordered so that one can assume that later economists’ assumptions about meas- variables do not influence earlier ones. urement error, the effect of lead expo- Even so, early in the 1990s, researchers sure on iq still could not be estimated. at the , Carnegie To estimate the effect of lead, Scheines Mellon, ucla, and Microsoft developed and his Dutch collaborators resorted to a algorithms and software for searching relatively new technique in Bayesian sta- for the class of diagrams that can ac- tistics. Bayesian statistics proceeds by count for any set of independence rela- assigning ‘prior probabilities’ to alterna- tions among variables. Since then many tive hypotheses, by computing for each related algorithms have been proposed hypothesis the probability of the data on and applied by others. These procedures the assumption that that hypothesis is search ef½ciently for information within true, and, from all this, computing a the huge space of alternative possible new, or ‘posterior,’ probability for each causal structures, but, unlike stepwise hypothesis or range of parameters con- regression, some of these procedures sidered. For a long time, because the come with a weak guarantee of reliabili- posterior probabilities often could not ty. For example, as the size of the sample be computed, Bayesian statistics was increases, according to the Markov con- chiefly a toy used only for simple prob- dition and one other further technical lems; computational developments in assumption, the Bayes net search pro- the last two decades have changed that

5 I oversimplify. For the general theory, 6 Ibid.; Clark Glymour and Gregory F. Cooper, caveats, and mathematical details, see Peter eds., Computation, Causation and Discovery Spirtes et al., Causation, Prediction and Search (Menlo Park, Calif.: aaai Press; Cambridge, (New York: Springer-Verlag, 1993; 2d ed., Mass.: mit Press, 1999); Judea Pearl, Causality Cambridge, Mass.: mit Press, 2000). (New York: Cambridge University Press, 2001).

Dædalus Winter 2004 73 Clark considerably. Scheines used the econo- copied into mrna within any interval of Glymour on mists’ judgments of the probability dis- time depends on several things, includ- learning tribution for values of parameters relat- ing the chemical sequence of the partic- ed to measurement error to assign prior ular dna piece (whether it is a coding probabilities to their measurement er- sequence, i.e., a gene), the chemical ror model. Then he and his collaborators sequences of other regions of the chro- computed the posterior probability dis- mosome that are physically close (regu- tribution for values of the parameter rep- lator sites), concentrations of small mol- resenting the influence of lead on iq. By ecules inside the nucleus of the cell, and this method, they found that low-level concentrations of proteins. Certain pro- lead exposure is almost certainly at least teins attach to the regulator sites of a two times more damaging to cognitive gene and cause the gene to be copied (in ability than Needleman had estimated.7 other terminology, ‘transcribed’ or ‘expressed’) into rna, which in turn Genetics is another ½eld in which sci- goes on to make proteins. An important entists are conducting research in new clue to fundamental biology and its ways by applying innovations in com- medical applications lies in this process puter software, lab techniques, and ob- of gene expression, in knowing which servational technology. genes respond to new chemical or physi- Every cell in your body has the same cal environments, and which cellular dna but cells in different tissues look functions are influenced by the proteins and function very differently–brains, those responding genes produce. after all, are not bones. The difference Traditionally, this kind of problem had comes from the proteins that make up been approached one gene at a time– the physical structure of a cell and regu- for instance by ½nding some of the pro- late–indeed, in some sense constitute– teins that regulate a gene, ½nding the its metabolism. The thousands of differ- protein or proteins the gene yields, iden- ent kinds of proteins are themselves pro- tifying some of the roles those proteins duced by a collaborative manufacturing play in cellular metabolism. But about process in the cell. Amino acids–any of ten years ago, biologists developed tech- twenty simple molecules provided to the niques for simultaneously measuring the cell from outside–are stitched together concentrations of each of the thousands to form a protein, which may then fold –and in some contexts essentially all–of and combine chemically or physically the distinct kinds of mrna molecules with other proteins. Each basic protein present in a collection of cells. Biologists originates along a template of ribonucle- could get a snapshot of how much each ic acid (rna) outside the nucleus, and gene in the cells had been copied or ex- different template molecules–different pressed. Multiple snapshots, moreover, kinds of rna molecules–make different could be taken at different times, as little proteins. Messenger rna (mrna), itself as a few minutes apart, so that research- copied from dna, generates the tem- ers could see the varying responses of plate rna. Whether a piece of dna is the cell genome to changing conditions. So what affects what genes? Answers to 7 Richard Scheines, “Estimating Latent Causal Influences,” in Glymour and Cooper, eds., Com- this question are coming in at an aston- putation, Causation and Discovery. Scheines’s ishing rate. software–freeware–was the tetrad III pro- About ½ve years ago, Tim Hammond, gram, .

74 Dædalus Winter 2004 lane, flew samples of kidney tissue in the can be seen in several sciences–a trend The automation space shuttle. When his samples, which to which we can also attribute the Virtu- of discovery had been chemically frozen while in al Observatory that is planned to enable microgravity, returned to Earth, Ham- astronomers to search and analyze vast mond and his collaborators measured data stores taken by remote instru- the expression of thousands of genes ments; and, in climate studies, the Earth within the tissue. They found that a large observation satellites that now send proportion expressed very differently down several gigabytes of data each from the genes within the Earth-bound day–data that is increasingly being used samples of the same tissue, no matter to monitor the state of the planet, to how the Earth-bound tissue had been locate causes of change, and to forecast mechanically treated. Acceleration or changes in the environment. Ever new low-gravity or something else as yet un- techniques make possible the measure- known about the shuttle environment ment of ever larger quantities of data; affected gene behavior. If, as seems like- data manipulation software makes pos- liest, the effect Hammond discovered is sible the selection of samples that are an essentially mechanical effect of low relevant to particular problems; auto- gravity, it has important implications mated search and statistical techniques for long-term habitation in space, on help guide researchers through the su- the Moon and Mars. perastronomical array of possible hy- Mechanical issues–flow and sheer potheses. over cellular surfaces–are known to in- Kuhn said that scienti½c revolutions fluence genes that are important to hu- generally meet ½erce resistance–and the man health. The cells that line the sur- automation of discovery in science is no faces of blood vessels play crucial roles exception. In some cases the animosity in lethal disorders–for example, in stems from nothing more than conser- aneurisms–and particular genes in vatism, an effort to preserve academic these cells have been known for some turf, or plain old snobbery. Above all, while to change their expression in re- automated science competes with a sponse to mechanical changes, in par- grand craft tradition that assumes that ticular to changes in liquid flow across science progresses only by scientists ad- their surfaces. Very recently, David vancing a single hypothesis, or a small Peters, a young biologist at the Universi- set of alternative hypotheses, and then ty of Pittsburgh, and his colleagues mea- devising a variety of experiments to test sured the change in gene expression in it. This tradition, most famously articu- response to changes of flow for almost lated by Sir Karl Popper, is championed all genes in living human cells lining by many historians and philosophers of blood vessels. In their experiment, more science, and resonates with the accounts than a hundred genes changed, includ- of science that many senior scientists ing some known to be involved in cellu- learned in graduate school. lar structure. Peters and his collabora- While the history of science can serve tors are now measuring all of the genes as an argument for norms of practice, for in such cells that respond to changes in several reasons it is not a very good argu- pressure and flow. ment. The historical success of research- ers working without computers, search The few cases I have briefly described algorithms, and modern measurement here are merely samples of a trend that techniques has no rational bearing at all

Dædalus Winter 2004 75 Clark on whether such methods are optimal, thing else out of the haystack and try Glymour on or even feasible, for researchers working again. Continue until you ½nd the nee- learning today. It certainly says nothing about dle or until civilization comes to an the rationality of alternative methods of end. inquiry. Neither was nor is implies ought. 2. Pick something you like out of the The ‘Popperian’ method of trial and haystack. Subject it to a test. If it error dominated science from the six- doesn’t pass the test, ½nd a weaker teenth through the twentieth century test (e.g., is the thing long and nar- not because the method was ideal, but row?) that it can pass. because of human limitations, including limitations in our ability to compute. 3. Try 1 for a while, and if no needle Historically, novel methods and strata- turns up, forget about needles and gems were devised from time to time start studying hay. to get round computational limitations. 4. Try 1 for a while, and if no needle For example, in the eighteenth century, turns up, change the meaning of nee- Leonard Euler, perhaps the most proli½c dle so that a lot of ‘needles’ turn up mathematician ever, could not reconcile in the haystack. seventy-½ve observations because the 5. Set the haystack on ½re and blow away calculations required far too many steps; the ashes to ½nd the needle. statistical estimation of theoretical pa- 6. Run a magnet through the haystack. rameters, introduced by Legendre in 1808 in a form known as ‘least squares,’ Method 1 is still the standard descrip- permitted the reconciliation of (for the tion of how science is and should be time) large quantities of data, such as conducted–the account we ½nd explic- the seventy-½ve that defeated Euler. The itly in the introductory chapters of sci- quick adoption of factor analysis in the ence textbooks and implicitly in the crit- 1940s was due in part to computational icisms some scientists and methodolo- tractability, and one could argue that the gists express toward other ways of doing same is true of the enormous influence things. of Sir Ronald Fisher’s statistical meth- Method 2 is practiced and effectively ods. advocated by many social scientists (you need only replace ‘something you like’ in When scientists seek to learn new, 2 with ‘theory’). interesting , to ½nd important pat- Methods 3 and 4 are the practices that terns hiding in vast arrays of data, they postmodernists claim science does and are often trying to do something like should follow. searching for a needle in a really huge Methods 5 and 6 are those made possi- haystack of falsehoods, for a correct net- ble by the automation of discovery. work among many possible networks, In principle, methods 5 and 6 are a lot for a robust pattern among many appar- smarter than the other methods, but ent but unreal patterns. they are not without limitations both So how does one ½nd a needle in a real and metaphorical. Burn the whole haystack? haystack and you might melt the needle. And that is a sound worry about auto- 1. Pick something out of the haystack. mating science: it may rush things, Subject it to a severe test, e.g., see if it sometimes too much. Because a proce- has a hole in one end. If so, conjecture dure for ½nding hypotheses is fast and it’s a needle; otherwise, pick some-

76 Dædalus Winter 2004 can be done by computer doesn’t mean very large correlations.8 A real advance The automation the procedure gives good results. Figur- in unraveling gene regulation networks of discovery ing out what a method can and cannot came recently–by chemical rather than reliably do requires hard work. by computer automation. Tong Ihn Lee Consider for example the problem of and his colleagues found a way to identi- identifying networks of gene regulation. fy a large fraction of the genes in yeast The ability to measure gene expression that are, in turn, directly regulated by simultaneously for thousands of genes genes known to be regulators. They did in normal and perturbed genomes (in so for more than a hundred regulator perturbed genomes, particular genes genes, effectively identifying a good have either been deleted or forced to piece of the regulatory structure in ‘wild over-express) invited the application of type’ yeast. computer methods that search for causal The automation of learning, whether networks. Algorithms were proposed for by computer or by new laboratory tech- piecing together networks from compar- niques, does not render human judg- isons of gene expression measurements ment obsolete, or marginalize scienti½c in cell lines with perturbed and unper- creativity. Nor does it cheapen the sweat turbed genomes; algorithms were pro- and effort, the insight and ingenuity of posed for ½nding networks from correla- human scientists, but shifts them to- tions with repeated measurements of ward the consideration of algorithms expression levels in unperturbed net- that can ef½ciently and reliably compare works–and they did very well on data many hypotheses with vast quantities produced by computer simulations of of data and toward laboratory methods gene expression. that answer many questions at once. It turns out, however, that much of this work proved to be illusory. The al- 8 Tiajaio Chu et al., “A Note on a Statistical Problem for Inferring Gene Regulation from gorithm for assembling a network from Micro-array Data,” Bioinformatics (in press); perturbation effects was incorrect. The David Danks et al., “Experimental Determina- algorithms for inferring networks from tion of Gene Regulation Networks: Complexity correlations of gene expressions over- and Statistical Realism” and Frank Wimberly looked the fact that measuring expres- et al., “Experiments on the Accuracy of Ma- chine Learning Algorithms for Discovering sion levels in aggregates of cells (rather Gene Regulation,” in the Proceedings of the than in individual cells, which is techni- Workshop on Bayes Nets and Gene Regulation, cally feasible but rarely done) creates International Joint Conference on Arti½cial correlations due entirely to the aggrega- Intelligence, 2003. tion itself rather than to the influence of particular genes on the expression levels of others. The simulations that seemed to work so well also turned out to be simulations of measurements at the lev- el of individual cells–measurements of a kind usually not made in reality. Un- doubtedly the automated procedures got some things right, but very likely what they got right was cherry picking– gene connections indicated by very large changes in expression levels or

Dædalus Winter 2004 77 Poem by David Ferry

October

The day was hot, and entirely breathless, so The remarkably quiet, remarkably steady leaf fall Seemed as if it had no cause at all.

The ticking sound of falling leaves was like The ticking sound of gentle rainfall as They quietly fell on leaves already fallen,

Or as, when as they passed them in their falling, Now and again it happened that one of them touched One or another leaf still on the tree,

Still clinging to the idea of being summer: As if the leaves that were falling, but not the day, Had read, and understood, the calendar.

David Ferry, a Fellow of the American Academy since 1998, is Sophie Chantal Hart Professor Emeritus at Wellesley College and a visiting lecturer in creative writing at Boston University. His books of poetry and translation include “His Epistles of Horace: A Translation” (2001), “Of No Country I Know” (1999), and “The Eclogues of Virgil” (1999).

© 2004 by David Ferry

78 Dædalus Winter 2004 Fiction by Chuck Wachtel

The Annunciation

The bear in the driver’s seat wasn’t was too young to walk on them, they made of flesh or any other three-dimen- hadn’t a trace of wear. When the car sud- sional substance, but of light and color, denly sped up and turned at the corner, like characters in animated cartoons. he became angry and frightened. He The car it drove had approached him woke then, with Nan’s hand on his from behind, pulled nearer to the side- chest. walk, and slowed to the pace of his walk. “You all right?” she asked him. “You The bear was purple, except for its ears, said something, kind of, and you were nose and mitten-like hands, which were rocking back and forth.” red, and as tall as a human, though “What did I say?” plumper around the torso and neck. “It was like a whole sentence, but it Holding its eyes ½xed to the street direct- didn’t really have words, just sounds.” ly ahead, it maintained the same slow speed just long enough for him to catch It had been six days since the morning a glimpse of the only other passenger: a he sat beside Nan, lying on an examina- little girl in a yellow dress, her legs ex- tion table, and watched, on the screen of tending to a point just beyond the edge a sonogram monitor, a thin tube enter of the backseat, her toes up, one foot her belly, then come so close to the fetus turned slightly inward. She was wearing inside her that the small hand actually blue and white high-tops; the colors reached toward it. “They do that,” Dr. were bright and clean, and because she Gisse said, as he af½xed a syringe to the end of the tube and withdrew a sample Chuck Wachtel teaches in the Graduate Program of amniotic fluid. A moment before that in Creative Writing at and the sonographer, an unshaven man in the mfa Program for Writers at Warren Wil- wearing a surgical cap, had been impa- son College. He is the author of the novels “The tient with Nan who’d begun to shiver Gates” (1994) and “Joe the Engineer” (1983), and cry. It was the kind of casually dra- “Because We Are Here: Stories and Novellas” matic impatience meant to tell the per- (1996), and two collections of poems and short son it is aimed at that they have made prose, “The Coriolis Effect” (1986) and “What your day harder. Happens to Me” (2000). “What’s your fucking problem?” he said to the sonographer, in one angry © 2004 by Chuck Wachtel breath.

Dædalus Winter 2004 79 Fiction by “Johnny,” Nan said, as if the man “The post of½ce?” Nan asked her Chuck wasn’t there. “Look,” she nodded to- back. Wachtel ward the screen. “I work there,” Johnny said. “They’ve been through this once be- “And it has the best medical plan,” fore,” Dr. Gisse said to the man, who Nan said, “in the whole damn country.” looked back at Johnny, but not at Nan– having understood the unspoken portion The day after Dr. Gisse’s assistant of the statement–and gave a nod that called with the results–normal, a girl– constituted an apology. they discussed how they’d announce the A nurse came in, labeled the vial of good news to the friends and relatives amniotic fluid and held it up to Nan. whom they hadn’t told about the preg- “You identify this as your name?” The nancy. Nearly all of their family, on both question was part of the same litigation- sides, lived at a distance, and Johnny and prevention protocol they’d gone through Nan had laid low during the last weeks the last time. Nan hesitated and the prior to the amniocentesis, at which nurse looked at Johnny for help. Johnny time she’d begun to show. The few lifted his gaze to the red-lit exit sign. friends and neighbors and coworkers Two of the four screws that held the who’d ½gured it out were sworn to secre- plate with the letters to the frame of cy. Two years ago, when they’d learned the ½xture were missing and it tilted a Nan was pregnant the ½rst time, they degree or two downward to the right, re- told everyone, even strangers, and the vealing a thin dash of white light over most dif½cult part was untelling them, the red I and T that made him think of undoing what the world around them the diacritical line that means a vowel was still expecting to happen. should be pronounced as it is spoken Since the day the at-home ept test af- when not inside a word. ½rmed their second pregnancy Nan had “My last name is Wilk,” Nan said. kept the test wand in a Ziploc bag in her She’d kept her own name when they got sock and underwear drawer, and so, the married. next afternoon, before she got home “But on the chart it says Rizzotti,” the from work, Johnny set their huge vol- nurse said. ume of the works of Leonardo da Vinci To Johnny the two missing screws on the living room floor, opened to the seemed cognate with the sonographer’s Annunciation they had seen at the Uf½zi lack of manners and unshaven cheeks. Gallery while on their honeymoon in Ex-eyet, he said to himself, without even Florence, and laid the wand across the moving his lips. I’m ready to head for the space between the hand of the archangel ex-eyet. Gabriel, with two ½ngers gently raised, “We use my husband’s plan,” Nan and the serene yet startled eyes of the explained. Virgin. Johnny then knelt over his com- The nurse pulled a strip of labels from position with his thirty-½ve-millimeter the pocket of her smock. camera, and from various angles and dis- “The post of½ce,” Nan said, then tances, and at slightly different foci, shot paused as she watched the nurse write two rolls of color ½lm. her name on a label, peel it off the strip, Eight years before, when Nan led and wrap it around the vial across the Johnny across the huge echoey, marble- part with Johnny’s last name. walled room to the painting, she had “The post of½ce?” the nurse asked her. said, improvising on an ad for Kentucky

80 Dædalus Winter 2004 Fried Chicken, “When it comes to an- During the last two days they’d been The gels, nobody does wings like da Vinci.” granting entry to feelings they’d held at Annunciation Johnny’s composition, reproduced as a abeyance for months. They’d reached postcard, would make a unique an- the top of a mountain so steep that the nouncement, a revelation of the knowl- labor of climbing had kept them from edge they had kept to themselves for taking notice of the scenery. Now they’d more than four months. stroll down the other side, enjoy every- The next day, on his lunch hour, John- thing, let gravity do the work. “Even so,” ny picked up the two rolls at the one- Nan had said, thoughtfully, “innocence hour photo counter at Rite Aid. He lost is never regained. And guess what?” opened the envelopes in the checkout she had begun to laugh. “I could give a line and by the time he’d flipped half- shit less.” That morning when he awoke, way through the second stack of photos Nan was sitting up, leaning against the his anticipation had eroded to disap- wall on her side of the bed, watching pointment: the collage he had con- him sleep. “You know what I just real- structed, that had looked perfectly clear ized?” she said. “We’ve been pregnant through the camera lens, was unrecog- more than nine months combined, and nizable in the images he held before him. now, ½nally, we’re in control.” Her exhil- The flash had bounced off the page aration and certainty frightened him, where it curved above the spine like a but he was much too happy to be wor- wave of parted hair, spilling a wide oval ried about anything. “Now we’re in con- of white light across half the photograph trol,” she repeated. “We control the hori- and leaving the other half too dark to zontal. Do do do do,” she sang the ½rst identify anything. four notes of the theme from The Twilight Later that afternoon, before Nan came Zone. home from school, he’d shoot another “That’s the wrong show,” he said. “It’s roll from different angles in the consis- The Outer Limits where they control the tent, nonviolent light of the overhead horizontal.” lamp. She slid her hand under the blanket, Although the next batch didn’t come gripped his penis. “And we certainly out much better, there were three shots control the vertical.” in which all the component parts were After they made love–the ½fth time in identi½able. If you knew what an annun- two days–Nan laid the back of her head ciation was, you would know this was on Johnny’s stomach and slid her feet one; the implausible object lying across up the wall. “I’m telling you right now, the composition was recognizable as an there’ll be none of that textbook-senti- ept wand and, most importantly, the mental-story-to-tell-later crap. No crav- red line that bisected the positive box ings for ice cream or shrimp dumplings, was clearly de½ned. It was time to show no belly-hiding muumuus, no sudden them to Nan, who had much more expe- mood swings, no sentimental platitudes, rience photographing art–she was a no storks on the birth announcement– professor of art history at City Univer- no fucking storks anywhere.” sity–and get her advice for the ½nal shoot. He left the three best ones faceup Johnny was sitting in the living room, on the kitchen table to see how she’d trying to read the paper, when he heard react to them when she came home from the door to the apartment open, then the work. sound of Nan’s footsteps crossing the

Dædalus Winter 2004 81 Fiction by kitchen, the clunk of her shoes, one after “I thought we could take a better shot, Chuck Wachtel the other, hitting the floor, and the then make a postcard. Nan and Johnny whoom of the bathroom door being have an announcement . . .” pulled shut, followed by the clack of the “At ½rst I thought it was some kind of door hook striking wood. weird submarine,” Nan said. He walked into the kitchen. Her brief- “Not in a better photograph. That’s case was on one of the chairs and a take- where you come in.” out bag with a widening grease blotch on Nan started laughing. “I like it. I like its side was sitting on top of the photos. that you want to tell everybody. I do Johnny moved the bag across the table, too.” and slid the photos to the side she would “I think it’s a work of art,” Johnny approach them from. said. “Not a spot,” Nan said, opening the Nan opened the bag and began setting bathroom door. “Not a spot all day.” the takeout containers on the table. “I’m She had been spotting since the ½fth starved,” she said. “Although the Virgin week of the pregnancy, and though they conceived in a very different manner had reached the middle of the second than I did, I know this: as her belly got trimester, it still hadn’t stopped. Dr. bigger, her appetite got bigger.” Gisse told them it probably wasn’t any- “Maybe it’s a good thing,” Johnny thing to be concerned about. He told said. “That it slowly reveals itself. I them they worried too much about mean, that’s how art works, no?” everything, “But don’t worry about wor- rying. That’s not unusual after what hap- Nan had been right about this preg- pened the last time.” The last time, when nancy not being ordinary. Although they the call came, they were sitting in front felt the anxieties of people becoming of the tv, watching Jeopardy, eating din- parents for the ½rst time, they felt, even ner. How could anything real happen at after reaching the point of being preg- such a moment? The genetics counselor nant longer than they’d been before, that told them he waited until evening to they would never feel the newness, the make such calls, when both partners constant surprise, that they remem- would most likely be at home: trisomy 21: bered. Down’s syndrome: three of the twenty- Once the news was out, Johnny’s ½rst chromosome instead of two, forty- mother, who lived in a senior housing seven in total instead of forty-six: odd, apartment in Florida, called often, usual- two parents, two of everything: odd ly to talk to Nan. When she called on the numbers are bad news in genetics. It morning of Johnny’s birthday, near the would have been a boy. end of the second trimester, he was in Johnny took Nan’s briefcase off the the shower. chair and motioned, like a maître d’, “We talk while I wait for him,” she for her to sit. “What do you think?” he told Nan. “And I tell you about forty-½ve asked when she looked down at the years ago today when I didn’t have to three photographs. She picked one of wait. He was in such a rush I still had my them up but still said nothing. shoes on.” He could no longer wait. “Da Vinci’s In the last weeks Nan had grown tired Annunciation. And that’s our ept test.” of her mother-in-law’s voice, annoyed “I get it,” she said, “but I didn’t get it at the endless childbearing stories from fast enough.” three generations of Johnny’s family told

82 Dædalus Winter 2004 as if they were instructions for how to nouncement and they still hadn’t pur- The conduct her own pregnancy. But this sued the idea. By this time the few peo- Annunciation was the ½rst time she had spoken of ple who hadn’t been told had gotten Johnny’s birth. word from those who had. The influx of “His father was at work, so when my notes and cards and phone messages time comes, my brother, Gianfranco– hadn’t tapered off and now baby gifts you know Johnny was named for him– had begun to arrive. drives me to the hospital. I’m seventy- One evening Nan walked in looking four but I remember like it was last pale and exhausted. Their plan was for week. We had to pass through the old her to take the next semester off, but neighborhood in Brooklyn and when he there were still four weeks left in this stops the car in front of Sal’s Fish Mar- one. ket I know what he has in mind. “You’re working so hard,” Johnny told “‘I’ll be quick,’ he says. ‘You wait in her. “I wish there was a way you could the car and we don’t get a ticket.’ What just stop now. Couldn’t they get a substi- am I gonna say? Since he never got mar- tute or something?” ried bacala was all he ever thought “How dare you,” she said, anger flash- about.” ing in her tired eyes. “Then we get to the hospital. Like a “What?” gentleman now, he opens the car door “How dare you accuse me of being and as soon as I stand up, it’s Niagara lazy?” Falls under my dress. Forty-½ve years “You got it all wrong.” ago today. I tell the doctor I can’t stand “You’re the one who got it wrong, up my back is hurting so bad and when buddy.” they put me on the rolling thing I tell Johnny walked into the living room, him, ‘No, no, I can’t lay down neither.’ sat on the couch, picked up the remote, The doctor examines me right there, and turned on the tv. He stared at the we’re not even in the room yet, and he Weather Channel, listened to a few bars says, ‘Why’d you wait till now?’ of the soft jingly music that accompa- “‘In a car out front,’ I tell him, ‘there’s nied the ½ve-day local forecast, then got ½ve pounds of cod ½sh that’ll answer up and walked back to the kitchen. your question.’” “Does your mother ever sleep?” Nan “That’s something,” Nan told her. asked. There were tears in her eyes now. “Wow.” He remained standing in the entryway. “Every year on his birthday the ½rst “Last night some movie star told thing I remember is getting out of that David Letterman that while she was car. That’s when it hurt, I can’t tell you pregnant she had the uncontrollable how much. That’s when I say to myself, urge to eat flowers. Daisies especially. he’s gonna get born–even then I knew Eight o’clock in the morning your moth- he was a boy–even if I’m gonna die.” er calls because she has to tell me this. “Oh, he’s dressed,” Nan said, waving Plus she keeps suggesting names. This Johnny into the room, “and he’s about morning’s suggestion was Ricardia, her to leave for work.” mother’s name. ‘Doesn’t Ricki sound “That’s all right,” she said. “You just nice?’ she said. How many times do I tell him I said happy birthday.” have to tell her we’re not discussing names yet?” Two months had passed since Johnny “That movie star, did she eat them?” showed Nan his photos for the an- “I can’t get enough sleep.” Dædalus Winter 2004 83 Fiction by “I’m sorry,” Johnny said. “Next time “And what about the thing Italians do Chuck Wachtel let the machine answer.” to girls?” Nan had said, smiling, but “You are such a gaping asshole.” tired of his persistence. He walked back into the living room, “What do they do to girls?” turned off the tv, and lay down on the “You know what I’m talking about.” couch. “I have no idea what you’re talking Five minutes later Nan came in. He about.” lifted his legs as she sat down, then low- “Prenatal ear piercing?” ered them onto the arm of the couch so Nan slid Johnny’s legs off her, turned they crossed the space above her lap like so she could lay her head on the side of the safety bar on a Ferris wheel seat. the couch, then laid her feet on his lap. “And you can tell your mother that After letting her weight settle she grunt- we’re not doing to our daughter what ed, arched her back, lifted her heavy Italians do to little girls.” torso just high enough to slip her hand Johnny laughed at this, though not too underneath, and pulled out the remote. much, since there had been no acknowl- She lowered herself back onto the edgment that the ½ght was over. She was cushions, turned the tv back on, and referring to an argument they’d had dur- skimmed the channels. She stopped at a ing the last pregnancy–during the wait shot of a beautifully pure blue sky, which for the amnio results–something they held only for a second before the camera had not spoken of since. The sonogram dropped and found two teenagers, a boy image had given some evidence that it and a girl, leaning against the fender of a was a boy. If the results con½rmed that, car. They were contemporary teenagers, Nan wanted him to be circumcised. but the car was a vintage, late-sixties Johnny did not. Corvette convertible, bright red. They “He’ll automatically be Jewish since appeared tired; they were sad and a little you are,” Johnny said. “It’s matrilineal. bored, yet sexy in an adult way. You told me that.” “You were with her,” the girl said, en- “I did.” ergized by her anger, though sleepy- “So can’t he be Jewish without having eyed. his little dick whacked?” The boy turned his head away. “Anatomically, he should look like his The girl, wearing dark red lipstick, father,” Nan argued. Johnny had been looked briefly at the camera, pouting, circumcised in the hospital, as had most then slowly lowered her gaze. male babies of his generation, and had The boy turned and directed his eyes never given it a thought. However, the downward, toward whatever the girl was idea of having it done to his son had looking at. caused him to imagine the pain for the “I bet it’s going to be Pepsi,” Nan said. ½rst time: it would be as if it were hap- “I was just . . . ,” the boy said, then pening to him all over again. He began paused. “I was just . . . there.” to envision the cutting of the foreskin as The camera slid down their slender an ongoing, constantly repeated process, bodies. They were both wearing jeans. like the bound Prometheus’s liver being One of the knees on the boy’s was eaten by an eagle, only to grow back ripped, showing the pale skin beneath. again overnight, then to be eaten again “The Gap,” Johnny says. “Five bucks by the same eagle, from his ripped-open says it’s the Gap.” torso. Nan, still holding the remote, turned off the tv before they could ½nd out. 84 Dædalus Winter 2004 Three weeks before their due date, there were no marigolds, only blue dai- The Johnny woke to the sound of Nan cry- sies that looked like they’d been watered Annunciation ing. He reached toward her before he with dye, ordinary yellow daisies, and even opened his eyes and found her side roses that looked morbid and inedible. of the bed empty. She was sitting in the Before he headed toward the twenty- chair across the room, leaning over, her four-hour greengrocer on Avenue A, he elbows on her knees. “Nan,” he said, bought the yellow daisies, just in case he then, “What?” He was afraid something couldn’t do better. According to the had gone wrong, or that she’d gone into thermometer on the Emigrant Savings labor early, but he knew, in the ½rst in- Bank it was nineteen degrees Fahrenheit, stant of full wakefulness, that it would minus seven Celsius. be best if he didn’t appear frightened. He He had met Nan in his ½rst and only responded as if the loud sobbing that year of graduate school. She was a stu- had penetrated his sleep was a question dent in the freshman composition he hadn’t fully heard or understood. course he taught. At the end of that year, “What?” he said again, softly. a cut in federal funds had forced nearly “It’s four-thirty in the morning,” she all teaching assistants at ccny to be laid said. Her breaths were sudden and shal- off. So Johnny, along with several other low, with a faint trace of voice in them. graduate students, quit school in what “I thought your idea for the pregnancy was both a statement of protest and an announcement was terri½c,” she said. act of necessity: he could not afford to “It was a work of art. I’m sorry we never continue without the teaching assistant- made the postcards.” ship that had paid his tuition. The ½rst “Who cares? Nan, are you all right?” job he found was at the post of½ce. One “I need something,” she said. She afternoon, more than eight years later, seemed angry now. “Why did your Nan handed him a yellow slip at the par- mother tell me about that stupid cel pickup and information window. fucking movie star?” “Mr. Rizzotti?” “What do you need, baby?” He didn’t recognize her. He assumed She started crying again, harder. she saw the name on his id tag. “You “Nan?” gave me an A. My ½rst in college. I never “Flowers.” She said this between thanked you.” gasps, in a whisper. In his entire adult life he had never felt “What?” anything like what he was feeling now, She covered her eyes and shook her walking east through the predawn head. Johnny helped her back into bed, morning. He’d carried mail for four then held her in his arms. “You want years, and drove a mail pickup route for flowers?” three more before becoming a supervi- “Marigolds,” she said. “I keep think- sor. You could see the city in a million ing of the thick part in the middle.” Her ways: during his workday he saw it as a breathing was slower now. It seemed she complex chain of mailboxes, with the might even be falling back to sleep. rest–the buildings, the cars and trucks He slid out from under her weight, and people–slightly out of focus. Now then got out of the bed and stood beside he saw this neighborhood, the one it. He pulled the blanket over her, leaned where their daughter’s ½rst home will down, kissed her hair. be, as a constellation of twenty-four- Just north of Houston Street, he found hour greengrocers, their lights glowing a greengrocer that was still open, but like stars. He did not feel the cold. He Dædalus Winter 2004 85 Fiction by only felt a steady current of elation. “Shit,” she said. “How dare you?” Chuck Wachtel They’d passed all the danger zones, now “Was it?” all they had to worry about was adoles- “You don’t understand.” cence and college tuition. He couldn’t “Was it?” wait to meet his daughter. He couldn’t “You’re a man. A mailman. How could wait. you fucking understand?” When he walked in Nan was sitting at He was furious, but knew he was still the kitchen table with a pencil in her happy underneath. “Tell me, professor. hand. What the hell is wrong?” “Since I was up,” she said, emphatical- “I married an idiot,” she said. “That’s ly casual, “I thought I’d get some work what’s wrong.” done.” He set an array of cone-shaped bou- When the baby was six days late, Dr. quets, wrapped in gift paper or clear Gisse sent Nan to the hospital for a non- plastic, before her, covering the entire stress test. Unfortunately, it was done in tabletop, including the student paper the same clinic as their second amnio- she’d been reading. Among them were centesis. two batches of marigolds. He’d got the As soon as they entered the ultrasound second batch at the last place because examination room, his eyes found the their tops were bigger than the ones exit sign. He was relieved to ½nd that the he’d already bought. missing screws had not been replaced: “Is it cold?” she asked, no trace of changing anything in that room might what she was feeling before he left in her indicate a change in their fate, perhaps voice or eyes. for the worse. He began unwrapping each bouquet. Though they had a hard time recogniz- When he laid them back down, they lost ing the parts of her anatomy, the baby their shapes. It was as if he’d amassed an appeared ½ne. At one point, the sonogra- entire flower garden on the table before pher–a different one, a woman–told her. “See anything you like?” them that the baby had just moved a foot She picked up three of the bigger mari- and a hand to her mouth, and pointed golds–the soft, orange centers inside with a little plus-sign-shaped cursor to the dense corollas of small petals were where this was happening. as big as marshmallows–and held them “Does that mean she’s hungry?” John- out to him. “You ½rst.” ny asked. “No,” he said. “You.” “It could,” the woman said. She moved the flowers closer, but did “Oh Hannah,” Nan said, trying out not speak a word. the name they had chosen, softly curv- He shrugged, slowly leaned forward, ing the second syllable downward. and took the head of the largest one in “What an appetite,” Johnny said, his mouth. shaking his hand Italian style, then be- She suddenly began to cry. gan to weep. “Nan,” he said. Everything looked ½ne, but nothing “This is crazy,” she said, anger ½lling would be certain until they got the her voice. results of the second test, which He did not know how he knew, but he involved Nan sitting in a room with knew what had happened. “That was the other beyond-due-date mothers, each one you wanted,” he said. “Wasn’t it?” with a fetal monitor strapped to her

86 Dædalus Winter 2004 belly, while various electronic bleeps address, they thought it was the infor- The recorded the baby’s movements and mation they’d asked the genetics coun- Annunciation vital signs and her own mild contrac- selor to send. The tone of its introduc- tions, most of which could not even be tion was sympathetic; it offered hope in felt. the form of knowledge. The persuasion Two days before the last pregnancy didn’t assert itself until the second page, had been terminated, the doctor insert- which began with the words Search and ed a branch-segment of laminaria, a Destroy, an anti-abortion catchphrase for kind of seaweed, into Nan’s cervix, to amniocentesis. dilate her in preparation for the abor- In the remaining two days of Nan’s tion. Once that was done there was no pregnancy, Johnny would have fantasies reversing the process. The following day so real they lifted him entirely out of the he would remove the insert and replace moment, out of the abrasive, fast-slow it with two branch-segments, widening dream of time: on the street, or in a the cervix further. Those two days, dur- hospital corridor, a crowd of strangers ing which Nan experienced the symp- would approach him and Nan, and even toms of early labor, were even darker before they spoke he would know they than the previous two weeks, when were the people who had sent that pam- they had lived each day with the news. phlet. He would lunge into them shout- They’d had to make a decision, as par- ing and throwing punches. He would not ents, as non-parents, and perhaps the stop until he had hurt them all. most dif½cult part was accepting that During that same visit the doctor also the decision had already been made, and told them that in second-trimester abor- that it resided inside them, always had, tions there are remains, and that now and would continue to, long after the might be a good time to think about how pregnancy was terminated. they wanted to handle them. The hospi- On the ½rst of those two visits, the tal could take care of it; forms would doctor told them there was a possibility, have to be signed. Or they could choose though an unlikely one, that they would cremation, even burial. His voice im- encounter anti-abortion activists on the plied, warmly, that it would be best to morning of the procedure. Legally, not make too big a deal of this part, to they’re not permitted to approach any- begin leaving the past behind as quickly one, he said, not even be on the same as possible. side of the street, but anyone can walk On the morning of the procedure they into the waiting room, and there’s no avoided the waiting room entirely. After telling who someone could turn out to helping Nan into her hospital gown, be. Records are con½dential, but they Johnny waited in the hall outside the have ways of ½nding out when late- recovery room, along with two Ortho- term abortions are scheduled. dox Jewish women who stood facing the “They know,” Nan had told him. “I wall that separated them from the ward think they knew before I did.” Less than in which their loved one would awaken, a week after getting the amnio results, a once whatever was being done to her pamphlet had arrived in the mail with a had been done. One of them opened a photo of Down’s children sitting in a cir- small book and held it between them. cle around a teacher, smiling and clap- They began to rock gently, chanting soft- ping their hands. Though the envelope it ly in Hebrew: the rhythm of their pray- came in had a post of½ce box as a return ing was the only thing that enabled the

Dædalus Winter 2004 87 Fiction by minutes to pass. They continued to pray the exit came between him and Nan. Chuck Wachtel when the doctor came through the door The rotunda was ½lled with people, most from the recovery room, his green surgi- of them standing in groups, talking, and cal mask hanging loosely from his neck, when Johnny reached the middle of the and approached Johnny. Everything had room he discovered that Nan wasn’t be- gone smoothly, he said. Nan would be side him. He spun around and caught a awake in a minute or two. glimpse of her walking into the of½ce. At Johnny felt relief. It was as if the that moment he realized that Nan was weighted matter within his own body, never more in the world than when she relentlessly subject to the pull of gravity, was pregnant, yet her grief had caused had been removed. For a breath’s time it her to withdraw from it to such an ex- was over, but he knew that the next mo- tent an onlooker’s casual gaze could not ment would begin a process of unbear- detect her presence; he, a father who able mourning. The doctor stepped into hadn’t been able to protect his child and the elevator. The two women had now his wife from danger, was cloaked in his stopped praying, and before Johnny own helpless anger. walked through the door into the ward The most direct route to the of½ce Nan he told them that he hoped the patient had just entered took him between two they were praying for would have a full men who were facing each other, per- and speedy recovery. haps two feet apart, talking. They looked familiar, and as they stepped back to al- A week later, when their taxi arrived low him to pass, he was certain he recog- at the Upper West Side funeral parlor nized them both. where they were to pick up the ashes, When he walked into the of½ce, the they had to wait while the three limos man sitting behind the desk rose, ap- ahead of them discharged their passen- proached him, and without introducing gers at the edge of the long awning that himself, motioned Johnny to one of the reached to the curb from the wall above two seats facing his desk, and said, “Your the entrance. It wasn’t until they had wife is in the rest room.” The man apol- stepped out that they noticed the police ogized for the crowd, pressed the ½nger- barricades holding back a crowd of on- tips of both his hands together, looked lookers on both sides of the street, and down at the desktop, and said nothing the network news trucks with telescopic more. antennae on top parked across the street. On the cab ride back downtown Nan Johnny took Nan’s hand, and they examined the white cardboard canister, walked at a quick, deliberate pace. He the same size and shape as a container of had no idea what was going on, and as Quaker Oats. There wasn’t a word or they approached the entrance he imag- number to identify what it contained. ined the things he’d shout at a police “How do they know it’s ours?” she of½cer if one tried to stop them. He asked, then said, “Give me your keys.” half hoped one would, but no one ap- With the penknife on his key chain she proached them. They were walking cut the tape encircling the middle, hold- through a different dimension: no ing the top and bottom halves together. one even noticed they were there. She tried to open it, but couldn’t get her They climbed three steps, walked nails into the small space between the through the entryway and into a wide two parts. rotunda where a woman walking toward

88 Dædalus Winter 2004 “Let’s wait till we get home,” Johnny small stuffed bear. The bear was purple The said. and red and reminded Johnny of the Annunciation Nan turned, looked out the window, chauffeur bear in his dream, which he and said, impatiently, “We’re only at did not remember as a dream, but as Sixty-eighth Street?” and then, as if it something that actually happened a long were part of the same thought, “It time ago. The girl’s mother, who was sounds crazy, but I think I saw Phyllis pregnant, was seated in one of the rows Diller coming out of the bathroom in the of chairs across the carpet from where funeral parlor.” Johnny was sitting, holding a smaller Johnny suddenly realized that the two child asleep on her lap. men he had walked between, less than A copy of People magazine lay faceup ½fteen minutes ago, were the ex-mayor on the seat beside him with a photo- David Dinkins and the comedian Alan graph of Vanna White, in a strapless, King. Nan and Johnny had been left no floor-length evening gown, on its cover. room for curiosity, no interest in looking The little girl held up the hula Barbie so through the window it opens on the Johnny could see it, and moved its arm proximate world. The enormous crowd, to wave hello. Johnny smiled and waved the police, the news trucks, were just back. She was a beautiful child, no more there. “I bet it was her.” Johnny said. than ½ve years old. She wore thick glass- A few blocks later Nan tried again, and es and had a yellow Band-Aid on her was still unable to open the canister. forearm covered with stars and planets. This time she handed it to Johnny. He The elevator door opened, and both he held the bottom, and with both hands and the girl watched as a man wearing a she loosened the top. Inside was a small business suit and yellow tie stepped out, plastic bag that contained less than a crossed the room, lifted the People maga- handful of pebble-hard, gray ashes, and zine off the chair, and sat down beside a scorched metal ring, perhaps an inch Johnny. in diameter, with the number ½ve “It’s raining,” he said. His damp suit stamped on it. jacket smelled like cigarette smoke. Nan closed it, embraced it and stroked “Your wife in there?” he asked Johnny, the smooth cardboard. who nodded. When they got home Nan fell asleep “Mine, too. This your ½rst?” Without on the couch and remained asleep for waiting for an answer he said, “I already the rest of the afternoon. After sunset have a six-year-old boy.” He lifted his Johnny went out to buy soup for their feet, one at a time, and inspected his dinner and on the way back noticed, on shoes, top to bottom. “She was two the front page of a Daily News on top of a weeks late with him. I hope we don’t stack at a newsstand, a picture of the have to wait that long for this one.” front of the funeral parlor and the head- The man fell silent then, and opened line: new york’s best, brightest the magazine. Johnny looked back at the and funniest say farewell to girl who had arranged the dinosaurs in henny youngman. rows as if they were an audience facing the bear and the biggest dinosaur, a A little girl, sitting on the carpet in the brontosaurus. waiting room, had set up in front of her “You know what I hate about Wheel of a collection of plastic dinosaurs along Fortune?” the man said to Johnny. He with a Barbie doll in a hula skirt, and a was holding up the magazine, pointing

Dædalus Winter 2004 89 Fiction by to the picture of Vanna White. “I hate it tosaurus and the bear were the bride and Chuck Wachtel that some really nice person, someone groom. She pulled off her Band-Aid and smart and nice, can get all the letters pressed it onto the back of Johnny’s except like maybe one or two, and then hand. “I don’t need it,” she said. “I was they go bankrupt.” He shook his head. just wearing it because it’s pretty.” He seemed genuinely angry. “And then He thanked her and looked admiringly some idiot dipshit who can barely read at it. Up close he could see that the stars gets the answer. Ever see that happen?” and planets had little faces. Johnny Johnny shook his head. wanted to ask her mother, who was smil- “That’s what happened last night.” ing at them, if he could pick the girl up, “Last night?” if he could hold her in his arms. “The clue was Theater complex of New “They’re going to have a baby,” the York and home of the Metropolitan Opera. girl said, pointing at the newly married Know what it is?” couple, at the bear’s fat little belly. “I don’t think so.” “The baby will be half bear and half “Lincoln Center. The dipshit wanted dinosaur,” Johnny said back, stupidly. to buy a vowel but they were all ½lled in “No, no, no,” she shook her head. “It’s already. You could see him moving his a girl.” lips as he sounded it out. The only letter “A beautiful one, I bet,” Johnny said. not there was the r and I swear, at one The girl turned, stretched, held herself point I thought he was going to say Lin- upright, but remained kneeling. “They coln Continental. He got it just as the just had a checkup,” she said, then buzzer went off, and this sweet, smart picked up the hula Barbie and held it young lady goes home with the parting out toward him. “This is the doctor.” gifts. You know, like carving knives and “I hope everything’s okay,” Johnny tickets to some shitty musical and din- said to the doctor. ner for two at a restaurant where the The girl rose to her feet, looked at him food’s so bad they have to give it away.” impatiently, held the doctor so close the Just then a very pregnant woman hard small face was touching his ear and passed through the doorway leading out whispered, “Of course it is.” of the examination rooms. She smiled at the man beside Johnny and gave him a thumbs-up. “All Raaaaiight!” he said, then got up from his chair, met the woman as she crossed the room, leaned over, and kissed her protruding belly. “And of course he didn’t win the bonus round,” the man said to Johnny as he was helping his wife into her coat. “The dipshits never do.” The hula Barbie waved at Johnny again, and this time he got up, walked over to the girl, and sat on the carpet beside her. “It’s a wedding,” she said. He now understood the arrangement. The bron-

90 Dædalus Winter 2004 those who want to maintain its roots in the humanities–and the new journal was, in part, meant to heal the rift. apsa acknowledged dissatisfaction after analyzing a 1998 survey of its mem- Jennifer Hochschild bers and ex-members. Over two-½fths of the current members who responded, on the social and half of the former members who science wars responded, criticized the Association’s flagship journal, the American Political Science Review (apsr); it headed the list of apsa activities with which respon- dents were unhappy. For example, indi- vidual respondents wrote that the apsr only “covers one small corner of the dis- cipline,” that it is “virtually useless for my teaching preparations and research specializations,” and that it is not “re- In the spring of 2003, as the founding flective of the range of research meth- editor of Perspectives on Politics, I helped ods and approaches in the discipline.” to launch the ½rst new journal sponsored The Association’s report concluded that many political scientists saw the by the American Political Science Asso- apsr ciation (apsa) in over a century. The as “too narrow, too specialized new journal grew out of the general dis- and methodological, and too removed affection that had been floating around from politics.” the discipline for years. In political sci- In short, some of the most prominent ence (as in other social sciences from members of the discipline, as judged by economics to anthropology) a cold war their appearance in its most selective has persisted for years between research- and prestigious journal, were developing ers who want to push the discipline in a new type of ‘science’ that left other the direction of the ‘real’ sciences and members of the discipline feeling angry, unimpressed, and disfranchised. Several years later the Association’s Jennifer Hochschild, a Fellow of the American governing council approved the creation Academy since 1996, is Henry LaBarre Jayne Pro- of a new journal and eventually selected fessor of Government at Harvard University, with me to serve as its ½rst editor. The new a joint appointment in the department of African journal’s mission would be to publish and African American studies. She is the found- “integrative essays” that are less special- ing editor of “Perspectives on Politics.” Her books ized than normal research articles and include “The American Dream and the Public that might “appl[y] . . . political science Schools” (with Nathan Scovronick, 2003), “Fac- to questions of public policy.” The com- ing Up to the American Dream: Race, Class, and mittee charged with implementing the the Soul of the Nation” (1995), and “What’s council’s directive added further man- Fair: American Beliefs about Distributive Justice” dates: the new journal should also in- (1981). clude “state-of-the-discipline type es- © 2004 by the American Academy of Arts says, book reviews, reviews of literature & Sciences

Dædalus Winter 2004 91 Note by in other disciplines with relevance to becomes a good deal greater than the Jennifer political science, conceptual and meth- sum of its parts. Hochschild odological essays, [and] a policy forum There is no intrinsic substantive con- for debates on current policy issues, tent to this claim about scienti½c ad- among other new materials” (italics vancement; it can hold for the study of added). Those other new materials canonical political philosophers, for a might, for example, include “articles particular area of the world, for explain- similar to those found in Science maga- ing how a speci½c institution conducts zine.” The implementing committee its business, or for the revelation of hid- concluded, in something of an under- den discrimination against disadvan- statement, that Perspectives on Politics taged groups or marginalized popula- “should publish a very wide range of tions. It is also not intrinsically opposed scholarship,” that is, it should both to engagement with political or policy widen the apsr’s conception of the concerns; small bits of cumulative, spe- ‘science’ in ‘political science’ and cialized knowledge may be just as impor- restore ‘politics’ to it. tant for determining how to combat ter- Although I was not involved in shap- rorism or reform tax law as for under- ing the journal’s mandate or design, I standing the median voter theorem in share its originators’ goals. Like similar legislative decision making. journals in other social science disci- Nevertheless, an alternative frame- plines–for example, The Journal of Eco- work sees increased specialization as nomic Perspectives and the new sociology insuf½cient to, or even the downfall of, journal Contexts–Perspectives on Politics is the social sciences. In this view, the com- a response, in part, to a widespread per- pilation of small, cumulative ½ndings is ception that the drive to be scienti½c boring to read and teach, and narrows risks distorting our purposes, and that one’s intellectual capacities. True sci- too many scholars are moving into nar- ence, de½ned now as real gains in knowl- rower and narrower specializations, edge and insight, consists in ½guring out divorced from the concerns of nonspe- how to ask the right question even if it cialists and the ‘real world.’ cannot be answered, understanding The respective merits of breadth and how people see the world from their depth are a complicated and old issue. own vantage point, and investigating To some, specialization is an essential large dynamics of change or stasis. Ab- virtue in the face of a wide range of wor- sent a broad vantage point, the ability to thy topics and the deepening of knowl- consider a problem from multiple per- edge about each; it is evidence of the spectives, and the recognition of one’s maturation of the social sciences. Only own inevitably partial and biased con- by specializing does an individual have a ceptual lenses, one cannot determine chance to acquire suf½cient substantive how and why the world works as it does. and methodological knowledge to devel- True science also entails knowing when op sharp hypotheses, test them de½ni- to abandon a given framework rather tively against alternatives, and pinpoint than to continue trying to re½ne it– their contribution to theoretical frame- but one cannot imagine alternative par- works. Science consists in the cumula- adigms without breadth of vision. tion of small advances built on previous Here also there is no intrinsic link small advances, all in the service of test- between the call for integrative breadth ing a larger theory–so that the whole and any particular topic of study, norma-

92 Dædalus Winter 2004 tive stance, or degree of policy relevance. cipline that has at its core the analysis of The social And in this framework too, the indeter- the exercise of power. But political sci- science wars minate signi½er of ‘science’ or ‘knowl- ence encompasses the canon of political edge’ is given a content intended to con- philosophy from Thucydides through fer status on a particular set of practices. Hannah Arendt, and also moves through Many political scientists do not aspire qualitative research via case studies and to the mantel of ‘science,’ however it is historical or institutional analysis to understood. They see their enterprise as highly technical quantitative analysis closer to that of the humanities or histo- and formal reasoning. No other social ry, in that they seek to give meaning to science covers such a wide epistemologi- a phenomenon rather than to provide a cal range so deeply; therefore it perhaps causal explanation for it. But they too makes sense that we argue over how to are involved in the methods wars that do our work more than over what our are roiling apsa and the social sciences. work is about. Of course, there is no need to insist The discipline of sociology, in con- that the study of politics be either a sci- trast, has largely avoided methods ence or an art, just as there is no logical wars, but at the cost of arguably even reason to pit breadth against depth: more painful disputes. In recent years, these are separate rather than conflicting battles among sociologists have revolved values. But every reader, writer, teacher, around the roles of race and gender in and journal editor must make trade-offs determining professional standing, and at the margins. Perspectives comes down the presumed association of race and on the side of integrative breadth rather gender with differing understandings of than cumulative depth, but less from a science and knowledge. In the late 1990s, deep commitment to the right way to for example, the American Sociological conduct our business than from a per- Association (asa) became embroiled in ception of the need for a counterweight a bitter dispute over the editorship of the to most high-status academic journals. American Sociological Review. The nomina- As I pointed out earlier, all social sci- tions committee proposed an African ences are facing this trade-off between American candidate and a slate of edito- breadth and depth in their publications, rial board members who collectively em- teaching, and graduate training. Most phasized qualitative and/or postmod- have begun a journal with a mission sim- ernist research, sustained attention to ilar to Perspectives’; in fact, political sci- issues of hierarchy and strati½cation, ence was a bit slow on the uptake, so we and a commitment to the view that the have been able to learn from the experi- pursuit of scienti½c and pre- ences of other disciplines. The underly- cision was a mistaken, or at least too ing conflict over the changing and con- narrow, way to understand the social tested meaning of ‘science’ or ‘knowl- world. But the governing council of edge’ has, however, taken a different asa chose a different set of candidates form in each of the four disciplines I (one of whom was also African Ameri- know best. can), amid vehement accusations of The fact that the nastiest ½ghts in po- racism against both speci½c named in- litical science are over methodological dividuals and asa as an organization. frameworks–not over competing po- There have been similar battles over litical values or desirable hierarchies of gender issues in asa, incorporating the power–might seem surprising for a dis- same underlying struggle over the mean-

Dædalus Winter 2004 93 Note by ing of science and the goals of social sci- The discipline of anthropology has, Jennifer ence analyses. like political science, engaged in disputes Hochschild Economists are much less likely to de- over methodology, but in this case the bate methods for conducting research or dominant position rejects the validity to challenge the ascriptive characteris- of and conventional under- tics of researchers; their central ½ght is standings of science. For several de- over the legitimacy of critiques of neo- cades, the most prominent anthropolo- classical orthodoxy. Dissident Euro- gists have argued that scholars need to peans have begun a movement for ‘post- attend ever more to the subjectivity of autistic economics,’ and in the United the researcher, the power dynamics and States a tiny tempest in a teapot at Har- subtle interplays of communication and vard University was deemed signi½cant between subjects and research- enough to be reported in The Weekly ers, the partiality of any claim to knowl- Standard and The Economist. At Harvard edge, and the context within which any a two-semester course of micro- and research endeavor takes place. Good an- macroeconomics taught by a senior thropological science, in this view, is a member of the department is the man- move away from the misguided search datory gateway course for all students for objective truth, precisely de½ned and who seek to do more study in econom- carefully tested causal hypotheses, and ics. This course is, everyone agrees, to- the cumulation of small ½ndings; it is a tally conventional; that is its purpose. move toward recognizing the inevitable A chaired professor in the department role of the investigator’s biases and flaws proposed an alternative gateway course at the center of the research process. In in microeconomics that would teach the anthropology, as in all disciplines, there same textbook but then explicitly ana- is disagreement, but there the backlash lyze the assumptions underlying the against the hegemonic paradigm is neoclassical model; the department swinging in the opposite direction from voted overwhelmingly not to permit it the concurrent backlash in political except as an elective. (Departmental science. faculty who were out of town took the In the end, I am reasonably optimistic almost unheard-of measure of voting by about the foreseeable outcome of the proxy, and the president of the universi- social science wars, at least for political ty spoke on behalf of the extant course; science. The apsr is becoming more this, despite a petition for the alternative eclectic in its assessment of what consti- course signed by hundreds of students tutes the best work, and other journals and alumni.) It is hard to conceive of a may follow its lead. Perspectives on Politics sociology or political science depart- is opening channels for communication ment collectively deciding that all of across sub½elds and rival frameworks. its majors must take one particular two- And the best graduate students and jun- semester course that is always taught in ior faculty are simply doing an end run the same way before taking any other around the boring old methods wars, by course in the discipline. In economics, learning how to combine diverse episte- in short, the meaning of ‘science’ is clear mologies and modes of analysis in new and widely shared; at issue is whether and flexible ways–and that is good news the mainstream can be overturned, rath- for the future of my discipline. er than how broadly it is to be de½ned.

94 Dædalus Winter 2004 Americans seem to insist, more than they have any right to, that even the most tragic situations must yield to a frightfully unreasoning optimism, so that all boats, in the end, are ‘uplifted.’) Gerald Early My mother had bought David and the Phoenix for me at a store in Philadelphia on literature called Laura’s, a second-hand shop that & childhood was down the street from our home. It cost ten cents. I don’t know why she bought it. Maybe Laura suggested it as something “your kids might like,” as she was wont to say to my mother about cer- tain items. My mother knew nothing about the book except that it was written for children. That fact alone seemed to make it acceptable, and potentially even ‘educational.’ My mother did not read The book that most deeply affected me books, but she respected them, as the as a child was David and the Phoenix by unliterary sometimes do, as a kind of tal- David Ormondroyd. First published in isman, conferring some strange virtue of 1957, the book is about a boy who be- mind. comes friends with a wise and some- I remember ½rst reading Ormon- times wisecracking phoenix, until it droyd’s book in the third grade. It was a burns and dies and then rises again, big step for me. Until then, I had mostly leaving the boy forever. The phoenix read picture books, things like Maj Lind- was especially appealing to me, since it mann’s Flicka, Ricka, and Dicka and Snipp, personi½ed resurrection, thus making Snapp, and Snorr series, and–my favor- death not death at all, but some sort of ites at the time–H. A. Rey’s Curious cosmic learning experience. (One fea- George books. By contrast, David and the ture of some American children’s litera- Phoenix was a chapter book, with per- ture is its third-rate Emersonianism, its haps one picture per chapter instead of remarkable mixture of childhood angst one per page. It looked like a novel (a and the regenerative power of pluck: very thin novel), not a kiddie book, so I felt rather grown-up when I tackled it, even if I was a little daunted. Gerald Early is Merle Kling Professor of Modern At the time, I was quite sick and out of Letters and director of the Center for the Hu- school. (I suffered several severe bouts of manities at Washington University in St. Louis. illness during my days in elementary The author of “The Culture of Bruising: Essays school.) For many weeks I lay in bed on Prize½ghting, Literature and Modern Ameri- with glasses of 7UP on a tray next to my can Culture” (1994) and “Tuxedo Junction” bed, beside a small stack of my favorite (1988), among other books, he has been a Fellow comics and David and the Phoenix. of the American Academy since 1997. His latest In between contemplating the patterns book is “This Is Where I Came In: Black Ameri- in my bedroom wallpaper, I read David ca in the 1960s” (2003). and the Phoenix over and over again. With each reading, I became more skilled as a © 2004 by the American Academy of Arts & Sciences Dædalus Winter 2004 95 Note by reader and was more moved. Indeed, the These are large questions. But study- Gerald Early book’s charms seemed to magnify the ing books like David and the Phoenix and more I read it. David and the Phoenix was the audiences they attract may help us to not only my favorite book–it had be- answer them. If we could understand, in come my favorite possession. some measure, what ‘children’s litera- Years later, having developed a schol- ture’ is supposed to be, then we might arly interest in children’s literature, I understand a bit better what ‘childhood’ learned that David and the Phoenix was a is supposed to mean. popular book. Until then, I had thought We sometimes suppose that books of it as my book, which is not unusual aimed at children are more imaginative with certain things from our childhood. than those aimed at adults. But when I Even though modern childhood in the was a child and wanted to read some- United States has been turned into a thing imaginative, I didn’t go to a chil- training ground for adult consumers, we dren’s book–I struggled with an adult are shocked to ½nd our solipsism violat- work, or read pulp ½ction. I thought ed by the reality of marketplace culture James Bond and Dickens novels were the and mass audiences. most thrilling stuff, far and away, that I Perhaps consumption is why no one, read as a kid. And while Dickens pro- in the end, really escapes childhood in duced some of the most memorable chil- our culture. We simply learn how to pro- dren’s characters in the history of Eng- long it, and reenact it. What was once, lish literature, his work, by and large, before the nineteenth century, a rather was not intended for children. (The less negligible phase of life, and for most said about my juvenile taste for Ian people surely not an especially pleasant Fleming the better.) And some imagina- period, has now become something that tive literature that is given to children to everyone has a right to enjoy, and is read, like traditional fairy tales, I found thought to be the best time of one’s life. more puzzling and disturbing, but not Just as reformers wept about child labor- more imaginative, than reading many ers in the nineteenth century, we weep adult books. In fact, has not experience today when we hear about the murder- taught us that adults are more suscepti- ous child soldiers in Liberia and the ble to make-believe than children, and Congo who have been denied a child- far more skilled at creating it? What is hood; they inhabit societies that lack the it that Hans Christian Andersen’s “The structures to support childhood as we Emperor’s New Clothes” is supposed to understand it. tell us, if not about the willful self-delu- On the other hand, Americans don’t sion of adults and the literal-mindedness mind trying children in our courts as of children? In any case, it is generally adults when they commit some heinous not children who create so-called chil- or grotesque crime, which, of course, dren’s literature. raises some questions: What separates a That is surely one reason why chil- child from an adult? How does a child dren’s literature is not always simpler cross that line? How can childhood end than literature aimed at adults, although while one is still a child? Can a child, it may be easier to read. There are a good through his or her own acts, lose the many formulaic books for adults (ro- right to a childhood? To what extent is a mance novels and many mysteries, for child responsible for his or her acts? instance) that are less complex and less

96 Dædalus Winter 2004 intellectually challenging than a book by toms and practices of the genre so well Literature Roald Dahl or Madeline L’Engle or E. B. that we can usually spot a children’s & childhood White. book without having to be told: it is a Some may wish to speak of children’s particularly remarkable and peculiarly books as more innocent, or as appealing conventionalized form of intergenera- to, or reflecting, the putative innocence tional communication, of intergenera- of children. This formulation assumes a tional art, that has become an especially characteristic about children that largely important form of expression, of educa- exists as a psychological tangle in the tion, of consumption in industrialized minds of most adults. Some of the chil- countries. It paradoxically socializes dren I went to school with were far more children by granting them a degree of brutal, petty, and cruel than any adults I independence in their powers of dis- have interacted with–and some of the cernment and in the indulgence of adults were pretty bad. The one comfort their taste. I take in thinking about my childhood is What it says to children is as important the assurance that I will not have to re- as what it says about children as a read- live it and be at the mercy of children ing public and about the adults who again–and I think my childhood was make this literature for them. Children’s pretty good and I have no especially literature is profoundly important–both bad memories or traumas to speak of! sociologically and artistically. At their As a child one rather expected the best, the books aimed at children ex- adults to be an arbitrary and disappoint- press how adults feel conflicted about ing lot: always resorting to the power of their childhood–and how this feeling their size or the power of their purse reflects an ambivalence that children when everything else failed, and general- also feel about childhood. ly acting with the whimsical authority That is why this literature speaks to one expects from a paternalistic ruling adults as well as to children. As adults– class that both loves and feels terri½cally at least if our rearing was reasonably inconvenienced by its subjects. I often normal–we never outgrow childhood. think it is a great misfortune that chil- We learn to live with what our child- dren have to be reared by beings who hoods have made us, as we learn to live used to be children. No one has more with the idea that, as Wordsworth sug- confounding views of childhood than gested in his “Intimations Ode,” “The former children. Nearly all of our ideas Child is the Father of the Man.” about how to relate to children as adults In the early 1970s, more than a decade stems from the experiences we had as after my infatuation with David and the children and from our efforts to ‘correct’ Phoenix had faded, I was sitting in a mistakes or to replicate the way we our- friend’s college dorm room one night, selves were raised or think we were listening to a song by Doug and Jean raised. Carn called “Power and Glory.” Doug, In short, I don’t believe we can specify a pianist and songwriter, and Jean, a useful criteria for de½ning children’s lit- singer, had put out a series of black con- erature by describing how it differs from sciousness jazz albums on the Black Jazz literature meant for adults. label. Often they wrote and recorded But of course a children’s book is not lyrics to famous jazz tunes like Col- an adult book. We understand the cus- trane’s “Acknowledgement” section

Dædalus Winter 2004 97 from the suite “A Love Supreme,” or Horace Silver’s “Peace,” or Wayne Shorter’s “Infant Eyes.” Their music was a frothy blend of black consciousness and Emersonian uplift that character- ized the Black Power movement as I Richard Stern experienced it–something wonderfully and richly aesthetic and moral. (When I on a writer’s was younger, I sometimes wondered if endgame black power wasn’t partly explained by the yearning of black people for a golden childhood for the race itself.) In a spirit of racial holiness, I heard Doug and Jean sing, “Those that were lost shall surely be returned”–and out of nowhere I recalled David and the Phoenix. I knew that book as well as I knew my own name, but as a child I could not, for the life of me, explain Haven’t I given specimen clues, if no what it meant to me. But when I heard more? At any rate I have written enough Doug and Jean’s song, I realized that to weary myself–and I will dispatch it to David and the Pheonix had taught me two the printers, and cease. But how much– contradictory yet complementary truths how many topics, of the greatest point about childhood. First, that some things and cogency, I am leaving untouch’d! about childhood are lost beyond recov- ery, and we are pained rightly or wrongly –Walt Whitman, “Last Saved Items” by the loss. Second, and more profound- ly, that most children’s literature is In January of 2002, I retired from ½fty- about lost children returning home. So three years of teaching, forty-six of them childhood is about the hope of recovery, at the . For tenured how everything that is lost is returned. professors of my time, the decision to Haunted by loss and return, who can retire is one’s own. I won’t go into the simply bid farewell to childhood? As pros and cons that weighed on me for Raymond Chandler wrote in another more than a year. One pro, though, was context, no one has learned a way to say that there would no longer be the slight- goodbye to that. Richard Stern’s most recent novel is “Paci½c Tremors” (2001). His ½fth ‘orderly miscellany,’ “What Is What Was,” was published in 2002. This year Northwestern University Press will pub- lish his collected stories, “Almonds to Zhoof,” and reissue three of his novels, “Stitch” (1965), “Other Men’s Daughters” (1973), and “Natural Shocks” (1978). Stern has been a Fellow of the American Academy since 1995.

© 2004 by the American Academy of Arts & Sciences 98 Dædalus Winter 2004 est academic obstacle to writing. Since the book and ½lm–but in a bar in South A writer’s I was in the midst of writing what I Chicago. The feeling that he had not endgame thought might be my strongest novel, taken proper care of his brother, that he this was a large pro. had, because of that, been partly respon- For about six postretirement months, sible for his death, fused with lyric, I did work reasonably hard on it. Then Hemingwayesque recollections of grow- the excellent assistant to whom I dictat- ing up ½shing and ½ghting in Montana ed–I won’t go into my compositional to create the groundswell of A River Runs habits–left for the summer. I was partly Through It. Another source was the pub- relieved, for I felt that I no longer needed lic and off-the-cuff sermonizing of his the stimulus of her pen poised over the benevolent but competitively critical pad of yellow paper waiting for my father whose reactions to Norman’s words. I’d spent the academic year early compositions may well have led 1999–2000 without such assistance to his decades of literary silence. and managed to write a great deal in the In Chicago, the father’s critical place manner of my ½rst twenty-½ve or thirty was taken by the strict senior scholars writing years. of the English department and once, I’m I did though take a pause, partly to re- afraid, by me. My second year at the uni- lax, partly to reflect on this span of expe- versity, Norman showed me the manu- rience in the lives of other writers. script of an un½nished book on General Retirement from teaching had pushed Custer, on which he’d worked for a long my old friend and longtime colleague time. He asked me to be as strict–“as Norman Maclean through a door he Stern”–with his manuscript as I–some- might otherwise never have gone. what notoriously–was with the student He’d not been regarded as an impor- essays I supervised. So I was strict, mark- tant scholar, had published fewer than ing the Custer manuscript heavily, lacing a handful of articles, but was thought into its phraseology, conception, and of as a great teacher. organization. Norman said he was grate- Then in his last seventeen years he ful for that, but since it was the last I, or, shifted into a very different gear and I believe, anyone else, heard of the man- produced two ½ne books at an age that uscript, I regretted my severity for many sees most people, let alone most writers, years. The question the Custer manu- rocking on the porch of recollection. script posed and answered was how a The ½rst was a best-selling memoir- pedestrian, semi-scapegrace of a general novel, A River Runs Through It; the sec- had turned into a national myth whose ond, the posthumously published Young picture hung in every saloon in the coun- Men and Fire, was brilliantly carved by try. It was clear to me that Custer was a Alan Thomas and, to a lesser extent, by version of, an objective correlative for, Wayne Booth, out of a much larger man- Norman’s feelings about himself. That uscript. Maclean became not only a cele- alone should have stirred me to the gen- brated writer, but a role model for retir- erosity Norman claimed not to want. ees (he was featured this way in such But it didn’t. I forgive myself now by periodicals as People). thinking that if he had ½nished, let alone His two books had been stirring in him published that book, he might not have for decades. The ½rst flowed from the written A River Runs Through It. death of his gifted daredevil brother My postretirement pause from writing Paul, not in a Montana bar ½ght–as in stretched from the summer of 2002 into

Dædalus Winter 2004 99 Note by fall. Only very occasionally did I work on where I’d never before stayed. The room Richard Stern my novel. we were ½rst shown was smaller than I wasn’t worried. I’d never experi- most jail cells, but for a few extra dollars enced writer’s block and wasn’t experi- we moved into a penthouse suite where encing it now. What I did experience was we were surrounded by sky and sky- a sort of creative lassitude, similar to the scrapers. My wife was too exhausted to sort Gustav von Aschenbach, the writer- go out, but I went for drinks and dinner protagonist of Thomas Mann’s Death in with the husband of my late sister and Venice, experiences. Perhaps mine was his companion, a woman whom, sixty more a disinterest, an it-doesn’t-matter years earlier, I myself had taken out. I-don’t-careism. When I got an idea for Meanwhile, I had a good conversation something to put in the book, I not only with my complicated oldest son and let it drift, I pushed it away. After a few arranged to meet him at noon the next months, I told myself, “Maybe it’s time day. I also spoke with my agent who told to round things off. I’ve ½lled enough me that he’d decided to retire and move pages. I’m not going to just keep going to . like an industrial plant turning out Three hours later, I was back in the clones of what I’ve already written.” I hotel, ready for sleep, but tingling. My thought of other writers, mostly my con- mind surged with thoughts that could temporaries, who I thought were doing not stop falling into patterns that sur- that. I wasn’t angry at or disgusted by prised, even thrilled me. I felt too tired their activity. “Let them do what they to get up and write them down–my wife want, I’ll do what I want.” was sleeping and I’d have to do this in I believe that I’ve never needed to the bathroom–but I realized that if I write; I’m not a driven writer. I wrote be- didn’t, I’d lose something I didn’t want cause I wanted to. Even when the work to lose. So I ½nally got up, took my note- was dif½cult, even agonizing, whatever book, and wrote down a rough version was necessary to keep going was there of one set of thoughts headed “Autumn for me. Now it wasn’t. I wrote a few 1962,” then, skipping a page, another set small pieces, kept writing my notebook, headed “Coda: 2003.” These dealt with read even more than usual, saw my the day’s rainy return to the town where friends, enjoyed the leisure of an unde- my parents and I were born and raised manding life, and that was that. and where my oldest son had spent most Then in March my wife and I went of his adult life. down to a seaside house in Georgia to Thinking about, more, feeling, this spend a mostly isolated ten days in the return somehow unleashed thoughts sort of ease we enjoy down there. I read about 1962, when I worked in Venice as a ½ve or six books, and, as is my habit, Fulbright professor while back home, in thought how well or poorly their authors America, the Cuban missile crisis boiled had put them together, but except for as close to annihilating the world as had writing my journal, wrote nothing. ever happened. The public and personal The eleventh day we flew into a bleak, events, the ten years of my ½rst marriage rainy during the evening preceding 1962 and the ten years follow- rush hour, were threaded by an ingen- ing that ended in divorce; the changing ious cab driver through Queens to the life of me and my children, cascaded in 59th Street Bridge, then to our hotel on my head. Forty years after that Venetian East 57th Street, an area I know well but year, I was about to see the son who part-

100 Dædalus Winter 2004 ly blamed me for what he–but not I– George Washington Bridge had been A writer’s regarded as the failure of his life. The built. There were the Palisades, the endgame morning after that, my wife and I would Cloisters, the bridge, Grant’s Tomb, the attend the bat mitzvah of our niece as mystic New York skyline, all more beau- another public crisis–the coming war tiful to me than ever. We headed for the with Iraq–boiled around us. These hap- day’s ½nal event–a raucous, rock- penings, feelings, and thoughts were charged, dj-led party of ninety thirteen- bubbling in my head in a way I recog- year-olds at the Columbia Faculty nized as the desire, maybe even the need, House, the girls and boys gawking at to write, one I hadn’t felt in months. home movies, raising eyebrows at adult The next forty-eight hours were rich speeches, screaming, dancing, and writ- with these and other events: ½rst, a ing politically incorrect slogans on wash- moving four and a half hours talking room mirrors with lipstick and mascara. and walking the packed, beautiful East My literary machinery was processing Side streets with my son, then dinner at as much of this as it could, even as my a good French restaurant with a very tal- stomach churned. Hours into the ented, very rich friend whose wealth had screams, music, proclamations, dancing, helped despoil her talent, then, on Satur- food, and champagne, I made it into the day, taking the train from Grand Central air to throw up, ½rst in a Department of Station–after coffee surrounded by cop- Sanitation trashcan, then for four hours rousted, fearful, fearsome bums–to back in the hotel. Chappequa where, in a beautiful syna- Sick, one pays attention to nothing gogue designed by the odd genius Louis but one’s misery, but later I realized that Kahn (who had, decades earlier, died in the vomiting was caused–in my writer’s Pennsylvania Station, his body uniden- view–not just by my reaction to the noi- ti½ed for two days), we rose and sat, rose some racket of the teenage party and my and sat, recited, chanted, and listened to complex feelings of pleasure, warmth, the unending bat mitzvah ceremony of anger, and disgust at the bat mitzvah, our dear niece. Its texts were full of a but by the wonderful Virginia Woolf peaceful, pastoral Israel where swords essay “On Being Ill,” which I’d read two were beaten into ploughshares; my head nights earlier and about which I’d been was full of tanks, barbed wire, blood, talking in connection with the ½lm The boys and girls not much older than my Hours, which my wife and I had seen the niece and grandchildren belting them- week before in Georgia. selves into human bombs. I will not add further to the ingredi- Into the pit of Chappequa teenagers, ents in my literary pot. During my trip the forceful young rabbi plunged, com- to New York, everything in that pot had manding them in vain to cease giggling, begun, like a good stew or, better, a pot- gossiping, gum-chewing. What a cere- au-feu, to affect everything else, and mony. Its ½nest moment came when my now it was my job–one I wanted–to brother-in-law spoke directly and mov- ½nish cooking, serving, and eating it. ingly to his tearing daughter. After, we It is now some months since my liter- went for lunch to their house, two hun- ary desires returned. Except for what dred yards from that of our ex-president I’ve put down here, I have not used the and his senator-wife, then, toward dusk, material conjured up in New York, nor drove down the Saw Mill River and Hen- have I yet returned to the novel I’d ry Hudson Parkways, a drive that I’d ½rst stopped writing eleven months ago. taken seventy years ago, before the Dædalus Winter 2004 101 Note by Here, all I want to add are some fessorial one, it revived, perhaps to live Richard Stern thoughts about the retirement compo- again on its own. nent of this endgame of mine. When one Perhaps. retires from a job one has had for a long time and begins living on what one has socked away, there is for someone like me a psychic gearshift. One is no longer earning, adding, accumulating; one is now subtracting, living off a compara- tively ½xed pile. It may be large enough so that one doesn’t worry excessively about it lasting; it may be managed not by one of the great corporations whose shenanigans, criminality, or mismanage- ment have broken the contract with– and the backs of–its workers, but by the good-as-gold managers of tiaa-cref– nonetheless there is the gearshift into dependence. Since I no longer have a university salary, I can no longer claim on tax forms expenses I used to be able to claim. This loss reminds me that, as Shakespeare’s Moor cries, “Othello’s occupation’s gone.” The armor of professionalism is large- ly gone, a relief in some ways–one is freer, lighter–but in others, not: one sometimes feels bare in the chill wind of the everyday world. Worst of all, the great force½eld of one’s classes is gone. Into that ½eld one was able to drop the ideas and observations that continue to bubble up in and around one. One can no longer say, “Oh yes, I’ll talk to the class about this when we discuss Kafka next week.” As a writer, of course, I still have a force½eld; I am still a professional, but, as I’ve indicated, after a few post-teach- ing months, I had more or less stopped being a writer. The writing life had for ½fty years been part of a double profes- sional life. Now, like a Siamese twin whose twin has died, it edged toward the void. Then, in New York, where the writ- ing life had started years before the pro-

102 Dædalus Winter 2004 Letter to the Editor of Dædalus

Cambridge’s ½rst African Ph.D.? November 26, 2003 To the Editor: I was struck by the statement in the Summer 2003 issue of Dædalus that Kwame Anthony Appiah “is reputed to be the ½rst African to have received a Ph.D. at Cambridge” (page 104). It is quite certain that this is false, and I would expect there to have been a num- ber even before 1939. However, I only need to ½nd one previ- ous to Appiah’s in 1982 to substantiate this, and I myself remember Ben Laing from Ghana who took his Ph.D. in 1958 from R. A. Fisher’s department of genet- ics. I believe he became a professor of botany in Accra. So far as I know he is African on both sides of his family.

Professor A. W. F. Edwards Gonville and Caius College University of Cambridge

Dædalus welcomes correspondence from its readers. Letters to the Editor should be sent to Dædalus, 136 Irving Street, Cambridge ma 02138, or emailed to [email protected].

Dædalus Winter 2004 103 President Patricia Meyer Spacks Executive Of½cer Leslie Cohen Berlowitz Vice President Louis W. Cabot Secretary Emilio Bizzi Treasurer John S. Reed Editor Steven Marcus Vice President, Midwest Center Martin Dworkin Vice President, Western Center John R. Hogness

Dædalus Board of Editors Committee Steven Marcus, Chair, Joyce Appleby, Russell Banks, Stanley Hoffmann, Donald Kennedy, Martha C. Nussbaum, Neil J. Smelser, Rosanna Warren, Steven Weinberg; ex of½cio: Patricia Meyer Spacks, Leslie C. Berlowitz

Committee on Publications & Public Relations Guy W. Nichols, Chair, Emilio Bizzi, Joan Bok, Jesse H. Choper, Denis Donoghue, Jerome Kagan, Jerrold Meinwald; ex of½cio: Patricia Meyer Spacks, Steven Marcus, Leslie C. Berlowitz

Inside back cover: Learning depends on neurons. This computer diagram of a shows what a sampling of neurons in about mm3 of early visual cortex might look like, based on light microscope data of actual tissue. The branch-like structures are den- drites, which generally receive signals from other neurons; the long tails are axons, which send signals to other neurons; and the spheres are the cell bodies, which contain the cell’s energy plant, dna, the structures for making proteins, and so on. See Patricia Smith Churchland on How do neurons know? pages 42–50. Image courtesy of Gred Hood, John Burkardt, and Greg Foss, Pittsburgh Super- computing Center.

Dædalus coming up in Dædalus: Dædalus

on happiness Martin E. P. Seligman, Richard A. Easterlin, Martha C. Nussbaum, Journal of the American Academy of Arts & Sciences Anna Wierzbicka, Bernard Reginster, Robert H. Frank, Julia E. Annas, Roger Shattuck, Darrin M. McMahon, and Ed Diener Winter 2004 Winter 2004: on learning on progress Joseph Stiglitz, John Gray, Charles Larmore, Randall Kennedy, on learning Howard Gardner What we do & don’t know 5 Sakiko Fukuda-Parr, Jagdish Bhagwati, Richard A. Shweder, and others Jerome Bruner The psychology of learning: a short history 13 Alison Gopnik Finding our inner scientist 21 on human nature Steven Pinker, Jerome Kagan, Vernon Smith, Richard Wrangham, Daniel John Povinelli Of apes & men 29 Patrick Bateson, Thomas Sowell, Jonathan Haidt, and Donald Patricia Smith Brown Churchland How do neurons know? 42 Michael Tomasello Learning through others 51 on race Kenneth Prewitt, Orlando Patterson, George Fredrickson, Ian Susan Carey On the origin of concepts 59 Hacking, Jennifer Hochschild, Glenn Loury, David Hollinger, Clark Glymour The automation of discovery69 Victoria Hattam, and others

poetry David Ferry October 78 on imperialism Niall Ferguson, Kenneth Pomeranz, Tzvetan Todorov, Anthony Pagden, Jack Snyder, Akira Iriye, Molly Greene, William Easterly, ½ction Robin Blackburn, Henk Wesseling, and others Chuck Wachtel The Annunciation 79

notes Jennifer Hochschild on the social science wars 91 plus poetry by Richard Wilbur, Franz Wright, Rachel Hadas, Gerald Early on literature & childhood 95 W. S. Merwin, Charles Wright, Richard Howard &c.; ½ction by Richard Stern on a writer’s endgame 98 Erin McGraw, Roxana Robinson, Victor LaValle, Sigrid Nunez &c.; and notes by Michael Wood, S. George H. Philander, Linda letter on Cambridge’s ½rst African Ph.D. 103 Hutcheon, Charles Altieri, Donald Green, Shelley Taylor, Robert Nagel, Philip L. Quinn, Michael Kremer, Hue-Tam Ho Tai, Peter Hohendahl, Morris E. Fine, Richard Sobel, Stephen Hubbell &c.

U.S. $13 www.amacad.org