Genes in Development: Re-reading the Molecular Paradigm 9780822387336

Citation preview

Genes in Development

science a n d c u lt u r a l t h e o ry

A Series Edited by Barbara Herrnstein Smith

and E. Roy Weintraub

Genes in Development RE-READING T H E M O L E C U L A R PA R A D I G M

Edited by

Eva M. Neumann-Held and

Christoph Rehmann-Sutter

Duke University Press Durham and London 2006

© 2006 Duke University Press All rights reserved Printed in the United States of America on acid-free paper  Designed by C.H. Westmoreland Typeset in Minion by Tseng Information Systems, Inc. Library of Congress Cataloging-inPublication Data appear on the last printed page of this book.

Duke University Press gratefully acknowledges the support of the Swiss Academy of Sciences, which provided funds toward the production of this book.

CONTENTS

Introduction eva m. neumann-held and christoph rehmann-sutter 1 I Empirical Approaches 1 Genome Analysis and Developmental Biology: The Nematode Caenorhabditis elegans as a Model System thomas r. bürglin 15 2 Genes and Form: Inherency in the Evolution of Developmental

Mechanisms stuart a. newman and gerd b. müller 38 II Looking Back into History 3 From Genes as Determinants to dna as Resource: Historical Notes on

Development and Genetics sahotra sarkar 77 III Theorizing Genes 4 The Origin of Species: A Structuralist Approach gerry webster

and brian c. goodwin 99 5 On the Problem of the Molecular versus the Organismic Approach in

Biology ulrich wolf 135 6 Genes, Development, and Semiosis jesper hoffmeyer 152 7 The Fearless Vampire Conservator: Philip Kitcher, Genetic

Determinism, and the Informational Gene paul e. griffiths 175 8 Genetics from an Evolutionary Process Perspective

james griesemer 199 9 Genes—Causes—Codes: Deciphering dna’s Ontological Privilege

eva m. neumann-held 238 10 Boundaries and (Constructive) Interaction susan oyama 272

Contents 11 Beyond the Gene but Beneath the Skin evelyn fox keller 290 12 Poiesis and Praxis: Two Modes of Understanding Development

christoph rehmann-sutter 313 IV Social and Ethical Implications 13 Developmental Emergence, Genes, and Responsible Science

brian c. goodwin 337 14 Nothing Like a Gene jackie leach scully 349

Contributors 365 Index 369

vi

INTRODUCTION

eva m. neumann-held and christoph rehmann-sutter In 2003, on the fiftieth anniversary of the discovery of the double helix, Carina Dennis and Philip Campbell, editors of a special issue of Nature, prefaced a collection of survey articles under the heading ‘‘The Eternal Molecule’’—a title that uses evocative religious language. This book is about the noneternal, changing understanding of the significance of dna. Significance belongs to the language of culture and subjectivity. None of the various views discussed here casts doubt on the notion that dna is tremendously important for life on Earth. But the authors’ ideas of how its role should be represented, explained, and evaluated differ considerably. This book explores central topics in ontology, philosophy of science and nature, and even ethics. Heredity is one aspect. Clearly, the sequence of dna is crucial in some way for the transmission of form and characteristics to the next generation, and thus for evolutionary change as well. But should dna always be viewed as the most significant component of transgenerational transmission and evolutionary change? 1 Here, development may be the key that helps us understand the role of dna in heredity by clarifying its involvement in the processes of the construction of organisms. In recent years, the discussion of dna’s role and significance has been broadened by the attention paid to development. In contexts of research, public understanding, and societal applications of genetic knowledge the issue has been how much causal influence can be attributed to dna and how much is due to other factors, for example, epigenetic inheritance, environmental influences, and cultural effects. Those who have attributed too much significance to dna and too little to extra-dna factors have been called gene centrists (other names have been used, too, not all so polite); the other faction is sometimes referred to as the developmentalists. The debates between these and other less polarized positions will no doubt continue; the results from the Human Genome Project will certainly not settle them. Each new piece of information about a gene opens up new

Eva M. Neumann-Held and Christoph Rehmann-Sutter

questions about causal interactions with extragenetic contexts en route to the phenotype: What function does dna perform within those processes that bring about the form and characters of the next generation? What is dna’s concrete agency within the differentiation of cells, or in organ formation? Has dna any ‘‘agency’’ or ‘‘activity’’ at all (very anthropomorphic terms)? Is it not instead a passive store of differences, whose effects on the organism are determined in very sophisticated ways by the cells—that is, by the organic context? What do the terms code, information, instruction, and so forth, really mean? Which contexts count? Which constellation of factors is most essential? Scientists have investigated the myriad details of the molecular interactions and biochemical events that occur around dna; they have also discovered some of the core principles of regulation and organization. But questions about the basic theoretical concepts used in understanding and interpreting those results still are open and deserve attention. They are the topic of this book. The relevance of these questions about the role of dna in development reaches far beyond research itself. Since its origin, molecular biology, and particularly molecular genetics—symbolized by dna—has had controversial historical, cultural, and social impacts. It has undoubtedly changed our image of the composition, development, and evolution of living organisms on Earth and of the principles of life itself. It has joined the chorus of voices that describe us, our bodies, our identities, our fates. More than that, molecular genetics has sometimes taken on a solo part. But beginning in the 1980s, together with the discovery of the details of how dna works in the context of the developing body, the words and melodies of that solo part started to diverge. The images generated by the molecular approach to the development and heredity of organic life have become subject to reinterpretation. Even the ideas and metaphors about the role of the genes in development continue to evolve. Originally titled ‘‘Genes and Development,’’ the present title ‘‘Genes in Development’’ shows that we came to see that our thinking encompasses both organic development and conceptual refinement. The genome has become a dominant theme of our time. And the genetic program, however critically discussed by scientists themselves, remains the most effective model for helping the public to understand genetics. Alternative metaphors are lacking: contextual thinking has not yet discovered a narrative and metaphoric language of comparable power to talk of programs. In some of the discourses about social and societal impact the only remaining questions seem to be whether or not we should embrace the biotechnologi-

2

Introduction

cal possibilities opened up by our knowledge of dna, and what the consequences will be if we do so? The scientific path is often said to be clear, needing nothing but pragmatic realization, while the open questions are located in the realms of ethics or technology assessment. But there are other questions. They are hidden, so to speak, in Walter Gilbert’s famous imaginary compact disc, which contains the data of our three billion dna base pairs and allows us to carry our personal genetic information in our pockets (Gilbert 1992: 96). What in fact would we have on such a cd? What would be its biological significance, its meaning for us? And how would people integrate the contents of their cds into their beliefs about the body, soul, time, the world, and so on? We understand the term molecular paradigm to mean the present strategy of research in the life sciences; that is, the study of developmental processes through the analysis and manipulation of molecular interactions at the level of gene regulation. A closer look at the scientific basis, however, reveals that even here the molecular paradigm, which explains wholes through molecular parts and which often identifies dna as the central determinant, is by no means generally accepted or unanimously construed. A suitable starting point for a critical review of the developments in molecular biology and developmental genetics would therefore be the search for scientific alternatives to the molecular paradigm. If such alternatives exist, then it is most likely that their evaluation depends—at least in part—on the tools of the philosophy of science. There is no doubt that the research strategy of the molecular paradigm has been scientifically fruitful. Its aim—to integrate genetics and developmental biology (which had already engulfed the older discipline of embryology)— is not, however, the product of the latest advances in molecular biology. A synthesis between genetics and embryology was attempted in the first decades of the twentieth century, in the early history of modern genetics. At that time neither the conceptual nor the empirical tools for such an integration existed. Thus, embryologists like Thomas Hunt Morgan turned away from embryology and devoted themselves to developing experimental genetics (see Morgan 1934). In the ensuing decades, new research objects (‘‘model organisms’’) and new theoretical tools became available and proved scientifically successful, generating new research agendas and facilitating the design of new explanatory strategies. The methods of classical genetics were applied to mutant bacteria and bacteriophages by analyzing the mutant phenotype in terms of differences in the production of proteins such as enzymes. In the early 1950s experimental evidence identified the biochemical molecule dna

3

Eva M. Neumann-Held and Christoph Rehmann-Sutter

as the material basis of heredity. Finally, the identification of the ‘‘coding relation’’ between the nucleic acid sequences of dna and rna, on the one hand, and the amino acid sequences of polypeptide chains, on the other, closed the explanatory gap between the material of inheritance and the construction of phenotypes (via polypeptides). These insights and the analytical tools of biochemistry led to the construction of enzymatic reaction chains, and François Jacob and Jacques Monod proposed their ‘‘operon model’’ of gene regulation along these lines. Researchers like Sydney Brenner and Lewis Wolpert outlined research programs that approached biological developmental processes as genetic networks (Brenner 1973; Wolpert and Lewis 1975; Lewin 1984). After the conceptual switch to viewing development as guided by an underlying genetic program, the empirical tools and technical processes available underwent tremendous advances. Today, the empirical investigation of systemic developmental processes is being conducted at the subcellular, molecular level. Major parts of the research agenda of the genetic program and many basic concepts, however, have remained largely unchanged. The divide between regulatory and structural genes introduced by Jacob and Monod, the regulatory cascades of Conrad Waddington, and the master genes in regulatory hierarchies of Brenner and Wolpert all seem to have found their empirical counterparts in specific, functionally definable dna sequences and networks of temporally regulated and causally interdependent gene activations.2 A glance at current textbooks suggests that, at least on the molecular level, the bridging of genetics and development has been a successful research strategy. On the basis of the widely accepted assumption that organisms can be viewed as assemblies of smaller pieces and lower levels of complexity (from the molecular to the cellular levels, and so on), the expectation seems justified that successes in understanding developmental processes at the molecular level will eventually result in an understanding of developmental processes at higher levels as well. In view of these patent successes, it might come as a surprise to be confronted with developments in science that may aptly be called an ‘‘irony of history’’ (Keller 1995: 21). The more deeply scientists have searched for the mechanisms by which individual genes control development, and the more details they have learned about the functioning of genetic programs and genetic blueprints, the less plausible the metaphors of genetic control, programs, and blueprints seem. The new evidence includes the alternative splicing of the rna transcripts of single ‘‘genes,’’ leading to the production of different polypeptides from the same stretches of dna, a phenomenon which may help

4

Introduction

to account for the surprisingly small number of genes in the human genome; the fact that even master control genes can have variable functions in other developmental contexts; and the phenomenon of mrna editing, by which the base composition of an mrna transcript is changed in vivo (by insertion, deletion, or exchange) prior to its use as a template for protein synthesis. Other examples of the context dependency of gene function in the field of the so-called epigenetic programming (Rideout, Eggan, and Jaenisch 2001) could also be cited. Richard Strohman (1999) saw reason to conclude, from the still incomplete data on the human genome, that: (1) there is not sufficient information in genomic databases to provide explanations for complex functional attributes of cells and organisms; (2) therefore, there must be other informational systems and operating rules that complement genomic systems; (3) epigenesis is one such system; and (4) the program rules by which regulation is produced are extragenomic and are most likely to be found not in molecular mechanisms per se but in their integration into complex gene circuits and, more peripherally, in their connectedness with regulatory networks (metabolic and other) of cellular dimensions. The roles of other molecules, of the architecture of the cell, of the organism-in-an-environment, and of the ‘‘whole system’’ (by whatever definition) are downgraded if we subsume them all under the notion of ‘‘necessary conditions.’’ They complement, or may even establish, what has been described as the informational content of the genes. What this means for the concepts ‘‘gene’’ and ‘‘development’’ must still be worked out. In all fields of contemporary life sciences, the molecular paradigm is being thoroughly reread. Although the image of development as the implementation of genetically encoded information (the narratives of the genetic blueprint or genetic program) still lies at the center of many theoretical and empirical explanatory approaches, a growing number of scientists are calling for a change in these basic views. There is an increasing awareness of the tension between the empirical results produced by ingeniously designed experiments, on the one hand, and an interpretation of the results based on the orthodox view of the causal significance of genes, on the other. The existence of different explanatory approaches raises the question of whether their validity can be tested by empirical analysis alone. A resolution of the tension we have described between evidence and interpretation requires a thorough analysis. The realm of developmental biology and genetic theory has thus become a field of research for both the theory of science and the philosophy of biology. Different

5

Eva M. Neumann-Held and Christoph Rehmann-Sutter

philosophical positions adopt different stances toward the epistemological and ontological assumptions made, with different consequences for assessing the relationship between scientific advances and their cultural environment. In principle, two general positions can be distinguished. Ontologically committed positions focus on the impact that different molecular or genetic theories have on ethics, anthropology, and metaphysics (Quine 1969). According to this model, the impact of science on our worldview is not merely a given fact, but is also theoretically justified. It is assumed that reading and interpreting ‘‘the genome’’ means both mapping the human body and configuring human identity. Alternative genetic theories can be viewed in terms of their consequences in the societal, anthropological, and ethical realms. By contrast, the methodical culturalist position (Hartmann and Janich 1996, 1998; Janich 1996) dispenses with ontological claims without denying the validity of scientific statements. Scientific manipulations and concepts are viewed as methodical developments from (successful) cultural practices. Scientific endeavors are thus seen as enterprises that are developed on the basis of the successful manipulation of the environment by particular means for specified ends. The existence—and further development—of different genetic theories (e.g., population genetics, classical genetics, and molecular genetics) is traced back to different ends and means, which render these theories valid within their individual frameworks. However, it does not follow that these theories must be reducible to each other, or that they describe ‘‘reality.’’ As a consequence, within the framework of this philosophical approach, any attempt to reconfigure human identity with reference to scientific concepts needs a justification that cannot be given by scientific results and theories themselves. In other words, the actual impact that science has on shaping our society and culture is more narrowly limited. Other differences beyond the relation between ontology and epistemology may also be helpful when trying to evaluate the diverse theoretical approaches to a critical reinterpretation of the ‘‘molecular paradigm.’’ The contributions in this book discuss several of them. Some of the new skeptical considerations have not yet made their way into the public domain. There is no doubt that in the realm of public discourse the molecular approach to the scientific study of organisms—including human beings—has already ‘‘geneticized’’ our views of ourselves and has shaped the predominant problem-solving strategies that we use even outside the domain of scientific research. The growing prominence of genetic approaches to medicine, agriculture, and technology cannot avoid affecting our worldview.

6

Introduction

Recombinant organisms, genomics, molecular medicine, predictive diagnostics, genetic engineering, nuclear transfer cloning, and even the reprogramming of stem cells have entered our everyday conversation. We spoke of ‘‘cultural meanings.’’ There is a basic tension between the universalist explanatory vision of genetics and the many diverse cultural determinations uncovered by cultural studies, the social sciences, and anthropology. While genetics points out causal relationships, the references to ‘‘culture’’ emphasize malleability and the historical task of innovative selfappropriation. Culture pertains to the domain of social subjectivity (Eagleton 2000: 39). But however differently socialized, all of us share the same molecular mechanisms, so that the geneticist John Sulston, a Nobel laureate, can legitimately state that dna is ‘‘our common thread’’ (Sulston and Ferry 2002); it runs through all our different cultures and subcultures. However, the account of dna as causal determinant leaves open the basic questions of what genes are in development, how their role in the contexts of other factors and relations can be explained, and how these explanations are related to other symbolic representations of the world. We do not see genetics as an antithesis to culture. Rather, it is a field in which basic components for cultural identities are formed and discussed. Applying the celebrated phrase of Charles Taylor (1989), genetic knowledge is among our modern ‘‘sources of the self.’’ Genes have no inherent subjectivity, but they are nonetheless introduced into the area of social subjectivity. Genetics in its broader sense is therefore an ambivalent resource for our explanations of existence and identity. The image of the ‘‘genetic code’’—possibly a cryptic narrative of development and becoming—is currently one of the most culturally forceful metaphors for the human constitution; it both dominates and restricts self-reflection and self-interpretation (Wils 2002: 188). With this said, we want to add that the different ‘‘culturifications’’ of dna urgently require study in their own right, as contributions to our forms of social life. They need clarification and philosophical critique. Recent years have seen the development of several distinct biological and philosophical alternatives to the gene-centric readings of the molecular paradigm. In 1999 we had the opportunity to explore some of these issues by convening an international symposium at the University of Basel, Switzerland (‘‘Genes and Development: Interacting Processes or Hierarchical Organization? New Theoretical Approaches to Developmental Biology and Their Ethical Implications’’). The papers presented and discussed there were the source for essays that form the basis for about half of this volume. To them

7

Eva M. Neumann-Held and Christoph Rehmann-Sutter

we have added additional points of view in the form of invited contributions. The essays in this volume present some of the most prominent explanatory approaches to genes in development—biosemiotics, developmental systems approach, epigenetics, hermeneutics, methodical culturalism, process structuralism, and others—and discuss their implications. Although these approaches diverge on many issues, importantly, they converge on others. The book opens with two essays by authors with a background in empirical research. Thomas Bürglin provides an introduction to the molecular genetics of development in the small nematode Caenorhabditis elegans, the first animal for which a complete genomic sequence became available. He explains the ‘‘gene’s-eye view.’’ Stuart Newman and Gerd Müller discuss the relation of developmental and evolutionary aspects in biological theory and develop ideas for a complementary epigenetic perspective. The next two chapters provide a historical dimension. Sahotra Sarkar traces anti-genecentric approaches to development from early ‘‘norm-of-reactions’’ concepts to current accounts in the age of genomics and proteomics, including his own ‘‘sequestered modular template model of the cell.’’ Historically, ‘‘process structuralism’’ has certainly been among the first and most influential approaches to a critical theory of evolution and development. Also in this section we reprint extracts from an influential 1982 paper by Gerry Webster and Brian Goodwin, a text that has stimulated many contemporary lines of thought within biology and philosophy, along with the authors’ current views of that work. A third set of essays is dedicated to different interpretative approaches to genes in developmental processes. Ulrich Wolf compares the organismic approaches, introduced on the one side by Cassirer and on the other by critical evolutionary theorists like W. Gutmann, and discusses concepts of gene function in the context of an epigenetic approach to the organism. Jesper Hoffmeyer applies ‘‘biosemiotics’’ to gain insights into a new philosophy of genetics. Arguing within the framework of the ‘‘developmental systems approach,’’ Paul Griffiths scrutinizes the information metaphor in genetic discourses. James Griesemer adopts an integrative approach by introducing and systematically comparing different empirical theories and philosophical evaluations. Eva M. Neumann-Held discusses the causality attributed to genes in different explanatory frameworks. She introduces the ‘‘process molecular gene’’ and criticizes realistic interpretations of the ‘‘genetic code.’’ In a pair of related essays Susan Oyama, whose name is synonymous with the ‘‘developmental systems approach,’’ and Evelyn Fox Keller discuss problems of body boundaries. Christoph Rehmann-Sutter adds a hermeneutic perspec-

8

Introduction

tive to genetic processes in development that can be described in the interpretative frames of ‘‘poiesis’’ or ‘‘praxis.’’ The two final chapters introduce social and ethical considerations. Brian Goodwin’s essay is a plea for a ‘‘responsible science’’ of development. Jackie Leach Scully examines the role of genetics in the social perception of identity and the ethical, medical, and social impact of different genetic theories. We thank the foundation Mensch-Gesellschaft-Umwelt at the University of Basel, Switzerland, whose generous financial and institutional support made the first years of our research possible (grant F 42/95: Genome and Organism. Philosophical Interpretation of Developmental Biology), especially Leo Jenni, Felicitas Maeder, and the members of the expert group, in particular Uwe Gerber. For additional financial and institutional support we thank the Europäische Akademie zur Erforschung von Folgen wissenschaftlicher Entwicklungen Bad Neuenahr-Ahrweiler GmbH, in particular Carl Friedrich Gethmann and Armin Grunwald; and the Kulturwissenschaftliches Institut Essen, in particular Lutz Wingert. The Swiss Academy of Science supported the editorial work. We are grateful to J. Reynolds Smith, Sharon Torian, and Pam Morrison from Duke University Press for their support in the editing and publishing process, and to Barbara Herrnstein Smith and E. Roy Weintraub for including this book in the Science and Cultural Theory series. Three anonymous referees made helpful comments on an early draft of the book manuscript. We thank Franziska Genitsch for indispensable administrative and typographic help, Rouven Porz for practical support in the production phase of the manuscript and motivating assistance, Mindy Conner and Carol Wengler for careful copyediting and proofreading, and Alexis Wichowski for compiling the index: without their help this book would not have been possible. We thank Jackie Leach Scully, Monica Buckland Hofstetter, Birgitta Fischer, and Louise Röska-Hardy for providing efficient and professional English revision and also for clarifying discussions. Last but not least, our very special thanks go to all our authors.

Notes 1 This has been questioned over the years by numerous researchers with different perspectives and research backgrounds. To name a few relevant works, in historical order: Gutmann and Peters 1973; Gutmann and Bonik 1981; Goodwin and Webster 1981, see this volume; Oyama 1985; Jablonka and Lamb 1995; and Avital and Jablonka 2000.

9

Eva M. Neumann-Held and Christoph Rehmann-Sutter 2 Historical accounts from different perspectives are given in Hudson 1996; Kay 2000; and de Chadarevian 2002.

References Avital, E., and Jablonka, E. 2000. Animal Traditions: Behavioural Inheritance in Evolution. Cambridge: Cambridge University Press. Brenner, S. 1973. The genetics of behaviour. Br. Med. Bull. 29: 269–271. Chadarevian, S. de. 2002. Designs for Life: Molecular Biology after World War II. Cambridge: Cambridge University Press. Dennis, C., and Campbell, P. 2003. The eternal molecule. Nature 421: 396. Reprinted in: J. Clayton and C. Dennis (eds.), 50 Years of dna (p. 82). New York: Macmillan. Eagleton, T. 2000. The Idea of Culture. Oxford: Blackwell. Gilbert, W. 1992. A vision of the Grail. In: D. J. Kevles and L. Hood (eds.), The Code of Codes (pp. 83–97). Cambridge: Harvard University Press. Gutmann, W. F., and Bonik, K. 1981. Kritische Evolutionstheorie. Hildesheim: Gerstenberg Verl. Gutmann, W. F., and Peters, D. S. 1973. Konstruktion und Selektion: Argumente gegen einen morphologisch verkürzten Selektionismus. Acta Biotheor. 22: 151– 180. Hartmann, D., and Janich, P. 1996. Methodischer Kulturalismus. Frankfurt: Suhrkamp. Hartmann, D., and Janich, P. 1998. Die Kulturalistische Wende. Frankfurt: Suhrkamp. Jablonka, E., and Lamb, M. J. 1995. Epigenetic Inheritance and Evolution. Oxford: Oxford University Press. Janich, P. 1996. Konstruktivismus und Naturerkenntnis. Frankfurt: Suhrkamp. Judson, H. F. 1996. The Eighth Day of Creation. Woodbury, N.Y.: Cold Spring Harbor Laboratory Press. Kay, L. E. 2000. Who Wrote the Book of Life? A History of the Genetic Code. Stanford: Stanford University Press. Keller, E. F. 1995. Refiguring Life: Metaphors of Twentieth-Century Biology. New York: Columbia University Press. Lewin R. 1984. Why is development so illogical? Science 224: 1327–1329. Morgan, T. H. 1934. Embryology and Genetics. New York: Columbia University Press. Oyama, S. 1985. The Ontogeny of Information. Cambridge: Cambridge University Press. Quine van Orman, W. 1969. Existence and quantification. In: Ontological Relativity and Other Essays (pp. 91–113). New York: Columbia University Press.

10

Introduction

Rideout, W. M., Eggan, K., and Jaenisch, R. 2001. Nuclear cloning and epigenetic reprogramming of the genome. Science 293: 1093–1098. Strohman, R. C. 1999. Five stages of the Human Genome Project. Nature Biotechnol. 17: 112. Sulston, J., and Ferry, G. 2003. The Common Thread: Science, Politics, Ethics, and the Human Genome. New York: Bantam Books. Taylor, C. 1989. Sources of the Self: The Making of the Modern Identity. Cambridge: Harvard University Press. Wils, J-P. 2002. Der Mensch im Konflikt der Interpretationen. In: A. Holderegger, D. Müller, B. Sitter-Liver, and M. Zimmermann-Acklin (eds.), Theologie und biomedizinische Ethik (pp. 173–191). Freiburg, Switzerland: Universitätsverlag; and Freiburg: Herder. Wolpert, L., and Lewis, J. H. 1975. Towards a theory of development. Fed. Proc. 34: 14–20.

11

1 G E N O M E A N A LY S I S A N D D E V E L O P M E N TA L B I O L O G Y The Nematode Caenorhabditis elegans as a Model System

thomas r. bürglin A general and basic tenet of scientific research is to simplify a complex problem to smaller, more tractable units that can be studied and unraveled. Depending on the biological question, scientists choose particular organisms as model systems. Each model provides researchers with particular biological or experimental advantages that help them in their quest to understand fundamental biological principles and problems. In 1963 Sydney Brenner wanted to proceed from bacterial and viral genetics to a more complex, multicellular animal. He proposed studying a small nematode that he thought would be eminently suitable for investigating many aspects of cell and nervous system development. After careful deliberation he chose Caenorhabditis elegans, and with this model he succeeded in establishing a whole new research field. As a result, ‘‘founding father’’ Sydney Brenner and two other C. elegans researchers, John Sulston and Bob Horvitz, were awarded the Nobel Prize in Physiology or Medicine in 2002. John Sulston made key contributions to elucidating the C. elegans cell lineage as well as to its genome project, and Bob Horvitz contributed significantly to the understanding of programmed cell death. I became attracted to this model system because of its elegance and other advantages I will outline below. The little worm will serve here as an introduction to how researchers study genes and understand their function in the context of a living organism. This chapter will first introduce the advantages of the C. elegans model system (see also the key textbooks by Wood [1988] and Riddle et al. [1997]), and then will proceed to the C. elegans genome, where the principle of gene function via

Thomas R. Bu¨rglin

proteins is introduced (for a key textbook on molecular biology, see Alberts et al. 2002). Subsequently, I will talk about a particular group of genes that regulate other genes—the homeobox genes—and demonstrate how we study the function of particular genes in the worm. I hope that this exposition will remove at least some of the mystical connotations that the term gene has acquired in recent times and reveal the true beauty of the gene and the genome.

The Biology of Caenorhabditis elegans C. elegans belongs to a group of animals called nematodes. Nematodes are roundworms or threadworms with smooth-skinned, unsegmented, longcylindrical bodies. There are both free-living and parasitic forms, and they can be found in both aquatic and terrestrial environments. Quite a number of parasitic nematodes are known to afflict human beings: it has been estimated that as much as 25 percent of the world’s population is infected by some type of parasitic nematode. C. elegans is a small, free-living nematode found in temperate regions in the soil, where it feeds on bacteria (fig. 1). There are two sexes: self-fertilizing hermaphrodite animals and male animals. Both are small, the adult hermaphrodite being a little larger and reaching a size of 1.2 mm when fully grown. Because of its small size and simple food requirements, C. elegans can be easily reared in the laboratory on agar plates seeded with a lawn of bacteria such as Escherichia coli. The life cycle of C. elegans is extremely fast: it takes only about three days from the time a young adult starts to lay eggs until the next generation has grown and starts laying its own eggs. Development proceeds through several stages: embryogenesis, four larval stages (termed l1, l2, l3, and l4), and the adult stage. This extremely rapid reproductive rate is unique among multicellular animals. Each hermaphrodite animal can produce up to three hundred offspring, so a single animal on an agar plate can produce thousands of first- and second-generation offspring in about a week. C. elegans is thus an ideal model system for genetic studies because a large number of offspring can be analyzed in a very short time. Apart from easy maintenance, small size, and fast and plentiful reproduction, C. elegans offers many other advantages that have made it an excellent model system for modern biologists. In part because of their small size, the embryos, larvae, and adult animals are transparent when viewed under a microscope, allowing researchers to identify all the different cells (in fact, the cell nuclei) and making it possible to follow the cell divisions of embryo-

16

Genome Analysis

1 • Caenorhabditis elegans: larval stage 1 (top) and adult hermaphrodite (bottom). Adapted from J. E. Sulston and H. R. Horvitz, ‘‘Post-embryonic cell lineages of the nematode Caenorhabditis elegans,’’ Dev. Biol. 56 (1977): 110–156.

Note: The gray ovals shown in the l1 phase are neurons, with the major concentration (‘‘the brain’’) being around the pharynx in the head. The gonad primordium consists of only four cells; these will develop during the larval stages into two arms, one toward the anterior and one toward the posterior, that will flex back as they grow until their ends reach the middle of the body again. rnai experiments and generation of transgenic animals involve microinjections into the two ovaries of adults.

genesis and larval development. The fact that the cell division patterns are remarkably reproducible from embryo to embryo permitted John Sulston and his co-workers to establish the complete cell lineage for C. elegans. That means we know exactly how each cell divides during development, what it gives rise to (e.g., a neuron, a muscle cell, or a gut cell), and what its relatives are. Thus, we know that the adult hermaphrodite animal has 959 somatic cell nuclei, and the male animal has 1,031. These cells can be classified into about 150 different cell types. The greatest complexity of cell types is found in the nervous system, which consists of 302 neurons and 56 glia and associated cells in hermaphrodites, and 381 neurons and 92 glia and associated cells in males. The 302 neurons of the hermaphrodite include at least 118 different types. John White and his co-workers examined serial sections of complete animals by electron microscope and analyzed their neuronal connections, and suc-

17

Thomas R. Bu¨rglin

ceeded in elucidating the complete wiring diagram of the nervous system. All of this information constitutes an unparalleled fund of knowledge about a complex multicellular animal. Despite its small size, the nervous system of C. elegans has an astounding complexity which permits the animal to respond to many different environmental cues. The animal has a large set of behaviors that allow it to survive and propagate: it pumps food with its pharynx (‘‘the throat’’), expels digested food through the anus, retracts when it bumps into an object with the tip of its head (the nose), and moves forward or backward in sinusoidal waves to move toward or away from particular stimuli such as chemical attractants or repellents, touch, heat, or fine temperature differences. The hermaphrodite lays eggs through the vulva, while the male animal has a specialized set of behaviors that enable it to find hermaphrodites and fertilize them with its specialized tail structure. Researchers are now using genetic and molecular tools to unravel the function of the nervous system to understand how the worm’s behavior is controlled by genes.

The Molecular Approach and the Genome of C. elegans The size of the haploid genome of C. elegans—that is, the dna contributed by one parent (an organism is usually diploid, having dna from both parents)—is fairly small, consisting of 100 million base pairs distributed on five autosomes (i.e., ‘‘regular’’ chromosomes not involved in sex determination) and one X chromosome (i.e., the sex-determining chromosome). Short pieces (about 30,000 to 300,000 base pairs) of C. elegans dna were cloned into bacterial or yeast vectors so that the dna could be amplified, mapped, sequenced, and distributed to researchers. Clones covering the complete genome are available from the Genome Center, either as cosmids (essentially, special types of plasmids which are grown in E. coli bacteria) or as cloned yacs (yeast artificial chromosomes, which are grown in yeast). These clones were mapped and used to establish a physical map of the genome of C. elegans. In mapping, the relative order of these clones was established to determine where precisely each clone lies with respect to other clones and where these clones are located on the chromosomes. After the clones had been placed in order they were sequenced. Since the end of 1998 the virtually complete genomic sequence of C. elegans has been available; and the finalized, totally gap-free sequence has been available since late 2002. At the time of

18

Genome Analysis

this writing C. elegans is the only multicellular animal for which this can be said. The delay in achieving the complete sequence lay in the fact that some pieces of dna are particularly difficult to clone and sequence, a general problem not restricted to C. elegans. Usually such dna does not contain many important genes; these hard to clone regions are often highly repetitive in nature and may be nonfunctional ‘‘junk’’ dna, or may contribute to chromosomal structure. Dispersed among the long strands of dna that make up the chromosomes are the genes. Most genes are pieces of dna that are transcribed and translated into proteins, although some genes are transcribed only into rna. A protein-coding gene usually consists of several stretches of chromosomal dna called exons that are separated by introns, although some genes have no introns (fig. 2). When a gene is to be activated in a particular cell, specific transcription factors sit down on the dna in the regulatory region, also referred to as ‘‘gene control region.’’ Transcription factors have the ability to recognize very specific short sequences of dna, and each type of transcription factor has a specific sequence (called a ‘‘binding site’’ or ‘‘enhancer element’’) that it recognizes. The extent of a gene control region is defined by all the different binding sites and other regulatory dna elements that are necessary for the correct regulation of the gene. Thus a control region can be small or large, depending how far the regulatory binding sites are spread out on the chromosome. The transcription factors binding in the control region make the gene accessible so that the enzyme rna polymerase can bind to the promoter and transcribe the dna of the gene into precursor rna (fig. 2). Thus, genetic information is converted from dna into rna. The precursor rna has to be spliced so that the intron sequences, which do not code for proteins, are removed and only the exons are left. The spliced rna—the messenger rna, or mrna—is then translated by the ribosome into a chain of amino acids: the protein. The ribosome binds to the start codon of the protein-coding region in the mrna and translates the mrna into a string of amino acids (the protein) until it reaches the stop codon. When a gene is ‘‘turned on’’ or ‘‘switched on’’ by the process outlined above, then this gene is said to be ‘‘expressed.’’ An enzyme called ‘‘reverse transcriptase’’ can convert mrna back into dna; this is mainly done in the laboratory for cloning purposes. The dna thus formed is called complementary dna (cdna). Randomly cloned and sequenced cdnas are also called expressed sequence tags (ests) because they can be used as ‘‘tags’’ for expressed genes. A mutation in the chromosomal dna can destroy or affect the function of

19

2 • The transcription and translation of a protein-coding gene. A small

piece of the chromosome containing one gene is shown; in the case of C. elegans the size is typically about one-half to one-fifth of the sequence contained in a cosmid. Classical definitions of a gene include the regulatory regions as well, which are mostly found upstream of the gene, although some important elements can also be found in introns or after the transcribed region. Transcription factors bind to the enhancer elements in the regulatory regions of the gene and activate the transcription of the gene by recruiting initiation factors and rna polymerase to the promoter. The gene is then transcribed by rna polymerase, which produces an rna copy from one strand of the dna. The opening of the dna is indicated by the separating thin lines that symbolize base pairs.

Genome Analysis

the final gene product, the protein, at many levels. For example, mutations in the regulatory region will lead to levels of the gene product that are either too high or too low (too much or too little activity). Mutations in introns may cause inappropriate splicing that results in mrna that will not be translated properly and thus produces aberrant proteins. Mutations within the exons can change individual residues in the protein or introduce a premature stop codon, while deletion mutations can remove whole parts of a gene. (This is done in the laboratory to produce gene knockout mutants in C. elegans, as discussed below). Large deletions often lead to a loss of function, but smaller deletions or changes in individual residues can lead to hyperactivity of the gene product. Proteins can have many functions: some are structural components, some catalyze chemical reactions, some fight disease, and some transport molecules through cells. Examples of structural genes are the components of hair or of the minute fibers that constitute the skeleton of a cell. Catalysts of chemical reactions are called enzymes; among other things, enzymes digest food, generate energy, and synthesize signal molecules in neurons (neurotransmitters). Proteins such as insulin are hormonal signals, while proteins called antibodies are produced by the immune system to inactivate microorganisms such as bacteria or viruses. The well-known protein hemoglobin in blood transports oxygen through the body. Proteins may not be active immediately after they are synthesized; some need to be chemically modified (e.g., by attachment of sugars or phosphate) or cleaved into fragments, and some must bind to other proteins to form an active complex. Although most genes produce proteins, there are some exceptions. A few produce only rna molecules; the biochemically active entity is the rna molecule itself, which is not translated into a protein. Examples are the ribosomal rnas. The ribosome, mentioned above, is a large complex made up of many proteins and ribosomal rna molecules. Finding genes in genomic dna sequences is not a trivial task. C. elegans has a considerable advantage over other organisms here, especially mammals, because the junctions between introns and exons (the splice sites) are quite highly conserved. In conjunction with other characteristic features, this allows the protein-encoding genes to be identified by computer methods (bioinformatics). While computer prediction alone is not always sufficient to identify genes, it is often an excellent starting point for experimental work. The analysis of thousands of ests has allowed the identification of expressed genes and helped refine the computer prediction programs for genes. We know now that C. elegans has about 20,000 genes in its genome.

21

Thomas R. Bu¨rglin

3 • Hypothetical model of how, after cell division, one daughter cell could be made different from the other daughter cell. P indicates phosphorylation.

How Cells Become Different The basic question developmental biologists seek to answer is: How do the different cell types develop from a single cell? All of the cells of a particular organism, whether muscle cells, skin cells, or gut cells, have the same number of chromosomes and thus the same number of genes. And yet, differentiated cells do different things. Neurons produce specific neurotransmitters, muscle cells make proteins that can contract to produce movement, blood cells make hemoglobin, and so on. How is this achieved? Earlier I described how genes are turned on (or off ) by specific transcription factors sitting down in the promoter region. Let us now consider a hypothetical case in which a ‘‘mother’’ cell divides into two daughter cells. The anterior one becomes a neuron, the other becomes a skin cell (fig. 3). Within the mother cell there is a protein, say n, which is an inactive transcription factor waiting to be activated (through phosphorylation, for ex-

22

Genome Analysis

ample). When the mother cell divides, the anteriormost daughter cell touches other cells located anteriorly. This releases a surface signal which activates (in this case phosphorylates) the inactive transcription factor n. The active factor n can now bind to the specific binding sites it recognizes in the regulatory regions of its target genes. The genes activated by factor n could be neuronal growth factors, neurotransmitters, or other genes which turn the cell into a neuron. In the posterior daughter cell, factor n is not active, but perhaps another factor, S, that turns on skin cell-specific genes is activated (not shown in fig. 3). While this scenario is rather simplistic, it illustrates the principle of how cells become different. In reality, hundreds of factors regulate cell fate and cell differentiation, some factors regulate others, and transcription factors can have feedback loops to their own regulatory regions. Thus, complicated cascades of transcription factor events turn different genes on and off during the development of an organism to generate the vast diversity of cell types found in the adult animal.

Homeobox Genes as Developmental Control Genes Genes that influence developmental events were first found almost one hundred years ago by geneticists studying mutations that caused abnormalities of development in the fruit fly Drosophila melanogaster, including a type of mutation called ‘‘homeotic.’’ Homeotic mutations lead to partial or complete transformations of particular body regions in the fly. For example, a segment can be transformed such that it resembles its anterior neighbor segment. In the most extreme case, mutations in a particular gene can give rise to a fly with four wings instead of two wings and two halteres, because the halteres of the third thoracic segment are transformed into wings. Dominant mutations in another type of gene can cause transformations of antennae into legs. Because of these findings, the homeotic genes were considered to be important for body patterning along the anterior-posterior body axis of the fruit fly. When molecular cloning techniques became available at the end of the 1970s, researchers embarked on the process of cloning these genes (cloning in this context means identifying the sequence of the particular gene; not a trivial task in the early days of molecular biology when no genome projects existed). Some of the first developmental control genes cloned were homeotic genes. When the first sequences for some of these genes were obtained, it was found that they share a conserved sequence element of about 180 base pairs;

23

Thomas R. Bu¨rglin

this was called the ‘‘homeobox’’ because it had been found in homeotic genes. We now know that not all homeotic genes are homeobox genes and, conversely, that not all homeobox genes are homeotic genes. Homeobox genes have been found to play important roles in all kinds of developmental events. The homeobox codes for a part of a protein (a domain) called the homeodomain. The homeodomain is a small protein domain—about sixty amino acids long—that binds dna. The amino acid chain that forms a protein folds into a complex three-dimensional structure; often, specific parts of a protein fold into domains with particular function(s), and within a domain, the chain of amino acids forms substructures such as rodlike alpha helices or sheetlike beta sheets. The structure of the homeodomain has been elucidated (for a review, see Gehring et al. 1994). It consists of three alpha helices that fold over each other, the third helix sitting in the major groove of the dna (fig. 4). There is now ample evidence that homeodomain-containing proteins act as transcription factors. Thus, because homeobox genes encode transcription factors and play pivotal roles in development, some of them have been called ‘‘master control genes’’ (Gehring 1998: 123ff.). Homeobox genes have been discovered not only in animals but also in fungi, plants, and slime molds. Thus we can safely conclude that all higher multicellular organisms have homeobox genes. At what point in eukaryote evolution the homeobox appeared we do not know, but it does not seem to be present in bacteria. Animals have quite a large number of homeobox genes. For example, C. elegans has more than 85 homeobox genes, and Drosophila melanogaster has about 120. The homeodomain proteins can be grouped into many different classes and families based on the sequence of the homeodomain and other conserved motifs flanking the homeodomain.

Function of Homeobox Genes Homeodomain proteins encoded by homeobox genes are a key group of transcription factors involved in many developmental events (Duboule 1994). Given the widespread occurrence of homeobox genes in higher eukaryotes, it is not surprising that the function of the proteins they encode is very diverse. Perhaps the best studied homeobox genes are those of the Hox cluster. The Hox genes of animals are involved in the patterning and specification of identity in regions along the anterior-posterior body axis. A striking aspect is that the order of the genes in the cluster is collinear with their function along

24

4 • The homeodomain structure. (A) Two views of the engrailed homeodomain bound to dna. The protein is highly simplified, with the three alpha helices represented as tubes and the rest of the protein as a line; the twisted ladder represents a short section of dna. Helix 2 and helix 3 form a helix-turn-helix motif that is also found in bacterial transcription factors. The third helix of the homeodomain lies in the major groove of the dna and makes the most important contacts. However, some dna contacts are also made by the amino-terminal arm of the homeodomain in the minor groove of the dna. (B) In this view, both the dna (light gray) and the homeodomain of Antennapedia (dark) are rendered as full space-filling molecules. Each ball represents an atom. While the structural organization of the protein backbone of the homeodomain is difficult to see in this picture, this representation makes it clear how tightly the homeodomain sits on the dna. View A is adapted from C. R. Kissinger et al., ‘‘Crystal structure of an engrailed homeodomain-dna complex at 2.8 Å resolution: a framework for understanding homeodomain-dna interactions,’’ Cell 63 (1990): 579–590.

Thomas R. Bu¨rglin

the anterior-posterior axis. Thus, the Drosophila gene labial (lab), which is located at one end of the Hox gene cluster on the chromosome, is expressed only in the very anterior of the animal, and the gene Abdominal-B (Abd-B), which is located at the other end of the gene cluster, is the most posteriorly expressed gene (expressed in the 5th to 8th abdominal segments). In vertebrates, the Hox genes are likewise involved in patterning along the body axis, and the organization of the Hox cluster has also been maintained. Homeobox genes are not involved only in pattern formation along the body axis. Some homeobox genes function very early in development, and others are involved in the final cell differentiation events. One of the earliest acting homeobox genes is found in Drosophila. The gene bicoid (bcd ) plays a key role in setting up the anterior-posterior axis in the early embryo. Mutant embryos lack head and thorax and develop tail structures at the head. The bcd rna is provided maternally, and the rna and the protein are localized at the anterior pole of the embryo. Despite its crucial role in the early development of Drosophila, the bcd gene is a relatively new gene in evolutionary terms. A case of a very late acting gene is the C. elegans prd-like gene unc-4, which is involved in the specification of motor neurons. Mutations in this gene lead to abnormal synaptic connections so that the motor neurons of a type called va receive synaptic input that is normally appropriate only for vb motor neurons. The consequence of these wiring defects is that the animals cannot move backward. The gene unc-4 is expressed in the va motor neurons, and the unc-4 protein confers va identity to these neurons. The regulatory role of unc-4 is thus one of the final steps in differentiating a subset of motor neurons.

C. elegans Methods Now that the complete genome sequence for C. elegans is known, and we know that there are about twenty thousand genes in this organism, we want to know what all these genes do. Quite a number of genes have been studied in the classical way; that is, they were first identified as a mutation and then cloned. But there are still thousands of unstudied genes. What methods do scientists use to study genes that are known only by their sequence? I will outline some commonly used C. elegans methods (see also Epstein and Shakes 1995) before describing results obtained with these methods for a homeobox gene.

26

Genome Analysis

Determining Where a Gene Is Expressed There are two major methods for determining where a C. elegans gene is expressed: antibodies and reporter constructs. Antibodies. The cdna for the gene is cloned into a bacterial expression plasmid and transferred into E. coli. The promoter region on the expression plasmid is then activated by adding an appropriate inducer (e.g., lactose), and the protein is produced in high quantities by the bacteria. The protein is separated from other E. coli proteins using biochemical methods and is then used to immunize an appropriate animal, such as a rabbit. Just as if the animal had been infected by a virus or bacterium, the rabbit’s immune system will start to produce antibodies against the foreign protein. Samples of the blood serum are collected from the rabbit after the antibodies have been produced. To study the expression of a particular protein in C. elegans, individuals are transferred into a mixture of methanol and cross-linking agents that preserve and fix their structures. The ‘‘fixed’’ worms are then incubated with diluted serum. The antibodies in the serum will penetrate into the worms and bind to the specific protein that they were made against, in those regions and cells of the worm where the protein is present. The location of the antibodies is then visualized. One way to do that is to use a second antibody that has been chemically coupled with a fluorescent dye, such as fluorescein 5(6)isothiocyanate (fitc). This second antibody binds to the rear part of the first antibody. The worms are then examined under a fluorescent microscope, and those areas of the worm containing the protein of interest will show green fluorescence. Reporter constructs. An alternative approach to determine where a gene is expressed is to take the regulatory region of the gene of interest and clone it in front of a suitable reporter gene, such as the gene encoding green fluorescent protein (gfp), which has the unusual property of fluorescing in green. The gfp gene, which occurs naturally in jellyfish, was cloned and is now widely used in molecular biology as a marker/reporter gene. The principle is to take the regulatory region of a gene of interest (e.g., a muscle gene, M) and merge this in front of the gfp gene in a bacterial plasmid. This regulatory region (gene M)-gfp fusion plasmid is grown and amplified in E. coli to produce many copies that can be injected into C. elegans. Transgenic C. elegans animals are then generated by injecting the plasmid dna containing the gfp fusion into the gonad of an adult hermaphrodite. A fraction of the offspring of the injected animal will then receive the plasmid dna, and this dna will be propagated like a little extra chromosome. The living transgenic animals

27

Thomas R. Bu¨rglin

are examined under a fluorescent microscope, and the gfp fluoresces green only in those cells where the regulatory region and promoter of our gene M is active. Why? The transgenic worms carry a small extra chromosome containing the regulatory region of gene M in front of the gfp gene. In muscle cells, the transcription factors that turn on gene M still turn on gene M, but they will also bind to the regulatory regions of gene M on the extra chromosome, and the gfp gene will be expressed in the muscle cells as well. Thus, the fluorescence of the gfp protein reveals the cells in which gene M is turned on.

Studying Gene Function It is not enough to know a gene’s expression pattern through space and time. To really understand the function of a gene and its product, we must also find out what goes wrong if the gene does not function properly. Again, there are two major ways to do this in C. elegans. rna interference (rnai). Double-stranded rna (dsrna) of the gene of interest is injected into the gonad of a hermaphrodite animal. The offspring often show phenotypes indicative of a mutation in this gene. Depending on the gene, some effects may even be seen in the injected animals themselves. It is often not even necessary to inject the dsrna: it is sufficient to soak the animals in it or to feed them bacteria that produce the dsrna. The molecular mechanisms through which this process works are still being elucidated. The dsrna triggers a process—perhaps a defensive mechanism originally against rna viruses—that destroys the dsrna. Since the introduced dsrna corresponds to an endogenous gene, the endogenous rna for that gene is also destroyed. rnai has limitations: for example, it is often impossible to see phenotypes in the nervous system. Nevertheless, due to the simplicity of the rnai method, large-scale projects are now under way to examine the gene function of all C. elegans genes by rnai, and a large number (more than sixteen thousand) have already been surveyed (Kamath et al. 2003). Gene disruptions. With this method a gene can be destroyed permanently. This is particularly appropriate for genes involved in nervous system function. Two variants of this method exist, one of which I will describe here. C. elegans are treated with chemical mutagens which induce small deletions of about 1 to 2 kilobases in the chromosomes. Since getting a deletion in precisely the gene of interest happens very rarely, thousands of populations of mutagenized worms (each grown on a separate small plate) are needed. Half the worms are washed off each plate, and the rest are frozen in liquid nitrogen for later use. The idea is that a mutagenized worm will produce many

28

Genome Analysis

offspring with the same genetic defect. Thus, if we can detect the gene deletion event in our total dna sample, we can go back to the frozen worms and recover more of these mutant worms alive. To detect deletions in the gene of interest, the polymerase chain reaction (pcr), a highly sensitive method of amplifying and visualizing specific regions of dna, is used. If pcr is performed on a normal worm, the pcr reaction will give a dna product with a length of, say, 3,000 bases. If a deletion has occurred, the product will be shorter, perhaps only 1,545 bases long. Thus, the principle is to test thousands of populations by pcr until a population is found in which a deletion has taken place. Once a positive population has been detected, the frozen sibling worms can be recovered and distributed to more plates so as to grow subpopulations, which are again tested by pcr. In this way, smaller and smaller subsets of the population are tested by pcr until individual mutant worms can be identified. The gene disruption method makes it possible to disrupt a gene and analyze the affected animals to determine their mutant phenotype—to see which functions or developmental processes are affected by this gene ‘‘knockout.’’ A specific example in which the above methods have been employed will help to elucidate the function of a C. elegans homeobox gene.

The C. elegans LIM Homeobox Gene ceh-14 The C. elegans homeobox gene ceh-14 (ceh stands for C. elegans homeobox) belongs to the lim class of homeobox genes. This class of homeobox genes has two domains, called lim, upstream of the homeodomain. The ceh-14 gene has homologues in flies as well as in vertebrates. In mice, the homologous genes Lhx3 and Lhx4 have been shown to be important for the growth, specification, and axonal outgrowth of motor neurons in the central nervous system. We first determined the expression pattern of ceh-14 using antibodies and gfp reporter constructs. The ceh-14 gene is expressed in several different tissues: the nervous system, the spermatheca in the gonad, and the hypodermis. Here I will focus on the nervous system. From larval to adult stages, neuronal expression is observed in one interneuron (called ala) and a pair of sensory neurons (called afd) in the head, and in a pair of interneurons in the anterior body (figs. 5 and 6). In addition, ceh-14 is expressed in the tail in the sensory phasmid neurons, and in several different types of interneurons (Cassata et al. 2000).

29

5 • The head (A) and tail (B)

of C. elegans. Some of the ceh14-expressing neurons on the left side (L) of the animal are shown with their processes. Drawing by H. Kagoshima and T. R. Bürglin.

6 • The expression of ceh-14 is visualized in the transgenic animal carry-

ing a ceh-14 promoter fused to gfp. The bright cells that fluoresce are the afd neurons and the ala interneuron. The thin dendritic processes of the afd neurons reaching to the tip of the nose can be seen, as can axonal processes that come from neurons in the tail of the animal. Photo by H. Kagoshima and T. R. Bürglin.

Genome Analysis

To study the function of ceh-14 we obtained a deletion in ceh-14 using a pcr-based screening approach similar to the one outlined above. Since ceh14 is expressed in a series of neurons, we examined behavioral defects in the mutant worms. The phasmid neurons (called pha and phb) in the tail are sensory neurons with processes (dendrites) that are exposed to the environment (fig. 5). It is possible to test to see if the phasmid neurons are still open to the environment. When worms are incubated in a fluorescent dye (e.g., fitc), sensory neurons such as the phasmids take up the dye through their sensory endings. In ceh-14 mutant animals this no longer happens. Thus, ceh14 is involved in regulating the outgrowth of the dendritic endings of the phasmid neurons. Another pair of neurons in which ceh-14 is expressed are the afd neurons, which have been shown by Ikue Mori and co-workers to be involved in thermosensation (Mori and Ohshima 1995). If C. elegans is grown at 20°C with plenty of food, the animals can remember these advantageous conditions: when placed on a plate with a temperature gradient from 15°C to 25°C (but no food), they will move toward the area of 20°C and in isothermal tracks at that temperature in search of food. Similarly, worms grown at 15°C will move toward 15°C, and those grown at 25°C will move toward 25°C. Although they cannot remember the original temperature precisely, once they have chosen a temperature, they can stay within 0.1°C of it so that they move on an isothermal track. Worms that carry a mutant ceh-14 gene, however, move randomly, chaotically, over the plate. This indicates that ceh-14 is important for the function of the afd neurons. Using gfp reporters that are specifically expressed in afd (e.g., the guanyl-cyclase gene, gcy-8), we observed that the afd neurons are still present in ceh-14 mutants and their morphology is normal at light microscopic levels. To study this in more detail, we analyzed serial sections under the electron microscope. The dendritic endings of the afd neurons normally form ‘‘fingers’’; in ceh-14 mutants the average diameter of the fingers was enlarged and fewer fingers were observed (Cassata et al. 2000). Thus, ceh-14 protein plays a subtle role in the finger morphology of the afd neurons, perhaps by regulating molecules located in these fingers. We think that ceh-14 regulates, probably directly, molecules involved in the sensation of temperature (Cassata et al. 2000). Although we have now determined several functions for the lim homeobox gene ceh-14, there remain outstanding questions. These include how ceh-14 is regulated (i.e., what are the genes that control ceh-14) and what the target genes of ceh-14 protein are, particularly the molecules involved in thermosensation. While much is known about the molecules involved in

31

Thomas R. Bu¨rglin

vision and chemosensation, relatively little is known about how temperature is perceived in higher animals. Further understanding will come once we can identify the transcription factors that regulate ceh-14 and the targets of ceh14 protein. I have described the molecular mechanisms of developmental control genes using homeobox genes as an example, but of course there are many more genes involved in developmental processes. Probably several thousand genes are necessary for the correct development of an organism such as C. elegans. Homeobox genes account for only about 0.5 percent of the genes in that organism’s genome.

Future Prospects: A Full Understanding of the Organism? C. elegans has become a very successful model system. For example, the work of Bob Horvitz and his colleagues has provided important insights into the process of programmed cell death during development. The underlying mechanisms are conserved in evolution, and are therefore highly relevant to the understanding of human disease as well. Programmed cell death plays a role in some forms of cancer and in various neurodegenerative diseases (Hanahan and Weinberg 2000). Other aspects of development, such as sex determination, vulval development, early embryonic cell divisions, and neuronal outgrowth, are also being studied in C. elegans, and important contributions have already been made in these fields. The simple genome of C. elegans has even made it possible to identify important genes in the molecular pathway involved in Alzheimer’s disease, which is of direct clinical importance for human medicine. Mutations in these genes cause developmental abnormalities in C. elegans, and the analysis of these genes has been instrumental to understanding the biological role of these molecules. Further, such analysis has allowed the identification of other molecules important in the pathway (Levitan et al. 1996; Baumeister 1999). The ceh-14 gene is only one of many developmental control genes now being studied. Information on dozens of genes is already available, and the large rnai screens will certainly speed up the identification of important control genes. The future of developmental biology and genetics will also be driven by technological advances. A recent methodology related to genome research is GeneChips, or dna microarrays, which works as follows. Gene fragments

32

Genome Analysis

are placed in a gridlike fashion on a solid substrate such as glass slide and then probed with rna or cdna prepared from the tissue or organ of choice. The more mrna copies of a gene are present in the sample, the stronger the signal on the chip will be. Thus, in a single experiment, expression levels of thousands of genes can be monitored. In the case of C. elegans the complete genomic sequence—and thus all putative genes—is known. GeneChips have been prepared containing all these genes. We can now probe these chips using rna from normal wild-type worms and rna prepared from our ceh-14 mutant worms. As we know, ceh-14 encodes a transcription factor. Thus, it is likely that some target genes will be expressed at lower levels in the ceh-14 mutant. By comparing the chip results of mutants and wild types, we hope to be able to identify target genes of ceh-14. And since the chips contain all the C. elegans genes, just a few experiments may reveal many target genes. This technique is used now for virtually every gene; one can, for example, compare cancer tissue with normal tissue to find which genes are specifically activated or repressed. The results from such experiments may be crucial for the diagnosis and treatment of different types of cancers. But gene chips, although promising, will not give us all the answers. There are sensitivity issues (genes expressed at low levels are not detected), and in the case of whole animals or organs, one is looking at a complex mixture of cell types: the chip results are an average over the whole, and do not reflect the expression levels of individual genes in individual cells. Nevertheless, molecular biology has made spectacular progress over the last couple of decades, and the amount of information already obtained is astounding. When I started my studies in 1978 I asked a well-known professor whether there were any computer programs for analyzing sequence information (the first sequences had been published at that point). His answer was, ‘‘Why would you need a program? You can write it down by hand.’’ Now we have complete genomes. At some point in the future it should be possible to describe all the components of an organism such as C. elegans with a limited number of genes and a defined number of cells. In the case of Drosophila, many of the genes involved in very early development have been described and analyzed. In fact, it has become possible to model the interactions and hierarchies of these genes by computer. The goal of achieving this for the complete development of C. elegans no longer seems so elusive. C. elegans is particularly well suited to attempts to model a complete organism with the computer. In each cell thousands of gene products are active (though not more than twenty thousand). The processing capacity of even a desktop computer should be capable of simulating the levels and activities of

33

Thomas R. Bu¨rglin

all the protein products within single cells in a reasonably short time. Thus, one thousand interconnected cheap computers should be able to simulate all of C. elegans. In reality fewer will probably be necessary, because there are only about 150 different cell types, and computing power is ever increasing. The only aspect of C. elegans that probably cannot be modeled yet at the molecular level is signaling in the nervous system. The signals of a neuron can act in milliseconds, and many channel-forming proteins must open and/or close for these processes to happen. Computers can now deal with one billion instructions per second, so there are a million instructions available per millisecond for a simulation. This may not be sufficient to track the opening and closing of hundreds of channels in addition to all the other molecules within a cell. But we should be able to model the behavior of individual neurons at a more simplistic level and perhaps mimic the neuronal signals that guide the behavior of the animal. Modeling the worm by computer will allow us to compare it with the living worm. If everything is correct, the model should behave like the real worm. Parts of the model that do not work will indicate where more experimental work is necessary to fill in the gaps in our knowledge. Although it still may take a long time, we should eventually be able to understand completely the functioning of a small animal like C. elegans. By extrapolation, we should also be able to understand more complex model organisms such as the fly, the zebrafish, or the mouse. The biggest riddle will be understanding human beings, with their large brains and complex behaviors. But only in model organisms can we perform the experiments necessary to understand the biological functions of thousands of genes.

Conclusion The information required to make a complete organism is contained within the genes of the genome. However, the genes alone are functionless; they need a complicated machinery of transcription and translation that is itself encoded in the genome. Thus, a genome can function only in the context of a living cell, which already has all the necessary molecules. This is very reminiscent of the chicken-and-egg question. Which came first? The answer, of course, is neither. Animals and all other living organisms have evolved over millions of years. There has been a continuous evolution from single-celled organisms to multicellular organisms. The advent of multicellular organisms created the need to regulate and control different cell types. This was solved

34

Genome Analysis

by means of new transcription factors, such as those encoded by homeobox genes, that became dedicated to regulating genes during development. Single-celled organisms have undergone a long process of evolution as well. But how did the genes in a unicellular organism come about? Theoretical and experimental evidence suggests that at the very origin of life, genes were probably coded not as dna but as rna, and furthermore there were no proteins: enzymatic and replication functions were carried out by rna molecules (Gesteland et al. 1999). Remnants of this rna world are still visible today. For example, rna molecules constitute critical components of the ribosomes. Also, some lower organisms have self-splicing introns which need no protein machinery but can catalyze their own removal from a precursor rna. We can imagine that at the very origin of life, there were simple rna molecules that had the capacity to replicate themselves. None of these ancient ancestral entities has survived through evolution, probably because the later dna and protein machinery-based cells were much more efficient. The pure rna world died out billions of years ago. How much of what is expressed in an organism is in the genes, and how much stems from the environment? While it is certainly true that the environment is important, it seems to me that many people place too much importance there. A recent metaphor compares the genome to a theatrical play in which the actors are individual proteins; the proteins are given rather autonomous qualities, suggesting that they can act and improvise freely like actors. However, I question the accuracy of this analogy. The gene products— that is, the proteins—can and do react to environmental influences, but they do not have autonomy. Each reaction to the outside is a direct consequence of the protein sequence, which in turn is derived from the gene sequence. Consider a hormone receptor; it reacts to the environment by reading the levels of hormone in the body. The function of this receptor can be influenced in multiple ways. For example, various drugs can interfere with the binding of the hormone or simulate the presence of hormone (remember all the doping scandals). But the receptor cannot be ‘‘trained’’ (like an actor) to do things it is not programmed to do by its genetic code; it cannot, so to speak, improvise. It behaves within the physical constraints of the structural and functional properties that are dictated by the sequence of its amino acids, which is ultimately encoded in the genome. Or let us take the case of cancer. Many cancers are caused by environmental influences; for example, when smoking causes lung cancer. But cancers develop not through changes in proteins per se, but because of specific mutations in the genomic dna. Some genes or their promoters are mutated so that abnormal proteins are produced, or too

35

Thomas R. Bu¨rglin

much or too little of a protein is generated. When several mutations that cause such abnormalities occur in the same cell, it can cause rogue, uncontrolled cell proliferation—cancer. Mutations in the dna occur all the time, but at a much higher frequency when carcinogenic substances or radiation cause extra damage. The genes that cause cancer encode normal proteins required for the proper function of a cell: usually they are involved in regulating the cell division cycle—that is, how often a cell divides. This also highlights why some individuals may have a higher than average risk of cancer: because they already have a mutation in some gene that predisposes them to cancer. And that is why cancer is so difficult to treat. The body’s own cells become abnormal, and there are hundreds of different types of cancers, depending on the cell type in which the disease arose and the particular mutations that occurred. Perhaps the best way to distinguish environmental from genetic influences is through the study of identical twins, which are natural clones of each other and share the same genome. Even twins who grow up in different environments share a remarkable set of common behaviors. An analogy for genes and environment that I find reasonably accurate is the soccer field. The genes (genome) define the borders (the field) within which the ball and the players can act. The proteins (ball and players) are constrained by their sequence and have only a limited range in which they can play. Only a mutation in a gene can change the range in which a protein can move. The protein, in other words, can now leave the field; in biological terms this could mean that a cancer may develop, or that over the long term a new species may evolve, in which the size and shape of the field will of course be different. I hope that I have succeeded in elucidating some basic aspects of the nature of the genome and how regulatory proteins such as homeodomain proteins can turn a gene on and/or off. To develop its potential the genome needs to be in the right environment—in the right cell. The homeodomain proteins help to create the right environment by activating the right genes, and consequently producing the right proteins in the right cells. Homeobox genes are, of course, only a tiny subset of all regulatory genes. And other types of molecules, such as signaling molecules and receptors, play key roles in creating the right cellular environment. Nevertheless, the ultimate reduction of the complicated process called life could be summarized as: everything is in the genome. But without the cell, the genome is nothing.

36

Genome Analysis

References Alberts, B., Johnson, A., Lewis, J., et al. 2002. Molecular Biology of the Cell. New York: Garland. Baumeister, R. 1999. The physiological role of presenilins in cellular differentiation: lessons from model organisms. Eur. Arch. Psychiatr. Clin. Neurosci. 249: 280–287. C. elegans Sequencing Consortium. 1998. Genome sequence of the nematode C. elegans: a platform for investigating biology. Science 282: 2012–2018. Cassata, G., Kagoshima, H., Andachi, Y., et al. 2000. The lim homeobox gene ceh-14 confers thermosensory function to the afd neurons in Caenorhabditis elegans. Neuron 25: 587–597. Duboule, D. 1994. Guidebook to the Homeobox Genes. Oxford: Oxford University Press. Epstein, H. F., and Shakes, D. C. 1995. Caenorhabditis elegans: modern biological analysis of an organism. Vol. 48. L. Wilson and P. Matsudaira (eds.), Methods in Cell Biology. San Diego: Academic Press. Gehring, W. J. 1998. Master Control Genes in Development. New Haven: Yale University Press. Gehring, W. J., Qian, Y. Q., Billeter, M., et al. 1994. Homeodomain-dna recognition. Cell 78: 211–223. Gesteland, R. F., Cech, T. R., and Atkins, J. F. 1999. The rna World. Cold Spring Harbor Monograph Series 37. Cold Spring Harbor, N.Y.: Cold Spring Harbor Laboratory Press. Hanahan, D., and Weinberg, R. A. 2000. The hallmarks of cancer. Cell 100: 57–60. Kamath, R. S., Fraser, A. G., Dong, Y., et al. 2003. Systematic functional analysis of the Caenorhabditis elegans genome using rnai. Nature 421: 231–237. Kissinger, C. R., et al. 1990. Crystal structure of an engrailed homeodomaindna complex at 2.8 A framework for understanding homeodomain-dna interactions. Cell 63: 579–590. Levitan, D., et al. 1996. Assessment of normal and mutant human presenilin function in Caenorhabditis elegans. Proc. Natl. Acad. Sci. USA 93: 14940–14944. Mori, I., and Ohshima, Y. 1995. Neural regulation of thermotaxis in Caenorhabditis elegans. 376: 344–348. Riddle, D. L., Blumenthal, T., Meyer, B. J., and Priess, J. R. 1997. C. elegans II. Cold Spring Harbor, N.Y.: Cold Spring Harbor Laboratory Press. Sulston, J. E., and Horvitz, H. R. 1977. Post-embryonic cell lineages of the nematode Caenorhabditis elegans. Dev. Biol. 56: 110–156. Wood, W. B. 1988. The Nematode Caenorhabditis elegans. Cold Spring Harbor, N.Y.: Cold Spring Harbor Laboratory Press.

37

2 GENES AND FORM Inherency in the Evolution of Developmental Mechanisms

stuart a. newman and gerd b. müller The close correlation between genotype and phenotype in many kinds of organisms has led to the widely held view that biological forms and features are produced by ‘‘genetic programs.’’ In the field of developmental biology this notion has gained seeming confirmation in findings that successive developmental steps are typically associated with episodes of new gene expression and that experimental alteration of gene expression frequently leads to a change in the developmental outcome. In evolutionary biology the tenet that genes determine form is essentially equivalent to the reigning neo-Darwinian paradigm: (1) evolution is the hereditary transmission of phenotypic change; (2) genes are the medium of heredity; (3) the sum of selected genes therefore specifies and determines the phenotypic differences between organisms. The evolutionary-theoretical version of the genetic program notion is generally uncontested even in glosses on neo-Darwinism that otherwise disagree over such issues as the tempo and mode of phenotypic evolution, the degree to which genetic change can result from selectively neutral mechanisms, and the universality of adaptation in accounting for complex traits. In previous work we have challenged the validity of the genetic program paradigm for the development and evolution of form in metazoan organisms (Newman 1994; Müller and Newman 1999; Newman and Müller 2001). We have suggested that a new way of looking at the causal relationship between genes and form can resolve some of the current debates in evolutionary theory as well as the apparent paradoxes in developmental biology that have arisen with recent findings of extensive functional redundancy in embryonic systems. In particular, we have proposed that the correlation of an organism’s form with its genotype, rather than being a defining condition of morphological evolution, is a highly derived property. This implies that morphological phenotypes of ancestral organisms were not originally determined pri-

Genes and Form

marily by their genotypes, and that in contemporary organisms they are also not exclusively genetically determined. We propose that nongenetic causal determinants of biological morphogenesis have been active over the course of evolution, and that a theory of morphological evolution based on neoDarwinian mechanisms alone must remain incomplete (Newman and Müller 2001; Müller and Newman 2003). Plasticity of form and developmental trajectory in many modern organisms provide an entry into our perspective. Organisms, including protists such as Dictyostelium discoideum (Bonner 1967), fungi such as Candida albicans (Magee 1997), plants (Hutchings and de Kroon 1994), and animals such as arthropods (Emlen and Nijhout 2000) and molluscs (Trussell 2000), may exhibit radically different forms in different microenvironments or ecological settings. Even in vertebrates organism-environment interactions can play a decisive role in morphological development. Amphibian metamorphosis, for example, can be influenced by environmental change as well as by intrinsic timing mechanisms (Gilbert et al. 1996). In mice, the number of vertebrae can depend on the uterine environment (McLaren and Michie 1958). The neo-Darwinian interpretation of these phenotypic polymorphisms is that they represent specifically evolved adaptations and are therefore sophisticated products of evolution. The different phenotypes consistent with a given genotype are thus considered to be programmed subroutines that have evolved as a result of distinct sets of selective pressures at different life-history stages (Stearns 2000), manifestations of an evolutionarily fine-tuned ‘‘reaction norm’’ (Van Tienderen and Koelewijn 1994; Pigliucci 1996; Schlichting and Pigliucci 1998), or products of evolution for ‘‘evolvability’’ (Gerhart and Kirschner 1997). In our alternative conception, however, rather than being the result of evolutionary adaptation, much morphological plasticity is held to reflect the influence that external physicochemical parameters exert on any material system and is therefore primitive and inevitable rather than programmed. Nonliving viscoelastic materials such as clay, rubber, lava, and jelly, for example, are subject (by virtue of inherent physical properties) to being molded and deformed by the external physical environment. Such materials have been called ‘‘soft matter’’ by the physicist Pierre-Gilles de Gennes (de Gennes 1992). Most living tissues are soft matter, and all of them are also what physicists term ‘‘excitable media’’ (Mikhailov 1990; Winfree 1994), materials that respond in active and predictable ways to their physical environments. It is clear that some, if not much, organismal plasticity results from such material properties. If physically based, nonprogrammed plasticity occurs in at least some in-

39

Stuart A. Newman and Gerd B. Mu¨ller

stances in contemporary organisms, it is likely that morphological determination based on an interplay of intrinsic physical properties and external conditions was even more prevalent in earlier multicellular forms. This is because ancient organisms undoubtedly exhibited less genetic redundancy, metabolic integration, and homeostasis than modern organisms do, and were thus more subject to external molding forces. Hence our proposal that morphological variation in response to the environment is a primitive, physically based property carried over to a limited extent in modern organisms from the inherent plasticity and responsiveness to the external physical environment of the viscoelastic cell aggregates that constituted the first multicellular organisms. The inference that ancient metazoans were even more developmentally plastic than modern ones implies that the correspondence of a given genotype to one morphological phenotype, as is frequently seen in higher animals, will be exceptional. Such close mapping can result from an evolutionary scenario in which the developmental mechanism by which a phenotype is generated changes from being sensitive to external conditions to being independent of such conditions (Newman and Comper 1990; Newman 1994; Salazar-Ciudad et al. 2001a, 2001b). If modern organisms are ‘‘Mendelian,’’ in the sense that genotype and phenotype are inherited in close correlation and morphological change is typically dependent on genetic change, then our hypothesis can be encapsulated in the postulate that there was a ‘‘preMendelian world’’ of polymorphic organisms at the earliest stages of metazoan evolution whose genotypes and morphological phenotypes were connected in only a loose fashion. In this exploratory period of organismal evolution the mapping of genotype to morphological phenotype would have been one-to-many rather than one-to-one. With the subsequent evolution of genetic redundancies (Tautz 1992; Picket and Meeks-Wagner 1995; Wagner 1996; Cooke et al. 1997; Nowak et al. 1997; Wilkins 1997) and other mechanisms supporting reliability of developmental outcome (e.g., Rutherford and Lindquist 1998), a closer linkage between genetic change and phenotypic change was established. In particular, with evolution under selective criteria favoring the maintenance of morphological phenotype in the face of environmental or metabolic variability (see Baldwin 1896; Schmalhausen 1949; Waddington 1961; Riedl 1978; SalazarCiudad et al. 2001a), organisms would come to be characterized by a closer mapping of genotype to phenotype, giving rise to the familiar Mendelian world. But even as body plans and other major morphological features such as the Bauplan of the vertebrate limb became locked in by the accumulation

40

Genes and Form

of reinforcing genetic circuitry, fine-tuning of details of organismal, and particularly organ, morphology continued (and continues) to occur through an interplay of genetic and nongenetic factors. Of the two main classes of evolutionary phenomena neo-Darwinism provides an adequate account for only one—the preservation by natural selection of advantageous phenotypes with some genetic component to their variability. The other, the causal origination of phenotypic novelty (Müller 1990; Newman and Comper 1990; Müller and Wagner 1991), is simply sidestepped. This much has been recognized ever since Darwin published his Origin of Species (e.g., Butler 1878), but the hegemonic notion of the genetic program essentially foreclosed a way of conceiving morphological continuity and change across generational lines in anything but genetic terms, and there are no consistent rules for mapping genetic mutations to morphological change. Without an understanding of the origination of embryonic and organ structures, however, including their nongenetic determinants, no description of molecular interactions involved in modern morphogenetic processes can provide an adequate causal account of organismal form (Müller and Newman 2003). Darwin recognized that biological forms evolve, but neither he nor his successors considered the possibility that the nature of the process that generated these forms could itself evolve in a systematic way. In our concept these mechanisms have evolved from a condition of more to less dependence on inherent material properties, and from less to more dependence on hierarchical genetic integration. By considering the emergence and transformation of developmental mechanisms as an evolutionary problem in its own right, we have arrived at the view that epigenetic mechanisms, rather than genetic changes, are the major sources of morphological innovation in evolution. In our usage epigenetic refers to the context-dependence of developmental mechanisms, not to dnaassociated mechanisms of inheritance such as methylation and chromatin assembly (for a review, see Müller and Olsson 2003). The epigenetic mechanisms that we consider are conditional, nonprogrammed determinants of individual development, of which the most important are (1) interactions of cell metabolism with the physicochemical environment within and external to the organism; (2) interactions of tissue masses with the physical environment on the basis of physical laws inherent to condensed materials—what we have termed ‘‘generic’’ processes (Newman and Comper 1990); and (3) interactions among tissues themselves, according to an evolving set of rules. We suggest that different epigenetic processes have prevailed at different stages

41

Stuart A. Newman and Gerd B. Mu¨ller

of morphological evolution, and that the forms and characters assumed by metazoan organisms originated in large part by the action of such processes. A number of authors have discussed the role of epigenetic factors in evolution, pointing to the intrinsically dynamic structure of developmental systems in accounting for a nonrandom variation of traits (Ho and Saunders 1979; Kauffman 1993; Goodwin 1994), the importance of developmental constraints in guiding the direction of phenotypic change (Alberch 1982; Maynard Smith et al. 1985; Stearns 1986), and the role of environmental continuity in propagation of developmental outcomes (Johnston and Gottlieb 1990). While the watchword of neo-Darwinism is contingency, we, along with these other commentators, emphasize inherency. Eckstein (1980), writing in a different context, has provided a useful formulation of the distinction between contingency and inherency in conceptualizing a complex developmental process: ‘‘Something is contingent if its occurrence depends on the presence of unusual (we might say aberrant) conditions that occur accidentally, conditions that involve a large component of chance,’’ while ‘‘something is inherent either if it will always happen (e.g., entropy) or if the potentiality for it always exists and actuality can only be obstructed.’’ In what follows we will provide examples that show that the inherent properties of metazoan organisms and the tissue masses they comprise extend beyond their genomes to encompass their physical identity as semisolid to solid excitable materials. Because the inherent physical properties—in their self-organizing capacities, but also conditioned by external parameters and extrinsic forces—can act as morphogenetic determinants, the dynamic, constraining, and environmental aspects of developmental causation can productively be analyzed in the framework of inherency and interaction, that is, epigenesis. Much of this essay will therefore consider an ‘‘a priori evolutionary developmental biology,’’ highlighting our divergence from the neoDarwinian approach, which deals primarily with the variational, selectional, and population genetic aspects of organismal evolution and thus sidesteps the generative domains on which we concentrate. Among the properties of organisms not shared with nonliving systems are genes, which have evolved to specify their rna and protein molecules, as well as evolutionarily generated activating and inhibitory relationships among these components (regulatory circuits). Once morphological characters emerged through the interplay of epigenetic and genetic factors, they served as templates for the accumulation, by natural selection, of stabilizing, reinforcing, and overdetermining regulatory mechanisms. The processes by which morphological characters were determined were therefore differ-

42

Genes and Form

ent at different phases of evolution, with genetic integration taking on a more prominent role after a morphological feature had been established (Müller and Newman 1999; Salazar-Ciudad et al. 2001b). As a result, the action of the originating epigenetic factors may be obscured or even superseded in a modern organism, with the succession of linked gene activities involved in development taking on the appearance (though not the reality) of a developmental program.

Generic Processes and the Origin of Body Plans Many forms that arise during early stages of embryogenesis are either produced by mechanisms common to cell aggregates and nonliving materials (generic mechanisms) or resemble the outcomes of such generic mechanisms. This section describes the physical attributes and behaviors of tissue masses and the forms that would be expected to arise from these inherent properties.

Diffusion and the Formation of Spatial Gradients The advent of cell-cell adhesion early in the history of multicellular life opened up possibilities for the molding of biological form that were unavailable to single-celled organisms. Diffusion of typical biomolecules, for example, is so rapid over the distances spanned by a single cell that in the absence of special docking or compartmentalization mechanisms (Agutter and Wheatley 2000), intracellular molecules would be well mixed. In contrast, on the size scale of a cell aggregate, the formation of gradients of released molecules is fostered, rather than undermined, by diffusion (Crick 1970). The capacity of a cell to secrete a product, which certainly predated multicellularity in the history of life, acquired a new meaning once multicellularity arose: it provided a means for establishing differences across an otherwise undifferentiated population of cells. For example, if a group of cells in one region of a cell mass releases a product at a higher rate than its neighbors—either by spontaneous, stochastic effects or because some nonuniformity in the environment induces it to do so—it can take on a privileged, ‘‘organizing’’ role in the aggregate (see fig. 1 below). This role would be particularly manifested if one effect of the secreted product was to inhibit surrounding cells from making the same thing —a kind of negative feedback regulation that must also have predated multi-

43

Stuart A. Newman and Gerd B. Mu¨ller

cellularity. The generation of molecular gradients by diffusion-dependent processes is a widely used mechanism for tissue regionalization in modern embryos (Gurdon and Bourillot 2001). The basic ingredients were already present in free-living cells, which had only to form primitive cell aggregates by any one of several means (Bonner 1998) for this inherent developmental mechanism to be realized.

Differential Adhesion and Compartment Formation Cell adhesion is the defining condition of multicellularity. Free-living cells prior to the origin of metazoan organisms undoubtedly contained proteins on their surfaces that had evolved to serve purposes other than adhesion. But protein-protein aggregation is a microenvironmentally dependent property, and under new conditions formerly nonadhesive cells could readily have acquired a tendency to aggregate (Kazmierczak and Degens 1986). Indeed, several different adhesion systems are used by modern organisms, including members of the cadherin superfamily (Takeichi et al. 2000), the immunoglobulin superfamily (Ranheim et al. 1996), and the ephrin-Eph receptor family (Klein, 2001), and evolutionary biologists generally agree that multicellularity probably evolved more than once (Bonner 1998). If an aggregate contains a mixture of cells with different types of adhesion molecules on their surfaces, or different amounts of the same adhesion molecule, the cell subpopulations will sort themselves out into islands of more cohesive cells within lakes composed of their less cohesive neighbors (fig. 1). Eventually, by random cell movement, the islands coalesce and an interface is established across which cells will not intermix (Steinberg 1998). Thus, when two or more differentially adhesive cell populations are present within the same tissue mass, multilayered structures can form automatically, comprising nonmixing ‘‘compartments’’: distinct spatial domains within a single tissue, in which no interchange or mixing of cells occurs across the common boundary (Crick and Lawrence 1975; Garcia-Bellido 1975). The physical basis of this sorting-out is similar to phase separation of two immiscible liquids, such as oil and water. Moreover, a more cohesive tissue (one with stronger bonds between its cells) will always be engulfed by a less cohesive one. Because the final relative configuration of the cellular subpopulations will always be the same, regardless of the initial conditions (e.g., whether the cells are randomly intermixed, arranged in discontinuous patches, or even arranged with the less cohesive tissue inside rather than outside), tissue morphogenesis driven by adhesive differentials is in

44

1 • The morphogenetic consequences of linking the regulation of cell-cell adhesion to various physical and chemical pattern-forming mechanisms.

Note: The central box denotes the effects of differential adhesion in causing the formation of a boundary within the tissue mass across which cells will not mix. All of the peripherally arranged boxes denote patternforming mechanisms which, when deployed in conjunction with differential adhesion, can lead to standard morphological outcomes. Polarized expression of adhesion molecules leads to cavities and other lumenal structures. Sedimentation of a dense cytoplasmic component and diffusion of a growth factor are ways in which an egg or morula can become spatially nonuniform; this can eventually be reflected in differentially adhesive cell populations, leading directly or indirectly to gastrulation. A chemical oscillation that regulates adhesive differentials can lead to segmentation, and reaction-diffusion coupling provides chemical prepatterns that can generate periodic structures.

Stuart A. Newman and Gerd B. Mu¨ller

effect ‘‘goal-directed’’ (Steinberg 1998), with the endpoint being dictated by a physical principle analogous to the approach to thermodynamic equilibrium. That differential adhesion alone is sufficient to generate tissue compartmentalization has been demonstrated by experiments in which populations of nonadhesive mouse l cells were genetically engineered to express differing amounts of p-cadherin. These populations sorted out and exhibited tissue engulfment behavior exactly as predicted by the differential adhesion hypothesis (Steinberg and Takeichi 1994). This, then, is another example of a physical property of a cell aggregate providing it with the inherent potential to undergo morphogenetic change. That is not to say that all compartmentalization in contemporary metazoan organisms is based exclusively on differential adhesion. For oocyte positioning during oogenesis in Drosophila differential adhesion appears to be the controlling mechanism (Godt and Tepass 1998; González-Reyes and St. Johnston 1998). In the case of rhombomeres in the mammalian hindbrain, however, while adhesive differentials in adjacent tissue domains play a part, the compartment boundaries are then reinforced by cell and extracellular matrix specializations (Heyman et al. 1995). In Drosophila body segmentation, local signaling at the segmental boundary appears to play a more important role than adhesive differentials (Rodriguez and Basler 1997). Any regionalizing influences (e.g., diffusion across tissues or gravity-driven gradients within the uncleaved egg; see below) that affected the expression of adhesion proteins, or even stochastic differences in the expression of such molecules in cells within a primitive aggregate, could lead to sorting-out of cellular subpopulations. Compartmentalization arising in this ready fashion during phylogenetic evolution was a likely basis for the origination of multilayering during gastrulation and analogous features of organogenesis.

Cell Polarity and Lumen Formation Tissues formed by cells bearing uniformly distributed adhesive molecules are said to be epithelioid. The morula stage of embryonic development is a case in point. When tissue cells begin to express adhesive proteins in a polarized fashion (Rodriguez-Boulan and Nelson 1993), there are morphogenetic consequences that would have been unanticipated when cell polarity initially appeared. Specifically, a tissue mass consisting of motile cells that are nonadhesive over portions of their surfaces would readily develop cavities, or lumina. As a result of random cell movement or death of cells that have become detached from their neighbors, such spaces will automatically come to

46

Genes and Form

adjoin one another (see fig. 1). The transformation of a morula into a hollow blastula is the first developmental manifestation of this process. Cell polarity in the evolutionary novel context of multicellularity led inevitably to hollow structures. The notion that lumen formation is an inherent property of tissue masses —the consequence of a delicate balance of adhesive interactions between cells and/or their extracellular substrata—is supported by experimental studies of various developmental and pathological conditions. In human autosomal dominant polycystic kidney disease (adpkd), for example, the formation of cysts (a closed-tissue arrangement similar to a blastula in which cells make maximal contact with one another apart from their nonadhesive portions), rather than tubules (an open-tissue arrangement in which the average cellcell contact area generally differs from that of a cyst), in the kidneys of severely affected patients involves the expression of mutated forms of polycystin-1 (Qian et al. 1996), an integral membrane glycoprotein which under normal conditions forms complexes with the cell-cell adhesive protein e-cadherin in kidney epithelial cells (Huan and van Adelsberg 1999). Mammary gland epithelial cells in culture provide a model for the formation of interior spaces in developing tissues, such as the blastocoel. When grown on an artificial substratum mammary epithelial cells adopt a flat ‘‘cobblestone’’ appearance. In the presence of the extracellular matrix (ecm) molecule laminin, however, they round up and cluster, and depending on the culture conditions may form hollow alveolar structures with well-defined apical and basal surfaces (Li et al. 1987). Here, formation of lumina in tissues composed of polarized cells depends not on a sorting-out process per se, but on apoptosis in the interior of the aggregate, apparently brought on by abrogated adhesion to the polarized surface layer (Lund et al. 1996). In contemporary organisms other cellular mechanisms, such as the contraction of apical actin filaments in a group of cells in a localized domain of an epithelial sheet, contribute to, and even initiate, lumen formation (Grant et al. 1991; Lincz et al. 1997). The mobilization of complex intracellular machinery for the formation of internal cavities in tissues would have been much easier in organismal forms in which this developmental step had already been established by ‘‘generic’’ means. Evolutionarily ancient metazoan organisms which were made up of adhesively polar cells, and therefore took the form of hollow sacs that resembled a modern blastula, were confronted with morphological ‘‘instability’’ once they reached a certain size: their overall shapes would have necessarily deviated from spherical (Drasdo and Forgacs 2000). Most of the new forms were

47

Stuart A. Newman and Gerd B. Mu¨ller

indeterminate and unstable, but in those cases in which an inherent patternforming mechanism caused adhesive or mechanical properties to become nonuniform across the cell mass, new stable forms emerged by invagination or ingression of part of the hollow sphere (Drasdo and Forgacs 2000; see fig. 1), providing templates and prototypes for modern gastrulae.

Excitability and Segmentation Because individual cells are chemically active, thermodynamically open systems, tissues composed of them are ‘‘excitable media.’’ With metabolic and genetic networks responsive to the external environment via permeability and membrane-based signal transduction and positive and negative feedback loops, cells will spontaneously develop chemical oscillations which can also be triggered by external cues (Goldbeter 1995; fig. 1). Such excitability will be exhibited at the level of a single cell or of a cell aggregate. The cell division cycle is a temporally periodic process driven by a chemical oscillation (Murray and Hunt 1993), but it has no morphological consequence in the world of single cells. That is to say, for free-living cells, cell division just leads to more cells, no matter what the process’s temporal dynamics may be. Even in a multicellular entity the division cycle typically acts only to increase the mass of the undifferentiated aggregate. But let us assume that the cells in the aggregate contain an additional biochemical oscillation, with a period different from the cell cycle, which happens to influence the adhesive properties a cell is born with (Newman 1993). The cell cycle could then act as a ‘‘gate’’ for the action of the regulatory cycle. Consider what would happen if a primitive organism’s ‘‘body’’ was being generated from a population of such oscillating and dividing cells. Each time a new line of cells emerged from the stem population it would have a set of adhesive properties different from the population that emerged just before it. But because the adhesion-regulatory oscillation would periodically return to its original state, tissue domains with identical adhesive properties would be separated from each other by domains with an alternative, but also identical, set of adhesive properties (Newman 1993; Salazar-Ciudad et al. 2001b). Such an arrangement of alternating differentially adhesive compartments is a hallmark of segmental organization (Wright and Lawrence 1981; Guthrie and Lumsden 1991; see fig. 1). The existence of regulatory processes for body segmentation based on chemical oscillation is well confirmed in modern vertebrates. In the beststudied case—the separation, after the embryonic axis has formed, of mesoderm on either side of the neural tube into discrete blocks of tissue called

48

Genes and Form

somites (Pourquié 2003)—the gating mechanism is actually a concentration wave front of the peptide factor fgf8. But the regulatory clock begins running even earlier, during gastrulation, when a series of periodically arranged cell aggregates, the somitomeres, which anticipate the placement of somites, are generated (Jouve et al. 2002). The gate for this early segmentation event must be something other than the fgf8 wave front—possibly the cell cycle, as suggested above. In ‘‘long germ band’’ insects such as Drosophila, segment formation is initiated in a simultaneous, rather than sequential, fashion. The Drosophila evenskipped transcription factor gene, for example, is expressed in a series of ‘‘chemical stripes’’ consisting of seven evenly spaced bands that form simultaneously across the early embryo (Small et al. 1991). The spatially periodic distribution of such factors imparts alternating identities to the embryonic nuclei which later specify cell states of differential adhesivity (Irvine and Wieschaus 1994), leading subsequently to the formation of overt segments. Consideration of tissues as excitable media can unify the understanding of sequential and simultaneous segmentation (Newman 1993). Specifically, the biochemical kinetics that give rise to a chemical oscillation were predicted (if one or more of the components is diffusible) to be capable of giving rise to standing or traveling spatial periodicities of chemical concentration (Turing 1952; Prigogine and Nicolis 1971). In the last decade this theoretical prediction has been realized experimentally in the form of nonliving chemical systems that produce spots and stripes (on the same spatial scale as embryonic primordia) in gelled media (Boissonade et al. 1994; see fig. 1). Significantly, in the Drosophila embryo, even-skipped protein and other segment-regulatory transcription factors first interact with their target genes in a syncytium—a common cytoplasmic environment in which the embryo’s nuclei are not yet separated by cell membranes. Some of these diffusible factors also positively regulate their own synthesis (Harding et al. 1989; IshHorowicz et al. 1989), a sine qua non of both chemical oscillators and Turingtype pattern-forming systems (Boissonade et al. 1994). Simultaneous segmentation, seen in long germ band insects, may thus have a common evolutionary origin, at the level of mechanism (i.e., oscillatory biochemical dynamics), with sequential segmentation, seen in short germ band insects such as beetles (Patel et al. 1989), in leeches (Wedeen 1995), and in the vertebrate segmental plate. The oscillatory dynamics of excitable media thus provide another set of inherent physical properties to cells which in the context of a multicellular aggregate can give rise to a fundamental morphological motif, in this case seg-

49

Stuart A. Newman and Gerd B. Mu¨ller

mentation. But while such inherent or generic processes may have been at the evolutionary origin of segmental patterning, it is not surprising that in modern organisms the regulation of segment formation can involve elaborate systems of multiple promoter elements responsive to preexisting, nonuniformly distributed molecular cues (Goto et al. 1989; Small et al. 1991). Such genetic networks can serve to ‘‘overdetermine’’ morphological outcomes during development, making them less subject to the vicissitudes of purely physical determination (Newman 1994; Salazar-Ciudad et al. 2001a, 2001b).

Origination and Development of Body Plans The epigenetic mechanisms described in the previous sections fall into two major categories: those that generate extracellular or intracellular chemical patterns (diffusion, cell polarity, the ‘‘two-oscillation’’ mechanism, reactiondiffusion coupling), and the mechanism of differential adhesion, which can potentially convert a chemical pattern into a morphological one, such as multilayering, lumina, and segments. There are several additional mechanisms by which biochemically distinct subpopulations of cells can arise. Gravity, acting on cytoplasmic components of different density, can create a chemical gradient within a single cell such as the egg. Indeed, a gradient of dense yolk platelets is one of the first positional cues in the cleaving amphibian egg (Kessler and Melton 1994; see fig. 1). Another such mechanism has been described by Kaneko and co-workers (Kaneko and Yomo 1999; Furisawa and Kaneko 2000), who demonstrated by computer simulations a previously unknown but inevitable effect in interacting dynamic systems (such as cells) which they have termed ‘‘isologous diversification.’’ Specifically, they showed that replicate copies of a complex chemical system (e.g., cells in a blastula) can force one another to take on distinct chemical states simply by virtue of exchanging materials. Since these chemical pattern-forming mechanisms can lead to aggregates containing different ‘‘cell types,’’ but ones whose division of labor has no particular relationship to the functional integration of the organism, additional steps would have had to occur before such differentiation processes could have resulted in a modern developmental system. Furthermore, allocation of resources between cell types generated in this manner would generally disrupt previously evolved cellular homeostasis. For these reasons, we have suggested above and elsewhere (Newman 1994; Newman and Müller 2001) that differential expression of adhesivity was among the earliest examples of biochemical differentiation to become estab-

50

Genes and Form

lished in the metazoan world. Adhesion, being an evolutionary novelty in the earliest multicellular forms, would have been tolerated by the constituent cells over a wide range of expression levels. The resulting compartmentalized forms would have constituted organisms with physically separate but otherwise equivalent cell populations, within which independently acting tissue primordia could evolve. Moreover, even without additional functional differentiation, differential adhesion would have generated a panoply of morphologically distinct forms—multilayered, hollow, segmented, separately and in combination. This early form of differentiation would have led readily to the establishment of novel organismal forms with the capacity to populate correspondingly new ecological niches. In principle, virtually all the features seen in early development of modern metazoan organisms can be attributed to the composition of one or more of the pattern-forming mechanisms with differential adhesion (Newman 1994; fig. 1). As noted above, however, we do not assert that modern development can thus be reduced to the composition of such epigenetic process. Rather, we have suggested that the combined effects of the various physical properties that were generic to the earliest multicellular aggregates considered as chemically excitable, viscoelastic soft matter ensured the production of a profusion of forms. This pre-Mendelian world of morphological prototypes provided the raw material for natural selection and was thereby converted into genetically routinized body plans (Newman 1994; Newman and Müller 2001). Because such forms based on inherent properties of tissue masses would have been relatively straightforward to generate but also limited in their variety (not all conceivable forms can be achieved using the described processes), our hypothesis has strong implications for the expected tempo and mode of morphological evolution. As we have discussed elsewhere (Newman 1994; Newman and Müller 2001), this view suggests a natural interpretation of the burst of morphological diversification recorded in the early Cambrian fossil record—including the capacity for a rapid elaboration (on the geological time scale) of virtually all the modern body plans. Even the Cambrian forms that failed to contribute to modern body plans have the appearance of being constructed of different combinations of the standard morphological elements (Conway Morris 1989; Gould 1989). This framework also suggests prior steps in the Precambrian era that possibly laid the basis for this ‘‘Cambrian explosion.’’ The first morphologically complex multicellular organisms, represented by the Vendian fossil deposits dating from as early as 700 million years ago, appear to have been flat, often segmented, but apparently solid-bodied creatures (Seilacher 1992; Conway

51

Stuart A. Newman and Gerd B. Mu¨ller

Morris 1993). It may have taken up to 100 million years after the appearance of the Vendian fauna for organisms to develop distinct body cavities, although some evidence suggests that this may have occurred more rapidly (Seilacher et al. 1998). Once these hollow forms arose, all the modern body plans burst onto the scene in short order. Thus the advent of polarized cells provides a plausible physical basis for the Precambrian origin of body cavities, a precondition of triploblasty. Whether cell polarity arose before or after any of the evolutionary events that led to multicellularity, lumen or cavity formation would have been an inevitable physical consequence of the conjunction of these two properties (Newman 1994).

Epigenetic Interactions and the Origin of Novelty We have indicated how the generic physical properties of tissues helped to determine the array of forms generated in early organismic evolution. Although the role of these purely physical processes in embryogenesis must have receded as stabilizing evolution led to more integrated and reliable developmental mechanisms, they clearly remain causally active in aspects of modern development. In particular, physical principles, including biomechanical factors, are active in secondary developmental fields. In such limited domains flexibility and conditionality of morphological outcomes are less likely to undermine overall organismal integration than similar lack of fixity at the body plan level. This has had important consequences for the capacity of morphological design to evolve in new directions. As evolution proceeded, cells acquired the capacity to differentiate and to produce, organize, and secrete increasingly sophisticated arrays of proteins and polysaccharides. Much as alloys and composite polymeric materials exhibit properties not readily predictable from the physics of simple liquids and solids, so will the outcomes of physical processes in secondary developmental fields be less ‘‘generic’’ and more idiosyncratic than those of the liquidlike cell aggregates that existed at earlier stages of metazoan evolution. Thus, while the basic body plans of metazoans, having been established in the earliest period of multicellularity, may bear the stereotypical stamp of physical molding forces acting on soft, excitable materials, later-evolving organs and appendages will exhibit more particularities. This is somewhat paradoxical since, in general, body plan organization will have become more genetically consolidated and overdetermined with time, while organs and appendages,

52

Genes and Form

being evolutionarily ‘‘younger,’’ may actually retain more physically based plasticity than the same organism’s general body plan.

Differentiation and Inductive Interactions As a consequence of adhesion-based compartmentalization, some organisms came to contain spatially discrete subpopulations of cells that were functionally similar except for their propensity not to mix across their common boundary. Once such organisms began to populate the biosphere, those that ‘‘bred true’’ would have left more descendants. That selective advantage would eventually have given rise to forms that generated tissue boundaries by reliable hierarchical genetic regulatory mechanisms rather than contextdependent, and therefore variable, self-organizational properties (SalazarCiudad et al. 2001a, 2001b). Moreover, once such mechanisms of spatially regulated gene expression existed they would have been mobilized for functional specialization beyond mere adhesive differentials. In this way, compartmentalization of tissue primordia, which gives rise to equivalent but nonmixing cell populations, can be looked at as a morphological analogy of gene duplication, which also provides initially identical modules which may eventually evolve into functionally differentiated entities. At this point in the evolution of the Metazoa, some organisms contained cell populations whose microenvironments now consisted of other cell populations in addition to the external world. As the biochemical properties of the cell types and the forms they produced began to depend on interactions between adjacent cell populations, embryonic induction came into existence. The conditionality of tissue interactions, along with the inherent morphogenetic properties retained from earlier periods, guaranteed that the resulting systems retained a significant degree of plasticity. Although it is impossible to reconstruct the detailed evolutionary history of a particular developmental mechanism with currently available knowledge, modern embryos are believed to bear the imprint of different evolutionary episodes in the establishment of functional spatiotemporal differentiation. One possible example of this pertains to mesoderm induction in the Xenopus blastula, in which the apical hemisphere is induced to form precursors of muscle and notochord by factors emanating from the cells of the vegetal hemisphere. Cells of the blastula appear to be organized into nonmixing compartments (Sheard and Jacobson 1987), but the forces keeping the cells from mixing are weak (Sheard and Jacobson 1990), and later developmental

53

Stuart A. Newman and Gerd B. Mu¨ller

mechanics seems not to make use of this organization. Mesoderm induction, however, depends on signaling across an interface between spatially distinct but initially equipotential cell populations (Kessler and Melton 1994). If, as we suggest, evolution of this process was based on the prior existence of nonmixing populations, modern blastula compartmentalization may be its relic. Once new functionalities were established, the original physical basis of the subpopulation difference may have been attenuated or dispensed with entirely.

Epithelia and Mesenchymes Among the distinct types of tissues that emerged from these diversification processes were epithelioid tissues, in which cells contact one another directly by means of membrane-attached proteins, and mesenchymal tissues, in which cells are bound to one another indirectly through hydrated ecms. Polarized expression of adhesion proteins will convert epithelioid tissues into epithelia, which enclose lumina (see above) or form sheets. Epithelial sheets can undergo bending, eversion, and invagination, and placode, cyst, and tubule formation, depending on the extent of polar expression of adhesive molecules as well as local variations within the sheet that may be influenced by inductive interactions (Newman 1998a). For epithelial tissue primordia in which the cells have relative mobility (i.e., in early embryos), the tissue behaves as a fluid confined to a plane, and the physics of soft matter can account for many details of epithelial sheet morphogenesis (Gierer 1977; Mittenthal and Mazo 1983; fig. 2A). In contrast to epithelioid and epithelial tissues, in which cells directly adhere to one another over a substantial portion of their surfaces, mesenchymal and other connective tissues consist of cells suspended in an ecm. There thus exists a set of additional morphogenetic mechanisms which depend on changes in the distance between cells, the effects of cells on the organization of the ecm, and the effects of the ecm on the shape and cytoskeletal organization of cells that typically occur in connective, but not epithelioid, tissue types (Newman and Tomasek 1996). As in epithelioid tissues, boundaries of immiscibility can occur in connective tissues, as in the interface between the flank of a developing vertebrate embryo and a limb bud emerging from the flank (Heintzelman et al. 1978). Since differential adhesion and sorting based on cell-cell contact are not relevant to cell populations in which cells do not directly adhere to one another, a different explanation must be sought for immiscibility of various embryonic

54

Genes and Form

mesenchymes. One possibility is that subtle differences in the organization of molecular fibers of the ecm can define distinct physical phases, as model systems have shown occurs (Newman et al. 1985; Newman 1998b). Because mesenchymal cells are dispersed in a matrix, the tissues they form have an inherent morphological propensity not available to epithelial and epithelioid tissues: the cells can move physically closer to one anther. This occurs, for example, in mesenchymal condensation, which is often a transient effect in development (see fig. 2B). Condensations generally progress to other structures, such as feather germs (Chuong and Widelitz 1998), cartilage, or bone (Hall and Miyake 1992, also 2000), or after conversion to epithelium, to kidney tubules (Ekblom 1992). A variety of cellular mechanisms have been suggested to cause mesenchymal condensation, including local loss of matrix materials, centripetal chemotactically driven movement through the matrix, cell traction, and absence of cell movement away from a center. The last of these, associated with haptotaxis, movement of cells up gradients of ecm materials such as fibronectin, is best supported by the experimental evidence. (See Newman and Tomasek 1996 and Miyake and Hall 2000 for reviews.) In vertebrates, echinoderms, and other gastrulating species, the formation of the trilaminar embryo places epithelial and mesenchymal tissues in direct contact with one another. The inductive interactions afforded by this proximity mobilize generic physical properties of both tissue types, leading to structures not possible with either type alone. The formation of the neural tube (Colas and Schoenwolf 2001), the kidneys (Ekblom 1992), exocrine glands (Kashimata and Gresik 1996), and the paired limbs (Newman 1988) in vertebrates are several examples of the profusion of complex forms that became possible after epithelial-mesenchymal interactions came on the scene. The formation of the vertebrate limb, for example, depends on the intrinsic propensity of the skeletogenic mesenchyme cells to form precartilage condensations (Frenz et al. 1989; Downie and Newman 1994). The growthpromoting and -shaping effects of the ectoderm (Saunders 1948) influence the arrangement of the forming elements. In agreement with earlier suggestions (Newman and Frisch 1979; Leonard et al. 1991; Newman 1996), recent experiments have provided evidence that the pattern of mesenchymal condensations in limb bud mesenchyme in vitro arises from a reaction-diffusion process (Miura et al. 2000; Miura and Shiota 2000a) that involves interplay between factors produced by the mesenchyme (Miura and Shiota 2000b) and epithelium (Moftah et al. 2002; see fig. 2C). Developmentally mature epithelial and connective tissues contain elabo-

55

2 • Epigenetic mechanisms of tissue morphogenesis and organogenesis. (A) Major modes of epithelial morphogenesis resulting from extrinsic alteration of cell parameters. In a, a pattern formation mechanism (e.g., a reaction-diffusion system) is activated in a flat epithelial sheet, possibly mediated by a subjacent mesenchymal layer, and marks a subset of cells to undergo alteration of one or more cellular functions (e.g., adhesive strength, cytoskeletal tension). In b–e, resulting epithelial morphologies are indicated. A placode (c) will form if the lateral regions of the epithelial cells become more adhesive than the apical and basal regions. An evagination (b), as in a developing intestinal villus, or an invagination (d), as in a developing hair or feather (Chuong and Widelitz 1998), will form if the change in cell property gives rise to a bending moment (Newman 1998) that destabilizes the flat configuration. Progressive cycles of patterning and invagination will give rise to a branched tubular structure (e), as in salivary gland morphogenesis (Kashimata and Gresik 1996). (B) Mesenchymal condensation, such as occurs during skeletal morphogenesis and many other developmental processes. Such condensations can be initiated by local patches of elevated production of extracellular matrix (ecm) molecules, and consolidated by cell-cell adhesion. (C) Morphogenesis of connective tissue elements, such as carti-

Genes and Form

lage rods and nodules, occurs by the regulation of the pattern of mesenchymal condensation formation. One way that this can occur is by the interplay of a positively autoregulatory diffusible activator of ecm production, such as tgf-beta (curved arrows), with a diffusible inhibitor of its activity (straight gray arrows). In the absence of the inhibitor (top) resulting cartilage forms as an amorphous mass; in its presence patterns of well-spaced nodules and rods can form as centers of activation become surrounded by domains of inhibition. (D) Origin of the fibular crest in archosauran hindlimbs by mechanical regulation of mesenchymal morphogenesis. Progressive evolutionary reduction of the fibula increases the mechanical load on the connective tissue between the tibia and the fibula, exerted by the pulling action of the iliofibularis muscle. A stressdependent cartilage (oval) forms in response and becomes incorporated into the ossifying tibia to form a prominent crest (stippled), a homologue shared by theropod dinosaurs and carinate birds. This tight fixation of the proximal fibula permits its further distal reduction in avian limbs (D adapted from Müller and Streicher 1989). From S. A. Newman and G. B. Müller, ‘‘Epigenetic mechanisms of character origination,’’ in: G. P. Wagner (ed.), The Character Concept in Evolutionary Biology (San Diego: Academic Press, 2001); reproduced by permission of Elsevier.

rate cell-cell or cell-ecm adhesive structures that permit physical forces originating within cells to contribute to tissue morphogenesis (reviewed in Troyanovsky 1999). In particular, intracellular forces necessary for cell shape changes and migration can be imparted onto the surrounding cells and ecm, resulting in mechanical stress in the tissue as a whole (Grinnell 1994; Ingber et al. 1994; Forgacs 1995; Beloussov 1999). Such cell-generated stresses can lead to contraction, orientation, or assembly of extracellular fibers or cytoskeletal filaments (reviewed in Newman and Tomasek 1996). In such cases the tissue no longer exhibits liquidlike behavior, since the cells are not independently mobile, but rather acts like an (excitable) elastic medium. The linking of cells into an elastic medium, which converts individual cell contractility into global mechanical stress, is another example (like diffusion and adhesion, above) of how a multicellular setting can mobilize physical processes that have only marginal significance on the single-cell scale. Because the resulting elastic media are also biochemically excitable, the stresses and strains generated within them can also regulate the active behavior of the component cells. Cells recognize and respond to mechanical stresses by

57

Stuart A. Newman and Gerd B. Mu¨ller

changing their shape, orientation, growth, expression of specific gene products, and cytoskeletal organization (Grinnell 1994; Ingber et al. 1994), as well as by remodeling their extracellular matrix (Unemori and Werb 1986; Lambert et al. 1992; Takahashi et al. 1998). This occurs in part by the transduction of mechanical forces into chemical signals within specialized structures such as the focal adhesion complex (Lambert et al. 1998; Parsons et al. 2000). Mechanical stress is an instance of interplay between inherent and extrinsic determinants that takes on increasing importance as a determinant of morphogenesis at later stages of development and, because it employs certain molecular components not characteristic of single-celled organisms, at later stages of evolution. The importance of mechanical stress in morphogenesis is seen in the role of mechanical loads in determining tissue pattern. Researchers almost a century ago, for example, showed that the organization of the cardiovascular system is influenced by mechanical forces arising from blood pressure and flow (Russell 1916). Similarly, ecm of trabecular bone was shown to be deposited along lines of eventual tension and compression (Koch 1917). This latter case is of interest in that the ecm organization arises during embryonic development, often before substantial mechanical stresses are placed on the bones. The basis for this apparent reversal in timing of cause and effect may be a reflection of the phenomena known as the ‘‘Baldwin effect’’ (Baldwin 1896; Simpson 1953), ‘‘genetic assimilation’’ (Waddington 1961), and ‘‘genetic integration’’ (see below), in which biochemical circuitry evolves that stabilizes or reinforces an outcome that was originally dependent on external forces to bring it about (see also Salazar-Ciudad et al. 2001a). Such pathways may come to be triggered earlier in the life history of the organism than the stage at which the external forces originally acted.

Morphological Novelties In the context of increasing consolidation and stabilization of developmental pathways, changes in system components that exceed certain thresholds can have dramatic effects on morphological outcomes. Such nonlinear responses can create unexpected byproducts, which may appear as phenotypic innovations at the subphylum level (Müller 1990, 2002, 2003a; Müller and Wagner 1991). Connective tissues and tendons, for example, have the capacity to react to biomechanical stimuli by forming cartilage and bone. Such skeletal elements,

58

Genes and Form

known as cartilaginous or ossified sesamoids, arise from a biochemical plasticity whereby fibroblasts modulate their ecm expression as a function of mechanical forces acting on them (Vogel and Koob 1989; Evanko and Vogel 1993; Giori et al. 1993). In the avian hindlimb, four sesamoids which are dependent on embryonic movement form during the course of normal development; experimental paralysis inhibits their formation (Wu 1996). Later, during ossification, these skeletal elements become incorporated into the long bones of the limb. The avian clavicle (Hall 1986) and cartilage arising from periosteum of membrane bone (Fang and Hall 1997) are similarly induced by interaction-dependent biomechanical processes. Evolutionary changes of bone proportion (Streicher and Müller 1992) will generate similar changes in embryonic biomechanics, with the resulting skeletal elements appearing as novel characters of long bones, such as the supratendinal bridge, the cnemial process, or the fibular crest of the avian tibiotarsus (Müller and Streicher 1989; see fig. 2D). These morphological side effects (Müller 1990) of selection affecting other characters, such as the size or the growth rate of the tibia, eventually become incorporated into the Bauplan of the limb, providing raw material for, and constraints on, subsequent evolution. Rather than being merely passive architectural byproducts of functionally selected processes (Gould and Lewontin 1979), these ‘‘nongeneric’’ epigenetic determinants of morphological novelty (like the generic determinants of body plan described above) can be major causal determinants of biological form. That the interplay between inherent tissue capacities and the physical environment can generate developmental novelties without the participation of either novel genes or incremental selection regimes raises profound questions for any analysis of development or evolution that posits simplistic correspondences between genes and form.

Developmental Integration and Homology The inherent properties of organisms and their tissues, in interaction with the physical environment and, increasingly (as the embryo becomes more complex), with one another, lead to stereotypical outcomes that are reflected in structural similarities in body plan and organ form across all metazoan taxa (Wake 1991; Sanderson and Hufford 1996; Moore and Willmer 1997). Metazoan development is largely a matter of molding clusters of dividing cells into physical shapes: layers, segments, sheaths, tubes, rods, spheres, and so on,

59

Stuart A. Newman and Gerd B. Mu¨ller

are formed by aggregates of cells, mobilizing a wide range of biomechanical forces that result from the different properties of different cell types and their extracellular products. Once these macroshapes have formed, they become important structural determinants of further development, not only creating geometric templates and barriers but also controlling gene activity. These morphological products of development become new inherent features of the embryo which in turn provide a platform for subsequent generation of form as development proceeds. In addition, through stabilizing and integrating evolution, such building elements serve as formal ‘‘attractors’’ of design around which more design is added (Müller and Newman 1999; Müller 2003b). Eventually they may become ‘‘autonomous elements of the morphological phenotype which are maintained in evolution because of their organizational roles in heritable, genetic, developmental, and structural networks’’ (Müller 2003b). These building units are known as ‘‘homologues’’ in taxonomic classifications. Homology, therefore, may find its causal explanation in the processes of developmental integration. One aspect of such integration lies in the evolutionary increase of conditional developmental interactions. With evolution of inductive interactions and emergence of architectural constraints, numerous structure-function interrelationships will evolve (Galis and Drucker 1996) and new characters will be integrated into the morphological phenotype. Over time, the position of a new character in the hierarchical design of an organism will gain increased importance as the number of structural and functional interdependencies grows larger and as more design and functional differentiation are added (Wagner 1989, 1995). Such integration, although it may render novelties indispensable to further development and organismal function, is not without developmental hazards. Galis and Metz (2001) have suggested, for example, that global interactivity occurring at certain early stages in the establishment of body plans and organ forms renders developing systems particularly vulnerable to perturbation by teratogenic agents at those stages. Clearly evolutionary pressures have also been at work to establish redundancy, modularity, and other means of resisting such morphological disruption. Evolution of genes and gene networks enters into this integration process at a number of levels. In certain cases, new forms arose from the mutation of a morphologically involved gene—the standard neo-Darwinian scenario—but when understood in relation to the foregoing discussion can no

60

Genes and Form

longer be conceptualized as the new gene determining or ‘‘programming’’ the new form. In other cases, existing genes or gene networks took on new roles in the production of conserved forms (Abouheif 1999; Salazar-Ciudad et al. 2001b; Wray 2001). Finally, acquisition of genetic redundancies (Tautz 1992; Picket and Meeks-Wagner 1995; Wagner 1996; Cooke et al. 1997; Nowak et al. 1997; Wilkins 1997), including duplication of developmental control genes (Holland 1999) and multiplication of their regulatory elements (Goto et al. 1989; Small et al. 1991), provided opportunities for evolution of parallel means to similar morphological ends. The epigenetic context in which such genetic redundancy arises influences which developmental interactions became genetically integrated (Müller and Wagner 1996), but the redundancy eventually leads to developmental systems that are canalized (Waddington 1942; Wilkins 1997), ‘‘generatively entrenched’’ (Wimsatt 1986), and overdetermined (Newman 1994). Perhaps the key point in the interplay between epigenetics and genetics in the course of integration of morphological features into the developmental repertoire is that epigenetic integration typically precedes genetic integration (Müller and Wagner 1996). The fact that the epigenetic integration mechanisms can persist even when more hierarchical, ‘‘hard-wired’’ genetic circuitry for producing the feature has been acquired makes it all the more difficult to deconstruct a modern developmental pathway into genetic programs or routines. That genetic change may occur despite morphological stasis, and indeed has continually been selected for in order to preserve integrated building units of the organism, attests to the fact that such constructional units can achieve autonomy (Müller and Newman 1999; Müller 2003b). The morphological identity of such homologues can eventually transcend all processes involved in their ontogeny, be they genetic, cellular, biochemical, or physical, since any of these can become reorganized over the course of evolution (Wagner 1989; Wray and Raff 1991; Hall 1994; Bolker and Raff 1996; SalazarCiudad et al. 2001a, 2001b; Wray 2001). Numerous studies have documented the decoupling of morphological from genetic divergence in both laboratory and natural populations (Atchley et al. 1988; Meyer et al. 1990; Sturmbauer and Meyer 1992; Bruna et al. 1996), contrary to neo-Darwinian expectations. The autonomization of morphological characters over the course of evolution by continual reorganization of epigenetic and genetic determinants must frustrate any attempt to discern a genetic program for development in modern organisms.

61

Stuart A. Newman and Gerd B. Mu¨ller

Conclusion: Genes, Form, and Inherency Dobzhansky’s famous dictum that ‘‘nothing in biology makes sense except in light of evolution’’ (Dobzhansky 1973) applies in a twofold sense to the relationship of genes to form. Forms have evolved, but so have the ways they are generated—the developmental mechanisms. The failure of the neoDarwinian synthesis to take account of such evolving generative mechanisms led to a situation in which, while it was acknowledged that not all genetic change had phenotypic consequences, all phenotypic change was viewed as due to genetic change. This was despite the facts that, as we have seen, the actual array of morphological phenotypes available to an organism depends on epigenetic as well as genetic determinants, and the extent to which nonprogrammed, conditional factors enter into the mix will differ at different points in evolution. A gene-centered approach to the generation of organismal form during evolution and development has long been recognized as lacking explanatory power (Ho and Saunders 1979; Oyama 1985, 2000; Nijhout 1990; Seilacher 1991; Goodwin 1994). Recent analyses have shown in detail that the term gene has been employed in different ways when applied to developmental and evolutionary questions and that the gene concept has continued to evolve in concert with new discoveries in cellular biochemistry and comparative genomics (Griffiths and Neumann-Held 1999; Neumann-Held 1999; Moss 2001; Wu and Morris 2001). Griffiths and Neumann-Held (1999) have gone so far as to distinguish between the ‘‘molecular gene’’ and the ‘‘evolutionary gene.’’ While the gene concept will undoubtedly remain useful even while it continues to mutate, we suggest that only by appreciating the centrality of nongenetic determinants in the evolution of developmental mechanisms, and in particular the inherent properties of tissues at different stages of the evolutionary process, will a synthetic, causal understanding of morphological form be achieved. Undeniably, random genetic change is an important factor of evolution, and programmed gene expression is a key aspect of embryogenesis. But if the organization of organisms is (in an evolutionary sense) a largely predictable consequence of the inherent material properties of their constituent tissues, then it is inaccurate to see genetic change as the driving force (as opposed to one necessary component) of hereditary morphological change. Similarly, if organismal forms were originally brought about by interactions of cell masses with one another and with their microenviron-

62

Genes and Form

ments based on their inherent properties, and only later integrated into developmental repertoires by stabilizing and canalizing genetic evolution, then the causal basis of developmental morphogenesis cannot simply be differential gene expression. The following analogy has proved useful. It is well recognized that the three-dimensional structure of a protein molecule is basically determined by its inherent distribution of charged and hydrophobic residues in interaction with its external physical environment, based on generic laws of thermodynamics. Nonetheless, the pathway by which this form is actually realized in a cell depends on a detailed series of free-energy-consuming steps utilizing a variety of coevolved accessory proteins (Grantcharova et al. 2001). The folding of the protein utilizing determinants different from those that define its form thereby integrates it into the cellular economy. Similarly, an understanding of the forms assumed by metazoan organisms during their development requires knowledge of the generative epigenetic processes that originally (in evolutionary history) produced those forms, not only the genetic circuitry that integrates them into the developmental repertoire (for further elaboration of these points, see Müller and Newman 2003). This view seeks the organism’s organizational basis not in any ‘‘blueprint’’ or ‘‘program,’’ but rather in an evolutionary process by which tissues take form by initially generic, and then increasingly specific, inherent material properties. At the earliest stages of this process, form generation was prolific and predictable in outcome. At later stages, it was more constrained and peculiar in outcome. At all stages, tissue-environment and tissue-tissue interactivity ensured that outcomes were plastic and that novelties were possible. At no stage was gene mutation or alteration in gene expression anything but a precipitating event for the mobilization of a morphogenetic potential. And however much development of modern organisms appears in the form of programmed behavior of machines, our considerations suggest that scientific investigation of organismal form must relinquish the ‘‘reverse engineering’’ approach which is compelled by the neo-Darwinian/functionalist paradigm and has become the standard in developmental biological research. Instead we need an ‘‘archaeological’’ perspective that recognizes that modern activities of form generation are based on principles whose origins have often become obscured by history.

63

Stuart A. Newman and Gerd B. Mu¨ller

References Abouheif, E. 1999. Establishing homology criteria for regulatory gene networks: prospects and challenges. In: Homology (pp. 207–225). Novartis Foundation Symposium 222. Chichester: Wiley. Agutter, P. S., and Wheatley, D. N. 2000. Random walks and cell size. BioEssays 22: 1018–1023. Alberch, P. 1982. Developmental constraints in evolutionary processes. In: J. T. Bonner (ed.), Evolution and Development (pp. 313–332). Berlin: SpringerVerlag. Atchley, W. R., Newman, S., and Cowley, D. E. 1988. Genetic divergence in mandible form in relation to molecular divergence in inbred mouse strains. Genetics 120: 239–253. Baldwin, J. M. 1896. A new factor in evolution. Amer. Nat. 30: 441–451, 536–553. Beloussov, L. 1998. The Dynamic Architecture of a Developing Organism. Dordrecht: Kluwer. Boissonade, J., Dulos, E., and DeKepper, P. 1994. Turing patterns: from myth to reality. In: R. Kapral and K. Showalter (eds.), Chemical Waves and Patterns (pp. 221–268). Boston: Kluwer. Bolker, J. A., and Raff, R. A. 1996. Developmental genetics and traditional homology. BioEssays 18: 489–494. Bonner, J. T. 1967. The Cellular Slime Molds. Princeton: Princeton University Press. Bonner, J. T. 1998. The origins of multicellularity. Integr. Biol. 1: 27–36. Bruna, E. M., Fisher, R. N., and Case, T. J. 1996. Morphological and genetic evolution appear decoupled in Pacific skinks (Emoia). Proc. Roy. Soc. Lond. B 263: 681–688. Butler, S. 1878. Life and Habit. London: Fifield. Chuong, C. M., and Widelitz, R. B. 1998. Feather formation: a model of the formation of epithelial appendages. In: C. M. Chuong (ed.), Molecular Basis of Epithelial Appendage Morphogenesis (pp. 57–74). Austin, Tex.: R. G. Landes. Colas, J. F., and Schoenwolf, G. C. 2001. Towards a cellular and molecular understanding of neurulation. Dev. Dyn. 221: 117–145. Conway Morris, S. 1989. Burgess Shale faunas and the Cambrian explosion. Science 246: 339–346. Conway Morris, S. 1993. The fossil record and the early evolution of the Metazoa. Nature 361: 219–225. Cooke, J., Nowak, M. A., Boerlijst, M., and Maynard Smith, J. 1997. Evolutionary origins and maintenance of redundant gene expression during metazoan development. Trends Genet. 13: 360–364. Crick, F. H. C. 1970. Diffusion in embryogenesis. Nature 225: 420–422.

64

Genes and Form

Crick, F. H. C., and Lawrence, P. A. 1975. Compartments and polyclones in insect development. Science 189: 340–347. de Gennes, P. G. 1992. Soft matter. Science 256: 495–497. Dobzhansky, T. 1973. Nothing in biology makes sense except in the light of evolution. Amer. Biol. Teach. 35: 125–129. Downie, S. A., and Newman, S. A. 1994. Morphogenetic differences between fore and hind limb precartilage mesenchyme: relation to mechanisms of skeletal pattern formation. Dev. Biol. 162: 195–208. Drasdo, D., and Forgacs, G. 2000. Modeling the interplay of generic and genetic mechanisms in cleavage, blastulation, and gastrulation. Dev. Dyn. 219: 182–191. Eckstein, H. 1980. Theoretical approaches to explaining collective violence. In: T. R. Gurr (ed.), Handbook of Political Conflict (pp. 135–167). New York: Free Press. Ekblom, P. 1992. Renal Development. New York: Raven Press. Emlen, D. J., and Nijhout, H. F. 2000. The development and evolution of exaggerated morphologies in insects. Annu. Rev. Entomol. 45: 661–708. Evanko, S. P., and Vogel, K. G. 1993. Proteoglycan synthesis in fetal tendon is differentially regulated by cyclic compression in vitro. Arch. Biochem. Biophys. 307: 153–164. Fang, J., and Hall, B. K. 1997. Chondrogenic cell differentiation from membrane bone periostea. Anat. Embryol. (Berlin) 196: 349–362. Forgacs, G. 1995. On the possible role of cytoskeletal filamentous networks in intracellular signalling: an approach based on percolation. J. Cell Sci. 108: 2131– 2143. Frenz, D. A., Jaikaria, N. S., and Newman, S. A. 1989. The mechanism of precartilage mesenchymal condensation: a major role for interaction of the cell surface with the amino-terminal heparin-binding domain of fibronectin. Dev. Biol. 136: 97–103. Furisawa, C., and Kaneko, K. 2000. Complex organization in multicellularity as a necessity in evolution. Artif. Life 6: 265–281. Galis, F., and Drucker, E. G. 1996. Pharyngeal biting mechanics in centrarchid and cichlid fishes: insights into a key evolutionary innovation. J. Evol. Biol. 9: 641–670. Galis, F., and Metz, J. A. 2001. Testing the vulnerability of the phylotypic stage: on modularity and evolutionary conservation. J. Exp. Zool. 291: 195–204. Garcia-Bellido, A. 1975. Genetic control of wing disc development in Drosophila. Ciba Found. Symp. 29: 169–178. Gerhart, J., and Kirschner, M. 1997. Cells, Embryos, and Evolution. Malden, Mass.: Blackwell. Gierer, A. 1977. Physical aspects of tissue evagination and biological form. Q. Rev. Biophys. 10: 529–593. Gilbert, L. I., Tata, J. R., and Atkinson, B. G. 1996. Metamorphosis: Postembryonic

65

Stuart A. Newman and Gerd B. Mu¨ller

Reprogramming of Gene Expression in Amphibian and Insect Cells. San Diego: Academic Press. Giori, N. J., Beaupre, G. S., and Carter, D. R. 1993. Cellular shape and pressure may mediate mechanical control of tissue composition in tendons. J. Orthop. Res. 11: 581–591. Godt, D., and Tepass, U. 1998. Drosophila oocyte localization is mediated by differential cadherin-based adhesion. Nature 395: 387–391. Goldbeter, A. 1995. Biochemical Oscillations and Cellular Rhythms: The Molecular Bases of Periodic and Chaotic Behavior. Cambridge: Cambridge University Press. Gonzalez-Reyes, A., and St. Johnston, D. 1998. The Drosophila ap axis is polarised by the cadherin-mediated positioning of the oocyte. Development 125: 3635–3644. Goodwin, B. C. 1994. How the Leopard Changed Its Spots. London: Weidenfeld and Nicolson. Goto, T., MacDonald, P., and Maniatis, T. 1989. Early and late periodic patterns of even skipped expression are controlled by distinct regulatory elements that respond to different spatial cues. Cell 57: 413–422. Gould, S. J. 1989. Wonderful Life. New York: W. W. Norton. Gould, S. J., and Lewontin, R. C. 1979. The spandrels of San Marco and the panglossian paradigm. Proc. Roy. Soc. Lond. B 205: 581–598. Grant, D. S., Lelkes, P. I., Fukuda, K., and Kleinman, H. K. 1991. Intracellular mechanisms involved in basement membrane induced blood vessel differentiation in vitro. In Vitro Cell Dev. Biol. 27A: 327–336. Grantcharova, V., Alm, E. J., Baker, D., and Horwich, A. L. 2001. Mechanisms of protein folding. Curr. Opin. Struct. Biol. 11: 70–82. Griffiths, P. E., and Neumann-Held, E. M. 1999. The many faces of the gene. BioScience 49: 656–662. Grinnell, F. 1994. Fibroblasts, myofibroblasts, and wound contraction. J. Cell. Biol. 124: 401–404. Gurdon, J. B., and Bourillot, P. Y. 2001. Morphogen gradient interpretation. Nature 413: 797–803. Guthrie, S., and Lumsden, A. 1991. Formation and regeneration of rhombomere boundaries in the developing chick hindbrain. Development 112: 221–229. Hall, B. K. 1986. The role of movement and tissue interactions in the development and growth of bone and secondary cartilage in the clavicle of the embryonic chick. J. Embryol. Exp. Morphol. 93: 133–152. Hall, B. K. (ed.) 1994. Homology. San Diego: Academic Press. Hall, B. K., and Miyake, T. 1995. Divide, accumulate, differentiate: cell condensation in skeletal development revisited. Int. J. Dev. Biol. 39: 881–893. Hall, B. K., and Miyake, T. 2000. All for one and one for all: condensations and the initiation of skeletal development. Bioessays 22: 138–147.

66

Genes and Form

Harding, K., Hoey, T., Warrior, R., and Levine, M. 1989. Autoregulatory and gap gene response elements of the even-skipped promoter of Drosophila. EMBO J. 8: 1205–1212. Heintzelman, K. F., Phillips, H. M., and Davis, G. S. 1978. Liquid-tissue behavior and differential cohesiveness during chick limb budding. J. Embryol. Exp. Morphol. 47: 1–15. Heyman, I., Faissner, A., and Lumsden, A. 1995. Cell and matrix specialisations of rhombomere boundaries. Dev. Dyn. 204: 301–315. Ho, M. W., and Saunders, P. T. 1979. Beyond neo-Darwinism—an epigenetic approach to evolution. J. Theor. Biol. 78: 573–591. Holland, P. W. H. 1999. The effect of gene duplication on homology. In: Homology (pp. 226–242). Novartis Foundation Symposium 222. Chichester: Wiley. Huan, Y., and van Adelsberg, J. 1999. Polycystin-1, the pkd1 gene product, is in a complex containing e-cadherin and the catenins. J. Clin. Invest. 104: 1459–1468. Hutchings, M. J., and de Kroon, H. 1994. Foraging in plants: the role of morphological plasticity in resource acquisition. Adv. Ecol. Res. 25: 159–238. Ingber, D. E., Dike, L., Hansen, L., Karp, S., et al. 1994. Cellular tensegrity: exploring how mechanical changes in the cytoskeleton regulate cell growth, migration, and tissue pattern during morphogenesis. Int. Rev. Cytol. 150: 173–224. Irvine, K. D., and Wieschaus, E. 1994. Cell intercalation during Drosophila germband extension and its regulation by pair-rule segmentation genes. Development 120: 827–841. Ish-Horowicz, D., Pinchin, S. M., Ingham, P. W., and Gyurkovics, H. G. 1989. Autocatalytic ftz activation and instability induced by ectopic ftz expression. Cell 57: 223–232. Jablonka, E., and Lamb, M. J. 1995. Epigenetic Inheritance and Evolution. Oxford: Oxford University Press. Johnston, T. D., and Gottlieb, G. 1990. Neophenogenesis: a developmental theory of phenotypic evolution. J. Theor. Biol. 147: 471–495. Jouve, C., Iimura, T., and Pourquié, O. 2002. Onset of the segmentation clock in the chick embryo: evidence for oscillations in the somite precursors in the primitive streak. Development 129: 1107–1117. Kaneko, K., and Yomo, T. 1999. Isologous diversification for robust development of cell society. J. Theor. Biol. 199: 243–256. Kashimata, M., and Gresik, E. W. 1996. Contemporary approaches to the study of salivary gland morphogenesis. Eur. J. Morphol. 34: 143–147. Kauffman, S. A. 1993. The Origins of Order. New York: Oxford University Press. Kazmierczak, J., and Degens, E. T. 1986. Calcium and the early eukaryotes. Mitt. Geol.-Palaeont. Inst. Univ. Hamburg 61: 1–20. Kessler, D. S., and Melton, D. A. 1994. Vertebrate embryonic induction: mesodermal and neural patterning. Science 266: 596–604. Koch, J. 1917. The laws of bone architecture. Amer. J. Anat. 21: 177–298.

67

Stuart A. Newman and Gerd B. Mu¨ller

Lambert C. A., Nusgens, B. V., and Lapiere, C. M. 1998. Mechano-sensing and mechano-reaction of soft connective tissue cells. Adv. Space Res. 21: 1081–1091. Lambert, C. A., Soudant, E. P., Nusgens, B. V., and Lapiere, C. M. 1992. Pretranslational regulation of extracellular matrix macromolecules and collagenase expression in fibroblasts by mechanical forces. Lab. Invest. 66: 444–451. Leonard, C. M., Fuld, H. M., Frenz, D. A., et al. 1991. Role of transforming growth factor-beta in chondrogenic pattern formation in the embryonic limb: stimulation of mesenchymal condensation and fibronectin gene expression by exogenous tgf-beta and evidence for endogenous tgf-beta-like activity. Dev. Biol. 145: 99–109. Li, M. L., Aggeler, J., Farson, D. A., et al. 1987. Influence of a reconstituted basement membrane and its components on casein gene expression and secretion in mouse mammary epithelial cells. Proc. Natl. Acad. Sci. USA 84: 136–140. Lieber, T., Wesley, C. S., Alcamo, E., et al. 1992. Single amino acid substitutions in egf-like elements of Notch and Delta modify Drosophila development and affect cell adhesion in vitro. Neuron 9: 847–859. Lincz, L. F., Buret, A., and Burns, G. F. 1997. Formation of spheroid structures in a human colon carcinoma cell line involves a complex series of intercellular rearrangements. Differentiation 61: 261–274. Lund, L. R., Romer, J., Thomasset, N., et al. 1996. Two distinct phases of apoptosis in mammary gland involution: proteinase-independent and -dependent pathways. Development 122: 181–193. Magee, P. T. 1997. Which came first, the hypha or the yeast? Science 277: 52–53. Maynard Smith, J., Burian, R., Kauffman, S., et al. 1985. Developmental constraints and evolution. Q. Rev. Biol. 60: 265–287. McGrew, M. J., Dale, J. K., Fraboulet, S., and Pourquié, O. 1998. The lunatic fringe gene is a target of the molecular clock linked to somite segmentation in avian embryos. Curr. Biol. 8: 979–982. McLaren, A., and Michie, D. 1958. An effect of the uterine environment upon skeletal morphology in the mouse. Nature 181: 1147–1148. Meyer, A., Kocher, T. D., Basasibwaki, P., and Wilson, A. C. 1990. Monophyletic origins of Lake Victoria cichlid fishes suggested by mitochondrial dna sequences. Nature 347: 550–553. Mikhailov, A. S. 1990. Foundations of Synergetics I. Berlin: Springer-Verlag. Mittenthal, J. E., and Mazo, R. M. 1983. A model for shape generation by strain and cell-cell adhesion in the epithelium of an arthropod leg segment. J. Theor. Biol. 100: 443–483. Miura, T., and Shiota, K. 2000a. Extracellular matrix environment influences chondrogenic pattern formation in limb bud micromass culture: experimental verification of theoretical models. Anat. Rec. 258: 100–107. Miura, T., and Shiota, K. 2000b. tgfbeta2 acts as an activator molecule in

68

Genes and Form

reaction-diffusion model and is involved in cell sorting phenomenon in mouse limb micromass culture. Dev. Dyn. 217: 241–249. Moftah, M. Z., Downie, S. A., Bronstein, N. B., et al. 2002. Ectodermal fgfs induce perinodular inhibition of limb chondrogenesis in vitro and in vivo via fgf receptor 2. Dev. Biol. 249: 270–282. Moore, J., and Willmer, P. 1997. Convergent evolution in invertebrates. Biol. Rev. Camb. Philos. Soc. 72: 1–60. Moss, L. 2001. Deconstructing the gene and reconstructing molecular developmental systems. In: S. Oyama, P. E. Griffiths, and R. D. Gray (eds.), Cycles of Contingency: Developmental Systems and Evolution. Cambridge: mit Press. Müller, G. B. 1990. Developmental mechanisms at the origin of morphological novelty: a side-effect hypothesis. In: M. H. Nitecki (ed.), Evolutionary Innovations (pp. 99–130). Chicago: University of Chicago Press. Müller, G. B. 2002. Novelty and key innovation. In: M. Pagel (ed.), Encyclopedia of Evolution. Oxford: Oxford University Press. Müller, G. B. 2003a. Embryonic motility: environmental influences and evolutionary innovation. Evol. Dev. 5: 56–60. Müller, G. B. 2003b. Homology: the evolution of morphological organization. In: G. B. Müller and S. A. Newman (eds.) Origination of Organismal Form: Beyond the Gene in Developmental and Evolutionary Biology. Cambridge: mit Press. Müller, G. B., and Newman, S. A. 1999. Generation, integration, autonomy: three steps in the evolution of homology. In: Homology (pp. 65–79). Novartis Foundation Symposium 222. Chichester: Wiley. Müller, G. B., and Olsson L. 2003. Epigenesis and Epigenetics. In: B. K. Hall and W. M. Olson (eds.), Keywords and Concepts in Evolutionary Developmental Biology (pp. 114–123). Cambridge: Harvard University Press. Müller, G. B., and Streicher, J. 1989. Ontogeny of the syndesmosis tibiofibularis and the evolution of the bird hindlimb: a caenogenetic feature triggers phenotypic novelty. Anat. Embryol. 179: 327–339. Müller, G. B., and Wagner, G. P. 1991. Novelty in evolution: restructuring the concept. Annu. Rev. Ecol. Syst. 22: 229–256. Müller, G. B., and Wagner, G. P. 1996. Homology, Hox genes, and developmental integration. Amer. Zool. 36: 4–13. Murray, A. W., and Hunt, T. 1993. The Cell Cycle: An Introduction. New York: W. H. Freeman. Neumann-Held, E. M. 1999. The gene is dead—long live the gene! Conceptualizing genes the constructionist way. In: P. Koslowsky (ed.), Sociobiology and Bioeconomics: The Theory of Evolution in Biological and Economic Theory (pp. 105–137). Berlin: Springer-Verlag. Newman, S. A. 1988. Lineage and pattern in the developing vertebrate limb. Trends Genet. 4: 329–332. Newman, S. A. 1993. Is segmentation generic? BioEssays 15: 277–283.

69

Stuart A. Newman and Gerd B. Mu¨ller

Newman, S. A. 1994. Generic physical mechanisms of tissue morphogenesis: a common basis for development and evolution. J. Evol. Biol. 7: 467–488. Newman, S. A. 1996. Sticky fingers: Hox genes and cell adhesion in vertebrate limb development. BioEssays 18: 171–174. Newman, S. A. 1998a. Epithelial morphogenesis: a physico-evolutionary interpretation. In: C. M. Chuong (ed.), Molecular Basis of Epithelial Appendage Morphogenesis (pp. 341–358). Austin, Tex.: R.G. Landes. Newman, S. A. 1998b. Networks of extracellular fibers and the generation of morphogenetic forces. In: D. Beysens and G. Forgacs (eds.), Dynamical Networks in Physics and Biology (pp. 139–148). Berlin: Springer-Verlag. Newman, S. A., and Comper, W. D. 1990. ‘Generic’ physical mechanisms of morphogenesis and pattern formation. Development 110: 1–18. Newman, S. A., Frenz, D. A., Tomasek, J. J., and Rabuzzi, D. D. 1985. Matrixdriven translocation of cells and nonliving particles. Science 228: 885–889. Newman, S. A., and Frisch, H. L. 1979. Dynamics of skeletal pattern formation in developing chick limb. Science 205: 662–668. Newman, S. A., and Müller, G. B. 2001. Epigenetic mechanisms of character origination. In: G. P. Wagner (ed.), The Character Concept in Evolutionary Biology (pp. 559–579). San Diego: Academic Press. Newman, S. A., and Tomasek, J. J. 1996. Morphogenesis of connective tissues in extracellular matrix. In: W. D. Comper (ed.), Molecular Components and Interactions (2: 335–369). Amsterdam: Harwood Academic. Nijhout, H. F. 1990. Metaphors and the roles of genes in development. BioEssays 12: 441–446. Nowak, M. A., Boerlijst, M. C., Cooke, J., and Smith, J. M. 1997. Evolution of genetic redundancy. Nature 388: 167–171. Oyama, S. 1985. The Ontogeny of Information. Cambridge: Cambridge University Press; 2d ed. Durham: Duke University Press, 2000. Palmeirim, I., Henrique, D., Ish-Horowicz, D., and Pourquié, O. 1997. Avian hairy gene expression identifies a molecular clock linked to vertebrate segmentation and somitogenesis. Cell 91: 639–648. Parsons, J. T., Martin, K. H., Slack, J. K., et al. 2000. Focal adhesion kinase: a regulator of focal adhesion dynamics and cell movement. Oncogene 19: 5606–5613. Patel, N. H., Kornberg, T. B., and Goodman, C. S. 1989. Expression of engrailed during segmentation in grasshopper and crayfish. Development 107: 201–212. Picket, F. B., and Meeks-Wagner, D. R. 1995. Seeing double: appreciating genetic redundancy. Plant Cell 7: 1347–1356. Pigliucci, M. 1996. Modelling phenotypic plasticity. II. Do genetic correlations matter? Heredity 77: 453–460. Pourquié, O. 2003. A biochemical oscillator linked to vertebrate segmentation. In: G. B. Müller and S. A. Newman (eds.), Origination of Organismal Form: Be-

70

Genes and Form

yond the Gene in Developmental and Evolutionary Biology (pp. 183–194). Cambridge: mit Press. Prigogine, I., and Nicolis, G. 1971. Biological order, structure and instabilities. Q. Rev. Biophys. 4: 107–148. Qian, F., Watnick, T. J., Onuchic, L. F., and Germino, G. G. 1996. The molecular basis of focal cyst formation in human autosomal dominant polycystic kidney disease type I. Cell 87: 979–987. Ranheim, T. S., Edelman, G. M., and Cunningham, B. A. 1996. Homophilic adhesion mediated by the neural cell adhesion molecule involves multiple immunoglobulin domains. Proc. Natl. Acad. Sci. USA 93: 4071–4075. Riedl, R. 1978. Order in Living Systems: A Systems Analysis of Evolution. New York: Wiley. Rodriguez, I., and Basler, K. 1997. Control of compartmental affinity boundaries by hedgehog. Nature 389: 614–618. Rodriguez-Boulan, E., and Nelson, W. J. (eds.). 1993. Epithelial and Neuronal Cell Polarity. Cambridge: Company of Biologists. Russell, E. S. 1916. Form and Function. Chicago: University of Chicago Press. Rutherford, S. L., and Lindquist, S. 1998. Hsp90 as a capacitor for morphological evolution. Nature 396: 336–342. Salazar-Ciudad, I., Newman, S. A., and Solé, R. 2001a. Phenotypic and dynamical transitions in model genetic networks. I. Emergence of patterns and genotypephenotype relationships. Evol. Dev. 3: 84–94. Salazar-Ciudad, I., Solé, R., and Newman, S. A. 2001b. Phenotypic and dynamical transitions in model genetic networks. II. Application to the evolution of segmentation mechanisms. Evol. Dev. 3: 95–103. Sanderson, M. J., and L. Hufford (eds.). 1996. Homoplasy: The Recurrence of Similarity in Evolution. San Diego: Academic Press. Saunders, J. W. Jr. 1948. The proximo-distal sequence of origin of the parts of the chick wing and the role of the ectoderm. J. Exp. Zool. 108: 363–402. Schlichting, C., and Pigliucci, M. 1998. Phenotypic Evolution: A Reaction Norm Perspective. Sunderland, Mass: Sinauer. Schmalhausen, I. I. 1949. Factors of Evolution. Philadelphia: Blakiston. Seilacher, A. 1991. Self-organizing mechanisms in morphogenesis and evolution. In: N. Schmidt-Kittler and K. Vogel (eds.), Constructional Morphology and Evolution (pp. 251–271). Berlin: Springer-Verlag. Seilacher, A. 1992. Vendobionta and Psammocorallia—lost constructions of Precambrian evolution. J. Geol. Soc. Lond. 149: 607–613. Seilacher, A., Bose, P. K., and Pfluger, F. 1998. Triploblastic animals more than 1 billion years ago: trace fossil evidence from India. Science 282: 80–83. Sheard, P., and Jacobson, M. 1987. Clonal restriction boundaries in Xenopus embryos shown with two intracellular lineage tracers. Science 236: 851–854.

71

Stuart A. Newman and Gerd B. Mu¨ller

Sheard, P., and Jacobson, M. 1990. Analysis of frequency of intermingling between labeled clones in Xenopus embryos. Ann. N.Y. Acad. Sci. 599: 141–157. Simpson, G. G. 1953. The Baldwin effect. Evolution 7: 110–117. Small, S., Kraut, R., Hoey, T., et al. 1991. Transcriptional regulation of a pair-rule stripe in Drosophila. Genes Dev. 5: 827–839. Stearns, S. C. 1986. Natural selection and fitness, adaptation and constraint. In: D. M. Raup and D. Jablonski (eds.), Patterns and Processes in the History of Life (pp. 23–44). Berlin: Springer-Verlag. Stearns, S. C. 2000. Life history evolution: successes, limitations, and prospects. Naturwissenschaften 87: 476–486. Steinberg, M. S. 1998. Goal-directedness in embryonic development. Integrat. Biol. 1: 49–59. Steinberg, M. S., and Takeichi, M. 1994. Experimental specification of cell sorting, tissue spreading, and specific spatial patterning by quantitative differences in cadherin expressions. Proc. Natl. Acad. Sci. USA 91: 206–209. Streicher, J., and Müller, G. B. 1992. Natural and experimental reduction of the avian fibula: developmental thresholds and evolutionary constraint. J. Morphol. 214: 269–285. Sturmbauer, C., and Meyer, A. 1992. Genetic divergence, speciation, and morphological stasis in a lineage of African cichlid fishes. Nature 358: 578–581. Takahashi, I., Nuckolls, G. H., Takahashi, K., et al. 1998. Compressive force promotes sox9, type II collagen and aggrecan and inhibits il-1beta expression resulting in chondrogenesis in mouse embryonic limb bud mesenchymal cells. J. Cell Sci. 111: 2067–2076. Takeichi, M., Nakagawa, S., Aono, S., et al. 2000. Patterning of cell assemblies regulated by adhesion receptors of the cadherin superfamily. Phil. Trans. Roy. Soc. Lond. B, Biol. Sci. 355: 885–890. Tautz, D. 1992. Redundancies, development and the flow of information. BioEssays 14: 263–266. Troyanovsky, S. M. 1999. Mechanism of cell-cell adhesion complex assembly. Curr. Opin. Cell Biol. 11: 561–566. Trussell, G. C. 2000. Phenotypic clines, plasticity, and morphological trade-offs in an intertidal snail. Evol. Int. J. Org. Evol. 54: 151–166. Turing, A. 1952. The chemical basis of morphogenesis. Phil. Trans. Roy. Soc. Lond. B 237: 37–72. Unemori, E. N., and Werb, Z. 1986. Reorganization of polymerized actin: a possible trigger for induction of procollagenase in fibroblasts cultured in and on collagen gels. J. Cell. Biol. 103: 1021–1031. Van Tienderen, P. H., and Koelewijn, H. P. 1994. Selection on reaction norms, genetic correlations and constraints. Genet. Res. 64: 115–125. Vogel, K. G., and Koob, T. J. 1989. Structural specialization in tendons under compression. Int. Rev. Cytol. 115: 267–293.

72

Genes and Form

Waddington, C. H. 1942. Canalization of development and the inheritance of acquired characters. Nature 150: 563–565. Waddington, C. H. 1961. Genetic assimilation. Adv. Genet. 10: 257–293. Wagner, A. 1996. Genetic redundancy caused by gene duplications and its evolution in networks of transcriptional regulators. Biol. Cybern. 74: 557–567. Wagner, G. P. 1989. The biological homology concept. Annu. Rev. Ecol. Syst. 20: 51–69. Wagner, G. P. 1995. The biological role of homologues: a building block hypothesis. N. JB. Geol. Paläont. Abh. 195: 279–288. Wake, D. B. 1991. Homoplasy: the result of natural selection or evidence of design limitations? Amer. Nat. 138: 543–567. Wedeen, C. J. 1995. Regionalization and segmentation of the leech. J. Neurobiol. 27: 277–293. Whittington, H. B. 1985. The Burgess Shale. New Haven: Yale University Press. Wilkins, A. S. 1997. Canalization: a molecular genetic perspective. BioEssays 19: 257–262. Wilkinson, D. G. 2000. Eph receptors and ephrins: regulators of guidance and assembly. Int. Rev. Cytol. 196: 177–244. Wimsatt, W. C. 1986. Developmental constraints, generative entrenchment, and the innate-acquired distinction. In: W. Bechtel (ed.), Integrating Scientific Disciplines (pp. 185–208). Nijhoff: Dordrecht. Winfree, A. T. 1994. Persistent tangled vortex rings in generic excitable media. Nature 371: 233–236. Wray, G. A. 2001. Resolving the Hox paradox. Science 292: 2256–2257. Wray, G. A., and Raff, R. A. 1991. The evolution of developmental strategy in marine invertebrates. TREE 6: 45–50. Wright, D. C., and Lawrence, P. A. 1981. Regeneration of the segment boundary in Oncopeltus. Dev. Biol. 85: 317–327. Wu, C. T., and Morris, J. R. 2001. Genes, genetics, and epigenetics: a correspondence. Science 293: 1103–1105. Wu, K. C. 1996. Entwicklung, Stimulation und Paralyse der embryonalen Motorik. Wien. Klin. Wochenschr. 108: 303–305.

73

3 FROM GENES AS D E T E R M I N A N T S TO D N A A S R E S O U R C E Historical Notes on Development and Genetics

sahotra sarkar H. J. Muller’s 1926 address to a symposium on ‘‘The Gene’’ at the International Congress of Plant Sciences on 19 August at Ithaca, New York, was titled ‘‘The Gene as the Basis of Life’’ (Muller 1962: 188–204). Because genes must have autocatalytic properties, Muller argued, the ‘‘gene . . . arose coincidentally with growth and ‘life’ itself ’’ (Muller 1962: 200). Not only were genes thus constitutive of life, but all of evolution was to be explained from a genetic basis: ‘‘In all probability all specific, generic, and phyletic differences, of every order, between the highest and lowest organisms, the most diverse metaphyta and metazoa, are ultimately referable to changes in . . . genes’’ (Muller 1962: 195).1 The same year, Muller’s mentor, T. H. Morgan, published The Theory of the Gene, summarizing fifteen years of research, primarily on the fruit fly Drosophila melanogaster, that established the hegemony of genetics in twentieth-century biology. Trained as an embryologist, Morgan had denied the full significance of both Darwinism and Mendelism at least until 1910 when he discovered sexlimited Mendelian inheritance of a trait (the mutant white eye in D. melanogaster; see Morgan 1910). That discovery spawned a path-breaking research program in genetics. By 1925, Morgan and his laboratory had investigated about four hundred mutant characters of D. melanogaster. Through the systematic use of linkage mapping, these were partitioned into four linkage groups corresponding to the four chromosome pairs. That nothing was known about the developmental genesis of these traits at the level of cell, tissue, or organ did not in any way impede these investigations. Thus, by 1926, Morgan had not only come to accept and insist on Mendelism as the theory

Sahotra Sarkar

of heredity, he was ready to demand a sharp divorce of genetics from development: Between the characters, that furnish the data for the [Mendelian] theory and the postulated genes, to which the characters are referred, lies the whole field of embryonic development. The theory of the gene, as here formulated, states nothing with respect to the way in which the genes are connected with the end-product or character. The absence of information relating to this interval does not mean that the process of embryonic development is not of interest for genetics . . . but the fact remains that the sorting out of the characters in successive generations can be explained at present without reference to the way in which the gene affects the developmental process. (Morgan 1926: 26)

Morgan was not the first to suggest such an analytic strategy; in 1914 William Bateson, in his presidential address to the British Association for the Advancement of Science, had also noted that the possibility of this separation is the feature that best characterized the new Mendelian genetics (Bateson 1914). Meanwhile, genes were slowly acquiring the material reality that most skeptics of Mendelism had long demanded of them. Muller’s successes at inducing mutations through physical processes, particularly X-rays, added confidence in the position that genes were associated with definite material objects (Muller 1927, repr. 1962: 245–251). The physical interpretation of Mendelism established what came to be called classical genetics. In the 1930s and early 1940s, genes were thought to be composed of protein; nucleic acids composed of only four nucleotide bases (A: adenine; C: cytosine; G: guanine; and T: thymine) as building blocks were believed not to be complex enough to provide the variability required by the hundreds of known genes. In 1944, however, Oswald T. Avery, Colin M. MacLeod, and Maclyn McCarty experimentally demonstrated that, at least in bacteria, genes were composed of dna (Avery, MacLeod, and McCarty 1944). The same year, the physicist, Erwin Schrödinger, in a book called What Is Life?, produced an ingenious combinatorial argument showing that even composites from a small number of building blocks can have more than the amount of variety required of genes (Schrödinger 1944). While the significance of this argument went largely unrecognized, Schrödinger’s book played a key role in encouraging physical scientists to tackle biological problems, leading to the rapid expansion of molecular biology in the 1950s. The most crucial development was the decipherment of the structure of dna by Watson and Crick in 1953 (Watson and Crick 1953a). While what gen-

78

From Genes as Determinants to DNA as Resource

erally gets emphasized is the double-helical structure of the model, what is critical to its eventual role in biology is the model of genetic specificity it incorporates. The term specificity was introduced in a genetic context by H. A. Timoféeff-Ressovsky and N. W. Timoféeff-Ressovsky only in 1926 (TimoféeffRessovsky and Timoféeff-Ressovsky 1926). However, the specificity of gene action was a presumption of genetics from its inception. Originally proposed as a one-to-one correspondence between gene and trait, the idea survived in an increasingly mitigated form throughout the twentieth century. The double helix provided a model of specificity entirely new in biology: specificity was achieved by the order of arrangement of nucleotide bases, on the possibility of which only Schrödinger had speculated. This model ushered in the age of biological information: information interpreted as a sequence or arrangement of bases became the model of specificity for genetics.2 Most important, it led to the view that genes were the sole purveyors of biological information. Crick summarized the view in what he called the ‘‘central dogma’’ of molecular biology: ‘‘This states that once ‘information’ has passed into protein it cannot get out again. In more detail, the transfer of information from nucleic acid to nucleic acid, or from nucleic acid to protein may be possible, but transfer from protein to protein, or from protein to nucleic acid is impossible’’ (Crick 1958: 153).3 The contrast here is with the physical model of specificity, stereospecificity, dating back to the immunologist Paul Ehrlich’s side-chain theory from the 1880s, which had begun to dominate structural studies in biology in the 1920s and 1930s (Silverstein 1989). The rise of the informational perspective also reified the view, articulated by Muller, that genes as determinants of biological features were special, different from the other resources used by organisms to carry out living processes. The specificity of the gene-gene product (nucleic acid or protein) relationship was informational and thus different from specificity at every other level of biological organization, which remained physical (or stereospecific). Thus arose the view of dna as the master molecule in charge of development (see ‘‘The Age of the Master Molecule,’’ below). Other views of development were displaced during the process of establishing the hegemony of genetics.

Evocators of Development Around 1900, for biologists in the field and in the laboratory it was far from obvious that organismic traits could be inherited through discrete units like

79

Sahotra Sarkar

Mendel’s factors. There were two problems: (1) discrete Mendelizing traits were rare. Most traits varied continuously (or were ‘‘quantitative’’) and were often normally distributed around a population mean, as hypothesized by the biometricians (Sarkar 1998). Moreover, their inheritance seemed to follow rules such as the Law of Ancestral Inheritance; and (2) developmentally, not only was there no one-to-one correspondence between traits and hereditary factors, there was also ample evidence that the relation between them was not even determinate. For instance, the German zoologist Richard Woltereck studied morphologically distinct strains of Daphnia and Hyalodaphnia species from different lakes. These were pure lines which maintained their form through several generations of parthenogenesis. Woltereck focused on continuous traits such as headheight at varying nutrient levels. The phenotype varied between different pure lines, was affected by some environmental factors such as nutrient levels was almost independent of others such as the ambient temperature, and showed cyclical variation with factors such as seasonality. Moreover, the response of the phenotype to the same environmental change was not identical in different pure lines. Woltereck drew ‘‘phenotype curves’’ to depict this phenomenon. These curves changed for every new variable that was considered. There were thus potentially an almost infinite number of them and Woltereck coined the term Reaktionsnorm to indicate the totality of the relationships embodied in them (Woltereck 1909: 135). Woltereck argued that what was inherited was the Reaktionsnorm, and that hereditary change consisted of a modification of that norm. Even Wilhelm Johannsen, who first made a sharp distinction between genotype and phenotype, endorsed the concept of the reaction norm, which he thought to be ‘‘nearly synonymous’’ with genotype (Johannsen 1911: 133). Only slightly later, Herman Nilsson-Ehle coined the term plasticity to describe the nonunique relation of the genotype to the phenotype and argued that this has general adaptive significance (Nilsson-Ehle 1914). This view found resonance in the Soviet Union where the norm of reaction (now understood as what Woltereck had called single phenotypic curves) emerged as a concept of central importance in genetics. Avoiding any genetic determinism was clearly concordant with the Soviet program of producing an interpretation of science based on dialectical materialism. However, in the West (that is, the United States and Europe outside what became the Soviet Union), where Johannsen’s sharp genotype-phenotype distinction became part of the standard picture of genetics, the subsequent decades witnessed a general trend to emphasize the constancy and causal efficacy of the genotype at the expense

80

From Genes as Determinants to DNA as Resource

of the complexity of its interactions. The norm of reaction (nor) remained a relatively ignored concept during this period (see Sarkar 1999). Ironically, the conceptual reticulation of classical genetics that helped maintain the primacy of the gene also emerged from developments in the Soviet Union. There, in the 1920s, an active genetics research group formed around the pioneering population geneticist, S. Chetverikov (Adams 1980). In 1922, one member of the group, Romashoff, discovered the Abdomen abnormalis mutation in Drosophila funebris which resulted in the degeneration of abdominal stripes (Romashoff 1925). There was individual variability in the mutant phenotype which Romashoff interpreted as a difference in the strength of the mutation’s effect. The manifestation of the mutation depended on environmental factors—in particular on the dryness and liquid content of food—but Romashoff could not rule out the possible influence of other genes. Another member of that group, N. W. Timoféeff-Ressovsky, studied the recessive Radius incompletus mutation of D. funebris (TimoféeffRessovsky 1925). In mutant flies, the second longitudinal vein does not reach the end of the wing. Timoféeff created different pure lines, each homozygous for this mutation. Descendants included phenotypically normal flies. The proportion of normals was fixed for each pure line but varied between lines. External factors had little influence; the differences between the lines were apparently under the control of genotypic factors. Some lines gave a large proportion of mutants but manifested the mutation weakly; in others, the converse was realized. The Soviet work was being carefully followed by the German neuroanatomist Oskar Vogt, who was a frequent visitor to Moscow because of a project to dissect Lenin’s brain to demonstrate his genius (see Laubichler and Sarkar 2002). Vogt, long committed to a genetic interpretation of psychoses, introduced two new concepts to describe Timoféeff ’s results: a mutation’s ‘‘expressivity’’ was the extent of its manifestation, and its ‘‘penetrance’’ was the proportion of individuals carrying it which manifested any effect at all. In Vogt’s definitions the differences between different lines was ignored. Expressivity and penetrance became properties of the gene rather than a property of a mutation relative to a constant genetic background. Timoféeff enthusiastically endorsed the new concepts (Timoféeff-Ressovsky and TimoféeffRessovsky 1926). What the original results of Romashoff and Timoféeff had shown was a predictable complexity in the genotype-environment interaction. Both data sets permitted the construction of nors though Vogt’s reinterpretation made such a move moot. Two related aspects of that reinterpretation

81

Sahotra Sarkar

deserve emphasis: (1) Vogt ignored the systematic differences between pure lines; and (2) he explicitly introduced expressivity and penetrance as properties of genes on par with, though different from, dominance. The introduction of expressivity and penetrance constituted a convoluted reticulation of the structure of Mendelian genetics by ad hoc extension of the concept of the gene: Besides having their standard transmission properties, genes were no longer only recessive or dominant (or displaying varying degrees of dominance), they also had degrees of expressivity and penetrance. There was no clear distinction between expressivity and dominance: expressivity was indistinguishable from the degree of dominance. The purpose that the new concepts served was to maintain a genetic etiology in the face of phenotypic plasticity induced by genotype-environment interactions. Variability in the phenotypic manifestation of a trait became a result of a gene’s expressivity and (indirectly) its penetrance. If the presence of a gene for a trait nevertheless failed to produce the trait, a genetic etiology for the trait was still maintained by simply positing that the gene had incomplete penetrance. If the presence of that gene led to the presence of the trait, but only to some variable degree, the gene was still responsible for the trait but had variable expressivity. The terms penetrance and expressivity were introduced into the English literature by Waddington in Introduction to Modern Genetics, where they were incorrectly attributed to Timoféeff (Waddington 1938). Waddington’s book, along with Timoféeff ’s growing prominence within Western genetics, made the terms common currency by the 1950s. Phenotypic plasticity —an almost inevitable outcome if development is the result of a suite of different factors rather than only of the genotype—was relegated to irrelevance by mystifying the concept of the gene. Waddington’s role in this story is curious. Though trained primarily as an embryologist, Waddington came to recognize the significance of the new genetics very early. In 1924, Spemann and Mangold had discovered the ‘‘organizer,’’ a region of the early embryo (gastrula stage) that seemed to direct subsequent development (Spemann and Mangold 1924). This led to an active research agenda by many embryologists to identify the ‘‘active principle’’ of the organizer. Committed reductionists believed this to be a chemical; Spemann himself had more holist leanings. Waddington was among those to demonstrate that dead ‘‘organizers’’ could induce cell differentiation. By 1938 he had come to view organizers as ‘‘evocators’’ of development: ‘‘[t]he factor which, in the development of vertebrates, decides which of the alternative modes of development shall be followed is the organiser, or, more specifically, the active chemical substance of the organiser which has been called

82

From Genes as Determinants to DNA as Resource

the evocator’’ (Waddington 1939: 37).Waddington argued that changes of this sort were discrete that is, there were definite developmental pathways with no intermediates between them. Because genes were also discrete, Waddington argued that ‘‘genes . . . act in a way formally like . . . evocators, in that they control the choice of alternative’’ (Waddington 1939: 37, emphasis added; elaborated in Waddington 1940). For Waddington, the aristopedia class of alleles (aristopedia, aristopediaSpencer, and aristopedia-Bridges) at the spineless locus (of the third chromosome) of Drosophila melanogaster provided an apposite example. The presence of the first two alleles from this class (aristopedia and aristopediaSpencer) lead to the transformation of the arista into a tarsus. In the case of the third (aristopedia-Bridges), the change is less marked but even in this case there is no true intermediate. Rather a smaller number of segments are altered thus showing that a discrete change has taken place. Waddington’s invocation of the language of control would be of critical significance after the advent of molecular biology (see ‘‘The Age of the Master Molecule’’). What is critical here is that his work marks the first serious attempt to synthesize genetics and development and it presumes, without argument, the primacy of the gene. Following through on this assessment of the importance of genes, in the 1940s Waddington shifted the focus of his research from classical embryology to the genetic control of tissue differentiation in Drosophila (Waddington 1962: 14). If Morgan had merely argued for a divorce of genetics from development, Waddington, in effect, demanded the subjugation of the former to the latter. A quote, though from a later period, emphasizes this point: ‘‘we know that genes determine the specific nature of many chemical substances, cell types, and organ configurations; and we have every reason to believe that they ultimately control all of them’’ (Waddington 1962: 4). Given the long dominance of developmental genetics in developmental biology since the 1960s, Waddington’s choice of ‘‘control’’ hardly seems unusual today. But, in the embryology of the 1920s and 1930s (and earlier periods), reproduction was an important component of development: a full developmental cycle included reproduction. From a developmental perspective, the one from which Waddington emerged, it makes just as much, if not more, sense to explicate and emphasize the developmental determination of genetics through the control of reproduction, rather than stipulate the genetic control of development. Yet, Waddington made that fateful move with far-reaching consequences for the study of development in the twentieth century.

83

Sahotra Sarkar

The Age of the Master Molecule As noted in the introduction, the construction of the double helix model for dna and the informational model of biological specificity in 1953 radically altered the conceptual terrain of biology, at least at the organismic and lower levels of organization. Schrödinger had already speculated on the existence of a ‘‘hereditary code-script’’ in 1943; starting in 1954, another physicist, George Gamow, began an explicit program of deciphering the ‘‘genetic code’’ (Gamow 1954a, 1954b; Sarkar 1996a, 1996b). The hope was to discover substantive properties of the code from simple formal rules incorporating functional assumptions about the efficiency and fidelity of information storage and transmission. As the mathematician Solomon W. Golomb put it: ‘‘It will be interesting to see how much of the final solution [of the coding problem] will be proposed by the mathematicians before the experimentalists find it, and how much the experimenters will be ahead of the mathematicians’’ (Golomb 1962: 100). As is often the case, Biology was not kind to the mathematicians: the theoretical program of deciphering the code was an unmitigated failure. The code that was experimentally deciphered in the early 1960s had none of the elegance envisioned by the theorists. Nevertheless, the theoretical research program led to the idea of the genome as a computer program; this idea was undoubtedly encouraged by the context in which it occurred: this was the period that saw the beginning of large-scale digital computation (see Kay 2000). Two papers from 1961 with radically different agendas explicitly introduced the idea of the genome as a blueprint and a program to be interpreted during development. In their classic paper laying out the details of the operon model for gene regulation, François Jacob and Jacques Monod concluded: ‘‘The discovery of regulator and operator genes, and of repressive regulation of the activity of structural genes, reveals that the genome contains not only a series of blue-prints, but a co-ordinated program of protein synthesis and the means of controlling its execution’’ (Jacob and Monod 1961: 354). What is critical about this passage is that agency resides in the genome: it controls the execution of the instructions in it. The fact that these instructions were already being interpreted as information gave credence to the metaphor of a genomic program. The operon model solved the decade-old problem of enzymatic adaptation through gene regulation. Later it became the standard model of gene regulation for most prokaryotic genes, as discussed below.

84

From Genes as Determinants to DNA as Resource

A much more extended and careful discussion of programming and computation appeared in a paper, the major purpose of which was to delimit the domain of molecular biology, that is, prevent its intrusion into organismic biology. In ‘‘Cause and Effect in Biology’’ Ernst Mayr notoriously distinguished ‘‘proximate’’ causes investigated by molecular biology from ‘‘ultimate’’ causes that are only provided by evolutionary biology. Evolution is the programmer producing a code that plays itself out in an individual, allowing individual behavior to be purposive: ‘‘An individual who—to use the language of the computer—has been ‘programmed’ can act purposefully. . . . Natural selection does its best to favor the production of codes guaranteeing behavior that increases fitness. . . . The purposive action of an individual, insofar as it is based on the properties of its genetic code, therefore is no more nor less purposive than the actions of a computer that has been programmed to respond appropriately to various inputs’’ (Mayr 1961: 1503–1504). Once again, agency resides in the genome, but because of natural selection and, in contrast to Jacob and Monod’s interpretation of the operon, not because of physical or chemical mechanisms. The critical feature of the operon model was that the regulation of gene activity apparently occurred at the genetic level. This was an unexpected development: while the problem of gene regulation was recognized as being critical to understanding development since a pioneering paper by J. B. S. Haldane in 1932, it was generally believed that the mechanism of control would operate from the cellular level (Haldane 1932). (Developmental holists believed that the mechanism would operate from even higher levels; for instance, that of the tissue or organ.) In 1962, in New Patterns in Genetics and Development, Waddington seized on the operon model to argue that regulation at the genetic level provided an explanation of tissue differentiation (Waddington 1962: 20–23). Differentiation was thus a matter of switching genes on or off. Even more controversially, Waddington interpreted other spatial developmental phenomena—histogenesis, morphogenesis, pattern formation, etc.—as special cases of differentiation (Waddington 1962: 1–3). Thus begun the program of a developmental genetics, of explaining development from a genetic basis, which took over the study of development in the 1970s. A detailed history of developmental genetics is yet to be constructed. From the perspective of that subdomain, the crucial developments were the discoveries in the 1980s of the homeobox sequence and hox genes, which were supposed to control much of morphogenesis.4 hox and similar genes do have significant regulatory roles in many species. Nevertheless, the confidence of

85

Sahotra Sarkar

geneticists in ‘‘master control genes’’ for development went far beyond what the data justified. This confidence was reflected in the initiation of the Human Genome Project (hgp) (and, later, other sequencing projects) in the 1990s, in which blind sequencing of genomes was supposed to reveal the mechanisms by which biological processes operate at all levels of organization. The trouble is that, by the late 1980s, there was ample reason to believe that dna sequences alone would reveal little about biology even at the cellular level, let alone at higher levels. The informational model for dna sequences as functional genes worked well provided that two conditions were satisfied: (1) a sufficiency condition—the presence of a dna sequence in a cell is sufficient to infer a capacity to produce the encoded protein (provided that the dna sequence is not a regulatory one, which, too, can be identified by inspection); and (2) a uniqueness condition—a single dna sequence produces exactly the encoded protein. If these two conditions are satisfied, the genetic code can be used—as a look-up table—to predict the amino acid sequence of the encoded protein. For prokaryotes, these conditions are satisfied: all dna sequences, besides regulatory ones, code for proteins and do so uniquely. However, for eukaryotes, this picture begins to unravel (see Sarkar 1996a, 1996b). Besides the standard genetic code, mitochondrial dna and even nuclear dna in some taxa use variant codes. The extent of such variation is at present unknown. Coding and regulatory regions of dna are interspersed with long strands of dna with no identifiable function.5 These nonfunctional regions, when occurring within structural (or coding) genes, are transcribed into mrna only to be spliced out before translation at the ribosome. Such noncoding regions are called ‘‘introns’’; coding regions are ‘‘exons.’’ 6 mrna is also routinely edited through a variety of other mechanisms, bases are added and removed, sometimes in the hundreds. Perhaps the most surprising—and, in retrospect, the most important (see ‘‘After the Human Genome Project,’’ below) discovery was that of alternative splicing: the same mrna transcript can be spliced in a variety of ways, leading to a set of different proteins. There is no evidence to suppose that the control of alternative splicing can be brought under the aegis of any simple genetic model such as the operon. Any informational model of biological specificity can survive without significant modification of the concept of biological information. The few attempts to rescue that model deny any claim that genes are the sole purveyors of biological information. But if they are not, developmental genetics, by itself, has no prospect of providing an adequate model of development.

86

From Genes as Determinants to DNA as Resource

After the Human Genome Project If the hgp is judged by the explicit promises that its proponents made in the late 1980s and 1990s to secure public support (and funding), it has been an unmitigated failure, the most colossal misuse ever of scarce resources for biological research. In 1992 Walter Gilbert claimed: ‘‘I think there will be a change in our philosophical understanding of ourselves. . . . Three billion bases [of a human dna sequence] can be put on a single compact disc (cd), and one will be able to pull a cd out of one’s pocket and say, ‘Here’s a human being; it’s me!’’’ (Gilbert 1992: 96). Today that claim seems laughable. None of the promises of Gilbert’s radical genetic reductionism have been borne out. Proponents of the hgp promised enormous immediate medical benefits. There have been none. Gilbert routinely promised the birth of a new theoretical biology. Instead, the emphasis now is on informatics: the design of computational tools to store and retrieve sequence information efficiently and reliably, with little expectation that any great theoretical insight is forthcoming. Commenting on the complexity of sporulation choice by an organism no more complex than Bacillus subtilis, C. Stephens recently pointed out: ‘‘Despite the explosive rate at which sequence databases are growing, and the concomitant increase in computing power available for sifting through them, sequence gazing alone cannot predict with confidence the precise functions of the multitude of coding regions in even a simple genome! Experimental analysis of gene function is still critical, a thought that brings with it the realization that the era of genomic analysis represents a new beginning, not the beginning of the end, for experimental biology’’ (Stephens 1998: R47). In one sense, from a socially responsive point of view, to the extent that basic research should provide tangible immediate social benefits, this failure of the hgp is no doubt unfortunate. However, it is not unexpected: in the late 1980s and early 1990s, scientific skeptics of the hgp routinely pointed out that it would not deliver on its promises (e.g., Sarkar and Tauber 1991; Tauber and Sarkar 1992). More important, social skeptics worried about the use of dna sequences for discrimination in health care and employment as well as social stigmatization. The failure of the hgp to deliver on its explicit promises provides an argument against the rationale for such uses of dna and thus assuages some of these social worries provided that the failure is publicly recognized. It must even have been abundantly clear to the propo-

87

Sahotra Sarkar

nents of the hgp that their promises were unrealistic, leaving them vulnerable to charges of fraud in their presentations to public funding bodies. Yet, no biologist, including the initial scientific skeptics, should now denounce the scientific results of the hgp. At the very least, the hgp has killed the facile genetic reductionism of the heyday of developmental genetics. There is little reason any more to suspect that claims of straightforward and irrevocable genetic determination of complex human traits will ever again be credible. hgp may even spell the extinction of molecular genetics itself, first transforming it into genomics, and then replacing it by proteomics. The reason is the startling properties of the human genome sequence when compared with other sequences: 1. The most important surprise from the hgp was that there are probably only about 31,000 genes in the human genome compared to an estimate of 140,000 as late as 1994 (see Hahn and Wray 2002 and references therein). Among the eukaryote genomes that have been fully sequenced, the human estimate remains the highest, but not by much; other plant genomes are expected to contain many more genes than in the human sequence. It is already known that the mustard weed, Arabidopsis thaliana, has 26,000 genes. Morphological or behavioral complexity is not correlated with the number of genes that an organism has. This has been called the g-value paradox (Hahn and Wray 2002). 2. The number of genes is also not correlated with the size of the genome, as measured by the number of base pairs. D. melanogaster has 120 million base pairs but only 14,000 genes; the worm Caenorhabditis elegans has 97 million base pairs but 19,000 genes; Arabidopsis thaliana has only 125 million base pairs while humans have 2 900 million base pairs (Hahn and Wray 2002). 3. At least in humans, the distribution of genes on chromosomes is highly uneven. Most of the genes occur in highly clustered sites.7 Most genes that occur in such clusters are those that are expressed in many tissues—the socalled housekeeping genes (Lercher et al. 2002). However, the spatial distribution of cluster sites appears to be random across the chromosomes. (Cluster sites tend to be rich in C and G whereas gene-poor regions are rich in A and T ). In contrast, the genomes of arguably less complex organisms, including D. melanogaster, C. elegans, and A. thaliana do not have such pronounced clustering. 4. Only 2 percent of the human genome codes for proteins; 50 percent of the genome is composed of repeated units. Coding and other functional regions (including regulatory regions) are interspersed by large areas of ‘‘junk’’ dna of no known function. However, hox gene clusters are never so invaded.

88

From Genes as Determinants to DNA as Resource

5. Hundreds of genes appear to have been horizontally transferred from bacteria to humans and other vertebrates, though apparently not to other eukaryotes (International Human Genome Sequencing Consortium 2001). 6. Once attention shifts from the genome to the proteome, a strikingly different pattern emerges. The human proteome is far more complex than the proteomes of the other organisms for which the genomes have so far been sequenced. According to some estimates, about 59 percent of the human genes undergo alternative splicing, and there are at least 69,000 distinct protein sequences in the human proteome. In contrast, the proteome of C. elegans has at most 25,000 sequences (Hahn and Wray 2002).

At the very least, except in rare cases, the presence of a particular dna sequence allows very little to be inferred about what happens in the proteome, let alone at higher levels of organization. At most, that piece of dna is a potential resource for use during development. Dethroned dna must find its place among other developmental resources. Some of these other resources are transferred intergenerationally through the material continuity of reproduction (e.g., through the maternal cytoplasm in most ‘‘higher’’ animals). Others are acquired from the environment (e.g., by accretion by some marine animals). dna may be special in many ways; as will be argued in the next section, ‘‘Concluding Remarks,’’ there is a strong case to be made for disparity between dna and other molecular constituents of cells. Nevertheless, dna and ipso facto, the gene, can no longer be the locus of agency responsible for the structural and behavioral repertoire of living forms including their remarkable diversity.

Concluding Remarks In the proteomics age, the most important problem in the philosophy of biology is to conceptualize the functional role of dna within the cell so as to explain the organization and other properties of the genome. This essay will end with a preliminary attempt to do so by explicating one speculative model which makes some novel predictions, though these have yet to be fully operationally disambiguated from predictions of other models: the sequestered modular template (smt) model of the cell. This model begins with the observation that the cell is probably the first spatially delimited living structure to have evolved. As biochemists realized in the 1910s and 1920s, its functions are primarily carried out by proteins, mainly enzymes. There are two types of

89

Sahotra Sarkar

such functions: those that maintain structural and behavioral integrity, and those that encourage reproductive proliferation. Evolutionary biology puts an emphasis on the latter type of function. But the former are as, perhaps even more, important for at least two reasons: (1) without the maintenance of structural and behavioral integrity at least up to reproductive age, there is no question of reproduction; and (2) in many organisms, especially sexually reproducing organisms, cellular functions continue beyond reproductive age. Maintenance of integrity, as well as reproduction, requires the production of replacement parts. Enzymes wear out (in spite of being catalysts); transport-enabling molecular moieties on cell membranes get damaged, as do the membranes themselves; they need to be replaced. There are two obvious ways to carry out replacement part production: (i) directly, by growth and fission of the relevant type; and (ii) indirectly, using a template. Whether or not the second strategy originally arose and got fixed entirely by accident, it has at least two advantages:8 (i) suppose that cellular processes are based on a small repertoire of basic chemical mechanisms. Then growth and fission will be catholic: the conditions under which one molecular type gets produced will very likely lead to the production of many other molecular types. Indirect reproduction permits preferential control; and (ii) templates can be sequestered from environmental insults in a way that the active molecules cannot. The latter must necessarily interact with the environment to maintain cellular functions. For the cell, it makes sense to have templates and then to sequester them. There is thus a critical disparity between the templates and the product molecules: dna and genes are thus special compared with the other molecular constituents of the cell. It makes even more sense to make these templates as physically stable as possible. It is again probably entirely an accident that the first templates were structurally simple molecules: probably rna, the variation in which was entirely combinatorial (that is, in sequence). But template integrity was better protected by a switch to a more stable form: dna. (For instance, the base uracil (U ), is easily transformed to C by deamination; dna uses the more stable T instead of U.) Enclosing templates by a membrane helps protection: eukaryotes achieve it by producing nuclei (and also enclosing genes in mitochondria and plastids). After enclosure, further tinkering to increase template protection would be evolutionarily advantageous. Thus, it makes sense to cluster genes when possible: protecting clustered sites is easier than protecting widely dispersed sites. Clustering happens in humans, as noted in the section ‘‘After the Human Genome Project.’’ The puzzle is why it does not seem to occur, or occur as much, in the other genomes that

90

From Genes as Determinants to DNA as Resource

have so far been sequenced. A possible resolution is that these genomes are smaller in size, resulting in less scope for clustering. It also makes sense that genes used as templates for many functions, and those that are critical resources early in development, should receive the most protection. hox genes deserve and get such attention. From the perspective of the smt model, resources should similarly be preferentially deployed to protect such genes from mutation. Here, the smt model makes a prediction partly in variance with the received gene-based evolutionary model. That model would explain the evolutionary conservation of such genes by the deleterious selective effects of such mutations. The smt model claims that, in addition, repair mechanisms preferentially target such sites. Modularity enters this model at two levels: (i) modularity of the genes themselves; and (ii) modularity of functional subregions (exons) of genes. At the genetic level, modularity is achieved because, by and large, genes are nonoverlapping and, much more important, they are separated from each other by long strands of nonfunctional dna which helps prevent gene disruption during recombination, a presumably physical inevitability of chromosome duplication. There is obviously a trade-off between this benefit and that of clustering. At the subgenic level, the benefits of modularity were clearly articulated by Walter Gilbert in 1978 (Gilbert 1978). Recombination in introns allows the combinatorial production of new proteins that are still likely to be at least partly functional because component parts have not lost their structural integrity. Gilbert also argued that point mutations at intron-exon boundaries can potentially alter splicing patterns and generate radically different proteins. According to the smt model, this would be undesirable. The smt model is inherently conservative: it predicts that such mutations are rare. If modules are being carefully protected—and therefore evolutionarily conserved—it makes sense to use the same modules for a variety of purposes. Alternative splicing makes sense. Two predictions of the smt model are that (1) there is an inverse correlation between genome-wide mutation rates over evolutionary time scales and the degree of alternative splicing in taxa, and (2) this is a result of mechanisms at the cellular level. A low number of genes and a high level of alternative splicing imply that organisms so constructed rely on the use of a large segment of the available combinatorial space at the level of modules within proteins. That organisms of this sort appear to be more structurally and behaviorally complex than others suggests a strong correlation between modularity at the proteomic level and evolvability. These arguments show the fallacy of any attempt at reading an organism off from its dna sequence alone.

91

Sahotra Sarkar

There are many other arguments for the smt model, which views dna as a resource. For instance, the same gene is often ‘‘co-opted’’ for different functions during the course of evolution. There is, however, one central unresolved issue: the model, as sketched above, assumes that the cell is the locus of agency, that is, the level at which it is appropriate to model and tabulate benefits, costs, and accidents. But is what is good, bad, or neutral for the cell also the same at higher levels of organization in multicellular organisms? Cancer trivially shows that it is not always so. Leo Buss and many others have noted the possibility of conflicts of interest between different levels of biological organization.9 It is far from clear that all the arguments given above will carry over to higher levels of organization than the cell. Finally, there is no reason to suppose that agency resides at exactly one level of organization. If the smt model is to be successful, it must be able to cope with the possibility of distributed agency.

Acknowledgments This work was supported by the U.S. National Science Foundation, grant no. ses-0090036, 2002–2003. For comments on an earlier version, thanks are due to Kelly McConnell.

Notes 1 A decade later, Dobzhansky (1937: 11) would echo the same sentiment, defining evolution to be a ‘‘change in the genetic [allelic] composition of populations.’’ 2 This started with Watson and Crick (1953a, 1953b); see Sarkar 1996a, 1996b for details. 3 Emphasis in the original. Thiéffry and Sarkar (1998) provide a critical history of the central dogma. 4 For a balanced analysis, noting both the importance of, and the interpretive excesses about, these discoveries, see McGinnis 1994. 5 The existence of such dna initially came as a relief because it resolved the C-value paradox, that genome size was not correlated with organismic complexity; see Cavalier-Smith 1985. 6 This terminology was introduced by Gilbert (1978). 7 These have often been likened to ‘‘urban centers’’ while the gene-poor regions have been likened to ‘‘deserts.’’

92

From Genes as Determinants to DNA as Resource 8 Haldane (1942) was the first to note the value of template-based reproduction and presciently suggested it as a model for gene duplication. 9 Buss (1987), in particular, has argued for the relevance of such conflicts for the evolution of patterns of development, especially germ-line sequestration. See also Falk and Sarkar 1992.

References Adams, M. B. 1980. Sergei Chetverikov, the Kol’tsov Insitute, and the evolutionary synthesis. In: E. Mayr and W. B. Provine (eds.), The Evolutionary Synthesis: Perspectives on the Unification of Biology (pp. 242–278). Cambridge: Harvard University Press. Avery, O. T., MacLeod, C. M., and McCarty, M. 1944. Studies on the chemical nature of the substance inducing transformation of pneumococcal types: induction of transformation by a deoxyribonucleic acid fraction isolated from Pneumococcus III. J. Exp. Med. 79: 137–157. Bateson, W. 1914. Address of the President of the British Association for the Advancement of Science. Science 40: 287–302. Buss, L. 1987. The Evolution of Individuality. Princeton: Princeton University Press. Cavalier-Smith, T. E. 1984. The Evolution of Genome Size. Chichester: Wiley. Crick, F. H. C. 1958. On protein synthesis. Symp. Soc. Exp. Biol. 12: 138–163. Dobzhansky, T. 1937. Genetics and the Origin of Species. New York: Columbia University Press. Falk, R., and Sarkar, S. 1992. Harmony from discord. Biology and Philosophy 7: 463–472. Gamow, G. 1954a. Possible mathematical relation between deoxyribonucleic acid and proteins. Biologiske Meddelser udviket af Det Kongelige Danske Videnskabernes Selskab 22(3): 1–11. Gamow, G. 1954b. Possible relation between deoxyribonucleic acid and protein structures. Nature 173: 318. Gilbert, W. 1978. Why genes in pieces? Nature 271:501. Gilbert, W. 1992. A vision of the Grail. In: D. J. Kevles and L. Hood (eds.), The Code of Codes (pp. 83–97). Cambridge: Harvard University Press. Golomb, S. W. 1962. Efficient coding for the desoxyribonucleic acid channel. Proc. Symp. Appl. Math. 14: 87–100. Hahn, M. W., and Wray, G. A. 2002. The G-value paradox. Evol. Dev. 4: 73–75. Haldane, J. B. S. 1932. The time of action of genes, and its bearing on some evolutionary problems. Amer. Nat. 66: 5–24. Haldane, J. B. S. 1942. New Paths in Genetics. New York: Random House.

93

Sahotra Sarkar

International Human Genome Sequencing Consortium. 2001. Initial sequencing and analysis of the human genome. Nature 409: 860–921. Jacob, F., and Monod, J. 1961. Genetic regulatory mechanisms in the synthesis of proteins. J. Mol. Biol. 3: 318–356. Johannsen, W. 1911. The genotype conception of heredity. Amer. Nat. 45: 129–159. Kay, L. E. 2000. Who Wrote the Book of Life? A History of the Genetic Code. Stanford: Stanford University Press. Laubichler, M., and Sarkar, S. 2002. Flies, genes, and brains: Oskar Vogt, Nikolai Timoféeff-Ressovsky, and the origin of the concepts of penetrance and expressivity. In: L. S. Parker and R. Ankeny (eds.), Medical Genetics: Conceptual Foundations and Classic Questions (pp. 63–68). Dordrecht: Kluwer. Lercher, M. J., Urrutia, A. O., and Hurst, L. D. 2002. Clustering of housekeeping genes provides a unified model of gene order in the human genome. Nat. Genet. 31: 180–183. Mayr, E. 1961. Cause and effect in biology. Science 134: 1501–1506. McGinnis, W. 1994. A century of homeosis, a decade of homeoboxes. Genetics 137: 607–611. Morgan, T. H. 1910. Sex limited inheritance in Drosophila. Science 32: 120–122. Morgan, T. H. 1926. The Theory of the Gene. New Haven: Yale University Press. Muller, H. J. 1927. Artificial transmutation of the gene. Science 66: 84–87. Muller, H. J. 1962. Studies in Genetics. Bloomington: Indiana University Press. Nilsson-Ehle, H. 1914. Vilka erfarenheter hava hittills vunnits rörande möjligheten av växters acklimatisering? Kgl. Landtbruks-Akad. Handl. Tidskr. 53: 537– 572. Romashoff, D. D. 1925. Über die Variabilität in der Manifestierung eines erblichen Merkmales (Abdomen abnormalis) bei Drosophila funebris F. J. Psychol. Neurol. 31: 323–325. Sarkar, S. 1996a. Biological information: a skeptical look at some central dogmas of molecular biology. In: S. Sarkar (ed.), The Philosophy and History of Molecular Biology: New Perspectives (pp. 187–231). Dordrecht: Kluwer. Sarkar, S. 1996b. Decoding ‘‘coding’’—information and dna. BioScience 46: 857– 864. Sarkar, S. 1998. Genetics and Reductionism. New York: Cambridge University Press. Sarkar, S. 1999. From the Reaktionsnorm to the adaptive norm: the norm of reaction, 1909–1960. Biol. Philos. 14: 235–252. Sarkar, S., and Tauber, A. I. 1991. Fallacious claims for hgp. Nature 353: 691. Schrödinger, E. 1944. What Is Life? The Physical Aspect of the Living Cell. Cambridge: Cambridge University Press. Silverstein, A. M. 1989. A History of Immunology. San Diego: Academic Press. Spemann, H., and Mangold, H. 1924. Über Induktion von Embryonalanlagen

94

From Genes as Determinants to DNA as Resource

durch Implantation artfremder Organisatoren. Wilhelm Roux’ Arch. Entwicksmech. Organis. 100: 599–638. Stephens, C. 1998. Bacterial sporulation: a question of commitment? Curr. Biol. 8: R45–R48. Tauber, A. I., and Sarkar, S. 1992. The Human Genome Project: has blind reductionism gone too far? Perspect. Biol. Med. 35(2): 220–235. Thiéffry, D., and Sarkar, S. 1998. Forty years under the central dogma. Trends Biochem. Sci. 32: 312–316. Timoféeff-Ressovsky, N. W. 1925. Über den Einfluss des Genotypus auf das phänotypische Auftreten eines einzelnes Gens. J. Psychol. Neurol. 31: 305–310. Timoféeff-Ressovsky, H. A., and Timoféeff-Ressovsky, N. W. 1926. Über das phänotypische Manifestieren des Genotyps. II. Über idio-somatische Variationsgruppen bei Drosophila funebris. Wilhelm Roux’ Arch. Entwicklungsmech. Organis. 108: 146–170. Waddington, C. H. 1938. An Introduction to Modern Genetics. London: George Allen and Unwin. Waddington, C. H. 1939. Genes as evocators in development. Growth 1: S37–S44. Waddington, C. H. 1940. Organisers and Genes. Cambridge: Cambridge University Press. Waddington, C. H. 1962. New Patterns in Genetics and Development. New York: Columbia University Press. Watson, J. D., and Crick, F. H. C. 1953a. Molecular structure of nucleic acids—a structure for deoxyribose nucleic acid. Nature 171: 737–738. Watson, J. D., and Crick, F. H. C. 1953b. Genetical implications of the structure of deoxyribonucleic acid. Nature 171: 964–967. Woltereck, R. (1909). Weitere experimentelle Untersuchungen über Artveränderung, speziell über das Wesen quantitativer Artunterschiede bei Daphnien. Verhandlungen der deutschen zoologischen Gesellschaft, 19: 110–173.

95

4 THE ORIGIN OF SPECIES A Structuralist Approach We have got on to slippery ice where there is no friction so in a certain sense the conditions are ideal, but also, just because of that, we are unable to walk. We want to walk: so we need friction. Back to the rough ground! —Wittgenstein, Philosophical Investigations

gerry webster and brian c. goodwin This essay is a critique of the system of concepts in terms of which biological organization is discussed in contemporary biology. The confusion we have experienced as to the nature and adequacy of the current view of organisms, arising both from reading the literature and from discussions with our colleagues, finally led us to the conclusion that the only way of achieving clarification was to abandon the system of concepts which we call the evolutionary paradigm and attempt to construct what seems to us a more satisfactory conceptual structure. While we are fully aware of the limited and preliminary nature of the work presented here, we hope that it may provoke a more general debate on these basic biological issues, which are so rarely discussed from a critical point of view. Our discussion is conducted entirely in terms of the problem of biological form—how to account for the production and reproduction of a diversity of specific morphologies which can be classified hierarchically—since this is an area with which we have some familiarity and, more important, is one which is distinguished by the absence of an adequate theory of one of its major phenomena, morphogenesis. The issue could equally well be discussed in other terms—for example, the problem of behavior—and it is hoped that those familiar with this area will be induced to attempt this. A requirement for any experimental or theoretical work in a science is an adequate conceptualization of the object or domain in which a phenomenon occurs. Such a conceptualization is not ‘‘given’’ by nature, nor is it an a priori

Gerry Webster and Brian C. Goodwin

‘‘projection’’ onto nature; rather it is progressively constructed as a result of theoretical and experimental work on a specific phenomenon. Nevertheless, this construction is constrained not only by the perceived phenomena but also by the currently accepted ‘‘tools for thought’’; modes of description and theoretical explanation are conditioned by the ontological and epistemological theories which are available or fashionable, and any conceptual system represents a historical choice from a number of possibilities. The complex of ontological and epistemological assumptions and of theoretical and descriptive concepts exerts constraints on the sort of problems which can be formulated and the way they are formulated, on the types of questions which are put to nature, on the nature of the theoretical explanations advanced, and on the kind of facts which are deemed significant. In Western thought, there have been two major conceptions of the structure of reality, whether physical, biological, social, or mental (see Collier 1979): on the one hand, a conception positing atomic events together with the ‘‘mechanical’’ interaction of autonomous units possessing certain intrinsic properties; on the other, a conception of a whole in which the parts have no autonomy or intrinsic properties, their nature following from that of the whole as the expression of some ‘‘central’’ unifying or directing agency, usually conceived as being nonmaterial, an ‘‘Idea’’ or ‘‘soul.’’ Both of these concepts, empiricist atomism and idealist holism, have been employed, often inconsistently it is true, in relation to organisms. In this essay we argue that contemporary biology relies on an unhappy marriage between atomism and a materialistic (and often mystical) holism in which a predominantly atomistic and functionalist conception of the organism per se is coupled with a holistic conception of a ‘‘central directing agency’’ conceived as a material entity—the so-called genetic program—which is supposed to determine, order, and unify the atomic units and events. The organism as a real entity, existing in its own right, has virtually no place in contemporary biological theory. We argue that the current conceptual system is derived from Darwinian evolutionary theory with its emphasis on historical and functional concepts, together with Weismann’s holistic theory of inheritance and development, and the atomism of classical Mendelian genetics. This ‘‘evolutionary paradigm’’ is expounded and defended in the two most influential discussions of biological organization produced in recent years, those of Monod (1972) and Jacob (1974); indeed Jacob’s account of the historical development of concepts of biological organization can be seen, in the last analysis, as a Whig history whose purpose is to justify the current concepts as the well-nigh in-

100

The Origin of Species

evitable and necessary outcome of a logical historical development (see Webster 1974). In this essay we question the adequacy of the evolutionary paradigm in relation to its failure to provide any satisfactory theory of the production and reproduction of biological form. We do not believe that this failure is a result of the supposed difficulty of the problem, but rather that it is a consequence of the intrinsic inadequacy of the current system of concepts; we contend that, without a change in the system, no progress can be expected in this crucial area. It follows inevitably from the absence of any coherent account of morphology or morphogenesis that the current theory of evolutionary transformations remains, at best, incomplete and unsatisfactory. . . .

Rational Morphology The biology of the late eighteenth and early nineteenth centuries made use of some important concepts which, while not necessarily denied in the twentieth-century ‘‘evolutionary paradigm,’’ tend to be ignored. It also introduced some concepts which, in a modified form, are central to the ‘‘evolutionary paradigm.’’ As pointed out above, we are not concerned to write history but to use it; we have therefore not hesitated to simplify the rich and complex preevolutionary paradigm and to reinterpret some of its key ideas in twentieth-century terms so that we can learn from it. In spite of the considerable complexity of thought and the vigorous controversy between competing ‘‘schools,’’ there seems to have been widespread agreement on certain fundamental points. The most important of these was that the individual organism and the biological domain as a whole were to be considered as systematic wholes or structures, that is, in terms of sets of internal relations. From this perspective the problem of biological organization, and therefore of form, was the primary problem and questions of material composition were secondary; as Cuvier pointed out, there is a constant flux of matter through organisms whose form remains constant, and the diversity of forms is greater than the diversity of materials out of which they are composed. The question of the external relations of organisms was not ignored, but the relevant concept, that of utility or functional adaptation to the environment, took its place as simply one of the several concepts in terms of which biological organization could be discussed. The work of the pre-Darwinian morphologists could be characterized as a classificatory preparation for a rational morphology; the method involved

101

Gerry Webster and Brian C. Goodwin

the search for empirical regularities in the diversity of adult and embryonic forms. These regularities appeared as invariant structural relations or ‘‘typical forms.’’ Hans Driesch (1914) succinctly summarized these researchers’ goal in comparison with their Darwinian successors: ‘‘they sought to construct what was typical in the varieties of form into a system which should not be merely historically determined but which should be intelligible from a higher and more rational standpoint.’’ The ‘‘system’’ they sought was, in the language of twentieth-century structuralism, a system of transformations, and we shall argue that such a system is potentially intelligible in terms of the ‘‘laws’’ or ‘‘rules’’ responsible for its generation. This approach to organismic form is exemplified in a developed state in the work of, for example, Reichert (see De Beer 1971) on the relations of homology between the reptile jaw and the mammalian middle ear, where an invariant relational structure can be discovered even though there have been such dramatic functional transformations that it is well hidden. Another example is the work of Geoffroy and, later, Owen on the rather more obvious homologies between the various forms of the vertebrate limb. In all these cases there are empirical regularities which point to the existence of a system of transformations. Similar empirical regularities can be discovered by comparing the parts of an individual organism—so-called serial homology (see De Beer 1971). In organisms which have metameric segmentation, such as annelids and arthropods, the segmental appendages possess an invariant structural pattern despite the considerable morphological diversity associated with their different functions (e.g., mouth parts versus walking or swimming limbs); these appendages therefore constitute a system of transformations. When they are made to regenerate, it is not uncommon for such organs ‘‘spontaneously’’ to undergo real transformations such that, for example, a walking limb is produced instead of a mouth part; this is known as heteromorphosis (see Needham 1952). In Drosophila, the so-called homeotic mutations can be seen as similar transformations during the course of embryonic development (see Ouweneel 1976). . . . The existence at an empirical level of invariant structural relations demanded explanation, and it is here that the most vigorous controversy developed. Initially, the principal controversy was over the primacy of ‘‘functional’’ or ‘‘structural’’ concepts in accounting for the constancy of relations. Cuvier argued that the observed structural relations were a consequence of more fundamental internal ‘‘functional’’ relations which were concerned

102

The Origin of Species

with the necessary ‘‘adaptation of the parts’’ to each other so that the whole could function as a unified and integrated entity, the existence and specific nature of each part providing the ‘‘conditions of existence’’ of all the others. The parts, which for Cuvier were functional components or units (‘‘organs’’), thus act upon each other—both functionally and in the sense of being mutually constitutive—and act together for a common end. Cuvier’s ‘‘Principle of Correlation’’ subsumes these concepts, and his conception of the organism is thus thoroughly Kantian. Kant had argued that what specifically distinguished organisms from mechanical devices like clocks was that they were self-reproducing and, therefore, self-organizing wholes. This means that, whereas in a mechanical device the ‘‘parts’’ exist only for each other—in the sense of being conditions of each other’s function in relation to a common end, so that a machine is a functional whole or unity—in an organism the ‘‘parts’’ not only exist for each other but also by means of each other in the sense of somehow producing each other, so that an organism is a functional and a structural unity. For Kant, therefore, an organism is that in which everything is both means and end (see Körner 1955). This was also Cuvier’s view, and it is in sharp contrast to the view held within the ‘‘evolutionary paradigm’’ where organisms are functional unities but structural dualities. Geoffroy, on the other hand, argued for a pure morphology uncontaminated by any functional considerations. The invariance of structural relations was a consequence of all organisms being constructed on the basis of one ‘‘plan’’ or pattern upon which all functional variation was, as it were, secondarily superimposed. . . . None of the rational morphologists considered the organismic domain as a whole or the individual organism to be irreducibly complex or diverse, for they believed that there were ‘‘rational’’ and general principles, either functional or structural, in terms of which all organismic forms—insofar as they were ‘‘typical’’—could be discussed; their initial goal was the formulation of general, albeit empirical, structural statements or ‘‘laws of form.’’ On the basis of these, it would be possible to employ ‘‘deductive’’ procedures to at least some extent: for example, to attempt the reconstruction of an entire extinct organism from a fossilized part. For all, the possible variety of organismic forms or transformations was restricted as a result of internal constraints: in the case of Cuvier by the necessity of ‘‘functional harmony,’’ in the case of Geoffroy by the necessity of conformity to the structural ‘‘plan.’’ Provided these requirements were satisfied, it was possible for an individual form to

103

Gerry Webster and Brian C. Goodwin

exist even if it was not useful, that is, did not satisfy the requirement of external functional adaptation, though the continued existence of a species was dependent on this requirement being satisfied also. . . .

Development and the Concept of a ‘‘Central Directing Agency’’ During the early nineteenth century the Enlightenment ideal was challenged and eventually displaced by a different one, the ideal of a historical or developmental science (Cassirer 1950) concerned with ‘‘becoming,’’ and therefore with questions of ‘‘origin,’’ ‘‘genesis,’’ and ‘‘causes.’’ This ideal had its effect on biology, concerning which von Baer could say in 1828 that the history of development is the true source of light for the investigation of organized bodies (Coleman 1971). The questions now came to be: How are organismic forms generated during both the history of the species and the history of the individual? What are the causes of the specific processes of development? The findings of the earlier morphologists were reinterpreted in the light of these new ideals and concerns, and in the course of this, the rationale behind the earlier approach to biological organization was obscured and eventually lost. The most famous of these reinterpretations is, of course, Darwin’s reification of the ‘‘type’’ into a concrete creature, the ‘‘common ancestor,’’ which plays a quasi-causal role in Darwin’s account of ‘‘origins.’’ But, as we point out in the next section, Darwin was rooted in a metaphysical tradition which is completely distinct from that of the predominantly Germanic tradition with which we are here concerned. For most of the biologists within this tradition who subscribed to the new ideals, ‘‘causes’’ in the biological domain were not to be considered as material and physical but as nonmaterial or ‘‘intellectual.’’ The reasons for these beliefs are extremely complex and are bound up with the whole development of German Romanticism and its various post-Kantian idealist philosophical systems, particularly those of Schelling (see Coleman 1971) and Hegel; all we can offer here is a tentative and schematic reconstruction. From the point of view of our immediate concerns with biological organization, some reasons for this view of causality are perhaps to be found in the failure of eighteenth-century biology to account for the phenomena of generation (including the development of the individual) in terms of a system of mechanical concepts, either Cartesian or atomistic. This failure was so complete that it led to the rejection of the idea of development as a real phenomenon and to the adop-

104

The Origin of Species

tion of various forms of preformationist interpretations of what appeared to be developmental processes (Gasking 1967). For the new philosophy of becoming, however, appearance and reality coincided; development was a real phenomenon, therefore the explanatory concepts of mechanics must be rejected as inadequate. The argument seems to be as follows: if development is a real process occurring in material objects and if the only concepts properly applicable to inert matter are those of mechanical causality, then matter, of itself, is incapable of development or self-organization. Two alternatives are then possible: either organized beings are the products of ‘‘chance’’ formation as in ‘‘Darwinism,’’ or some nonmaterial controlling agency or formative power must be at work (see Cannon 1961). There are some apparently ambiguous remarks in Kant’s Critique of Judgement which could be read as implying the latter view. Indeed, philosophical justification for this position as a whole could have been derived from a reading (or a misreading) of the difficult section on teleological judgment and biology in this critique, since Kant spells out in some detail the necessity of using teleological principles if biological organization is to be made intelligible. However, for Kant, teleological principles are not ‘‘constitutive’’ of the empirical world in the way that mechanistic causal principles are; they are only ‘‘regulative principles’’ or ‘‘maxims’’ for our reflections on this world. A teleological metaphysics, such as that employed by the nineteenth-century German biologists (and possibly by Driesch), was, for Kant, an impossibility (Cassirer 1950; Körner 1955). The notion of a teleological or ‘‘intellectual’’ cause can be developed by analogy with human intentional or ‘‘planned’’ (and therefore ‘‘organized’’) action; as Kant put it: ‘‘the idea of the effect [is made] into the condition of the possibility of the cause’’ (Körner 1955). The abstract, relational concept of the ‘‘type,’’ with its possible connotation of a ‘‘plan,’’ can be reified into a concrete ‘‘idea’’ distinct from its material embodiment and can serve as an ‘‘intellectual’’ cause conceived as either transcendent or immanent (see Driesch 1914). The ‘‘type’’ reified and considered as a transcendent cause, an ‘‘Archetype’’ or ‘‘Idea,’’ seems to be present in the thought of some of the followers of Geoffroy and in that of Agassiz (1857), biologists who were only peripherally associated with the new ideals. Of more interest from our point of view are the concepts of biologists concerned with the dynamic and apparently unified phenomena of growth and development where we find the reified ‘‘type’’ apparently conceived as an immanent Idea or ‘‘soul’’ which controls or guides organismic activity and development into ‘‘typical’’ forms. This view can be found in thinkers as different as von Baer and Claude Bernard (1865); as von Baer put it, ‘‘it is the essence (the Idea according to the new school) of the de-

105

Gerry Webster and Brian C. Goodwin

veloping animal form which controls the development of the germ’’ (quoted by Coleman 1971). This is idealist holism. This view results in a conception of the organism radically different from that held by Cuvier and Kant which we outlined above. It is worth spelling out the differences in some detail because they are relevant to our subsequent discussion. First, its conception of the organism as a whole is dualist in terms both of ‘‘composition’’—there are two substances, material and spiritual— and ‘‘activity’’—the spiritual substance is active or controlling, the material substance inert and controlled (with respect to its organization at any rate). It follows that the organism is no longer considered as a functional and a structural unity, as it was by Cuvier and Kant, since its unified nature is no longer conceived of in terms of the self-organizing activity of the whole as whole, but in terms of the organizing of unifying activity of one ‘‘part,’’ a ‘‘spiritual’’ or ‘‘intellectual’’ organizing center—a ‘‘central directing agency.’’ This means that, insofar as the ‘‘material part’’ of the organism can be considered as a unity, it is solely as a functional unity; that is, as the unity of a mechanical device in the sense outlined above, in which the ‘‘parts’’ exist only for each other and no longer by means of each other, for they are produced by the nonmaterial center. The organism is, therefore, the phenomenon of its Idea; it is no longer conceptualized as a ‘‘self-organizing totality’’ but as an ‘‘expressive totality’’—an expression of the nature and activity of the Idea. As far as we are aware, this is the first appearance in modern biology of the view that organisms possess ‘‘organizing centers,’’ and that their organization is to be understood in terms of them. It is a view that is prominent in the ‘‘evolutionary paradigm’’ of modern biology, thanks to Weismann, and is a product of the German holistic tradition. It is probably not coincidental that the second appearance of a broadly similar view regarding the organization of embryonic development—Spemann’s (1938) concept of the embryonic ‘‘organizer’’—also rose within the German tradition. We may also observe that a mechanical or atomistic conception of the structure of organisms goes hand-in-hand with a holistic conception of the ‘‘origin’’ of their specific organization. At least as far as these biological phenomena are concerned, atomism and holism are mutually constitutive concepts; they are two sides of the same coin. Second, within this new paradigm, the old program of the abstract reduction of the diversity of organismic morphology to laws of form, with its associated concept of transformation, becomes problematic. Whereas previously the diversity of forms of biological organization had been seen as conceivably intelligible in terms of a relatively small number of general structural state-

106

The Origin of Species

ments (‘‘laws of form’’), in this new view each specific form could be seen as the product of a single, ‘‘simple,’’ and therefore irreducible, Idea. Form and organization therefore become irreducible, and have to be taken as ‘‘given.’’ If the Idea is seen as static and species-specific, then any concept of gradual transformation becomes impossible, and this seems to be the case with transcendental idealists such as Agassiz. However, it is also possible to see the Idea as both universal and dynamic; capable of change or self-development in a quasi-Hegelian fashion, so that actual and gradual evolutionary transformations are possible during the course of history, different species being ‘‘momentary expressions’’ of the historical self-development of the Idea, and therefore reducible to it. The general structure of this conceptualization of the evolutionary process is retained by Weismann, though it is transformed in a number of important ways we shall outline below.

Darwinism We have suggested that the advent of German holist and idealist conceptions of the organism marked the beginning of the end of rational morphology. Its final demise was a consequence of the appearance and eventual acceptance of ‘‘Darwinism.’’ Although Darwin was influenced to a considerable extent by the new ideal of a historical science, he was rooted in a completely different metaphysical and conceptual tradition than that of the German idealists and holists. This tradition was that of Protestant natural theology as initiated in the seventeenth century by Boyle and Ray and developed and transmitted to the nineteenth century in the writings of Paley, Malthus, and the Bridgewater Treatises. Our thesis in this section is that Darwinian theory developed within, was conditioned by, and transformed this scientific tradition, which was characterized by its emphasis on the ‘‘content’’ rather than the ‘‘form’’ of the organismic domain, and, moreover, concentrated on one aspect of this ‘‘content’’ to the virtual exclusion of all others. Allied to this tradition was an epistemology, empiricist positivism in the tradition of Hume and the utilitarians, which was complementary to it. The choice of this particular system of concepts and epistemology gave rise to a biology in which the goal of rational morphology, insofar as it remained intelligible, was judged to be no part of science. We have Darwin’s own admission (Darwin 1889) of the extent to which he was influenced by the concepts of natural theology. Indeed, the similarity be-

107

Gerry Webster and Brian C. Goodwin

tween Darwin’s worldview and that of the natural theologians is so great that at least one historian of science (Cannon 1961) has characterized Darwin’s ‘‘world’’ as effectively a secular version of the world described in Christian thought; hence the antagonism between the two traditions. Although the German and British traditions are very different, they are not, especially when transformed, incompatible, and our contention is that the basic system of concepts concerning organismic form which is current in contemporary biology (the ‘‘evolutionary paradigm’’) can be seen as the descendant of a marriage between German romantic idealism and Protestant natural theology. The scientific tradition of natural theology is remote from the Enlightenment tradition of Cuvier and Kant, which, as Cassirer (1950) noted, was concerned above all to develop a clear system of guiding concepts concerning the organismic domain (of which the concept of the ‘‘typical’’ was one) so that we do not lose ourselves in the contemplation of its variegated and multiform character. Its aim was to see the (abstract) wood, not just the (concrete) trees, to discover the real unity or ‘‘form’’ hidden in the diversity of appearances which constitute the ‘‘content.’’ Natural theology, on the other hand, concerned as it was with ‘‘design,’’ was predominantly ‘‘contemplative’’ and preoccupied with appearances, with the individual ‘‘event,’’ the ‘‘content’’ of the natural world. In consequence, it was very strong on detailed empirical observation but relatively weak in systematic theoretical concepts. Both Cassirer (1950) and Russell (1916) have remarked on the conceptual impoverishment, vis-à-vis the problem of form, that occurred in nineteenth-century biology subsequent to the acceptance of Darwinian theory. Since within natural theology organisms are considered to have been constructed by some agency (God) they are—implicitly—conceptualized as mechanical devices; that is, as functional unities in which the only structural relations are those of spatial contiguity. This exclusively functional approach is further restricted by a tendency to concentrate on the external functional relations of organisms considered in terms of the utility of the ‘‘part’’ to the organism in relation to a particular mode of life in a particular environment. Internal functional relations, while by no means ignored, are relegated to secondary status. One may speculate that it was because of the easily observable and often dramatic nature of the adaptation or ‘‘fitness’’ of an organism that the external relation was especially singled out to demonstrate the providence of the creator and his capacity for perfect design: the contemplation of a caterpillar with the exact form of a leaf or a dead twig, or an orchid which resembles a bee, concentrates the mind wonderfully. This sort of ap-

108

The Origin of Species

proach naturally tends to emphasize the diversity of the biological domain: the specific peculiarities of organisms and the way that species differ from each other in relation to their particular environments and modes of life. It thus encourages an ‘‘atomistic’’ rather than a ‘‘systematic’’ conception of the totality of organismic forms, a conception which is at one with the conception of the individual organism as an ‘‘atomistic’’ or ‘‘mechanical’’ aggregate of parts, each with a primary functional relation to the external world. Thus at both the individual level and at the level of the ‘‘totality’’ there is a tendency to see the biological domain as irreducibly complex and ‘‘given.’’ The concern of natural theology with observable ‘‘content’’ together with its ‘‘atomistic’’ approach situates it in relation to the philosophical tradition of empiricist positivism deriving from Hume, which can also be seen in Darwin’s thought. The empiricists distinguished between an objective physical reality—consisting of observable objects and events and observable relations of temporal succession and spatial contiguity—and the subjective constructions placed on the perceptions of this reality. Causal relations, for example, are not observable and therefore are not real; they are subjective constructions placed on observable constant conjunctions of events. Similarly, for the utilitarians, society, for example, since it is not observable, cannot be objectively real; what we subjectively interpret as society is just the ‘‘sum’’ of observable individuals and their observable acts. For empiricist positivists, therefore, theories which posit the existence of objects, events, and systems of relations which are not ‘‘of the order of fact’’ must be regarded as subjective fictions having dubious scientific status. Examples of theories subject to such categorization would be those of Newton, Faraday, Maxwell, Marx, Freud, Lévi-Strauss, Saussure, and Chomsky, among others. For the nineteenthcentury ‘‘Darwinists’’ the rational morphologists’ theory of a ‘‘Plan of Creation’’ falls into this category. From this perspective, Darwin (1859) found it natural to reify Cuvier’s abstract concept of the ‘‘type’’ into a concrete creature—the ‘‘common ancestor.’’ Since the empiricist positivist admits the real existence only of that which is observable, his accounts of complex phenomena must necessarily be statistical models of the behavior of populations of observable individuals interacting in observable ways under specified observable conditions. Empiricist positivists who are less intransigent may grant scientific status to accounts of this type which involve entities and relations which are ‘‘in principle’’ or potentially observable given the appropriate technology; certain forms of statistical mechanics might be so categorized. At its worst, the antitheoretical stance of an empiricist-positivist philosophy encourages a science of the

109

Gerry Webster and Brian C. Goodwin

type so wittily parodied by Borges in his parable Of Exactitude in Science, in which the goal is simply to reduplicate in thought the appearance of the natural world in its entirety. The system of concepts regarding organismic form which had been developed within natural theology was taken over by Darwin (1859) with only small modifications and served to define the problems which were to be given naturalistic explanation in empiricist-positivist terms. While it is true that Darwin pays some attention to the structural aspects of form (‘‘Correlations’’) and concedes at various points in his argument that they are probably important, his discussion is not extensive, and structural concepts play no role in the final theory which is exclusively functionalist and in which the external relation of adaptation (now regarded as relatively rather than absolutely ‘‘perfect’’) is of paramount importance. The problem of form is thus effectively ‘‘reduced’’ to the problem of functional adaptation and its specificity. With the eventual acceptance of ‘‘Darwinism,’’ this conception of the problem became that of mainstream biology, and all other aspects tended to be ignored. . . . The insight which this theory, as it stands, provides with regard to the problem of form is clearly not great, since being essentially a theory of the differential growth of ‘‘given’’ components of a population, it is effectively ‘‘preformationist.’’ Selection, being an a posteriori principle, can account only for changes (or the absence of changes) in the composition of a population of ‘‘given’’ variant forms under specified conditions. Thus, provided the conditions could be specified, the theory could in principle account for, say, the transformation of a population in which variation was continuous into one in which it was discontinuous. This would be a case of the ‘‘origin of species,’’ for Darwin characterizes species as simply ‘‘well marked varieties.’’ The theory is therefore consistent with the fact that taxonomists often experience difficulty in constructing empirical taxonomies, that is, in deciding whether a given form is to be regarded as a distinct species or merely as a variety of a species. Assuming therefore (1) that a continual supply of appropriate variant forms is taken as given, (2) that observable variation is causally related to variation in fitness, and (3) that an appropriate mechanism of inheritance (nonblending) exists, then the theory is, in principle, capable of accounting for the aspect of the ‘‘content’’ of the organismic world which had been emphasized by natural theology, viz. the fact that species of organisms differ from each other in ways that can be accounted for in terms of adaptation to different modes of life in different environments. One says ‘‘in principle,’’ of course, because an account of the origin (in the sense employed by Darwin) of any

110

The Origin of Species

one particular adaptation would demand a knowledge of the specific, historically contingent, conditions under which the composition of a known, historical, population underwent the changes leading to the present state of affairs which is to be explained. Such knowledge is usually unobtainable, so that accounts are bound to be speculative and largely untestable. . . .

Weismann and the Evolutionary Paradigm We have seen that in Darwinian theory the historical concept of inheritance plays a central and crucial role in all explanation of form. It therefore became critical to develop a way of thinking about inheritance which was not only compatible with the central tenets of natural selection and environmental functionalism but would actually serve to reinforce the role of these factors (rather than undermining their centrality as was the case with Darwin’s own theory of pangenesis). This task was undertaken by Weismann (1883, 1885, 1904; McSherry 1975), who reformulated the problem. The solution he provided not only established the conceptual structure of mainstream thinking on this problem but also reinforced the ‘‘Darwinian’’ tendency to isolate arbitrarily one instance of invariance—the temporal or historical invariance displayed within a species between parent and offspring—from all other instances. Once again we should stress that we are not concerned to write history but merely to use the historical source of a widely accepted system of concepts to clarify the nature of that system and of the ‘‘choice’’ that was made; consequently we have ignored the specific details that are relevant only to Weismann’s own thinking, which, in many respects, was a good deal more critical than that of the ‘‘Weismannists.’’ A rational reconstruction of Weismann’s argument takes the following form. The problem of inheritance is not primarily to be thought of in terms of how the structure of the parent is transmitted to the offspring (as in Darwin’s theory of pangenesis), but rather in terms of the control of growth and development; ‘‘the task is to trace heredity back to growth’’ (Weismann 1883). Implicit in this statement is the claim, with which we agree, that reproductive invariance is to be understood in terms of invariant developmental—that is, generative—processes. However, in developing his specific conceptualization of the developmental process, Weismann reintroduced into biological thinking a transformed idealist and holist conception of the organism which, we will argue, does not represent a progressive move but rather a step back-

111

Gerry Webster and Brian C. Goodwin

ward. Weismann argued that offspring resemble parents because both are the results of identical processes of growth and development, and constant effects imply constant causes; this is a classical empiricist conception of causality (see Bhaskar 1978). This cause, Weismann argued, is to be found in a distinct entity, the germ cell, which contains a specific substance with a highly complex structure, the germ plasm, which has the power of developing into a complex organism (Weismann 1885). The supposed constant nature of the cause, and hence of the effect—that is, the phenomenon of inheritance—is simply a reflection of the continuity of the substance of the germ cell or germ plasm, for ‘‘the germ cells are not derived at all, as far as their essential and characteristic substance is concerned, from the body of the individual; they are derived from the parent germ cell. This substance has remained in perpetual continuity from the first origin of life’’ (Weismann 1883), and during the course of time successive changes have occurred in it such that there have been successive changes in the forms of organisms which are its ‘‘effects’’; there has been evolutionary change. In a graphic metaphor, Weismann (1904) later likened the germ plasm to a stolon growing and ramifying underground and periodically throwing up leafy shoots which, being visible above ground, indicate its presence. In Weismann’s view, therefore, organisms do not really reproduce themselves; they are a kind of artifact produced by the germ plasm, which is an irreducible absolute given by history. This view of inheritance is completely consistent with Darwinian theory. We might note in passing that it is also a view of reproduction which resulted in confusion for decades about the process in multicellular forms which reproduce asexually and have no obvious germ cells. From the start, Weismann’s view of inheritance was based on a particular conception of the morphogenetic process; as the germ plasm theory was elaborated, so also was the theory of development. The acceptance of the cell theory and the consequent work in cytology in the latter half of the nineteenth century led to the conclusion that the continuity between organisms was essentially a consequence of nuclear and probably chromosomal continuity, so that by the end of the century it was widely believed that the nucleus contained the ‘‘material substratum’’ of heredity. In 1884, Strasburger posited the central controlling role of the nucleus in growth and development: ‘‘nuclei determine the specific direction in which an organism develops’’ (quoted by McSherry 1975). Weismann adopted an identical view and localized his germ plasm (now with slightly modified properties) in the nucleus, from which it actively directed the course of growth and development; as he put it, ‘‘the essence of heredity is the transmission of a nuclear substance of specific structure’’ (1885). The germ plasm was now supposed to

112

The Origin of Species

consist of active (Weismann’s term), particulate units (determinants), each of which stood in a specific causal relation to a particular part of the organism; there were to be as many of these specific causal units as there were independently variable and heritable parts of the organism, and the germ plasm exercised its control over development as a result of the determinants being activated or liberated in a definite and regular temporal and spatial order (McSherry 1975). Weismann also confronted the problem of the apparent determination of specific forms by the environment—he examined regional and seasonal polymorphism in butterflies (see McSherry 1975). In order to account for this phenomenon he postulated the existence of alternative sets of determinants, each set responsible for a particular form, a specific set being activated by an environmental variable, in this case temperature. We have already remarked that Weismann’s concept of the germ plasm can be seen as a transformation of the concept of an Idea or soul which was postulated by the German holists to account for form, and the similarity in the two conceptual systems is clear. Both regard the organism as essentially mechanical and incapable of self-reproduction and self-organization. Both regard the reproduction of specific organized form as due to the action of an immortal (or potentially so), historically given ‘‘agent’’ which can change its ‘‘nature’’ during the course of history and which is not really a part of the material and visible organism it produces. For the German holists, of course, the ‘‘agent’’ was a nonmaterial Idea or plan, hence they subscribed to a dualism of passive matter and active ‘‘spirit.’’ Weismann viewed the Idea as materialized or embodied in matter—and in this respect the view is monist—but the germ plasm could not really be regarded as a part of the organism per se; as Weismann (1883) emphasized: the germ plasm and the substance of the body, the somatoplasm, have always occupied different spheres. So once again we have a dualism of structure and function, and the material and visible organism is, effectively, regarded in holistic fashion as an ‘‘expressive totality’’— the expression of the activity of a ‘‘central directing agency’’ or material Idea. The essentially mechanical or atomistic conception of the organism, which was characteristic both of natural theology and the German holists, is retained and clarified in Weismann’s scheme. As far as form is concerned, the organism is thought of as an aggregate of parts each of which stands in functional relation to the external environment (and is therefore subject to natural selection) and in constitutive relation to a set of ‘‘determinants’’ in the germ plasm. With this transformed concept of a ‘‘central directing agency,’’ what we refer to as the ‘‘evolutionary paradigm’’ is essentially complete as regards the basic system of concepts.

113

Gerry Webster and Brian C. Goodwin

In using Kuhn’s (1970) term, we do not imply (as he sometimes seems to) that this ‘‘system’’ is closed and monolithic so that thought is constrained absolutely. What we mean is that there exists a system of concepts which serve to give a certain ‘‘form’’ to the ‘‘content’’ of experience and in terms of which both theoretical and experimental activity and the natural object of that activity, considered as a ‘‘thought object,’’ are constituted, but these activities themselves result in changes to the system. For anyone who is not an extreme empiricist, it is evident that the existence of such a system is a prerequisite for any thought or activity whatsoever. Our discussion of ‘‘Darwinism’’ and ‘‘Weismannism’’ is an attempt to demonstrate the constraining effects of one such system on thought about biological form. We would tentatively define the ‘‘evolutionary paradigm,’’ in relation to the problem of the form, in terms of the following three characteristics. 1. It is an arbitrary isolation of one instance of invariance, that displayed in inheritance, as paradigmatic, and a consequent belief that one set of primary concepts relevant to the problem of form are those of a historical nature. This is combined with a predominantly functionalist conception of the organism, so that concepts concerning the relations of organisms to the external environment provide a second explanatory set. Genetic and environmental concepts exhaust the explanatory repertoire of the paradigm vis-à-vis form. 2. It is a basically mechanical or atomistic conception of the structure of the organism in which the only structural relations are those of spatial contiguity; the organism is thought of, figuratively speaking, as a ‘‘Humean Bundle’’ rather than a ‘‘Kantian Structure’’ (Collier 1979). This means that when considering the reproduction of this structure, the organism cannot be considered as a self-organizing system in the fullest sense (i.e., Kant’s) in which the ‘‘parts’’ are reciprocally constitutive (so that some structural relations are constitutive relations) but must be regarded holistically as some kind of artifact or ‘‘expressive totality.’’ This leads to two possible views of morphogenesis which emphasize, respectively, the atomistic and the holistic aspects: on the one hand a conception of preformed parts which spontaneously assemble on the basis of their intrinsic properties which are specified by the ‘‘central directing agency’’ (atomism); on the other a conception of the ‘‘central directing agency’’ as organizing the process of formation in accordance with some preexisting and species-specific Idea, Plan, or Program whose execution is regulated by nonspecific variables in the internal environment (a form of holism). 3. It exhibits a bias toward an empiricist-positivist epistemology which is consistent with all of the above (although the concept of a ‘‘central directing

114

The Origin of Species

agency’’ is in some respects anomalous). This bias results in a preoccupation with the observable particularities and concrete ‘‘content’’ of the organismic world and a belief that explanation consists of redescription in terms of the observable (or ‘‘in principle’’ observable) behavior of microscopic entities— ‘‘reductionism’’ in the popular sense of the term. There is distrust of abstract theory which postulates the existence of unobservable entities and relations or which attempts ‘‘conceptual’’ reduction in terms of abstract and general principles. Work subsequent to Weismann can be regarded as largely a matter of giving empirical flesh to this conceptual skeleton. Of course, this activity has resulted in some changes in the substantive content of the system, but the basic structure has remained intact. The most significant changes in the content of the ‘‘evolutionary paradigm’’ came about as a consequence of the rediscovery of atomistic, analytical Mendelian genetics; the development of the chromosome theory of the gene; and the identification of Weismann’s germ plasm with Mendel’s hereditary factors. These changes resulted on the one hand in neo-Darwinian theory, and on the other in the development of molecular biology. Weismann’s distinction between germ plasm and somatoplasm and their relations to the environment, the former ‘‘active’’ and able to reproduce itself, the latter relatively inert and possessing no reproductive powers, is identical with the distinction and form of dependence between genotype and phenotype which is central to neo-Darwinian theory. A typical statement of this position is that of Dunn: ‘‘We may start from the fact that all organisms, considered as individuals or as populations distributed in space or in time, represent the outcome of interactions between factors intrinsic to the organism and factors external to it. The usual way of stating this is to say that any individual is a joint production of his heredity and his environment. The intrinsic factors consist of an assemblage of genes derived from the parents’’ (1963). This view continues to be taken by the majority of contemporary population biologists, as far as we are aware. From this perspective, the organism is regarded as the ‘‘expression’’ of, and hence reducible to, a collection of Mendelian genes which, in an organism developing in a given environment, determine the characters which constitute the phenotype. When environmental influences are taken into consideration, this genotype is considered to determine the reaction norm (Dobzhansky 1951) as in Weismann. While the latter’s original postulate of a one-to-one correspondence between a gene or genes and each independently variable and heritable ‘‘part’’ of the organism seems to be no longer accepted by most geneticists [Huxley (1963), although Dobzhansky (1951) notes that it continues to influence the thinking of many geneticists],

115

Gerry Webster and Brian C. Goodwin

we have the impression that among nongeneticists the expression ‘‘the gene for X’’ is often used in a Weismannist sense. On the other hand, a statement like that of Dunn makes sense only if it is believed that there is a one-to-one correspondence between the genotype as a whole and the phenotype. It is within molecular biology or in partial reaction to it (which amounts to the same thing) that the ‘‘evolutionary paradigm’’ has been most fully and explicitly articulated, and it is here that the holist conception of the organism as an ‘‘expressive totality’’ and the related concepts of the ‘‘central directing agency,’’ atomism, functionalism, and environmentalism are crucial constitutive elements. We take Jacob (1974) and Monod (1972) to be typical molecular biologists and Wolpert (1969, 1971) as the most notable and influential representative of views developed in partial reaction to molecular biology. From one point of view molecular biology can be seen as nothing more than the theoretical and empirical fleshing out of Weismann’s conceptual skeleton. This is apparent if, for example, we consider the empirical questions which naturally arise from Weismann’s concept of the germ plasm: What is its chemical nature? How is its historical continuity assured? How does it ‘‘express’’ itself ? These are precisely the questions Crick (1964) addresses in an account of the molecular biology of the fifties and early sixties. Within molecular biology, Weismann’s fundamental distinction appears in terms of the self-replicating properties of the dna, on the one hand, and, on the other, the somatic, phenotypic role of the proteins which are specified by the dna. The modern version of Weismann’s concept of a ‘‘central directing agency’’ or ‘‘genetic program’’ is expressed by Jacob (1974) as follows: ‘‘The whole plan of growth, the whole series of operations to be carried out, the order and the site of syntheses and their coordination, are all written down in the nucleic acid message.’’ While it is true that this statement, like many similar ones in Jacob’s book, is subsequently qualified, the qualification in our view deprives it of any clear and unequivocal meaning. In many respects, Monod (1972) provides the clearest and least equivocal contemporary expression of the ‘‘evolutionary paradigm,’’ and it is remarkable how close he is conceptually to the late nineteenth century and Weismann, the empirical detail and cybernetic and semiological terminology notwithstanding. Like Weismann, he isolates historical or reproductive invariance as the most fundamental biological invariance, which is to be explained, via cell theory, ‘‘in terms of the genetic text—the sole repository of the organism’s hereditary structure [Monod’s emphasis]—and the ne varietur reproduction of the text, written in the form of dna nucleotide sequences. And to the extent that all the structures and performances of organisms re-

116

The Origin of Species

sult from the structures and activities of the proteins composing them, one must regard the total organism as the ultimate epigenetic expression of the genetic message itself.’’ Monod’s atomistic conception of epigenesis and morphogenesis which arises from this position we consider below. Although Wolpert’s (1969, 1971) views can be regarded as having developed in partial reaction to molecular biology, they remain firmly within the evolutionary paradigm and rely entirely on the concepts of a ‘‘central directing agency’’ and cellular atomism together with a form of internal environmentalism. The concept of organism as an ‘‘expressive totality’’ is clearly stated. For example, ‘‘the expression of genetic information in terms of pattern and form is a central problem’’ (1971), as is the role of a ‘‘central directing agency’’: ‘‘genetic information does not contain a description of the adult, but a set of instructions on how to make it’’ (1971). Wolpert’s views will be considered in our critique of the ‘‘evolutionary paradigm,’’ to which we now turn.

A Critique of the ‘‘Evolutionary Paradigm’’ At this point it should be evident that a critique of the ‘‘evolutionary paradigm’’ cannot be a simple matter of dramatic refutations or falsifications. What we are dealing with is not simply a collection of atomic theoretical propositions (though of course such propositions can be extracted) which stand or fall on the basis of atomic facts, but a structured system of concepts —a paradigm—which constitutes its object—the organism—in a particular way and in which a very large amount of empirical data ‘‘finds its place’’ and makes sense. The widespread acceptance of the paradigm is not without reason. The undoubted (if partial) power of the system has given it almost the status of a system of natural interpretations (Feyerabend 1975). An important part of any critique, therefore, is to show that the system is not natural or inevitable, but the result of a number of specific choices: conceptual, ontological, epistemological, and phenomenological. This we have attempted to do in a preliminary fashion in the historical sections. As we have emphasized, the point of our historiography is not to understand the past but to liberate ourselves from the present. Although we have suggested that the evolutionary paradigm has remained basically unchanged since Weismann, it is possible to point to certain results arising from its progressive articulation which appear to be inconsistent with propositions which can be extracted from the paradigm. Whether such inconsistencies or anomalies are conceived as contradictions will obviously de-

117

Gerry Webster and Brian C. Goodwin

pend on whether they are regarded with the eye of faith or that of skepticism. The faithful will always be able to find explanations in terms of the paradigm; for example some, at least, of the anomalies we consider below can be explained in terms of additional hidden variables and other ad hoc assumptions. From our skeptical perspective, many such explanations appear to be instances of explaining away since there is often no independent evidence for the explanation and a petitio principii is therefore involved; a substantial part of this section is concerned with a discussion of these anomalies. We argue, first, that the concept of a ‘‘central directing agency’’ is ill-founded; second, that morphological invariance cannot be explained in terms of invariance in this ‘‘center’’; and third, that theories of morphogenesis based on the concept are problematic. In essence these arguments rest on the demonstration that, in a given environment, the relation between ‘‘atomic composition’’ (either genetic or chemical) and form is not that of simple determination or one-to-one correspondence. Further, we suggest that the articulation of the ‘‘evolutionary paradigm’’ (especially within molecular biology) has resulted in the development of a conceptual scheme whose full significance has not been realized and which, when generalized, provides the foundation for an alternative conceptualization of biological organization and morphogenesis in terms which are related to those outlined by Driesch (1914, 1929) and Waddington (1957). These conclusions lead naturally to the final part of the essay, which is also the final part of the critique, in which we outline this alternative conceptualization which constitutes the organism in structuralist terms. . . . We turn now to the empirical evidence relating to the molecular and cellular assembly processes which have been invoked as the means whereby specific genetic information is translated into specific organismic structure. Monod (1972) is very explicit about the causal logic involved, and we now assess the validity of his arguments. For Monod, organisms are chemical machines whose structure is generated by a process strictly comparable to molecular crystallization. He describes a hierarchy of these crystallization processes as follows: 1. Folding of the polypeptide sequences culminating in globular structures provided with stereospecific binding properties. 2. Associative interactions between proteins (or between proteins and certain other constituents) so as to build cellular organelles. 3. Interaction between cells, so as to constitute tissues and organs.

118

The Origin of Species

Monod makes it quite clear that genetic determination requires that the structure generated at each level of the hierarchy be uniquely specified by the properties of the constituents at the lower level, so that any form is causally reducible to the primary protein structure of its constituents and hence to the historically given information in the nucleic acid sequence of the dna. He is, of course, aware that environmental conditions must be specified in order to realize a particular structure or form, but he restricts the definition of these constraints to purely thermodynamic variables such as temperature, aqueous phase, and ionic composition. Let us see if these arguments are universally valid for biological processes, as Monod contends. Consider the example of bacterial flagellum formation as described by Oosawa, Kasai, Hataro, and Asakara (1966). These investigators showed that a homogenous solution of the protomer, flagellin, obtained by the disaggregation of flagella from wild-type Salmonella, could be induced to reaggregate into either wavy (wild-type) or curly (mutant) flagella by adding to the solution wavy or curly seeds, respectively. These seeds, or nucleation centers, were fragments obtained by sonication of flagella of the two different types. Thus the same protein can, under identical thermodynamic conditions, make different forms according to the way in which they are assembled under the direction of different crystal seeds, so that there is not a unique relationship between composition and form at this level of organization. At a higher level of organization, another example is provided by Sonneborn’s (1970) studies of the inheritance of altered form in the unicellular ciliate protozoan, Paramecium. Rotation of a ciliary row on the cell surface of an individual resulted in the appearance of a reversed row in all its progeny (observed up to the eight-hundredth generation). No change in the genetic information accompanied this heritable change of form, nor was there any change in the external environment, so that it is clear that more than one cell structure can be associated with a particular genotype. Sonneborn’s (1970) interpretation of the observation is that the rotated ciliary row acts as a local nucleation center for the assembly of gene products during preparation for cell division and hence the altered form is transmitted from generation to generation by a mechanism like that involved in flagellin assembly into flagella. Again it is clear that the specification of form requires more than the specification of the genotype, thus falsifying Monod’s genetic determinism. Finally let us consider evidence relevant to Monod’s third level of the hierarchy, interactions between cells to produce tissues and organs. Observations of the type made by Zwilling (1964) provide an instructive example. He took

119

Gerry Webster and Brian C. Goodwin

limb-bud mesenchyme from stage 19–20 chick embryos, disaggregated the tissue into cells, pelleted them by centrifugation, and replaced them under limb ectoderm. No limbs of typical form were produced, although limblike structures of diverse morphology were frequently formed. This shows that mesenchyme cells can form structures different from a normal limb, suggesting that cellular interactions alone are insufficient to specify a unique morphology. Another example of this is the different forms of the fore and the hind limbs in tetrapods, which consist of the same cell types differently arranged in space to give distinctive morphologies. The basic limitation of Monod’s reasoning about morphogenesis, which he characteristically calls molecular ontogenesis, is his assumption that the units or ‘‘atoms’’ which constitute a structure contain in themselves all the necessary information for specifying a unique form. This view tends to ignore a basic lesson that has emerged from molecular biology itself: nucleotide polymers or polypeptides can be produced by spontaneous polymerization in reaction mixtures consisting of nucleotides or amino acids only. Thus molecular interactions governed by purely kinetic (mechanical) processes are sufficient to generate nucleic acids and proteins, but the products of such reactions are random mixtures comprising all the diverse members of the set of linear polymers which are combinatorial transformations of one another. Because no specific order is imposed on the reaction, the product has no specific structure. The concept of atomistic association is perfectly adequate to account for such phenomena. The nucleic acids and proteins found in living organisms, however, have specific forms; that is, they comprise a particular selection from the total set of possible forms (i.e., primary structures), and a major triumph of molecular biology was the discovery of the mechanism whereby these specific structures are generated. Basic to this is the existence of one-dimensional polymeric templates which, together with a specific generative process (i.e., covalent bond catalysis), specify the primary structure of the products; that is, the template selects a particular sequence from the set of possible sequences which could be generated by polymerization. The demonstration that such mechanisms exist in organisms means that atomistic models of the generation of molecular form have been invalidated by molecular biology, for atomism does not allow for the emergence of spatial order by any process other than that arising from the interaction of the atoms themselves (the ‘‘atoms’’ in this case being nucleotides or amino acids). To postulate that organismic morphology arises by the self-assembly of protein monomers is another expression of atomism, the atoms being in this case

120

The Origin of Species

proteins. The existence of spatial-organizing nucleation centers in cells invalidates this as a general hypothesis, as we have seen. It would seem, therefore, that the attempt to articulate in molecular terms Weismann’s concept of ‘‘determinants’’ in the germ plasm causally responsible for the generation of organismic form has not been successful. However, many models have been advanced which propose various additions to this scheme in an attempt to provide sufficient information to give an adequate morphogenetic theory. These all involve postulates about spatial information in the organism in the form of segregated cytoplasms, gradients, of ‘‘morphogens,’’ or unspecified determinants which, together with the ‘‘genetic program,’’ are intended to provide a complete specification of the morphogenetic process. One such theory which has been particularly influential is that developed by Wolpert (1969, 1971). In his theory of positional information, Wolpert assumes the absolute minimum required for spatial order in a developing domain, namely a co-ordinate system which assigns a unique positional label to every point within it. The boundaries of the domain are defined by extremal values (maxima or minima) of the coordinate system, generated by a mechanism which is not specified but is assumed to be based on chemical reactions producing ‘‘morphogens’’ whose diffusion through the domain produce spatial continuity of state in the form of gradients. A cell located at any point within such a monotonic gradient is then characterized by a unique positional value specified by one, two, or three morphogens according to the dimensionality of the domain. This provides positional information which the cell’s genome then ‘‘interprets’’ in a manner specific to the species and to the cell’s developmental history, the interpretational process being described by the switching on or off of particular genes. Wolpert realized that by locating all the specificity of positional response in the genome, the coordinate system or positional information could be universal: that is, every distinct domain of a developing organism could be specified by the same type of coordinate system, and furthermore this universality could be extended over all organisms. Universality thus became a basic postulate of the theory, and was used by Wolpert to explain a variety of phenomena observed in developing organisms. A further elaboration of the theory of Wolpert and Lewis (1975) contains the following statement: ‘‘A theory of development would effectively enable one to compute the adult organism from the genetic information in the egg.’’ The theory is thus thoroughly Weismannist in conception, being based on a sharp distinction between an ‘‘active’’ decision-making genome (‘‘germ

121

Gerry Webster and Brian C. Goodwin

plasm’’) and a ‘‘passive,’’ nonspecific coordinate system (‘‘somatoplasm’’). Wolpert is explicit that there is no relationship between the pattern of positional information and the specific morphology which is generated by the genomes within it. His solution of the problem of morphogenesis in the individual organism, conceived as the spatially patterned determination of cell states, is thus formally identical with Weismann’s solution of the problem of temporal and regional polymorphism and employs the same conceptual scheme. Weismann argued that a temporally or spatially organized variable in the external environment selected which particular set of determinants in the germ plasm would be expressed, and hence which of the possible alternative forms would be produced. Wolpert argues that a spatially organized variable in the internal environment (‘‘positional information’’) determines which particular set of genes in each member of a (necessarily) spatially distributed set of genomes will be expressed and hence, ultimately, which component part of the total form will be realized in that part of space. Wolpert’s scheme is completely Weismannist in its reliance on the historically given ‘‘central directing agency’’ as the determinant of form, and morphological diversity is inevitably, therefore, seen as irreducible. There can be no general ‘‘laws of form’’ in such a theory since the only universal, positional information imposes no constraint on the form which is generated other than that it be spatially extended. The question arises whether this form of Weismannism is consistent with the empirical evidence. We now argue that it is not. First, the theory is unable to account for morphogenetic processes in unicellular, or more accurately, acellular organisms such as the ciliate Protozoa and algae (e.g., Acetabularia) because it involves the necessary assumption that developing systems consist of many cells, form emerging from the spatially specific behavior of genomes within these cells. But not only can complex form be generated or regenerated in many unicellular species (Morgan 1901; Frankel 1974), the regeneration of apical structures in Acetabularia can occur in the absence of a nucleus so that complex and large-scale morphogenesis occurs independently of the presence of a genome (Hämmerling 1963); nuclear products only are required. Wolpert has tended to discount such examples as special cases which do not belong within the realm of developmental processes proper, which he associates only with multicellular organisms. So let us now turn to these. All embryos start as single cells, the egg, and in many species this single cell has, after fertilization and before cleavage, the same type of regulative behavior as the unicellular ciliate Protozoa. For example, the amphibian egg

122

The Origin of Species

prior to first cleavage can be induced, by rotation, to change its axes of symmetry so that the plane of bilateral symmetry of the future embryo occurs in a different position from the normal (Pasteels 1964). Such embryos can develop completely normally, so the single cell behaves as a regulative unit which we shall later refer to, in accordance with classical usage, as a field. It is also possible, by altering the time of such rotations, to induce the formation of a second axis within the egg, so that twins are produced. Double axes can also be induced by surgical means at later stages of development, when there are many cells (Spemann 1938). These manipulations are analogous to the production of doublets in ciliate Protozoa by surgical manipulation (Tartar 1961). It would appear, therefore, that it is irrelevant whether or not an organism is multicellular as far as its behavior as a regulative developing entity is concerned. We return to this point in a different context later. . . . Let us abstract from the lessons we have learned from the above critique of genetic theories. In the biological domain, more than one form is possible for a given composition (genetic, molecular, or cellular), always assuming a fixed external environment. Therefore, in order to account for the existence of a particular form, we need to specify both the means of generating the set of possible forms and a means of selecting a specific member of this set for actualization at all levels of the biological hierarchy. Although exclusively atomistic conceptions of the organisms are untenable, this does not mean that atomistic association plays no role in biological phenomena; it takes its place in a larger scheme as a limiting case in which the set of potential forms has only one member. This larger scheme is a structuralist conception of the organism. The second conclusion relates to the role of the nuclear genome. As molecular biology has shown, this specifies the primary structure of proteins and some temporal sequences in which these are made; in other words, it defines the potential ‘‘atomic’’ composition of the organism in terms of molecular materials and molecular events. However, as we have argued above, ‘‘atomic’’ composition does not determine form. The genome, we suggest, is no more the ‘‘directing center’’ of organismic structure than a lexicon or a dictionary is the directing center of a sentence or a text. The concept of a ‘‘central directing agency’’ or ‘‘genetic program’’ and of the organism as an ‘‘expressive totality’’ is the last vestige of a mystifying holism which should have died as a consequence of the discoveries of molecular biology but, paradoxically, appears to live on most vigorously in the molecular biologist’s conception of how organismic form is generated. But perhaps it is only fitting that those whose work allows us to give a decent burial to a concept that has dominated biology for so long should render it a last and elaborate homage.

123

Gerry Webster and Brian C. Goodwin

Organisms are ‘‘wholes’’ or ‘‘totalities’’ but they are not ‘‘expressive totalities’’; they do not have ‘‘centers’’ or ‘‘souls,’’ whether material or otherwise. They are ‘‘self-organizing totalities’’ or, to use a fashionable phrase, ‘‘decentered structures.’’

Driesch and the Foundations of Structuralist Biology We have suggested above that a fundamental lesson to be learned from studies of molecular and cellular form is that there must exist both a means of generating a specific set of potential structures and a means of selecting (actualizing) specific members of this set. We now argue that this is also true of higher levels of biological organization. It will be remembered that the work of the nineteenth-century rational morphologists emphasized two important features of biological organization that were obscured by the development of Darwinian theory. First, organisms are structural wholes in which the parts are to be understood in terms of their relation to each other and to their place in the overall structure; second, the various adult forms comprise empirically recognizable systems of transformations. These conclusions, based on a static anatomical study of organisms, were elaborated in a dynamic developmental context by Driesch (1914, 1929). His analysis of morphogenesis, subsequently confirmed in its essential features by many investigators, provides, we believe, the means of decisively breaking with both the atomistic and holistic traditions and establishing the conceptual framework for a structuralist biology. Driesch’s achievements in this direction came about primarily as a result of his rejection of empiricist and historical explanations (‘‘Darwinism’’) and his recognition that the organism must be conceived as a reality in its own right and hence as the proper object of biological study. From this came his vitalism and his entry into biological folklore as that stock character, the ‘‘woolly-minded vitalist,’’ whose work can be safely ignored. However, as Driesch (1914) himself pointed out, vitalism, at least in the first instance, is nothing more than the recognition of the specific and real nature of the organism as the object of study in biology and its difference from other objects. It is a form of methodological antireductionism, which in the present instance consists of a refusal to assimilate organisms ab initio to, for example, chemical or mechanical systems, and an insistence on their specific reality. As Mepham (1973) has noted, the demand for specificity is of quite general

124

The Origin of Species

epistemological significance in relation to the development of the science of a given domain, and our position in this essay is, like that of Driesch, antireductionist in this sense. In his analysis of morphogenesis, Driesch, like his nineteenth-century predecessors, emphasized structural wholeness and elaborated it in terms of the following organismic properties. 1. Equipotentiality—the fact that the potentiality for the development of a particular part of the form exists in a larger part of the organism than that in which it is normally actualized. 2. Equifinality—the fact that a developing organism can reach the same endpoint from apparently different starting points and by apparently different trajectories. 3. Self-regulation—a whole of typical or specific form is conserved or restored, following perturbation, by compensatory reorganization.

These characteristics of living systems are also the formal characteristics which Piaget (1971) used to define structure; we consider this further below. Equipotentiality and self-regulation are concepts under which can be subsumed the phenomena of regeneration and of generation—that is, reproduction—since both involve the transformation (under conditions of isolation, either physical or physiological) of a part of the organism (and not necessarily a particular part as in the Weismannist conception) into a new whole. Moreover, they are concepts which are applicable to both unicellular (Frankel 1974) and multicellular organisms, to sexual and asexual forms. The concepts, therefore, subsume universal and fundamental features of living organisms. The experimental demonstration of these properties involves the isolation and transplantation of parts of organisms, that is, changing their relations with other parts. The conclusions are often summarized in the aphorism ‘‘developmental fate is a function of position,’’ but this tends to obscure the fact that what is important is not position per se, but relations within the whole. Thus a part of the organism is constituted as a specific part in consequence of its relations with other parts of the organism. This does not mean that the part possesses no intrinsic properties of its own; for example, in interspecific transplantation experiments, the part develops in a manner appropriate to its relations but into the structure characteristic of the species from which it came. Thus, embryonic frog ectoderm transplanted to the head region of an embryonic salamander develops into frog head structures (Spemann and Schotté 1932). Similar phenomena can be observed in certain sorts of intraspecific transplantations carried out at relatively late stages of devel-

125

Gerry Webster and Brian C. Goodwin

opment. For example, in transplantations between organs, organ specificity is retained; thus, in the chick, if embryonic thigh mesoderm is transplanted to the wing tip it induces scales and claws in the overlying ectoderm and itself develops into foot structures (Saunders, Cairns and Gasseling 1957). . . .

Organisms as Structures Structuralism is concerned with order, its generation and transformation. It rejects both atomism and holism. Following Piaget (1971), we may characterize it in terms of three key concepts: wholeness, transformation, and selfregulation. Structures are wholes, first, in the sense that they have the property of maintaining themselves in being while their elements change; hence they are not reducible to the sum of their elements. Second, the structure ‘‘controls’’ its elements in the sense of giving them specific properties by virtue of the relations they are assigned in the structure. Thus the structure cannot be understood atomistically, that is, in terms of the intrinsic properties of the elements considered as individuals. However, the structure is not ‘‘all-powerful’’; it does not constitute the elements, which also have intrinsic properties and obey their own laws in addition to the laws of the structure; holism is thus rejected (Collier 1979). Structures are internally constrained, that is, they are law-governed. The laws of the structure are essentially laws of composition (in the dynamic sense of making) or laws of form; that is, they are generative laws and what they generate is the specific relational structure of the system. We have seen that biological structures appear to be wholes in precisely the structuralist sense. A given structure is a member of a set or system of transformations. Because structures obey laws there is a restriction on the ‘‘coherent’’ forms which are possible; that is, the potential set of transformations is a logically closed set, though not necessarily finite. We have seen that species of organisms and their parts appear to comprise systems of transformations and that individual organisms can undergo specific transformations as a consequence of ‘‘internal’’ or ‘‘external’’ perturbation. We suggest that these empirical transformations should be understood as transformations in the structuralist sense, that is, as representative members of a set of potential forms generated by specific laws. In other words, the diversity of forms should not be regarded as irreducible. In this connection, Piaget (1971) has suggested that the development or genesis of structures is always to be viewed as a transfor-

126

The Origin of Species

mation of existing and ‘‘simple’’ structures into other ‘‘more complex’’ structures. The existing structures are taken as ‘‘given’’ in the sense of being the starting point of the analysis, but they are not ‘‘primitive’’ in any absolute sense, being merely the result of previous transformations. The final characteristic property of structures is that of self-regulation, which refers to the power of the structure to maintain a given member of the set of transformations in the face of perturbation. We have noted that biological structures can resist perturbation due to changes of composition, either of genetic origin or as a result of mechanical removal of parts. The fact that structures are conserved is a consequence of the existence of laws. It is appropriate in discussing biological forms to consider them, as Waddington (1957) did, in terms of Bateson’s concept of ‘‘positions of organic stability.’’ The general aim of structuralist theory is to make the order of a unified system intelligible. It aims to express a formal system in which ‘‘the actual is explained or interpreted as an instance of the possible’’ (Piaget 1971). Theoretical explanations, therefore, involve on the one hand formulating a set of laws or rules with generative power which can account for the range of possible forms belonging to a particular set (i.e., the transformations), and on the other formulating what principle of selection is used to actualize particular instances of the possible. If, as we have argued, organisms are to be conceptualized as structures, then any theory of biological organization or form must be a generative theory of transformations. Such a conceptualization is, of course, compatible with a variety of theories and does not necessitate any one particular theory. At this point it is necessary for our argument only that we consider the general type of theoretical explanation that such a conceptualization of biological form might allow, and this we will attempt by considering as a ‘‘pedagogic analogy’’ the explanation of the forms of motion provided by Newtonian mechanics. While this theory cannot be regarded as a structuralist theory in the full sense of the term, it will serve our purpose insofar as it emphasizes the relational structure of systems of moving bodies and attempts to account for the actual in terms of the possible. Prior to the systematic theoretical work of Galileo and Newton, the existence of some empirical correlations between different initial conditions (‘‘causes’’) and different forms of motion (‘‘effects’’) was well known. This recalls the situation in biology, which we have discussed above, of the empirical correlations studied by classical genetics. Newton provided a theory which accounted for these correlations and allowed the generation of all the forms of motion which are possible for a body moving under a central attractive

127

Gerry Webster and Brian C. Goodwin

force. These comprise a closed set of infinite size, the members of which fall into distinct ‘‘morphological’’ classes: circle, ellipse, paraboloa, hyperbola, and straight line (i.e., the conic sections). Selection from this set of possibilities (i.e., actualization of a particular member) is accomplished by specifying the initial conditions, and any change in the initial conditions will result in a transformation of the form to another member of the potential set. The set defined by Newtonian theory consists, therefore, of forms which are transformations of one another, and in this sense the set comprises a system of transformations. A further characteristic of Newtonian theory which is relevant to our argument is that it ‘‘integrates’’ the universal and the necessary together with the particular (or historical) and the contingent. Thus the ‘‘hidden’’ relations described by the equations of motion are invariant and pertain to all the observable forms of motion; the diversity within this unity is explained in terms of contingent variations in relevant parameters (‘‘causes’’) such as initial velocity. The universal and the particular can, of course, be regarded as relative concepts since it is conceivable that other universes can exist— nevertheless, the distinction is theoretically important for it is the ‘‘hidden’’ relational structure which determines the repertoire of possibilities for the system (its ‘‘competence’’) and, therefore, the ‘‘effect’’ of a given ‘‘cause.’’ We have noted above that in organisms, the same genetic ‘‘cause’’ can have different ‘‘effects,’’ and conversely, that the same morphological ‘‘effect’’ can be produced by different ‘‘causes.’’ Although Newtonian mechanics is obviously much simpler than any generative theory of biological form could possibly be, we believe that it provides an instructive illustration of the explanatory power of such theories, which, if nothing else, at least allows us to keep in focus the type of theory which is required in biology. One of the major consequences of a conceptualization of organisms as structures or self-organizing totalities is, obviously enough, the reinstatement of the organism as the proper object of biological research, as a real object, existing in its own right and to be explained in its own terms. This, of course, was the concept of the organism held by the rational morphologists but which during the subsequent decades was progressively eroded; first, by cell theory with its attempt to conceptualize the organism in atomistic terms as a population of virtually autonomous individuals, and, second, by Weismann’s concept of a ‘‘central directing agency’’ in which the organism, conceived as an ‘‘expressive totality,’’ becomes a kind of epiphenomenon of the genome. This progressive ‘‘disappearance of the organism’’ and its replacement by microscopic entities, to which are attributed some or all of the properties of the organism, reaches its most absurd and degenerate stage in the

128

The Origin of Species

concept of the ‘‘selfish gene,’’ in which the entities are freely invented and endowed with whatever properties are required to ‘‘explain’’ biological phenomena (Dawkins 1976). . . . With a monistic conceptualization of organisms as self-organizing and law-governed structures, the genome (i.e., effectively the chemical composition) takes its place as a part of the totality of the constraints on the generative process along with others such as preexisting organization and environmental factors. A ‘‘random’’ change in any of these factors will result not in a ‘‘random’’ change of structure but in an orderly change to another possibility, another member of the system of transformations, and typical form will be conserved. It is striking that both genetic and mechanical perturbations, if they are not lethal, generally result in new forms which are typical, orderly, and structurally harmonious even if functionally disharmonious (i.e., nonadaptive); one thinks, for example, of the homeotic mutants and the various heteromorphic transformation referred to above. Such phenomena are unintelligible within a conceptualization of the organism as an ‘‘expressive totality’’ generated by a ‘‘central directing agency’’ which is subject to ‘‘random’’ changes. In a structuralist conceptualization, there is a reasonably precise sense in which one can speak of genetic (or any other) ‘‘information’’ since chemical composition can act to select (actualize) one of the possible transformations. However, no explanation of form can be provided solely in terms of selection (i.e., information); it is necessary also to account for how the particular form is possible, and this requires knowledge of the generative processes which ‘‘make’’ the forms. . . . It is clear that, in general terms, the picture of the morphogenetic process presented above is similar to that which Waddington (1957) arrived at on the basis of his analysis of genetics and experimental embryology. Waddington pictured the orderly change in ‘‘competence’’—the set of possible transformations available at any given time—and the capacity for self-regulation in terms of a spatial metaphor, the epigenetic landscape. What is now required is to transform this metaphor into a specific theory.

History and Structure It is therefore far from being the case that the search for intelligibility comes to an end in history as though this were its terminus. Rather it is history that serves as the point of departure in any quest for intelligibility . . . history

129

Gerry Webster and Brian C. Goodwin

leads to everything, but on condition that it be left behind. (Lévi-Strauss 1966)

It will be apparent that we regard the theory of evolution, and in particular neo-Darwinism, as having extremely limited explanatory power with respect to the problem of form to which it was originally addressed. This limitation arises as a consequence of the absence of any adequate theory of the means of production of ‘‘typical forms’’ and is such, we would maintain, as to render debatable the claim that neo-Darwinism is the unifying theory in biology (Maynard Smith 1975). These remarks are not to be construed as a rejection either of Darwinist functionalist explanations in biology or of a historical theory of descent, the claim that existing species are ‘‘blood relations’’ which have arisen by the evolutionary transformations of a smaller number of preexisting species. The structuralist position we adopt is perfectly consistent with a theory of historical descent (though not dependent on it) and, by providing a conceptual framework within which a theory of transformations can be produced, both clarifies some of the issues raised by such a theory and provides a picture of the evolutionary process which is somewhat different to that of the neo-Darwinists. . . . A structuralist conception of living organisms, with its emphasis on the logical, the universal, and the necessary, implies that the organismic domain as a whole has a ‘‘form’’ and is, therefore, intelligible (which does not mean predictable) and that the ‘‘content’’—the diversity of living forms, or at least their essential features—can be accounted for in terms of a relatively small number of generative rules or laws. This claim will no doubt strike many as grandiose, and to neo-Darwinists it will seem absurd. For, from the perspective of neo-Darwinism, with its empiricist concern with the particular and the contingent and its reliance on the historically given, there can be no ‘‘Plan of Creation,’’ and the domain of the living is, in the last analysis, fundamentally irrational and therefore unintelligible. Yet, as we have argued, this inadequate conceptualization is neither inevitable nor necessary; it is itself the result of a historical ‘‘selection’’ from the domain of possible conceptualizations. Unlike those of other organisms, our ‘‘selections’’ can be deliberate choices. Let us now choose to leave history behind.

130

The Origin of Species

The Origin of Species: A Structuralist Approach, Twenty Years Later The editors of the present volume have suggested that we write a short afterword to this essay, which was our first extended attempt to argue for a revival and development of the tradition of ‘‘rational morphology’’; that is, the attempt to study form scientifically as opposed to historically. The origins of this essay were diverse. On the one hand there were internal factors. These included, first, our dissatisfaction with attempts, subsequent to the revolution in molecular biology, to explain form in terms of a ‘‘genetic program’’; we found such attempts not only simplistic but also ignorant of the findings of classical genetics and experimental morphology and embryology. We also regarded the attempts of some scholars to attempt a rapprochement between molecular biology and the classical discoveries as unsatisfactory. Second, we were not sympathetic to claims that evolutionary theory was the central theory in biology and we were dismayed by the development of evolutionary sociobiology, especially as applied to the human world. We regarded the claims of the sociobiologists as intellectually facile and as morally and politically pernicious—indeed nihilistic. On the other hand there were external factors. During the 1970s the University of Sussex was an exciting place to be. Scholars from different disciplines met both formally and informally to exchange ideas, and it was by this means that we came into contact with the various forms of structuralism that were fashionable at the time. It is gratifying that some people consider our paper to be worth reprinting after twenty years. We consider that the major criticisms of genetic reductionism and historical explanation in biology that we presented then are as valid today as they were in 1982. However, it would be both surprising and depressing if we could now read our first effort with complete satisfaction. Although in broad terms we still subscribe to the position outlined in this essay, it now seems to have a number of defects. Perhaps the most significant of these concerns the ‘‘structuralism’’ we advocate. From the present perspective it might seem to be an uneasy and perhaps not even entirely coherent mixture of Piaget and Lévi-Strauss. But at that time the adaptation of their theoretical contributions proved helpful for setting free our own views. Furthermore, from our present perspective, the historical sections leave something to be desired, and our frequent references to Kant (though important to us at the time) now seem somewhat gratuitous and not particularly helpful to the reader.

131

Gerry Webster and Brian C. Goodwin

Since the publication of this paper our ideas have developed considerably (see Webster and Goodwin 1996). We now have a much better understanding of the conceptual structure of evolution theory, which has enabled us to considerably sharpen our critique. This is almost entirely due to our reading of the penetrating analyses of David Hull. We also have a much better understanding of structuralism in relation to rational morphology and, especially, rational systematics. This is a consequence of our improved understanding of the writings of Hans Driesch (arguably the most profound thinker about biology of the last century), an understanding gained via the writings of Ernst Cassirer and considerably assisted by the work of Ron Brady and Margaret Hagen. Moreover, this has resulted in an appreciation of the prescient views of Goethe, which, although sometimes assimilated to those of Darwin, are, in fact, radically different and a good deal closer to modern positions. Finally, the theoretical models of morphogenesis which have been developed (by bg) in the light of our conceptual arguments are considerably more sophisticated than those available twenty years ago.

Note This essay is reprinted (and abridged by Christoph Rehmann-Sutter) from G. Webster and B. C. Goodwin, ‘‘The origin of species: a structuralist approach,’’ Journal of Social and Biological Structures 5: 15–47. Copyright 1982, Academic Press, Inc. (London) Limited; with permission from Elsevier.

References Agassiz, L. 1857. Essay on Classification. Reprinted in: P. Appleman (ed.), Darwin. New York: W. W. Norton, 1970. Bernard, C. 1865. Introduction to the Study of Experimental Medicine. Reprint. New York: Dover, 1957. Bhaskar, R. 1978. A Realist Theory of Science. Brighton: Harvester. Cannon, W. 1961. The bases of Darwin’s achievement: a revaluation. Victorian Stud. 5: 109–134. Cassirer, E. 1950. The Problem of Knowledge. New Haven: Yale University Press. Coleman, W. 1971. Biology in the Nineteenth Century. New York: Wiley. Collier, A. 1979. In: Issues in Marxist Philosophy, vol. 3. J. Mepham and D. H. Ruben (eds.). Brighton: Harvester. Crick, F. 1964. Proceedings of the Plenary Sessions. Sixth International Congress of Biochemistry. New York.

132

The Origin of Species

Darwin, C. 1859. The Origin of Species. 1st ed. Harmondsworth: Penguin. Darwin, C. 1889. The Descent of Man. 2d ed. London: Murray. Dawkins, R. 1976. The Selfish Gene. Oxford: Oxford University Press. De Beer, G. 1971. Homology, an Unsolved Problem. Oxford: Oxford University Press. Dobzhansky, T. 1951. Genetics and the Origin of Species. 3d ed. New York: Columbia University Press. Driesch, H. 1914. The History and Theory of Vitalism. London: Macmillan. Driesch, H. 1929. The Science and Philosophy of the Organism. 2d ed. London: Black. Dunn, L. C. 1963. Genetics. Encylopaedia Brittanica. Feyerabend, P. 1975. Against Method. London: New Left Books. Frankel, J. 1974. Positional information in unicellular organisms. J. Theor. Biol. 47: 439–481. Gasking, H. 1967. Investigations into Generation 1651–1828. London: Hutchinson. Hämmerling, J. 1963. Nucleo-cytoplasmic interactions in Acetabularia and other cells. Annu. Rev. Plant Physiol. 14: 65–92. Huxley, J. 1963. Evolution: The Modern Synthesis. 2d ed. London: Allen and Unwin. Jacob, F. 1974. The Logic of Living Systems. London: Allen Lane. Körner, S. 1955. Kant. Harmondsworth: Penguin. Kuhn, T. S. 1970. The Structure of Scientific Revolutions. 2d ed. Chicago: University of Chicago Press. Lévi-Strauss, C. 1966. The Savage Mind. London: Weidenfeld and Nicholson. Maynard Smith, J. 1975. The Theory of Evolution. 3d ed. Harmondsworth: Penguin. McSherry, G. M. (1975). August Weismann: the methodology of a research programme. M.Sc. thesis, University of Sussex. Mepham, J. 1973. The Structuralist Sciences and Philosophy. In: Structuralism, D. Robey (ed.). Oxford: Clarendon Press. Monod, J. 1972. Chance and Necessity. London: Collins. Morgan, T. H. 1901. Regeneration. London: Macmillan. Needham, A. E. 1952. Regeneration and Wound Healing. London: Macmillan. Oosawa, F., Kasai, H., Hataro, S., and Asakara, S. 1966. In: G. E. W. Wolstenholme and M. O’Connor (eds.), Ciba Foundation Symposium: Principles of Biomolecular Organization. Boston: Little, Brown. Ouweneel, W. J. 1976. Developmental genetics of homoeosis. Adv. Genet. 18: 179– 248. Pasteels, J. J. 1964. The morphogenetic role of the cortex of the amphibian egg. Adv. Morphogenet. 3: 363–388. Piaget, J. 1971. Structuralism. London: Routledge and Kegan Paul. Russell, E. 1916. Form and Function. London: Murray.

133

Gerry Webster and Brian C. Goodwin

Saunders, J. W., Cairns, J. M., and Gasseling, M. T. 1957. The role of the apical ridge of ectoderm in the differentiation of the morphological structure and inductive specificity of limb parts in the chick. J. Morphol. 10(1): 57–87. Sonneborn, T. 1970. Gene action in development. Proc. Roy. Soc. Lond. B 176: 347– 366. Spemann, H. 1938. Embryonic Development and Induction. New Haven: Yale University Press. Spemann, H., and Schotté, O. (1932). Über xenoplastische Transplantation als Mittel zur Analyse der embryonalen Induktion. Naturwissenschaften 20: 463– 467. Tartar, V. 1961. The Biology of Stentor. Oxford: Pergamon Press. Waddington, C. H. 1957. The Strategy of the Genes. London: Allen and Unwin. Webster, G. 1974. Jacob as historian and biologist. Nature 251: 81–82. Webster, G., and Goodwin, B. 1996. Form and Transformation: Generative and Relational Principles in Biology. Cambridge: Cambridge University Press. Weismann, A. 1883. On Heredity. Reprinted in: T. S. Hall (ed.), A Source Book in Animal Biology. New York: Hafner, 1964. Weismann, A. 1885. The Continuity of the Germ Plasm as the Foundation of a Theory of Heredity. Reprinted in: J. A. Moore (ed.), Readings in Heredity and Development. New York: Oxford University Press, 1972. Weismann, A. 1904. The Evolution Theory. London: Edward Arnold. Wolpert, L. 1969. Positional information and the spatial pattern of cellular differentiation. J. Theor. Biol. 25: 1–47. Wolpert, L. 1971. Positional information and pattern formation. Curr. Topics Dev. Biol. 6: 183–224. Wolpert, L., and Lewis, J. 1975. Towards a theory of development. Fed. Proc. 34: 14–20. Zwilling, E. 1964. Development of fragmented and of dissociated limb bud mesoderm. Dev. Biol. 9: 20–37.

134

5 ON THE PROBLEM OF THE MOLECULAR VERSUS THE O R G A N I S M I C A P P R OAC H I N B I O L O G Y

ulrich wolf During the last decades of the twentieth century the success of molecular biology has to a large extent supplanted other approaches, particularly organismic biology. In the molecular biological concept, genes have displaced the organism, and the term organism no longer has a place therein. This raises a question: Can the organism be understood from the genes and their functions? And, in a wider sense, can the organism be reduced to molecular processes? Both questions are discussed in this essay. The confrontation between molecular and organismic biology is a sign that the two approaches operate on different levels and that it is impossible to completely resolve the repertoire of notions of organismic biology in molecular terms. In the organismic approach problems occur which, for conceptual reasons, cannot even be proposed under the molecular approach, and vice versa. It follows that the two ways of thinking do not coincide, and therefore they must complement each other in order to set forth an adequate idea of the organism as the subject of biology. Ernst Cassirer’s analysis of biological thought in his time reveals similar contrasting positions and addresses problems and endeavors of surprising relevance to the present. In particular it ascribes specific significance to formtheoretical considerations in biology which did not at the time, and do not today, receive enough attention. Indeed, Cassirer’s understanding of biology can still set the tone of present-day discourse on the nature of the organism, and therefore, the questions I deal with in this essay are discussed against that background.

Ulrich Wolf

The Notion of Wholeness Cassirer’s voluminous work Das Erkenntnisproblem in der Philosophie und Wissenschaft der neueren Zeit (The problem of knowledge in the philosophy and science of modern times) includes a chapter on biology entitled ‘‘The ideal of knowledge in biology and its changes.’’ 1 The chapter deals with the history of ideas from the eighteenth century and Carl von Linné to the third decade of the twentieth century and Ludwig von Bertalanffy; that is, from a descriptive and classificatory science to the beginnings of system theory. Those unfamiliar with this history may be surprised at the many concepts of biological thought produced during less than two hundred years. Cassirer characterizes his own position, his ‘‘ideal of knowledge,’’ clearly in his work, referring to the Critique of Judgement, in which Kant claims an autonomy, a methodological independence of biology, but without abandoning its connection with the exact sciences. According to Cassirer, this was the first time in the history of biological thought that this problem had been addressed, and it was Kant who detected it. Cassirer comments on Kant’s work as follows: ‘‘With this [statement] a new question was put forward which could not be ignored by future biological research, irrespective of the school and view to which it belonged’’ (1957: 137).2 Cassirer’s position focuses on understanding the phenomena of life as a category of its own whose investigation requires its own rational methods and logical devices: ‘‘The knowledge of a holistic order of processes in the organism is a knowledge sui generis, which does not become dispensable or replaceable by the detection of causal connections. . . . We must absolutely retain the term of wholeness because it is the doorway, so to speak, through which we must pass in order to access the sphere of biological problems at all. . . . With this aim, an approach other than a causal analysis of single processes is necessary’’ (Cassirer 1957: 249–250).3 These statements reflect the state of knowledge in 1932. Almost seventy years later, molecular biologists Eric Lander and Robert Weinberg would claim that ‘‘twentieth century biology triumphed because of its focus on intensive analysis of the individual components of complex biological systems. The 21st century discipline will focus increasingly on the study of entire biological systems by attempting to understand how component parts collaborate to create a whole. For the first time in a century, reductionists have yielded ground to those trying to put forward a holistic view of cells and tissues’’ (Lander and Weinberg 2000: 1781). Cassirer’s postulate, formulated in the early twentieth century from an organismic point of view, has acquired

136

Molecular versus the Organismic Approach

fresh interest as the task at hand for molecular biology on the threshold of the twenty-first century. Cassirer did not live to see the rise of molecular biology, which has decisively shaped today’s interpretation of the organism. And the success of molecular biology in the second half of the twentieth century has widened the gap to organismic concepts. Lander and Weinberg, however, point to a change in the concept of molecular biology that may indicate that the molecular approach is beginning to take the organismic one into consideration. Before proceeding further, let us first characterize the two positions.

The Molecular and Organismic Approaches In 1985, Stent addressed this confrontation in his already classical paper on ‘‘Thinking in one dimension: the impact of molecular biology on development.’’ He explains: ‘‘The tenet that the gene is a one-dimensional description of the primary structure of a particular protein molecule was turned . . . into the doctrine that the genome is a one-dimensional description of the whole animal. In particular, it came to be believed that the genome embodies not merely a protein catalogue but also a genetic program for development, from zygote to adult.’’ As for his own position, he states that ‘‘development is a historical rather than programmatic phenomenon, by which each stage in the progressive structural and compositional reshaping of the embryo is both the effect of earlier and the cause of later developmental transactions.’’ In its most reductionistic version, the molecular view is gene centric; that is, the genes represent a code of preformed programs containing information for the entire development of the organism, and calling up this code results in the realization of this development. Ontogenesis is thus preformistic, and development is a deterministic process. Concerning biological evolution, Sydney Brenner (cited in Lewin 1984: 1327) summarizes the molecular position by stating that ‘‘ultimately, the organism must be explicable in terms of its genes, simply because evolution has come about through alterations in dna.’’ Thus, evolution is based on changes in the genetic program caused by random mutations, and the course of evolution is shaped by adaptive selection, resulting in the ‘‘survival of the fittest.’’ A quote from Richard Lewontin characterizes the organismic view: ‘‘A living organism at any moment in its life is the unique consequence of a developmental history that results from the interaction of and determination by internal and external forces. The external forces, what we usually think of

137

Ulrich Wolf

as ‘environment,’ are themselves partly a consequence of the activities of the organism itself as it produces and consumes the conditions of its own existence. Organisms do not find the world in which they develop. They make it. Reciprocally, the internal forces are not autonomous, but act in response to the external’’ (Lewontin 1992: 34). The classical dualism of two independent determinants, genes and environment, according to which the environment confronts the organism with definite problems and opportunities to which it must adapt, becomes an interdependency. Oyama (2000) has coined the term ‘‘constructivistic interaction’’ to describe the mutualistic relationship by which the conditions of development are continuously generated. These conditions are not informative, but rather formative; the living system with its internal and external environment is not only a forma formata but is itself a forma formans. Oyama notes, ‘‘Evolution is change in the constitution and distribution of developmental systems, organism-environment complexes that change over both ontogenetic and phylogenetic times’’ (2000: 112). The roles of mutation and selection are limited. Stability and variability are systemic properties; the organism as the product of its development is the subject, rather than its genotype being the object, of evolution (Weingarten 1993). The variability is not only the consequence of mutations; it depends on the internal and external contexts of the system, and in different contexts a mutation may have different effects. Similarly, selection is not the consequence of the influence of an external agent on a passive object, but is rather the result of ‘‘constructivistic interactions’’ between the organism and its milieu. Therefore it is misleading to speak of a ‘‘selective agent’’ as a driving force in evolution. If evolution is considered as a change of developmental systems, then stability and variability depend on the relevance of developmental processes. When one compares these two positions it becomes evident that they represent not only different research approaches for explaining biological phenomena but differences in the conception of the organism as well. Whether an organism develops according to a preformed program or is the result of a historical process, whether evolution is the consequence of changes in the program or of systemic interactions—these are conceptual differences. Nevertheless, the subject of these conceptions is one and the same: the organism as a whole. Referring to this identity of the subject of research and knowledge, Cassirer (1957: 250) writes: ‘‘The ‘organismic view’ . . . does not oppose the attempt to explain life processes physico-chemically; but it is aware that

138

Molecular versus the Organismic Approach

other logical aids are required in order to tackle the problem of the wholeness successfully.’’ 4 Since 1932, a number of logical aids for analyzing complex systems have been added—in particular the principle of self-organization and cybernetics, including feedback mechanisms, circular causality, and self-referentiality— which have contributed much to our understanding of biological processes (e.g., Freeman 2000). While cybernetics was not foreseeable at the time he wrote, Cassirer recognized that ‘‘the distinctiveness of an organism cannot be characterized by a special attribute, but is based on the correlation between all its individual determinants. From this it follows that biology, if it wants to achieve its aim and realize its empirical ideals, cannot do without the rational moments, but it needs to develop its own rational methods. For its way does not simply lead from the particular to the general; it has to presuppose the whole already in considering each part, the general in the special’’ (Cassirer 1957: 152).5 Thus, the subsystems are to be considered as components for the functioning of the entire system, and any investigation of their action has to consider the performance of the entire system. Evidently, principles like self-organization and self-referentiality meet Cassirer’s postulates.

Range and Limits of Molecular Analysis The research program of molecular biology consists of ‘‘the causal analysis of single processes’’ (Cassirer 1957: 250; see quotation with note 3). The biology of molecules, e.g., the genes, is the biology of parts of the organism. Certainly, molecular analyses are directed toward understanding connections, proceeding to insights into developmental pathways, and eventually to the unraveling of the mechanism underlying a morphogenetic process. This process, however, obeys definite principles of organization, and it can take place only under certain initial and boundary conditions which are themselves contingent. Invariably, a particular subsystem being analyzed proves to be in the environment of other subsystems that are constitutive for the system being analyzed (Luhmann 1984). The molecular biological approach, emphasizing first the structural analysis, has advanced to the investigation of the correlations between the components, and this is also a main subject of the organismic approach. In molecular biology, however, the correlations are defined by the functions of the components, and in the system-theoretical concept of the organism the

139

Ulrich Wolf

correlations follow from the functions required from the components. Cassirer supports the latter view when he states that the order of the processes is the clearest—if not the only—difference of vital functions from ordinary physicochemical processes (1957: 250). Goodwin (1994) presents abundant examples showing that pattern formation during development depends more on the relational order of interactions in space and time than on the nature of the components involved (e.g., cells). Our increasing knowledge of molecular modifications and epigenetic and constructivistic interactions taking place within the organism and with its environment has further weakened the reductionist position. In this context, the genome becomes only one among numerous factors constituting life processes. Thus, such processes cannot be explained by molecular genetic mechanisms only. However, the molecular biological position is restricted not to genetic mechanisms but rather to molecular ones. Therefore, the question follows as to what extent life phenomena, including those on higher levels of complexity, are accessible to and satisfactorily explained by the molecular approach. How this question is answered determines whether organismic biology can be reduced to molecular biology; whether, in other words, the phenomena described and studied by the organismic approach can be given a molecular explanation. In any case a practical border of molecular analysis is to be reckoned with, depending on the complexity of the phenomena studied in organismic biology. This complexity, already in developmental processes, and more so at the level of neurobiological performance and the formation of social systems, will bring the molecular approach to the analysis of partial processes with all other conditions remaining contingent. Therefore, organismic biology uses a repertoire of terms and methods which apply at a higher level of complexity than that of molecules and their interactions. Alberch explains, ‘‘To successfully analyze and understand a phenomenon we must define our variables at the appropriate level of complexity. . . . Reductionism very often does not increase our ability to understand phenomena of higher order interactions’’ (Alberch 1980: 664, 665). These are challenging statements for the molecular biologist. Now that the molecular level has become accessible, it becomes almost a postulate that we must analyze, and thus explain, biological phenomena on that level. This postulate originates from the fact that molecular analysis has a higher resolution capacity than analyses performed on elements at higher levels of complexity. The methodological development in biology has thus shifted generally to molecules as the ‘‘appropriate’’ variables. Cassirer’s postulate that the subject of biology

140

Molecular versus the Organismic Approach

is the organism as a whole does not and could not yet consider the extent to which the biological phenomena and processes studied by the organismic approach can be derived from molecular mechanisms. It may be doubted whether all biological phenomena can be explained in terms of molecular mechanisms, and it may not be adequate to describe complex organismic phenomena in molecular terms. And some phenomena may not be accessible to molecular analysis at all because they follow principles of a different nature. Since the introduction of molecular biological methods, a comprehensive repertoire of biological notions has become interpretable in molecular terms. The rules of Mendelian genetics are an instructive example, which already showed a high explanatory power before the underlying mechanisms were known. These rules refer to phenotypic traits whose distinct modes of inheritance (dominant, recessive, sex-linked, etc.) are mirrored by regular segregation patterns within pedigrees. The elucidation of the molecular basis of Mendelian inheritance has revealed that the phenotype cannot always be assigned to a particular genotype (Wolf 1995, 1997). The genotype-phenotype relationship is, in fact, highly heterogeneous, and in many cases the phenotype cannot be reduced to the genotype. With growing knowledge of the molecular basis of the Mendelian rules, their validity became considerably restricted. The key terms of Mendelian genetics encompass a multitude of molecular mechanisms, e.g., those which result in phenotypic dominance (Wilkie 1994). Nevertheless, these terms remain useful on the phenotypic level. The gene, conceived as an elementary entity, has proved to be unexpectedly complex as well. The term gene was introduced by Johannsen in 1909 replacing various terminological precursors such as pangene, biophore, gemmule, and the like. Johannsen himself understood the gene first as a kind of unit of calculation, and even in 1933 T. H. Morgan stated: ‘‘It does not make the slightest difference whether the gene is a hypothetical unit, or whether the gene is a material particle’’ (quoted in Falk 1984). Molecular conceptions of the gene became possible only after the structure of the dna had been elucidated (Watson and Crick 1953), but the term itself has remained elusive, and the discussion about its meaning or definition continues (e.g., Falk 1984; Neumann-Held 1999; see also the essays in this volume). Molecular analysis has shown that the gene cannot be defined as a structural unit or as a unit of function. It is neither discrete nor continuous; its borders are variable, and its localization varies as well. Also, it cannot be defined by its product in the form of rna or a polypeptide because these products can have sequence de-

141

Ulrich Wolf

viations from a direct projection of the dna sequence. Therefore, the meaning assigned to the term gene either emerges from its context or is directed by the requirements of the investigator. Gene is thus an example of a term which has outlasted changes in conception on all levels of analysis because it has proved not to be reducible and is therefore irreplaceable. It is still valid to designate a unit on a higher level of complexity. The field of developmental biology in the premolecular era produced a detailed terminology. Many of these terms and concepts have survived and have been at least partially explained by molecular analysis; examples include induction, polarity, fields of self-organization or of inhibition, cell determination, and pattern formation. Wilkins (1997), for example, attempted a molecular interpretation of canalization, a term coined by Waddington (1942) which addresses the concept that developmental pathways are buffered against internal and external interferences in such a way that the variation in the formation of a trait is reduced and the phenotypic norm is guaranteed. Disturbances can be of genetic or epigenetic nature or caused by environmental cues. Wilkins, in his derivation of self-stabilizing mechanisms, focuses on genetic canalization, compensation for the effects of mutations. He refers in particular to coexpressed paralogous genes providing at least a partial functional redundancy, and he also mentions other possibilities like modifying genes. It is to be expected, however, that a number of other complex and central notions of developmental biology cannot be reduced to molecular mechanisms; for example, the term developmental constraint, which addresses certain restrictions of the developmental potential by organizational fixations that materialized during evolution. This phenomenon is usually understood to be the consequence of constraints exerted by the epigenetic system. The development of the aortic arch in the rabbit (Alberch 1980), for example, occurs in six different types with different frequencies, and the most frequent type occurs in more than 80 percent of cases. The variation is assumed to be due to slight fluctuations in the temporal course and growth rates during morphogenesis. The restricted range of variation, the nonrandom frequency distribution, and the absence of a regular mode of inheritance speak in favor of epigenetic constraints. Because of the autodynamics of epigenetic processes, including stochastic effects (Kurnit et al. 1987), processes of this kind obviously exclude a mere molecular interpretation. The extent to which concepts of the premolecular developmental biology can be interpreted in molecular terms aside, the molecular approach in analyzing morphogenetic processes has been extraordinarily successful. The

142

Molecular versus the Organismic Approach

multitude of examples runs from segmentation and development of limbs and inner organs to the formation of the eye and the nervous system. Usually, the molecular analysis proceeds via mutations in genes involved in developmental processes and their phenotypic effects. The preferred model organism is the mouse, and directional mutations are produced by homologous recombination of transgenes allowing for deductions about their functional role by their expression pattern and pathological consequences. A prominent example is a group of genes controlling axial morphogenesis (segmentation), the homeobox (Hox) gene family (Maconochie et al. 1996). Studies of these genes have provided information on the regulation of gene expression, cooperation between individual genes, and compensation for defective mutants by functional redundancy. Molecular analysis has also contributed to the understanding of morphological variation and diversification during evolution. Carroll (2000) considers changes in gene regulation during development to be responsible in the first place for evolution, noting that ‘‘changes in the cis-regulatory systems of genes more often underlie the evolution of morphological diversity than do changes in gene number or protein function.’’ Genes and gene families that control morphogenetic processes of phylogenetically far distant species show a high degree of sequence homology. The Hox genes, which occur as gene clusters in tandem array in organisms as different as annelids, arthropods, and vertebrates, may serve again as an example. A correlation has been found between differences in axial morphology and the positional regulation of these genes. It appears that the expression pattern of individual Hox genes has shifted along the body axis through evolutionary changes of their cisregulatory elements. These and similar findings have led Carroll to conclude that genetic changes of regulatory dna are the main source of the genetic differences underlying morphologic evolution, up to changes in body plan. Research projects of this kind refer to the dna sequence as a basic parameter, and the observations they produce are interpreted in terms of the genome. On the next higher level of analysis, that of proteins, in principle a correspondence between nucleotide and amino acid sequences operates. However, properties of proteins give rise to highly ordered processes which cannot be derived simply on the basis of dna sequence. As discussed by Alberts (1998), all significant cellular processes follow their course by selfassembly of at least ten protein molecules which interact with several other such complexes. The driving force of these processes is energy-dependent conformational changes which in general are unidirectional and committed, thereby producing a high degree of order. However, at the protein level,

143

Ulrich Wolf

macromolecular sequences remain the parameter of reference for understanding life phenomena, and therefore the question arises as to whether the analysis of the genes, their function, and their products can explain the organism as a systemic whole.

Self-Organizing Processes In a paper with the allusive title ‘‘Molecular ‘vitalism’’’ Kirschner et al. (2000) ask: ‘‘Will true understanding of living systems come from further annotating the database of genes, or must we explore the physicochemical nature of living systems?’’ And further: ‘‘Although the units we consider, proteins, cells, and embryos are manifestly the products of genes, the mechanisms that promote their function are often far removed from sequence information.’’ This statement refers to processes of macromolecular assembly and self-organization for which the genome does not contain the blueprints: ‘‘Self-assembly is an extension of the central dogma of molecular biology, bringing us from the realm of linear information to the realm of protein assemblies.’’ Structures of higher order are generated which are ‘‘uniquely determined by size, number of components, geometry, and strength of interaction,’’ reaching ‘‘equilibrium, a state of minimum free energy.’’ Self-organization extends self-assembly, allowing for ‘‘a wider set of conditions’’ and greater structural variability. Even if size and composition of the components vary, self-organizing systems can still generate an ordered structure. A classical example from the premolecular era is the experiments by Fankhauser (1945) that show the influence of the degree of ploidy on the development of newts (Triturus). With increasing ploidy, cell size increases and cell number decreases, resulting in organs and entire individuals of normal size. The development of the zygote to a multicellular organism is brought about at many steps by self-organizing processes. An example put forward by Kirschner et al. (2000) is dorsoventral axis formation in the frog egg, which does not depend on prior organization. The random position at which the sperm enters the egg and produces an aster determines the dorsoventral axis, which is then formed by the self-organization of components of the cytoskeleton. Axis formation also takes place if the egg is artificially activated, thus excluding the existence of specific inducers transmitted by the sperm. They note, ‘‘Since the initial symmetric arrangement of contents is lost when a new axis is formed, the process is a symmetry-breaking process’’ (Kirschner et al. 2000: 83). The subsequent steps of differentiation are performed in relation to this

144

Molecular versus the Organismic Approach

axis, and here again self-organizing processes take place. They depend on cellular competence—the capacity of the embryonic cell to follow two or more developmental pathways depending on the local milieu and resulting in the selection of one while the other potentials become suppressed. ‘‘The dependence of the cellular responses on local conditions underlies the mechanism of self-organization’’ and ‘‘increases the separation of the cellular phenotype from the pattern of gene expression’’ (p. 85). Gene expression determines the multipotent precursor, which by its competence and the local conditions becomes fixed in a specific developmental direction. The genetic expression pattern thus offers a range of possible phenotypes, and the actual phenotype cannot be predicted. It can be assumed that processes of self-organization are analyzable in principle at the molecular level, and this may even be required to understand them. Indeed, the molecular basis of pattern formation is already understood to a considerable extent in a growing number of cases, Drosophila representing the model example (Nüsslein-Volhard 1992). Thus, molecular biology has already gone deep into the problems of morphogenesis, prompting Fox Keller (1992: 29) to say: ‘‘Molecular biologists, it appears, have discovered the organism.’’

Form-Theoretical Considerations The question of how form is brought about in development and how it changes in evolution was a main concern of Cassirer, who felt that formtheoretical thinking was missing in biology and considered formative processes as unique to the organism and not reducible to physics: ‘‘Considering organic phenomena . . . we finally always come across a specific difference [to physics] in the elements [of biological research]: a difference in the ‘form.’ If we try to describe an organism and if we try to follow its development, we always have to reckon with this form, an original ‘Anlage’ [primordial mold]’’ (Cassirer 1957: 220).6 Considering methodology, he writes: ‘‘Within life phenomena it can be shown on purely causal grounds how a subsequent stage of development originates from the previous one: however, we finally always arrive at an initial state of ‘organization’ which we must admit as a prerequisite, independent of how far we proceed backward. The causal consideration teaches us the rules by which one structure passes over to another one; however, that such individual ‘nuclei’ exist at all, that there are primordial entities under-

145

Ulrich Wolf

lying development which are different from each other in a specific way, cannot be made understandable [by causal consideration], but [can] only be addressed as a fact’’ (Cassirer 1921: 368–369).7 No doubt Cassirer refers here to ‘‘other logical aids’’ (Cassirer 1957: 250, see quotation with endnote 4) than the causal analysis of single processes in order to approach the problem of form in biology. Form-theoretical problems have recently been taken up again after having been overshadowed for decades by molecular biology. Stephen J. Gould (1970) postulated a new ‘‘science of form,’’ and at the Senckenberg Museum at Frankfurt a ‘‘construction morphology’’ was conceived on the basis of the ‘‘methodical constructivism’’ (see, e.g., Janich 1996). Interestingly, this construction morphology refers to the form-theoretical considerations of Cassirer (e.g., Weingarten 1993; Gutmann 1996; Janich and Weingarten 1999). In this conception, morphology is taken as the basis of biological organization. Starting from the organism as a whole, its construction principles are analyzed and reconstructed using machines as models for organisms. Mechanical coherence of organismic constructions is considered to be fundamental. This coherence is not based on an aggregation of parts with preexisting defined qualities; but the qualities of the parts can be determined only in view of the functional requirements of the coherent construction (Weingarten 1993: 293). Organisms as mechanically coherent constructions must be open energetically but are closed mechanically and operationally. Therefore, a construction is not directed from outside, and its self-preservation is not due to external conditions. From these determinations, Weingarten states, it follows that organisms ‘‘as developing constructions can only exist if they retain the invariant construction principles required. . . . However, at the very least those construction principles where a (possible) change allows for optimizing the capabilities and functions of the organism, must be able to vary. By this relation between invariance and variability, a field of possibilities is constituted for developmental processes. . . . neither the construction nor the invariant construction principles are predetermined for the organism in any way, but they can be understood only as self-produced by the activity and the dynamics of the organism and resulting in self-preservation’’ (Weingarten 1993: 295, 298).8 Therefore, the construction also determines ‘‘what an environment can be for it, namely that section of the outer world in which the organism finds all that it needs for life’’ (Weingarten 1993: 281).9 In construction morphology, the construction principles are derived from model-theoretical considerations. Weingarten notes that ‘‘at least some of these construction principles determine invariants which must be realized

146

Molecular versus the Organismic Approach

by all organisms at any time if they are to be able to exist at all. . . . The construction principles for an organism are the a priori requirements of empirical reconstructions of ‘real’ organisms’’ (Weingarten 1993: 298, 299).10 It is the conception of the construction morphology that its determinants and discrimination criteria are not empirically derived from real existing organisms, but rather gained by model-theoretical considerations. Model building is used to explain the natural system on which the model is based. The model does not depict the natural system exactly, but it should fulfil the purpose of explanation which itself was conceived first by the describer of the system (Janich and Weingarten 1999: 97). Construction morphology applies to a level of complexity which at least practically excludes the molecular approach. On the other hand, it gives access to biological phenomena whose analysis contributes considerably to the understanding of processes in development and evolution. Construction-theoretical considerations have also gained entrance into the microstructural field, such as Ingber’s (1998, 2000) theory of the ‘‘architecture of life.’’ Ingber constructed models based on the analysis of the architecture of biological cells and microstructures like dna, cell nuclei, filaments of the cytoskeleton, and ion channels. Because the construction principles are independent of the scale, they can also be applied to tissues and entire organisms. These constructions reach a mechanical equilibrium of the participating components by tensegrity, a term referring to ‘‘a system that stabilizes itself mechanically because of the way in which tensional and compressive forces are distributed and balanced within the structure’’ (Ingber 1998: 40). The combination of topological and energetic constraints to which the various components in different systems are exposed is the driving force of progressive self-assembly and therefore also of organization. ‘‘The question how living things form has less to do with chemical composition than with architecture,’’ Ingber explains (1998: 49). The way nature uses tensegrity for self-assembly at various levels up to more complex hierarchical structures prompted Ingber to assume that the formation of tensegrity systems also guided evolution from the origin of cellular life onward. ‘‘While changes in dna may generate biological diversity, genes are merely one product of evolution and not its driving force’’ (Ingber 2000: 1161). As driving force, he prefers the requirement to meet the fundamental architectural and energetic constraints. If molecular biology is seen from a construction-theoretical viewpoint, structural considerations have played an important role from the very beginning. The structural model of the dna was the key for the understanding of

147

Ulrich Wolf

its functions of stable self-replication and transcription. The way genes are arranged within the chromosome is not only of functional significance but also implies potentials of evolution, as the example of the Hox genes shows. At the protein level, Alberts (1998) speaks metaphorically of ‘‘protein machines.’’ Structural research at the molecular level has become a discipline of its own because it has proved to be indispensable for the understanding of functional processes. Nonmolecular approaches like construction morphology give us access to knowledge which cannot be obtained with the methods of molecular biology. These approaches, while compatible with molecular biology, demonstrate that the latter covers only one section of life phenomena and that other considerations and methods are required to analyze mechanisms which follow nonmolecular principles. Cassirer’s ideal of knowledge of an order of processes which he considered the clearest differential characteristic for life processes can now be encountered in organismic as well as molecular biology, and it emerges that the two concepts complement rather than exclude each other, contributing to one and the same ‘‘ideal of knowledge’’ in biology.

Notes 1 Cassirer (1957). In the following text the original German versions of the quotations appear in the endnotes. 2 ‘‘Damit war eine neue Frage gestellt, an der künftig die biologische Forschung, gleichviel welcher Schule und Richtung sie angehörte, nicht mehr vorbeigehen konnte.’’ 3 ‘‘Die Erkenntnis der ganzheitlichen Ordnung des Geschehens im Organismus ist eine Erkenntnis sui generis, die durch die Feststellung kausaler Zusammenhänge nicht entbehrlich gemacht oder ersetzt werden kann. . . . Den Ganzheitsbegriff können wir und müssen wir unbedingt festhalten, da er gewissermaßen die Eingangspforte bildet, die wir durchschreiten müssen, um überhaupt zu den biologischen Problemen vorzudringen. . . . zeigt es sich, dass hierfür eine andere Betrachtungsweise notwendig ist, als die kausale Analyse der Einzelabläufe.’’ 4 ‘‘Die ‘organismische Auffassung’ . . . tritt keinem Versuch, die Lebensvorgänge physikalisch zu erklären, entgegen; aber sie ist sich andererseits bewusst, dass es noch anderer logischer Hilfsmittel bedarf, um das Problem der Ganzheit mit Erfolg in Angriff zu nehmen.’’ 5 ‘‘Die Eigentümlichkeit eines Organismus lässt sich nicht durch irgendeine

148

Molecular versus the Organismic Approach

besondere Eigenschaft ausdrücken, sondern sie beruht auf der Korrelation, die zwischen all seinen Einzelbestimmungen waltet. Es ergibt sich hieraus, dass die Biologie, wenn sie ihr Ziel erreichen und ihre empirischen Ideale verwirklichen will, die ‘rationalen’ Momente nicht nur nicht entbehren kann, sondern dass sie hierfür der Ausbildung eigener rationaler Methoden bedarf. Denn ihr Weg führt nicht schlechthin vom ‘Einzelnen’ zum ‘Allgemeinen,’ sondern sie muss schon in der Betrachtung jedes Teils das Ganze, sie muss im Einzelnen das Allgemeine voraussetzen.’’ 6 ‘‘Bei der Betrachtung organischer Phänomene . . . stoßen wir in den Elementen zuletzt stets auf einen spezifischen Unterschied: einen Unterschied der Form. Versuchen wir einen Organismus zu beschreiben und versuchen wir seine Entwicklung zu verfolgen, so müssen wir stets mit dieser Form, mit seiner ursprünglichen ‘Anlage’ in ihm rechnen.’’ 7 ‘‘Denn innerhalb der Lebenserscheinungen kann freilich rein ursächlich gezeigt werden, wie das folgende Glied der Entwicklung aus dem vorhergehenden wird und entsteht: aber wir gelangen, soweit wir hierbei auch zurückgehen mögen, zuletzt immer nur auf einen Anfangszustand der ‘‘Organisation,’’ den wir als Voraussetzung zugeben müssen. Die kausale Betrachtung lehrt uns, nach welchen Regeln die eine Struktur in die andere übergeht; dass aber überhaupt solche individuellen ‘Keime’ vorhanden sind, dass es ursprüngliche, voneinander spezifisch-verschiedene Bildungen gibt, die der Entwicklung zugrunde liegen, vermag sie nicht mehr weiter begreiflich zu machen, sondern nur als Tatsache auszusprechen.’’ 8 ‘‘Sondern Organismen können nur als sich entwickelnde Konstruktionen bei Beibehaltung notwendiger invarianter Konstruktionsprinzipien existieren. . . . Variabel müssen aber zumindest diejenigen Konstruktionsprinzipien sein, mit deren (möglicher) Veränderung Fähigkeiten bzw. Leistungen von Organismen optimiert werden können. Durch genau dieses Verhältnis von Invarianz und Variabilität wird ein Möglichkeitsfeld konstituiert für u.a. auch Entwicklungsvorgänge. . . . dass weder die Konstruktion insgesamt noch die invarianten Konstruktionsprinzipien den Organismen in irgendeiner Weise vorgeordnet sind, sondern nur verstanden werden können als durch die Aktivität und Dynamik der Organismen selbst erzeugt und selbst erhalten.’’ 9 ‘‘[Die Konstruktion legt fest] was für sie eine Umwelt sein kann, nämlich derjenige Ausschnitt aus der Außenwelt, in dem der Organismus das vorfindet, was er zum Leben braucht.’’ 10 ‘‘Zumindest einige dieser Konstruktionsprinzipien bestimmen Invarianten, die von allen Organismen jederzeit realisiert werden müssen, wenn sie überhaupt existenzfähig sein sollen. . . . Die Konstruktionsprinzipien für Organismen sind die apriorischen Grundlagen empirischer Rekonstruktionen von ‘wirklichen’ Organismen.’’

149

Ulrich Wolf

References Alberch, P. 1980. Ontogenesis and morphological diversification. Amer. Zool. 20: 653–667. Alberts, B. 1998. The cell as a collection of protein machines: preparing the next generation of molecular biologists. Cell 92: 291–294. Carroll, S. B. 2000. Endless forms: the evolution of gene regulation and morphological diversity. Cell 101: 577–580. Cassirer, E. 1921. Kants Leben und Lehre. 2d ed. Darmstadt: Wissenschaftliche Buchgesellschaft. Cassirer, E. 1957. Das Erkenntnisideal der Biologie und seine Wandlungen. In: Das Erkenntnisproblem in der Philosophie und Wissenschaft der neueren Zeit. Band 4, Von Hegels Tod bis zur Gegenwart. Quoted from Ernst Cassirer, Gesammelte Werke, Hamburger Ausgabe, Band 5. Hamburg: Meiner, 2000. Falk, R. 1984. The gene in search of an identity. Hum. Genet. 68: 195–204. Fankhauser, G. 1945. Maintenance of normal structure in heteroploid salamander larvae, through compensation of changes in cell size by adjustment in cell number and cell shape. J. Exp. Zool. 100: 445–455. Fox Keller, E. 1995. Refiguring Life. New York: Columbia University Press. Freeman, M. 2000. Feedback control of intercellular signalling in development. Nature 408: 313–319. Goodwin, B. 1994. How the Leopard Changed Its Spots. New York: Charles Scribner’s Sons. Gould, S. J. 1970. Evolutionary paleontology and the science of form. Earth-Sci. Rev. 77–119. Gutmann, M. 1996. Die Evolutionstheorie und ihr Gegenstand. Berlin: Verlag für Wissenschaft und Bildung. Ingber, D. E. 1998. The architecture of life. Sci. Amer. 278: 48–57. Ingber, D. E. 2000. The origin of cellular life. BioEssays 22: 1160–1170. Janich, P. 1996. Konstruktivismus und Naturerkenntnis. Frankfurt: Suhrkamp. Janich, P., and Weingarten, M. 1999. Wissenschaftstheorie der Biologie. Munich: Wilhelm Fink Verlag. Johannsen, W. 1909. Elemente der exakten Erblichkeitslehre. Jena: Gustav Fischer. Kirschner, M., Gerhart, J., and Mitchison, T. 2000. Molecular ‘‘vitalism.’’ Cell 100: 79–88. Kurnit, D. M., Layton, W. M., and Matthysse, S. 1987. Genetics, chance, and morphogenesis. Amer. J. Hum. Genet. 41: 979–995. Lander, E. S., and Weinberg, R. A. 2000. Genomics: journey to the center of biology. Science 287: 1777–1782. Lewin, R. 1984. Why is development so illogical? Science 224: 1327–1329. Lewontin, R. C. 1992. The dream of the human genome. N.Y. Rev. Books, 28 May.

150

Molecular versus the Organismic Approach

Luhmann, N. 1984. Soziale Systeme. Grundriß einer allgemeinen Theorie. Frankfurt: Suhrkamp. Maconochie, M., Nonchev, S., Morrison, A., and Krumlauf, R. 1996. Paralogous Hox genes: function and regulation. Ann. Rev. Genet. 30: 529–556. Neumann-Held, E. 1999. The gene is dead—long live the gene. In: P. Koslowski (ed.), Sociobiology and Bioeconomics: The Theory of Evolution in Biological and Economic Theory (pp. 105–137). Berlin: Springer. Nüsslein-Volhard, C. 1992. Musterbildung im Drosophila-Embryo. Nova acta Leopoldina, N.F. 67: 281, 267–283. Oyama, S. 2000. Evolution’s Eye. Durham: Duke University Press. Stent, G. S. 1985. Thinking in one dimension: the impact of molecular biology on development. Cell 40: 1–2. Waddington, C. H. 1942. Canalization of development and the inheritance of acquired characters. Nature 150: 563–565. Watson, J. D., and Crick, F. H. 1953. Molecular structure of nucleic acids: a structure for deoxyribose nucleic acid. Nature 171: 737–738. Weingarten, M. 1993. Organismen—Objekte oder Subjekte der Evolution? Darmstadt: Wissenschaftliche Buchgesellschaft. Wilkie, A. O. M. 1994. The molecular basis of genetic dominance. J. Med. Genet. 31: 89–98. Wilkins, A. S. 1997. Canalization: a molecular genetic perspective. BioEssays 19: 257–262. Wolf, U. 1995. The genetic contribution to the phenotype. Hum. Genet. 95: 127– 148. Wolf, U. 1997. Identical mutations and phenotypic variation. Hum. Genet. 100: 305–321.

151

6 G E N E S , D E V E L O P M E N T, AND SEMIOSIS

jesper hoffmeyer The capacity for prediction has always been a highly esteemed value in science, but biology has not fared well in this respect. Only rarely has it been possible to formalize the description of a living system to the extent that its future fate could be deduced in more than the most general way. Recently, however, even physics has encountered nonpredictability in cases of nonlinear dynamic systems where initial conditions cannot—in principle—be measured with sufficient precision. Since living systems are necessarily far from equilibrium systems and are characterized by complex nonlinear dynamics, the nonpredictability of such systems should no longer surprise us. But opinions differ as to whether this unpredictability is rooted solely in epistemological problems or reflects a deeper reality pertaining to our universe. This question concerns the ontological status of chance. If chance is interpreted as uncertainty, that is, as a lack of knowledge about true causative agencies, then chaos dynamics remains inside the secure deterministic vision of classical physics. If, however, chance is seen as reflecting a real indeterminacy of the world, as has been suggested by among others Karl Popper (1990), then initial conditions are indeterminate in principle and chaos dynamics becomes an explanatory tool for our understanding of irreversibility and history. Such a view fits well into modern interpretations of the second law of thermodynamics (Swenson and Turvey 1991; Salthe 1993; Ulanowicz 1997).

Formalization and Realization I believe there are strong reasons to adopt the concept of chance as expressing a real indeterminacy inherent in our world, and one major route of reasoning which has led to this view is a growing understanding of the ‘‘inverted’’

Genes, Development, and Semiosis

temporal relation between formalizations and realizations in evolutionary systems. Science, in the spirit of engineering, has tended to claim that digital blueprints always precede and specify the construction of real systems. Outstanding examples are the belief that dna originated before cells, that the accumulation of new genetic material (mutations) is the essence of speciation and evolution; that cognition is based on the manipulation of symbolic objects; that human language is pure syntax insulated from meaning, perception, and emotion and ultimately coded for in the genes; and that efficient robots can be constructed through advance programming. It now seems likely that evolutionary systems in general work the other way around. Formal aspects do not specify the systems beforehand, but on the contrary have to be established in the very process of their evolution. Thus the view of rna as the ‘‘magic molecule’’ behind the origin of life has been severely criticized by Stuart Kauffman, among others, who showed that the formation of a complex, autocatalytically closed chemical system was not only a possible but in fact a likely precursor for life (Kauffman 1993, 1995). Others have emphasized the importance of membranes for the origin of life (Weber, Depew, et al. 1989; Morowitz 1992; Hoffmeyer 1998b; Weber 1998). The view of speciation as founded on a gradual accumulation of genetic novelties is presumably still held by a majority of biologists, but it has been strongly challenged from many directions. A more pluralistic understanding seems to be penetrating evolutionary biology in recent years, allowing the organismic level a bigger share of evolutionary agency (Depew and Weber 1995). Developmental systems theory, for example, claims parity between genes and other organismal or environmental factors that collectively influence the future of the species (Oyama 1985; Griffiths and Gray 1994; further references in Sterelny and Griffiths 1999). In cognitive science the ‘‘symbolic paradigm’’ is now challenged by a ‘‘dynamic approach’’ (Van Gelder and Port 1995), and the Chomskyan idea of language as inherent in our computational brains is strongly opposed by a growing understanding of the ‘‘embodied nature’’ of natural language (see, e.g., Lakoff and Johnson 1999 for a recent summary). Michael Polanyi should be credited as an early proponent of this view, which was implicit in his ideas of ‘‘tacit knowledge’’ more than forty years ago (Polanyi 1958). The American neurobiologist Terrence Deacon, who has dealt thoroughly with the evolution of the human brain, claims that even human children have no inborn mastery of symbolic reference but have to learn it by growing up in a linguistic culture (Deacon 1997). His surprising thesis is that language and brain have coevolved in the human lineage, but that language did most of the evolv-

153

Jesper Hoffmeyer

ing by adapting its syntactic structure to the nature of children’s guesswork. Finally, let me refer to the very persuasive work on robotics that confirms this general trend toward ‘‘embodiment.’’ Bottom-up programming through ‘‘situated emergence,’’ rather than top-down programming, appears to hold the key to the construction of efficient autonomous agents (Hendriks-Jansen 1996; Clark 1997). In sum, a convergence of new insights from different areas seems to point to a new conception of formalizations. Apparently, real systems do not have formal origins; formalizations become possible only after realizations, never before. Thus the need for deterministic causation fades away and it becomes more plausible to see irreversible time as a real phenomenon. Something just happens in evolutionary systems that no formula could have predicted, and this strongly supports the conception of chance fluctuations as an ontologically real aspect of our universe. The combination of chance fluctuations and chaos dynamics can be seen as responsible for the making of history in the true sense of this word, as a unique and unpredictable series of temporal events on a macroscale. But while the combined effect of chance and chaos dynamics is what makes history possible, it does not explain history. Chaos dynamics may legitimize the concept of living systems as creative agents, but it does not tell us how creativity actually works in these systems. This is where the concept of semiosis (signification or sign-processes) becomes a useful tool in our descriptions. Once strict determinism is abandoned, we are left with a more open causative universe where cohesiveness and self-organization are ensured through semiotic emergence (Hoffmeyer 1997) consistent with the second law of thermodynamics (see below).1

Semiosis and Information in Living Systems That even molecular biology cannot escape a communicative or informational dimension is implicit in Francis Crick’s postulate of the central dogma, which stipulates that something, termed information, can be passed on in only one direction, that is, from dna to rna to protein, and never in the other. Crick certainly saw no parity here, which is of course due to the narrow focus of his statement.2 But what about this entity, information, which is so elegantly transported away from master (dna) to pupil (protein)? What is the nature of informa-

154

Genes, Development, and Semiosis

tion? I will not exhaust the reader’s patience by entering into the endless discussions on the proper understanding of this term, but I will note that (1) no agreement has been reached or is even in sight, and (2) for the practicing biologist the central dogma has always meant simply that ‘‘instructions’’ are passed on from dna to protein. In the absence of any rigid definition of the term information the instructional conception of this word imperceptibly slid into the matrix of tacit metaphors nourishing the minds of modern biologists. But the adoption of an instructional understanding of the ‘‘something’’ transmitted from dna to protein is no innocent move. It immediately raises questions about how we should understand the relation between senders and receivers of this ‘‘information’’: are they supposed to be connected through a causal relation? A yes to this question would seem to violate Norbert Wiener’s famous statement that ‘‘information is information, not substance or energy. No materialism that fails to admit this can survive today’’ (Wiener 1962). The concept of the transfer between molecules of something which is neither substance nor energy feels foreign to materialistic science. John von Neumann touched on this peculiar aspect of living systems in a somewhat different way (von Neumann 1966). Von Neumann claimed that replication in real existing cells (as opposed to cellular automata) requires not only dynamic fabrication (i.e., active enzymatic rate control of synthesis) but also nondynamic (‘‘quiescent’’) constraints (i.e., genetic memory description).3 The fact that such functional organization could arise was to him ‘‘a miracle of the first magnitude,’’ notes Howard Pattee (Pattee 2000). Howard Pattee has further developed this ‘‘complementarity’’ between ‘‘dynamic and linguistic modes of complex systems’’ (Pattee 1973, 1977). Living systems are inherently engaged in what Pattee calls measuring processes— that is, processes whereby the living system observes its environment—and this creates an epistemological problem: ‘‘We must define an epistemic cut separating the world from the organism or observer’’ (Pattee 1997; italics mine). The concept of semantic information requires the separation of the knower from the known. Pattee emphasizes that he is not suggesting a Cartesian dualism here but only a ‘‘descriptive dualism,’’ for although a measuring process depends on choices which cannot be derived from laws, such choices are seen by Pattee as functions coded in dna and ultimately generated by natural selection. Measurement and observer are epistemic concepts, and as such they presuppose at least an organization that can construct the measuring device and

1 55

Jesper Hoffmeyer

use the results for its survival. Subcellular entities such as enzymes are not sufficiently sophisticated to count as measuring devices: ‘‘To qualify as a measuring device it must have a function,’’ Pattee writes, ‘‘and the most primitive concept of function implies improving fitness of an organism. Thus, observation and measurement require an organization that (1) constructs the measuring device and (2) uses the results of the measurements for survival. This requirement I have called the semantic closure principle. This provides an objective criterion for distinguishing measurements and observations from other physical interactions. Only organizations with this semantic closure property should be called observers. The cell is the simplest natural case of an observing system’’ (Pattee, in press). A major difficulty with this ‘‘epistemic cut position’’ is that it makes it very difficult to see how life could ever have evolved at all. How did physical constraints become semiotic controls (Hoffmeyer 2001a)? I think the central dogma is to blame for the whole epistemological misery. The essential idea of ‘‘something’’ which is passed on from dna to protein is a misunderstanding and a false reification. Nothing is passed on, I suggest. Instead the cell interprets the dna as a set of instructions. Note that this explanation not only gets rid of the mysterious ‘‘something/information’’ package, it also immediately changes ‘‘the burden of agency’’ from the passive, hermetically closed dna molecule to the very active biochemical machinery of the cell. Obviously the cell is the ‘‘doer,’’ not the dna. But what do we mean by ‘‘cellular interpretation’’? Answering this question brings us to the core of the new field of biosemiotics, which will be presented in the next section; but a preliminary answer might refer to the formation of an interpretant in the form of the specific activity of that complex of finely regulated and membrane-associated enzymatic reactions which collectively are known as transcription and translation. While these processes per se remain firmly inside Pattee’s ‘‘dynamic mode,’’ their historically appropriated functional organization reflects the semiotic dynamics of the organism-environment interface (Hoffmeyer 1998b). The regulated transcriptional and translational enzyme systems are tuned to the semiotic history of the organism. Accordingly I have suggested that the ‘‘epistemic cut’’ may be transcended through the application of the semiotic distinction between different sign categories; that is, iconic, indexical, and symbolic signs (Hoffmeyer 2000b).

156

Genes, Development, and Semiosis

Biosemiotics According to biosemiotics, all processes that take place in animate nature —at whatever level, from the single cell to the ecosystem—should be analyzed and conceptualized in terms of their character as sign processes (Hoffmeyer 1998a). Biosemiotics does not contradict well-established physical and chemical laws; it simply claims that life processes are part of—and are organized in obedience to—a semiotic dynamic. Biosemiotics, however, is still in a rather vague state, and diverse interests and viewpoints have come to express themselves under its umbrella, as witnessed by the collection of articles edited by Sebeok and Umiker-Sebeok (Sebeok and Umiker-Sebeok 1992) and a recent special issue of Semiotica devoted to biosemiotics (Hoffmeyer and Emmeche 1999; Sebeok 1999). As linguistic animals, human beings are inescapably suspended from early childhood in a world of signification. But linguistic signification is just one very peculiar and highly developed form of signification. Although much of semiotics, following the paths laid down by the Swiss linguist Ferdinand de Saussure at the beginning of the twentieth century, has traditionally been exclusively concerned with the sign processes of language, modern semiotics has widened its scope to embrace the much broader understanding of semiosis originally suggested by the American scientist and philosopher Charles Sanders Peirce (1839–1914). The existence of signification as a dimension of the human condition challenges biological theory to uncover the roots for this phenomenon in prehuman nature. This challenge has often been met by a reductive strategy, in which semiotic processes are identified as signals unambiguously releasing well-defined effects. While such a strategy may help in affirming deep-rooted beliefs in the omnipotence of traditional explanatory strategies of science, it does so only at the cost of turning our subjective experience of what it feels like to be conscious into an illusion. Worse yet, it leaves us with a mind-body dualism, which from an intellectual point of view, at least, cannot decently be believed. As the philosopher John Searle points out, the experience of being a first-person singular cannot possibly be explained through third-person discourse (Searle 1992); or to state it a little more pointedly: an ‘‘I’’ cannot be derived from ‘‘its,’’ no matter how complicated the patterns of ‘‘its’’ one constructs. Biosemiotics can be seen as an alternative to this approach. It accepts that signification in the true sense of this word did not suddenly appear with the

157

Jesper Hoffmeyer

human creature; rather, it has a natural history of its own, beginning with modest unpersonalized forms of signification such as the kind of natural preferences we once called natural laws (e.g., the propensity of massive bodies to approach each other). This was the route to understanding our existence as sentient beings in the universe suggested more than a century ago by Peirce. Peirce was strongly opposed to the ‘‘necessitarians’’ of his time. In 1891 he wrote: Uniformities are precisely the kind of fact that need to be accounted for. That a pitched coin should sometimes turn up heads and sometimes tails calls for no particular explanation; but if it shows heads every time, we wish to know how this result has been brought about. Law is par excellence the thing which wants a reason. Now the only possible way of accounting for the laws of nature and the uniformity in general is to suppose them results of evolution. This supposes them not to be absolute, not to be obeyed precisely. It makes an element of indeterminacy, spontaneity, or absolute chance in nature. (CP 6.12–13)4

Thus, according to Peirce, natural laws cannot explain evolution because they are themselves a product of evolution. And behind our evolving universe we find in his vision one fundamental principle or law, which he sometimes called nature’s tendency to form habits. As this terminology suggests, he saw the human mind and its associative or generalizing power as just one peculiar instantiation of a much broader principle pertaining to our universe as such. Peirce’s ‘‘cosmogonic philosophy’’ appears strangely well suited to cope with many challenges of modern science (Christiansen 1999), but there is no reason for a biosemiotic understanding of life to commit itself to a strict dependence on this particular metaphysical system. The idea of nature’s tendency to form habits is, however, central to biosemiotics. As Peirce himself observed, a habit is the most general form of an interpretant because the formation of a habit implies that an event will (nearly) always provoke the same response, so that the response is therefore not just accidental but must be related to the event. Habit formation is thus the core of semiosis (sign-process). Biosemiotics holds that semiosis is a fundamental aspect of our universe, a principle which, however, manifests itself in its genuine triadic form only through the birth of life. Many scientists may think that these ontological commitments place biosemiotics far away from decent natural science. And yet, recent interpretations of the second law of thermodynamics seem very close to the assumptions of biosemiotics. Thus, according to the maximum entropy production

158

Genes, Development, and Semiosis

principle suggested by Rod Swenson (Swenson 1989), thermodynamic fields will behave in such a fashion as to reach the final state—minimize the field potential or maximize the entropy—at the fastest possible rate permitted by the constraints. This implies that progressive evolutionary ordering entails the production of increasingly higher ordered states. The point is this: every time an ordered state is produced in some system, implying a lowering of the entropy at that location, even more entropy must be produced at other locations in order to obey the thermodynamic dictate that the total entropy in the universe always increases. It thus appears that the best way to produce entropy as fast as possible is by producing order as fast as possible. Or in the words of Swenson and Turvey: ‘‘The world is in the order production business, including the business of producing living things and their perception and action capacities, because order produces entropy faster than disorder’’ (Swenson and Turvey 1991: 345; my italics). Now, as Swenson and Turvey go to great lengths to show, ‘‘perception and action cycles,’’ which are, of course, key processes in the semiotic dynamic of living systems, are even more instrumental than physical systems in furthering entropy production: ‘‘Living things with the capacity to perceive how and where they are moving, and with the coordinate capacity to move in ways that allow them to perceive more and to perceive better, expand the patches of the planet in which energy degradation can take place. In the terms introduced above, they expand the Earth’s dissipative space. Thus the purpose of living things are differentiations or productions, literally higher order symmetry states, of the environment itself towards its own ends’’ (Swenson and Turvey 1991: 345). Biosemiotics suggests that our universe has a built-in tendency (originating in the second law of thermodynamics) to produce organized systems possessing increasingly more semiotic freedom in the sense that the semiotic aspect of the system’s activity becomes more and more autonomous relative to its material basis (Hoffmeyer 1992, 1996). The semiotic dimension of a system is always grounded in the organization of its constituent material components, and cannot exist without this grounding, but evolution has tended to create more and more sophisticated semiotic interactions which are less and less constrained by the laws of the material world from which they are ultimately derived. It seems obvious that science must somehow come to terms with the necessity of understanding our universe in such a way that it is not absurd to claim that this universe has itself produced us, as observed by Ilya Prigogine and Isabelle Stengers (Prigogine and Stengers 1984). Accepting the Peircean

159

Jesper Hoffmeyer

idea that semiosis is a general aspect of our universe—or, in his own words, that nature has a tendency to form habits—seems like a modest step to take in order to produce a description of the world which does not preclude our own existence as real persons, as ‘‘I’’s, within it.

Sign Action ‘‘Every reasoner,’’ Peirce says, ‘‘has some general idea of what good reasoning is. This constitutes a theory of logic’’ (CP 2.186); and since ‘‘all thought [is] being performed by means of signs, logic may be regarded as the science of the general laws of signs’’ (CP 1.191). This is how Peirce comes to his conception of the sign as a triadic unit; that is, an undecomposable unity of three interdependent relations. Peirce thought traditional logic, based on two-factor, dyadic, relations too limited. Bound as this logic is to the single dimension of the linear chain, it cannot be made to branch out. He believed that logical processes ought rather to be regarded as a multidimensional network. Such a network can be arrived at by combining three-factor relations, triads, or signs. So the sign in Peirce’s understanding is a logical node. The following quote captures the essence: ‘‘I define a Sign as a thing which is so determined by something else, called its Object, and so determines an effect upon a person, which effect I call its Interpretant, that the latter is thereby mediately determined by the former. My insertion of ‘upon a person’ is a sop to Cerberus, because I despair of making my own broader conception understood’’ (Peirce 1966, 404). Although Peirce derived his concept of the sign from his understanding of logic, he did not think of signs (or logic) as bound to the mental sphere of humankind, rather—as already noticed—he saw human mind as a particularly highly elaborated instantiation of nature’s general tendency to form habits, or as he sometimes called it, ‘‘the law of mind.’’ Putting the word person into his definition thus was ‘‘a sop to Cerberus,’’ a limitation of the concept’s true reach which he conceded to avoid turning people against the general idea. As a proponent of a biosemiotic understanding I am obviously not in a position to throw the same sop to Cerberus, however much it might still be needed. Instead I endorse the Brazilian Peirce scholar Lucia Santaella’s interpretation of the Peircean sign: ‘‘Since the three elements, sign, object, and interpretant, by themselves, or better, by their existential nature, may belong to various orders of reality as single objects, general classes, fictions, mental representations, physical impulses, human actions, organic activities,

160

Genes, Development, and Semiosis

Object

Sign

y-chromosome

a

b

Embryonic processes

Interpretant

Unknown signal

Maleness

tdf-gene

c

Binding of transcription complex to tdf-gene

Activated tdf-gene

tdf protein

d

tdf-gene expression in sex cord cells

High level of tdf protein

Male sex determination

e

Further embryonic processes

1 • a. Graphic representation of the Peircean conception of a sign; b. the human Y-chromosome sign relations; c-e. web of semiotic relations including the tdf gene expression in epithelial sex cord cells.

or natural laws, what constitutes the sign relation in its logical form is the particular way in which this triad is bound together’’ (Santaella-Braga 1999: 514–15). As illustration let us consider the human Y chromosome. Analyzed in an overall process view, this chromosome may be seen as a sign for maleness. Growing human embryos encountering the Y chromosome in their genomic setup normally ‘‘know’’ what to do about it, namely construct a biologically male baby. Graphically we can depict this as a Peircean triadic sign process (fig. 1a) in which the Y chromosome occupies the position of the primary sign, maleness stands in the position of the object, and the embryonic reading of the chromosome is the interpretant (fig.1b). Looked at in more detail, what happens is that around the seventh week of development certain embryonic cells called epithelial sex cord cells begin for unknown reasons to express a gene (termed tdf ) located on the Y chromosome. This results in the production of the so-called testis determining factor (tdf) (Gilbert 2000). From there on, apparently, the male sex determination process is taken care of by these transformed cells. What we see is that the organism acquires its male determination through

161

Jesper Hoffmeyer

a series of steps whereby semiotically competent cells ‘‘read the messages’’ made available to them in part from their internal genetic makeup and in part from the external biochemical context set by a multitude of cues (molecular signs) derived from neighboring cells or from other embryonic tissues. These contextual cues are received at specific receptors located in the plasma membrane. Stanley Salthe uses the term system of interpretance to indicate the integrated character of this web of often poorly understood interpretative mechanisms involved in even the simplest cases (Salthe 1993). Even though all these processes may sooner or later be fully characterized at the biochemical level, this will not by itself exhaust our need for explanation. For, obviously, we are dealing here not with a haphazard mixture of biochemical processes but with a delicately organized system. What we really want to know is the logic of the organization of these biochemical processes, and this logic has to do with their developmental function. Figure 1c–e shows three chained semiotic steps in the matrix of male sex determination. First, an unknown biochemical event, perhaps the activation of a specific inducer, refers the epithelial sex cord cells to the tdf gene (the object) by way of the binding of a transcription complex to the gene. Here the binding process occupies the position of an interpretant to the signal (c). This activates tdf to guide the formation of a new interpretant in the form of tdf gene expression, whereby testis determining factor (tdf) is produced (d). With increasing concentrations of tdf a level is reached such that further steps of embryonic processes are released in the chain of male sexdetermining semiosis (e). In each step the interpretant becomes a sign in yet another semiotic triad. Basically the triadic nature of this functional logic derives from the fact that the expression of the tdf gene is undertaken only because some unknown signal provokes the cellular machinery to relate to the tdf genes in a way which represents the historical relation of the unknown signal to those same tdf genes. The sign points out a historically created relational logic of the macromolecular events. History is the essence of the semiotic analysis. It should be noticed that semiosis as described here is actually sign action (Deely 1990). Scientists are accustomed to talking about actions in a dyadic sense of the word, as dynamic interaction or (coercive) ‘‘brute force’’ relating cause and effect. Action in this sense is also always involved as a central element in the actions of signs (while reading this, for instance, numerous neural actions are performed in your brain, or so I hope), but the particular effect of sign action is the pointing out of an interpretant; that is, signs do not just

162

Genes, Development, and Semiosis

act, they point out actions which are related to something else (the object) in ways which are inherent in the system of interpretation. The Peircean sign, in other words, unites the formal and the functional aspects of action; it is both logical node (i.e., a triadic relation) and material process (i.e., semiosis). And for this reason the Peircean sign cannot be decomposed into three dyads. The positions in the triad are not logically equivalent because the interpretant is related both to the sign and to the object, but in such a way that the relation to the object is grounded in the sign’s relation to that same object.5 Biosemiotics insists on a view of process in which agency cannot be left out of analysis. Without denying the physical and chemical nature of biological processes biosemiotics claims that these processes, when occurring in living systems, are always organized into functional systems with an agential character, and that neither development nor evolution can be adequately accounted for as long as this aspect is not integrated in the conceptual structure of our explanations.6 The developmental process depends on the proper function of a highly tuned set of semiotic checkpoints; that is, cellular bifurcations controlled by recognition processes whereby tissues and organs communicate in time and space. Reducing such semiotic checkpoints to dyadic relations of cause and effect not only complicates matters unnecessarily (since the organizational logic is triadic, not dyadic), but also tends to blind us to the contextual character of the recognition processes. How cells respond to given signals often does not depend on the signal itself or on the cell, but on cellular history as well as the whole communicational setting into which the signal becomes situated. The proper level of description therefore is the triadic level of sign processes rather than the dyadic level of cause and effect.

The Extended Membrane The semiotics of dna is inherently tied to the semiotics of the organismenvironment interface. The surface of an organism (or an embryo) is the locus where it meets its environment and through which pass the cues that must be correctly interpreted if the organism is to enjoy life and survive. I use the phrase ‘‘enjoy life’’ deliberately because I want to stress that, apart from highly peculiar human animals, survival does not normally enter the perceptual (or cognitive) universe of organisms, their Umwelten.7 Organisms may feel fear, they may flee, they may search for food, but they never try to survive; that

163

Jesper Hoffmeyer

is just the eventual outcome of their behavior. In other words, ‘‘survival’’ is a descriptive category in evolutionary theory and should never be ascribed to actual behavior, whether genetic or organismic. For the organism the environment is that ‘‘reality’’ which it senses through the respective receptors situated at its surface or projected out into the world around it. The environment, in other words, is a kind of virtual reality or model of the world. As biologists we can try to unravel the characteristics of this virtual reality and eventually to understand its evolutionary role as a tool for survival. As such, the virtual reality, the Umwelt, has itself become molded by the evolutionary process tuning it to the motoric means of the organism as well as to its physiological and morphological needs (Hoffmeyer 2001b). If we move inside the surface of a multicellular organism we run into new surfaces, enveloping tissues, organs, or single cells. Here again we meet virtual worlds constructed by the macromolecular machinery of the cells in order to cope with cellular environments. The cell membranes act as interfaces through which signs from the outside are converted into cascading processes on the inside. Moreover, in animals, specialized cells from the immune system are relentlessly patrolling the body fluids. Located in the cell membranes of each of these cells are millions of specific receptor molecules ready to grasp whatever molecular sign might travel their way and prepare the cell for its changing situations in the milieu interieur of the body. Should we move yet another step inward, through the cell membrane, we immediately run into new membranes belonging to cellular organelles like mitochondria, chloroplasts, lysozomes, endoplasmic reticulum, Golgi apparatus, and so on. The endoplasmic reticulum is composed of one continuous membrane which ramifies throughout most of the cell. More significant in our context is that the space enclosed by the endoplasmic reticulum—as well as that enclosed by the Golgi apparatus, lysozomes, and transport vesicles—is topologically the same as the extracellular space (Barrit 1996). To these structural elements of the cell can be added the multitude of filaments forming the cytoskeleton, microtubules, intermediate filaments, and actin filaments. It has been suggested that an even finer network of filaments, the microtrabecular lattice, penetrates the residual space between the intracellular membranes. As a consequence of this, the topological ordering of biochemical processes becomes of key importance for understanding cellular activity. Cellular membranes never form de novo by self-assembly of their constituents; they always grow, in an essentially homomorphic fashion, by accretion, that is, by the insertion of additional constituents into preexisting membranes. The corresponding patterns are transmitted from generation to gen-

164

Genes, Development, and Semiosis

eration by way of the cytoplasm (e.g., of egg cells) which contains samples of the different kinds of cytomembranes found in the organism (de Duve 1991). The ordinary textbook talk of dna as governing cellular or even organismic behavior is therefore rather misleading. In fact, if any entity should be thought of as a governor of cellular activity it is the membrane. dna contains the recipes for constructing the one-dimensional amino acid chains which form the backbones of enzymes, including the enzymes needed for catalyzing the formation of the constituents of lipid bilayers and assembling them. But whether these recipes are actually ‘‘read’’ and executed by cellular effectors depends on membrane-bound activity. All major activities of cells are topologically connected to membranes. In the prokaryotes (bacteria) the plasma membrane (the active membrane inside the cell wall) is itself in charge of molecular and ionic transport, biosynthetic translocations (of proteins, glycosides, etc.), assembly of lipids, communication (via receptors), electron transport and coupled phosphorylation, photoreduction photophosphorylation, and anchoring of the chromosome (replication) (de Duve 1991). In eukaryotic cells, the execution of these are topologically connected to the membranes of mitochondria, chloroplasts, the nuclear envelope, the Golgi apparatus, ribosomes, lysozomes, and so on. Many—if not all—of these membranes are themselves descendants of once free-living prokaryotic membranes which perhaps a billion years ago became integrated into that cooperative or symbiotic complex of prokaryotic membranes which is the eukaryotic cell. Membranes are also the primary organizers of multicellular life. The topological specifications necessary for the growth and development of a multicellular organism cannot be derived from the dna for the good reason that the dna cannot ‘‘know’’ where in the organism it is located. Such ‘‘knowledge’’ has to be furnished through the communicative surfaces of the cells. Morphogenesis is mostly a result of local cell-cell interactions in which signaling molecules from one cell affect neighboring cells. Animal cells, for instance, are constantly exploring their environments by means of tiny cytoplasmic feelers called filopedia (filamentous feet) that extend out from the cell. ‘‘These cytoplasmic extensions that drive cell movement and exploration are expressions of the dynamic activity of the cytoskeleton with its microfilaments and microtubules that are constantly forming and collapsing (polymerizing and depolymerizing), contracting and expanding under the action of calcium and stress,’’ explains Brian Goodwin (Goodwin 1995: 36). But not only are membranes involved in all the organized activities of the life sphere, the membrane can actually be seen as the principal locus of life

165

Jesper Hoffmeyer

itself (Hoffmeyer 1998b, 2000b).8 It is the membrane that creates the potential inside-outside asymmetry from which the organism-environment asymmetry must have grown. The origin of life is by necessity also the origin of the environment, and lack of concern for this aspect of the origin problem has seriously hampered much theorizing on prebiotic evolution. Somehow the world became divided into organism and environment, and the formation of a closed membrane must have been part of this process. Here the membrane not only ensures the necessary topological closure, but, more significant, it takes on the role of an interface facilitating a flow of messages between its interior and exterior domains. Considered from the point of view of the membrane, prebiotic evolution is essentially a process of ‘‘interiorization’’ (Hoffmeyer 2000b). Prebiotic membranes colonized the interior space and thereby scaffolded themselves through the formation of a multitude of autocatalytic metabolic loops and finally of replicative molecules mapping constituents of the internal autocatalytic system. Thus persistent architectures appeared as entities engaged in the trick of conjuring up a virtual reality on the inside for the purpose of coping effectively with the outside. Against the background of this discussion it might be fruitful to introduce the term extended membrane as the inner locus for life. The extended membrane comprises all of the membranes that make up an organism (including its skin, plumage, etc.) and is responsible for the actual execution of life as process, semiotic agency. It is the extended membrane that directs ontogeny in a self-organized process scaffolded by an internal system of ‘‘labels,’’ genes, kept orderly in the genome.

Code-Duality Claus Emmeche and I have suggested the term code-duality to encompass the overall semiotic logic of reproduction and ontogenesis (Hoffmeyer and Emmeche 1991). Here the term code is used simply in the sense of a set of signs acting in concerted ways in a given context. Thus the genome may be seen as a digital code embracing the totality of dna-based—and thus sequential— signs, whereas most other sign processes in the natural world are analogically coded; that is, their codification relies on the principle of likeness, such as the key-and-lock principle involved in the passage of messages across cell membranes. Analog coding necessarily implies spatial and temporal continuity. Whereas the notion of the genetic code may have paved the way for our

166

Genes, Development, and Semiosis

vision of the genome as a digitally coded string of labels—an inventory control system, so to say, in which single genes operate like menu topics whose activation occasions the delivery of specific protein resources—the idea of body dynamics as dependent on analogically coded sign systems may perhaps seem less obvious. And certainly Pattee’s distinction between a linguistic and a dynamic mode bears this out. It is, however, important to stress the interdependence of the analog and the digital as two equally necessary forms of referential activity arising like twins in the individuation of that logic which we call life. The digital code is the seat of self-referential activity—the redescription in a sequential alphabet of all the macromolecular constituents of the organism—whereas the analog codes are engaged in non-selfreferential activity, that is, the semiotic looping of organism and environment into each other through the activity of their interface, the closed membrane.9 To claim that only the digital twin is semiotic whereas the analog twin remains in the sphere of classical dynamics is to block the only possibility for transcending Pattee’s semantic cut position. Having emphasized above the dynamic role of the extended membrane (i.e., the analog coded system), let me here clarify the distinctive role of the genetic or digital code system. As Emmeche and I have pointed out (Hoffmeyer and Emmeche 1991), digital codes have at least three characteristic advantages that make them the obvious instruments for life’s self-referential tasks. These three characteristics are freedom from the constraints of nature, objectivity and temporality, and abstraction.

Freedom from the Constraints of Nature Digital codes allow for impossible messages because there is no strict binding between the code itself and the message it carries. In linguistic messages anything goes; Socrates may have lunch with Meryl Streep and the wives of pilots may give birth to children with wings. The same is true of genomes. Impossible genetic instructions are created all the time through processes of genetic recombination such as crossing over, resulting in early abortions or the birth of nonviable descendants. The incredible combinatorial capacity of living systems for creating endless chains of novelties is the result of this freedom.

Objectivity and Temporality Digital codes are codes for memory. We know Socrates’ dialogues today only because Plato wrote them down. Had the dialogues been coded only in pieces

167

Jesper Hoffmeyer

of mime, they would probably have died with Socrates himself or would have survived a few generations at most. The key here is that digital codes are objective in the sense that they depend on a shared convention. Anthony Wilden points out the following distinction between analog and digital codes: ‘‘a digital code is ‘outside’ the sender and receiver and mediates their relationship; an analog code is the relationship which mediates them’’ (Wilden 1980: 173). Genomes are not normally thought to be outside the sender or the receiver, because they are normally exchanged via processes of mitosis or meiosis. But if we see life, as we try to do here, as unfolding across membranes, then even the inside of the membrane is in a way an outside. The important thing in this context, however, is that dna is actually protected from the vicissitudes of life by its relative chemical inertness and by a sophisticated apparatus of enzymatic repair systems. In many sexually reproductive species, furthermore, the germ line is kept separate from the somatic cell lines so that no Lamarckian inheritance is believed normally to take place. Digital codes thus are necessary to ensure the temporal semistability needed for evolution— nothing can evolve if it is not remembered (because then we talk about substitution).

Abstraction Digital codes are eminent tools for the construction of meta-messages; that is, messages necessary for interpreting other messages. Gregory Bateson has shown that meta-messages may also be communicated in the analog. When he observed young monkeys engaged in so-called play, an activity in which they exchanged signals similar to those seen during combat, Bateson observed that when the monkeys snapped at one another while creating an imaginary combat situation, the snap actually signified the following metamessage: ‘‘this is not a bite’’ (Bateson 1972: 177–193). The absence of a bite, in other words, is announced by the presence of the snap. The snap is an indication of something that is not there. But Bateson also comments that this is probably as far as an analogically coded communication can go in the direction of the abstract category of ‘‘not.’’ For real abstractions to take place, digital codes are needed. We still do not know the full syntactic structure of the genetic code, but regulatory genes are examples of meta-messages, and the occurrence of atavisms such as the three-toed horse seems to indicate that exclusion is not the only way to get rid of outmoded ontogenetic instructions. Negation may suffice. Abstraction thus furnishes plasticity in the

168

Genes, Development, and Semiosis

absence of which the evolutionary process might perhaps not have been as rich as it actually is. These three essential aspects of digital codes not only make these codes indispensable tools for the evolutionary process, they also explain why digital codes are fundamentally passive. We do not believe in spells because there is no consistent way that the mere pronunciation of words could cause desired physical events to take place, and likewise we should not believe that genes by themselves do anything. In both cases it takes an interpretant to mediate between the message and the active world; in both cases large amounts of ‘‘tacit knowledge’’ (in the sense of Polanyi 1967) are required by the system if the digital message is to be of any use (and the never-failing availability of this ‘‘tacit knowledge’’ was of course taken for granted when the message was first coined in dna). The invention of ‘‘digitality,’’ I have suggested, was the step which some four billion years ago allowed certain swarms of communicating closed membrane systems floating in the prebiotic mud to escape the indifference of the mere moment and to enter a temporal world of genuine selfhood (Hoffmeyer 2001a).

Demystifying Genetic Information Current theories need ‘‘information flow’’ to re-create the next generation, writes Susan Oyama, ‘‘because we have been taught to see phenotypes themselves as evolutionary dead-ends, that information must pass by means of the germ cells, diminutive reproductive lifeboats, that must, if they can, abandon that doomed body before it goes down’’ (Oyama 2003). What really went wrong with the Weismannian separation of the germ line from the somatic line was not, however, so much the separation itself as the implied reification of the digital code.10 Later generations of biologists repeated this fundamental reification when they saw chromosomes, genotypes, dna segments, or so-called replicators as exerting causal power over heredity. The inherent preformationism of this view requires a causal power to reside in the replicator/genome, and this is where the metaphor of ‘‘information flow’’ enters the scheme. The preformationist assumption needs a causal agent to account for the ontogenetic miracle, and since this agent obviously cannot be the gene as a substantial structure, the agent must reside in the gene in a hidden form, that is, as information. By postulating that ontogeny is then caused by

169

Jesper Hoffmeyer

a transport process whereby this information ‘‘flows’’ from the genes to the unfolding body, the mystification is consummated. The inbuilt inconsistencies of the information concept as this concept is presently used in molecular biology have been thoroughly analyzed by Sahotra Sarkar (Sarkar 1996, 1997).11 As we can now see, Weismannism did in fact have a healthy core, namely code-duality: the presence inside living entities of a separate, fairly wellprotected, digitally organized system of signs directed specifically toward the need to control the production of the constituents of the entity itself. The appearance of multicellularity implied that no single cell would any longer need the whole self-referential system; rather, differentiation implied a tissuespecific inactivation of parts of the system (e.g., via chromatin modification such as methylation). The Weismannian separation of a specific totipotent germ line at an early embryonic stage was one of several solutions to this problem and became adopted by the species of most animal taxa.12 While this fact supports the anti-Lamarckian intuition of Weismannism, it does not by itself imply that the organism ends up as a dead end in the evolutionary process, although that was what Weismannism came to suppose. Code-duality implies no such thing, because it maintains that the ontogenetic process is the result of the interpretive activity of embryonic or somatic cells or of semiotically integrated living systems. Thus the deterministic ascription of causative power to the genome depends on an unjustified reification of this code as a replicator unit whereby its causative power is transformed from the indeterminate semiotic domain to the determinate biochemical domain. I therefore suggest that the existence of a digitally coded and relatively independent system for self-reference should be recognized as a very significant property of living systems. The role of dna is not to govern the construction or activity of living systems, but it certainly does play a very special role in the semiotics of life. By seeing dna as a truly unique but semiotic resource for the ontogenetic process it becomes possible to balance its uniqueness against its dependence on the semiotic competence of the extended membrane and thereby its dependence on organism-environment interaction.

Notes 1 Semiotic emergence is the emergence through situated semiotic activity of patterns of activity whose high-level structure cannot be reduced to specific sequences of low-level activity.

170

Genes, Development, and Semiosis 2 Parity in the sense of a two-way flow of information requires the introduction of longer time scales (life cycles) and more complex levels of organismenvironment interaction. 3 By ‘‘nondynamic’’ I mean time independent. 4 References to the work of Peirce are from The Collected Papers of Charles S. Peirce (1931–1935), abbreviated CP, followed by volume number and section number. 5 The interpretant belongs to Peirce’s category of Thirdness, whereas the sign is Firstness, and the object Secondness. 6 By ‘‘agential’’ I mean systems exhibiting agency in the sense that the activities of the systems are tied to internal settings reflecting history of survival. The origin of agency is analyzed in Hoffmeyer 1998b. 7 The term Umwelt was introduced by the German biologist Jakob von Uexküll to denote the fundamentally subjective character of the animal’s own perceptual world (Uexküll 1928). 8 The total area of membranes in the human body (including subcellular membranes) is of the order of one or several square kilometers (Hoffmeyer 2000a). Most of a person’s metabolism at rest is spent on upholding these inner surface areas. 9 Here nonself should be understood as a hierarchical concept; that is, at the cellular level other cells are ‘‘nonself ’’ whereas at the organismic level other organisms are ‘‘nonself.’’ 10 As shown by Leo Buss, strict Weismannian separation does not in fact occur in most taxonomic groups (Buss 1987). 11 Sarkar concludes that the information concept has not been of any help to molecular biology and that there is in fact no need to transcend the wellestablished conceptual system of biological chemistry. I draw the opposite conclusion and present an explicit semiotic understanding by which so-called information is understood as signs or sets of signs, encoded messages. 12 That it was animals and not plants or fungi that adopted early segregation as a strategy for ensuring the interests of the multicellular system toward the potentially conflicting interests of somatic cell lines may, according to Buss (1987), have been due to the absence of rigid cell walls in animal cells, which makes these cells mobile and thus capable of invading favorable positions in the embryo. Early segregation could not, of course, prevent mutant cell lines from exploiting this opportunity, but by monopolizing totipotence the eventual success of mutant cell lines in thus getting access to next generation would be eliminated.

171

Jesper Hoffmeyer

References Barrit, G. J. 1996. Communication within Animal Cells. Oxford: Oxford University Press. Bateson, G. 1972. Steps to an Ecology of Mind. New York: Ballantine Books. Buss, L. 1987. The Evolution of Individuality. Princeton: Princeton University Press. Christiansen, P. V. 1999. Downward causation: from macro- to micro-levels in physics. In: P. B. Andersen, N. O. Finnemann, and C. Emmeche (eds.), Downward Causation (pp. 51–62). Aarhus: Aarhus University Press. Clark, A. 1997. Being There: Putting Brain, Body, and World Together Again. Cambridge: mit Press. Deacon, T. 1997. The Symbolic Species. New York: W. W. Norton. Deely, J. 1990. Basics of Semiotics. Bloomington: Indiana University Press. Depew, D. L., and Weber, B. H. 1995. Darwinism Evolving: Systems Dynamics and the Genealogy of Natural Selection. Cambridge: mit Press. Duve, C. de. 1991. Blueprint for a Cell: The Nature and Origin of Life. Burlington, N.C.: Neil Patterson. Gilbert, S. F. 2000. Developmental Biology. 6th ed. Sunderland, Mass.: Sinauer. Goodwin, B. 1995. How the Leopard Changed Its Spots. London: Phoenix Giants. Griffiths, P. E., and Gray, R. D. 1994. Developmental systems and evolutionary explanations. J. Philos. 91: 277–304. Hacking, I. 1990. The Taming of Chance. Cambridge: Cambridge University Press. Hendriks-Jansen, H. 1996. Catching Ourselves in the Act: Situated Activity, Interactive Emergence, and Human Thought. Cambridge: mit Press. Hoffmeyer, J. 1992. Some semiotic aspects of the psycho-physical relation: the endo-exosemiotic boundary. In: T. A. Sebeok and J. Umiker-Sebeok (eds.), Biosemiotics: The Semiotic Web 1991 (pp. 101–123). Berlin: Mouton de Gruyter. Hoffmeyer, J. 1996. Signs of Meaning in the Universe. Advances in Semiotics. Bloomington: Indiana University Press. Hoffmeyer, J. 1997. Semiotic emergence. Rev. Pensée d’aujourd’hui 25–7(6): 105– 117 (in Japanese). Hoffmeyer, J. 1998a. Biosemiotics. In: P. Bouissac (ed.), Encyclopedia of Semiotics (pp. 82–84). New York: Oxford University Press. Hoffmeyer, J. 1998b. Surfaces inside surfaces: on the origin of agency and life. Cybernet. Hum. Knowing 5(1): 33–42. Hoffmeyer, J. 2000a. The biology of signification. Perspect. Biol. Med. 43(2): 252– 268. Hoffmeyer, J. 2000b. Code-duality and the epistemic cut. In: J. L. R. Chandler and G. Van de Vijver (eds.), Closure: Emergent Organizations and Their Dynamics (pp. 175–186). New York: New York Academy of Science.

172

Genes, Development, and Semiosis

Hoffmeyer, J. 2001a. Life and reference. In: L. Rocha (ed.), The Physics and Evolution of Symbols and Codes: Reflections on the Work of Howard Pattee. BioSystems 60 (1–3): 123–130. Hoffmeyer, J. 2001b. Seeing virtuality in nature. Semiotica 134, special issue: Jakob von Uexküll: A Paradigm for Biology and Semiotics, ed. Kalevi Kull): 1/4. Hoffmeyer, J., and Emmeche, C. 1991. Code-duality and the semiotics of nature. In: M. Anderson and F. Merrell (eds.), On Semiotic Modeling (pp. 117–166). New York: Mouton de Gruyter. Hoffmeyer, J., and Emmeche, C. 1999. Biosemiotica 2. Semiotica 127(1–4): 133– 695. Kauffman, S. A. 1993. Origins of Order: Self-Organization and Selection in Evolution. New York: Oxford University Press. Kauffman, S. 1995. At Home in the Universe. New York: Oxford University Press. Lakoff, G., and Johnson, M. 1999. Philosophy in the Flesh. New York: Basic Books. Morowitz, H. 1992. Beginnings of Cellular Life: Metabolism Recapitulates Biogenesis. New Haven: Yale University Press. Oyama, S. 1985. The Ontogeny of Information. Cambridge: Cambridge University Press. Oyama, S. (2003). On having a hammer. In: B. Weber and D. Depew (eds.), Learning, Meaning, and Emergence: Possible Baldwinian Mechanisms in the Coevolution of Language and Mind (pp. 169–191). Cambridge: mit Press. Pattee, H. H. 1973. Discrete and continuous processes in computers and brains. In: W. Guttinger and M. Conrad (eds.), The Physics and the Mathematics of the Nervous System (pp. 128–148). New York: Springer Verlag. Pattee, H. H. 1977. Dynamic and linguistic modes of complex systems. Int. J. Gen. Syst. 3: 259–266. Pattee, H. H. 1982. Cell psychology: an evolutionary approach to the symbolmatter problem. Cognit. Brain Theor. 5(4): 325–341. Pattee, H. H. In press. The problem of observables in models of biological organizations. In: E. L. Khalil and K. E. Boulding (eds.), Evolution, Order, and Complexity. London: Routledge. Pattee, H. H. 1997. The physics of symbols and the evolution of semiotic controls. Presented at Workshop on Control Mechanisms for Complex Systems, Las Cruces, N.M., 8–12 December. In: Santa Fe Institute Studies in the Sciences of Complexity, Proceedings Volume. Pattee, H. H. 2000. Personal communication. See www.ws.binghamton.edu/ pattee/semiotic.html. Peirce, C. S. 1908. Letter to Lady Welby. In: P. P. Wiener (ed.), Charles S. Peirce: Selected Writings, chap. 24. Reprint, New York: Dover Publications, 1966. Peirce, C. S. 1931–1935. Collected Papers of Charles Sanders Peirce, vols. 1–6. C. Hartstone and P. Weiss (eds.). Cambridge, Mass.: Harvard University Press. Polanyi, M. 1958. Personal Knowledge. London: Routledge.

173

Jesper Hoffmeyer

Polanyi, M. 1967. The Tacit Dimension. London: Routledge. Popper, K. 1990. A World of Propensities. Bristol: Thoemmes Antiquarian Books. Prigogine, I., and Stengers, I. 1984. Order out of Chaos. London: Heinemann. Salthe, S. 1993. Development and Evolution: Complexity and Change in Biology. Cambridge: mit Press. Santaella-Braga, L. 1999. A new causality for the understanding of the living. Biosemiotics. Semiotica, special issue, 126: 497–519. Sarkar, S. 1996. Biological information: a skeptical look at some central dogmas of molecular biology. In S. Sarkar (ed.), The Philosophy and History of Molecular Biology: New Perspectives (pp. 187–231). Dordrecht: Kluwer. Sarkar, S. 1997. Decoding ‘coding’: information and dna. Eur. J. Semiot. Stud. 9 (2): 227–232. Searle, J. R. 1992. The Rediscovery of Mind: Representation and Mind. Cambridge: mit Press. Sebeok, T. A. 1999. Biosemiotica 1. Semiotica 127: 5–131. Sebeok, T. A., and Umiker-Sebeok, J. 1992. Biosemiotics: The Semiotic Web 1991. Berlin: Mouton de Gruyter. Sterelny, K., and Griffiths, P. E. 1999. Sex and Death: An Introduction to Philosophy of Biology. Chicago: University of Chicago Press. Swenson, R. 1989. Emergent attractors and the law of maximum entropy production. Syst. Res. 6: 187–197. Swenson, R., and Turvey, M. T. 1991. Thermodynamic reasons for perceptionaction cycles. Ecol. Psychol. 3(4): 317–348. Uexküll, J. v. 1928. Theoretische Biologie. Berlin: Julius Springer Verlag. Ulanowicz, R. E. 1997. Ecology, the Ascendent Perspective: Complexity in Ecological Systems. New York: Columbia University Press. Van Gelder, T., and Port, R. 1995. It’s about time: overview of the dynamical approach to cognition. In: R. Port and T. Van Gelder (eds.), Mind or Motion: Explorations in the Dynamics of Cognition (pp. 1–43). Cambridge: mit Press. von Neumann, J. 1966. Theory of Self-Reproducing Automata. Urbana: University of Illinois Press. Weber, B. 1998. Emergence of life and biological selection from the perspective of complex systems dynamics. In: G. Van de Vijver, S. Salthe, and M. Delpos (eds.), Evolutionary Systems: Biological and Epistemological Perspectives on Selection and Self-Organization (pp. 59–66). Dordrecht: Kluwer. Weber, B. H., Depew, D. J., Dyke, C., et al. 1989. Evolution in thermodynamic perspective: an ecological approach. Biol. Philos. 4: 373–405. Wiener, N. 1962. Cybernetics: Or Control and Communication in the Animal and the Machine. Cambridge: mit Press. Wilden, A. 1980. System and Structure. New York: Tavistock.

1 74

7 T H E F E A R L E S S VA M P I R E C O N S E R VATO R Philip Kitcher, Genetic Determinism, and the Informational Gene

paul e. griffiths Genetic determinism is the idea that significant human characteristics are strongly linked to the presence of certain genes—that it is extremely difficult, for example, to attempt to modify criminal behavior or obesity or alcoholism by any means other than genetic manipulation. Recent discussion of human cloning has revealed how real a possibility genetic determinism seems to many people. Surveying this discussion, the eminent developmental biologist Lewis Wolpert was amused to see so many ‘‘moralists who denied that genes have an important effect on behavior now saying that a cloned individual’s behavior will be entirely determined by their genetic make-up’’ (Wolpert 1998: 18). His observation is accurate, and the vehemence of many attacks on behavioral genetics probably reflects an underlying belief that if genes affect behavior at all, then they must determine it. In fact, it is now well known that genes are very unlikely to be deterministic causes of behavior. But if genetic determinism is unlikely to be true, why are we as a community so afraid of it? Wolpert seems to think that moral and political commentators on biology are simply ignorant, but the facts of which they are supposedly ignorant have been widely available for a very long time. Perhaps there is more to the strange persistence of genetic determinism. In a famous analogy the psychologist Susan Oyama compares arguing against genetic determinism to battling the undead: ‘‘But wait,’’ the exasperated reader cries, ‘‘everyone nowadays knows that development is a matter of interaction. You’re beating a dead horse.’’ I reply, ‘‘I would like nothing better than to stop beating him, but every time I think I am free of him he kicks me and does rude things to the intel-

Paul E. Griffiths

lectual and political environment. He seems to be a phantom horse with a thousand incarnations, and he gets more subtle each time around. . . . What we need here, to switch metaphors in midstream, is the stake-in-the-heart move, and the heart is the notion that some influences are more equal than others, that form, or its modern agent, information, exists before the interactions in which it appears and must be transmitted to the organism either through the genes or by the environment.’’ (Oyama 1985: 26–27)

Oyama suggests that genetic determinism is inherent in the way we currently represent genes and what genes do.1 As long as genes are represented as containing information about how the organism will develop, they will continue to be regarded as determining causes no matter how much evidence exists to the contrary. Proof that developmental information is not localized in the genes is the ‘‘stake in the heart’’ that will lay the vampire of genetic determinism to rest. In a recent paper Philip Kitcher strongly disputes Oyama’s diagnosis and Richard Lewontin’s related call for a ‘‘dialectical biology,’’ arguing that the conventional ‘‘interactionist’’ perspective on development is the correct framework for understanding the role of the genes in development. The persistence of genetic determinism, Kitcher argues, is not caused by any conceptual problem in current representations of genetic causation, but by two much simpler facts: the universal human preference for simple explanations over complex ones and the sheer difficulty of communicating complex science to a wider audience (Kitcher 2001). Kitcher agrees that the widespread acceptance of genetic determinism reflects ‘‘the tendency to draw certain kinds of pictures on the basis of woefully inadequate evidence’’ (p. 399) and that the resulting misunderstandings of genetic causation are likely to have a deleterious effect on public policy decisions (pp. 409–411). He understands the motivation that leads Oyama, Lewontin, and others to call for a wholly new approach to understanding developmental causation: ‘‘It is small wonder then, that people appalled by the sloppy thinking . . . yearn for the ‘stake in the heart’ move’’ (p. 396). Kitcher fears, however, that calls for a radical new approach may serve to entrench genetic determinism rather than help to displace it. Calls for a radical new biology will merely alienate working scientists from practical efforts to take account of nongenetic factors in development: ‘‘critics of conclusions about the important effects of genotype on phenotype will be seen as taking refuge in nebulous appeals for a new general view of causation of behavior and as driven to this predicament solely by their sense of outrage at the determinist claims’’ (p. 408).

1 76

The Fearless Vampire Conservator

While acknowledging the legitimacy of many of Kitcher’s observations, I believe that Oyama’s diagnosis of the problem is substantially correct. In this essay I try to support her view in three main ways. First, I examine the fallacious ways of thinking about genetic causation that make up genetic determinism and argue that these are a natural consequence of attributing semantic properties to the gene. Second, I use data from an empirical study of biologists to document an apparent association between endorsing informational representations of the gene and being relatively uninterested in contextual effects on gene expression. I do not want to place too much weight on this one preliminary result, but it does suggest that efforts to determine whether Oyama is correct need not be confined to philosophical argument: the claim that genetic determinism is caused by a certain representation of the gene can be bolstered by documenting a correlation between determinist thinking and that representation. Finally, I suggest that Kitcher is mistaken in thinking that ‘‘neither Lewontin’s ‘dialectical biology’ nor Oyama’s ‘developmental systems theory’ offer anything that aspiring researchers can put to work’’ (Kitcher 2001: 408). There is a substantial research tradition in developmental psychobiology that fits the prescriptions of developmental systems theory (dst) for the simple reason that dst is an attempt to abstract a theoretical framework from research in that tradition. Philosophers of science and other commentators on the biological sciences need to become more aware of this tradition and its achievements. Popular presentations of those achievements may also offer a practical route to improving public understanding of the role of the genes in development.

Genetic Determinism and the Informational Gene

What Is Genetic Determinism? In contemporary popular discourse, a trait that is supposedly characteristic of one sex, of some ethnic group, or of humanity in general is said to be ‘‘in the genes,’’ just as in previous centuries such traits were said to be ‘‘in the blood.’’ Individual differences that might once have been said to ‘‘run in the family’’ are now attributed to genes. The popular concept of a genetic trait is the latest expression of the ancient idea that some traits are inborn and unalterable expressions of an organism’s nature while others are acquired, malleable effects of experience. Opening today’s New York Times I read that human cartilage is too weak for the demands of American football and ‘‘that’s something that

177

Paul E. Griffiths

is genetically predetermined. It’s the way God made us’’ (Roberts 2002: 23). The scientific validity of a sharp distinction between ‘‘nature’’ and ‘‘nurture’’ has long been debated. The founders of modern animal behavior studies, particularly Konrad Lorenz, attempted to provide a scientific basis for the distinction between nature and nurture with the ‘‘deprivation experiment.’’ By his definition, behaviors that develop when the organism is experimentally deprived of the opportunity to learn are innate, while those that fail to develop are acquired. Innate behaviors result from evolutionary adaptation and are transmitted in the genes. Acquired behaviors result from learning. It soon became clear, however, that all behaviors have both genetic and nongenetic causes. For any behavior there will be some genetic modifications that prevent its development and some nongenetic modifications that prevent its development. On the one hand, social deprivation of young rhesus monkeys will prevent them from displaying their ‘‘innate’’ sexual behaviors as adults (Harlow, Dodsworth, and Harlow 1965). On the other hand, a rat and a bird will emerge from an identical program of conditioning having learned very different behaviors: their genetic endowment affects what is ‘‘acquired’’ (Garcia, McGowan, and Green 1972). Such examples formed the empirical core of Daniel Lehrmann’s influential critique of Lorenz’s innateness concept (Lehrman 1953), a critique that was widely accepted. Ethologists came to realize that questions about the development of a behavior and questions about its evolution should not be conflated (Tinbergen 1963). Many evolutionary adaptations require complex and highly specific inputs from the environment, and not all traits that are robust in the face of variations in the developmental environment are evolutionary adaptations. Even Konrad Lorenz grudgingly admitted that he had offered an overly simplistic ‘‘understanding of the relations between phylogenetic adaptation and adaptive modifications of behaviour. It was Lehrman’s (1953) critique,’’ he wrote, ‘‘which, by a somewhat devious route, brought the full realisation of these relations to me’’ (Lorenz 1965: 80). Once it is accepted that all traits develop as a result of the interaction of genetic and nongenetic factors, genetic determinism becomes a view about how these factors interact. This view can be conveniently represented using ‘‘norms of reaction’’—graphical representations of a phenotypic variable as a function of genotypic and environmental variables. The strongest form of genetic determinism claims that norms of reaction show no response to the environmental variable. An organism needs an environment for the trait to develop, but it does not matter which environment (fig. 1). Kitcher suggests that some modern genetic determinists think norms of reaction have this

1 78

The Fearless Vampire Conservator

form, but only in some limited, but perhaps contextually important, range of environments. Someone might claim, for example, that ‘‘genetic diseases’’ develop in any environment except environments which contain specific clinical interventions designed to cure the disease. A more moderate form of genetic determinism claims that genetic and environmental factors interact additively. Genotype makes a constant difference across some range of environment. A determinist picture of the relationship between genetic factors (G) and education (E) in the determination of iq (P) might appear as shown in figure 2. Perhaps the single most influential contribution to the literature on the interpretation of behavioral genetics is Richard Lewontin’s paper ‘‘The analysis of variance and the analysis of causes’’ (Lewontin 1974). Lewontin argues that the empirical evidence suggests that actual norms of reaction are likely to be nonadditive (fig. 3). In that case, it makes no sense to talk of a particular genotype ‘‘determining’’ a particular phenotypic difference. Genotype and environment jointly determine the outcome in the straightforward sense that the effect of each factor on the outcome is a function of the particular value taken by the other factor. Nor, as Lewontin further points out, does it make any sense in the context of figure 3 to perform an analysis of variance on trait differences in a population and interpret the resulting statistic as indicating the percentage contribution of genes and environment to the trait. The very same causal system can produce a pattern of trait differences that correlate 100 percent with the environmental factor (if everyone lives in the environment where the lines cross) or correlate strongly with genotype (if everyone lives at one extreme of the graph). Lewontin and many others insist that because gene-environment interactions are typically nonadditive, heritability studies do not yield information about the relative importance of genetic and environmental developmental factors in the actual causal process that gives rise to a phenotypic trait.2 Nonadditive interaction between developmental causes is a critical element of the case for Oyama’s ‘‘developmental systems’’ approach. The point at which interaction becomes nonadditive is the point at which it becomes impossible to think of development as the determination of a ‘‘phenotypic resultant’’ by a number of causal ‘‘vectors.’’ Instead, the causal significance of the actual value of each causal variable becomes a function of the actual values of other variables, which is as much as to say that the different factors have to be understood as part of a single system. A developmental system in Oyama’s sense can be thought of as a norm of reaction in which multiple genetic and multiple nongenetic variables interact, often nonadditively, and in

1 79

p g1

g2

1 • A norm of reaction vindicating simple genetic determinism. P = phenotypic variable; E = environmental variable; G = genotype.

p = phenotypic variable e = environmental variable g = genotype

e

p g1

g2

p = phenotypic variable e = environmental variable g = genotype

e

2 • Pure additive interaction between genotype and environment.

p g1

g2 p = phenotypic variable e = environmental variable g = genotype

e

3 • Nonadditive interaction between genotype and environment.

The Fearless Vampire Conservator

which the role of developmental timing has been made explicit with additional axes, so that the very same cause acting at two different stages in development can have two different effects (Gray 1992). There is, however, no objection to using traditional norms of reaction figures to represent the interaction of two specific developmental causes, with the other relevant causes held constant to allow the experimental study of that relationship. Figures 1 and 2 represent two senses in which a genotype can be said to ‘‘determine’’ a trait or a trait difference. For Kitcher, genetic determinism as a general intellectual position is simply the claim that many norms of variation have ‘‘determinist’’ shapes. If this claim is true, then for many scientific purposes the role of the environment in producing traits or trait differences need not be considered. Hence, according to Kitcher, genetic determinism arises from the widespread and understandable human desire for simple, monocausal explanations. His antidote to genetic determinism is the careful caseby-case investigation of how genetic and environmental factors interact to determine phenotypes. Kitcher is confident that in many cases genetic determinism will prove to be false. He sees behavioral genetic research as the most scientifically tractable approach to such investigations. Specific genetic loci that are shown to correlate with behavioral traits in certain environments provide valuable entry points to the complex molecular pathways that construct the behavior. Kitcher is also confident that the importance of nongenetic factors will become evident once these pathways are understood. Below, I discuss some other scientific approaches to developmental interactions that I take to be at least equally tractable, but which Kitcher and other commentators seem to have overlooked. It is hard to disagree with Kitcher that the careful elucidation of specific developmental pathways will provide evidence bearing on the issue of genetic determinism. I also share his confidence that in many cases the norm of reaction will turn out not to have a ‘‘determinist’’ shape. But I am less confident that simply publicizing more examples of nonadditive gene-environment interaction will lay the specter of genetic determinism. Unlike Kitcher, Oyama sees the persistence of genetic determinism as a puzzling phenomenon that requires special explanation. Consider, for example, a footnote to Kitcher’s paper containing an anecdote about a leading population geneticist’s irritated response to the assertion by a behavioral geneticist that heritability figures reveal something about the role of genes in the development of behavioral traits. Heritability measures, Kitcher comments, are ‘‘irrelevant’’ and the fact that behavior geneticists persist in using them is ‘‘an unfortunate tic

181

Paul E. Griffiths

from which they cannot free themselves’’ (Kitcher 2001: 413). It is this sort of anomaly that sends Oyama in search of a cause. Why do so many intelligent scientists appear to ignore facts that are well known to them, such as the likely nonadditive interaction of genotypes and environment? For Oyama, ‘‘genetic determinism’’ refers to something deeper than a pattern of interaction between genotype and environment that may or may not hold in any particular case. It is an underlying attitude to genes and their role in development that makes it hard to assimilate certain facts and easy to assimilate—or to assume—others. Genetic determinism is whatever it is that causes otherwise sensible people to draw various inappropriate inferences from evidence suggesting that genes have a causal role in the development of a trait. Some typical inferences include the following: The prevalence of the trait in the population can never be reduced below the proportion of variance in the trait that is found to be correlated with genetic factors. Development of the trait will be insensitive to environmental factors in development in rough proportion to ‘‘how genetic’’ the trait is (the proportion of variance in the trait in some study population which is due to genetic factors). A given genetic change will make a constant difference irrespective of the values of other developmental variables. Consequently, the variance accounted for by genetic factors in one population can be safely extrapolated to other populations.

It is these and similar inferences that Oyama takes to result from an underlying conception of the gene as a source of information about the phenotypic outcomes of development.

The Informational Gene Although biologists think of genes as key parts of the molecular machinery that assembles a protein product, they also think of them as instructions or programs for the production of particular phenotypic traits.3 In popular science writing this second representation of the gene predominates, leading to assertions like the following: ‘‘An organism’s physiology and behaviour are dictated largely by its genes. And those genes are merely repositories of information written in a surprisingly similar manner to the one that computer scientists have devised for the storage and transmission of other information. . . . [Biology] is itself an information technology’’ (Economist 1999: 97).

182

The Fearless Vampire Conservator

This way of thinking about genes has its roots in Mendelian or transmission genetics—the discipline that first postulated genes. In the absence of any molecular understanding of the gene a tractable theoretical and experimental framework was constructed in which genes were identified by the phenotypic characters with which they correlated in breeding experiments. Developmental biology—the investigation of how characters seen in the parent are reconstructed in the offspring—was put to one side in favor of a black box strategy in which genes, identified in the manner described, were treated as if the transmission of a chunk of chromosome explained in and of itself the ‘‘transmission’’ of the phenotypic character. Making use of metaphors from the new sciences of information theory, cybernetics, and computing, biologists came to describe genes as containing ‘‘blueprints,’’ ‘‘programs,’’ and ‘‘instructions’’ concerning the traits with which they correlate in breeding experiments (Keller 1995; Kay 2000). The results of the molecular revolution in biology have been explained to the general public almost entirely in these terms. The popular understanding of the nature of the molecular revolution, and the common metaphors used by scientists themselves when explaining their work, are in stark contrast to the views of many contemporary philosophers of biology. The biologist and philosopher of science Sahotra Sarkar has noted that ‘‘there is no clear, technical notion of ‘information’ in molecular biology. It is little more than a metaphor that masquerades as a theoretical concept and . . . leads to a misleading picture of possible explanations in molecular biology’’ (Sarkar 1996: 187). The leading philosopher of biology, Peter GodfreySmith, concludes that ‘‘[a]ll the genes can code for, if they code for anything, is the primary structure (amino acid sequence) of a protein’’ (Godfrey-Smith 1999: 328). The point is not that there is no useful way to apply formalisms from the information sciences to the study of molecular developmental systems—there are many such ways. The point is that the facts of molecular developmental biology do not correspond to the popular idea that the genetic code is a language in which the genome contains instructions about phenotypes. Kenneth Schaffner makes this point by saying that there are no tiny ‘‘traitunculi’’ living in the genome (Schaffner 1998). The slippage from a code for protein structure to a language for specifying phenotypes embodies a systematic confusion about the meaning of the term information. Concepts of information can be divided into two very broad classes, which Kim Sterelny and I have called ‘‘causal’’ and ‘‘intentional’’ (Sterelny and Grif-

183

Paul E. Griffiths

fiths 1999: 101).4 Causal notions include the measure of quantity of information that is at the heart of the mathematical theory of communication as well as the measures found in algorithmic complexity theory and various notions of information content inspired by these mathematical measures of information quantity (Dretske 1981). The simplest causal accounts of the information content of a signal—what the signal is about—define the content as whatever the signal is reliably correlated with. Smoke contains information about fire because, as the saying goes, ‘‘where there’s smoke there’s fire.’’ The weakness of this causal account of information content—and of many of its more complex relatives—is that it makes information ubiquitous. As John Maynard Smith has noted, ‘‘With this definition, there is no difficulty in saying that a gene carries information about adult form; an individual with the gene for achondroplasia will have short arms and legs. But we can equally well say that a baby’s environment carries information about growth; if it is malnourished, it will be underweight’’ (Maynard Smith 2000: 189). Maynard Smith concludes that a definition of information that can be used to capture the traditional idea that genes carry information while other developmental causes merely support the expression of that information will have to be a definition that includes an element of what philosophers refer to as ‘‘intentionality.’’ Intentional information is information in the sense derived from human thought and language. The distinctive feature of intentional information is that it can be false (Godfrey-Smith 1989). The utterance ‘‘There are fairies at the bottom of my garden’’ and the thought that accompanies it have never occurred in response to the presence of fairies in someone’s garden, because fairies do not exist. But this has no effect on what the utterance means or on the content of the thought. The idea that genes have meaning in something like the way that human thought and language have meaning is lurking in the background in many discussions of genetic information. This can be seen from the way in which cases of gene-environment interaction are described. For example, it has been suggested that under starvation conditions, human mothers methylate growth-enhancing genes in their children, and thus block transcription of those genes. Children with identical genomes and identical nutrition will reach one adult height and weight if they have well-nourished mothers and another if they have starving mothers. When, as in this case, such a gene-environment interaction is thought to be an adaptation, the environment is said to trigger a ‘‘disjunctive genetic program’’ (fig. 4). The genes contain the instruction ‘‘grow fast if your mother was adequately nourished, grow slowly if she was starved,’’ and the environment specifies which disjunct is to be obeyed.

184

The Fearless Vampire Conservator p

4 • Norm of reaction corp = phenotypic variable e = environmental variable

e

responding to a ‘‘disjunctive genetic program.’’

There are innumerable cases, however, where the norm of reaction for a trait resembles that in figure 4 but people do not talk of disjunctive programs or regard the norm of reaction as an expression of conditional information in the genome. These cases are the pathological or merely quirky effects that are revealed by abnormal interventions in development, either those made by developmental biologists or those made by nature. The claim that the Drosophila genome contains the conditional instruction ‘‘develop a second thorax if given an ether shock’’ sounds like metaphor mingled with hyperbole, as does the claim that the macaque genome encodes the conditional instruction that a mother should kill her babies if she is raised in social isolation. If the concept of information in use here were a causal concept, then the contrast between these cases and the case of infant growth rates would be puzzling because the causal information in the genome is more or less the same thing as the genome’s norm of reaction. If, however, the concept in use is that of intentional information, then it is clear why some outcomes are regarded as part of the informational content of the genome and others are not. The intentional content of an instruction is the behavior it is intended to produce, not the behavior it actually produces. No matter how many students ignore the instruction to write a term paper based closely on the set texts, the meaning of that instruction remains the same. That is why it is legitimate to deduct grades for not following the instruction even when it is predictable that most students will ignore it! Likewise, if the human genome contains the intentional information specifying a normal human phenotype, then the information content of the genome is unaffected by the cases of what Lorenz used to call ‘‘bad rearing,’’ even if these cases are in the majority. When the ‘‘wrong’’ outcome occurs, the phenotype simply misrepresents the information in the genes.

185

Paul E. Griffiths

In various contexts I and others have argued that any analysis of intentional information that makes it part of the natural world will reveal that if genes contain intentional information, so do many other, nongenetic developmental inputs. The molecular biologist Robin Knight and I have called this the ‘‘parity thesis.’’ 5 In this essay, however, the question is not whether the intentional concept of information content is legitimately restricted to the genes but only whether it is as a matter of fact used when discussing genes and not used when discussing other developmental causes. This latter claim is relatively uncontroversial. The fact that the intentional concept of information is used in this asymmetric way explains why the proposal that all developmental resources contain developmental information has been so controversial. As mentioned above, Maynard Smith has argued that only an intentional concept of information can capture the intent of the many biologists who have used the idea of information to distinguish the role of genes in development from the roles of nongenetic causes (Maynard Smith 2000). I will rest my case, therefore, after just one more example of the asymmetric treatment of genetic and nongenetic causes. A critical temperature range in the nest plays a role in sex determination in crocodiles strikingly similar to that played by the sry gene on the Y chromosome in mammals. Both initiate a biochemical cascade that masculinizes the fetus. Both causal factors are brought into existence by a complex system that has evolved to ensure that the masculinizing factor is present often enough to generate the correct sex ratio. Despite this, people are intuitively reluctant to describe the temperature using locutions that suggest intentional information. Like the sry gene, the nest temperature can ‘‘cause,’’ ‘‘determine,’’ and even ‘‘signal’’ the fetus to masculinize, but it sounds odd to say that the molecular kinetic energy in the nest provides the fetus with information about masculinity. It seems natural to say that the sry gene contains the ‘‘instruction’’ to masculinize the fetus, but this would seem forced in the case of nest temperature. One might try to justify this asymmetry by arguing that the effect of the temperature is strongly context dependent. It is only in the very precise context of a crocodile fetus that this temperature has this effect. The same, however, is true of the sry gene, whose effects can be blocked by mutations affecting receptors for its products or by the environmental conditions in the womb whose tragic results create work for gender reassignment units around the world (Money 1993). Both the temperature and the gene act as switches, causing certain other genes to be transcribed. Neither has any connection to a phenotype outside of a specific class of developmental systems.

186

The Fearless Vampire Conservator

Intentional Information and Genetic Determinism In the last section I tried to document the asymmetric use of information talk in biology. This asymmetry can help to explain the persistence of genetic determinism. The predominant vernacular meaning of information is ‘‘intentional information’’: the truism that the Internet contains a lot of information means not that it has a high degree of entropy, but that it contains a large number of intentional representations. It is a central feature of intentional information that it retains its identity in the face of misrepresentation or, in the case of imperative representations, noncompliance. This is what makes it possible for intentional representations to be false and for intentional imperatives to be disobeyed. The relationship between an intentional imperative and its effect is thus quite different from that between a material cause and its effect. If we describe a gene as a switch that initiates a cascade of gene transcription leading to, for example, an initial state of the brain which under some range of environmental conditions produces a behavioral preference for homosexual relationships, it is evident that the link between the switch and its final effect is a function of the complex causal system in which the switch is embedded. If the context is changed, the gene is no longer a switch that controls homosexuality, any more than a light switch remains a light switch when it is wired to an exhaust fan. If, however, we describe the same gene as a genetically encoded instruction to be a homosexual, then, intuitively, the presence of different genes at other loci, or prenatal environments that do not support the cascade of gene expression, or postnatal environments that lead the brain to mature differently, all merely cause the organism to misinterpret or disobey the instruction contained in the gene. Furthermore, the gene retains its identity as a ‘‘gay gene’’ even in an individual to which it has made some other biochemical contribution and who is, phenotypically, a heterosexual. In other words, intentional information is intrinsically context insensitive and thus intrinsically unsuited to express the causal link between genes and complex phenotypes, because that link is intrinsically context sensitive.6 Representing genes as intentional imperatives that contain a representation of a phenotype supports genetic determinism because it allows genes to retain a link to a specific phenotype when they are moved from one context to another. A causal intervention that removes the causal pathway between a gene and the phenotype with which it was previously associated does not change the ‘‘meaning’’ of the gene; it merely prevents that instruction from being obeyed. It does not put the old phenotype on a par with all the other

187

Paul E. Griffiths

phenotypes that form part of that gene’s norm of reaction. The new phenotype with which the gene is associated as a result of the intervention is not the new meaning of the gene; it is merely a misrepresentation of the information embodied in the gene. Allowing genes to retain their imperative link to a particular phenotype across changes in causal context creates a background assumption that if the gene were expressed, it would produce the phenotype about which it contains information. The intentional representation of the gene also makes it natural to think that environments in which the gene does not ‘‘express’’ its meaning are qualitatively different from those in which it does; such environments are somehow abnormal or pathological because they create a mismatch between gene and phenotype. In all these related ways, the intentional representation of the gene supports the idea that genes have a constant effect across context, and hence the idea that genetic and environmental factors interact additively. If genes contain intentional information, then changing the environment either facilitates the expression of this information or hinders expression of the same information. This view is naturally represented by something like figure 2 rather than figure 3. If genotype G1 in figure 2 contains the instruction ‘‘be intelligent,’’ for example, changing the environment merely determines the extent to which this instruction is obeyed. It cannot turn G1 into an instruction to be unintelligent, as would seem to happen in figure 3. Thus, intentional representations of genes lead almost inevitably to one of the central fallacies identified in Lewontin’s critique of behavioral genetics: the default assumption that associations between a gene and a phenotypic difference observed in one environment can be extrapolated to any other ‘‘normal’’ or ‘‘healthy’’ environment. The intentional representation of the gene is connected to the other, more vulgar fallacies described above (see ‘‘What Is Genetic Information’’) by various simple misunderstandings of the sort that Kitcher and others have observed to bedevil the public’s understanding of genetics. If the claim that a behavior is 30 percent genetic is understood to mean that in 30 percent of cases studied the behavior can be traced back to the presence of a particular gene, then the intentional representation of the gene suggests that these form a ‘‘hard core’’ of cases which will be insensitive to environmental variation. These are the 30 percent in which homosexuality is caused not by a specific developmental environment, but by an instruction to be homosexual that would be present in any environment. It is all too easy to imagine the claim that schizophrenia or same-sex preference is 30 percent genetic being understood in this way. Alternatively, if a continuous trait such as height or obesity is described as being 30 percent genetic, and if this is understood as

188

The Fearless Vampire Conservator

partitioning the actual trait into a portion that can be ascribed to the genes and a portion that can be ascribed to the environment, then the intentional representation of the gene will suggest that the part due to the genes can be suppressed only by inducing individuals in each generation to misrepresent their genetic nature.

Can the Influence of ‘‘Information Talk’’ Be Tested Empirically? Oyama claims that representing genes and gene action in terms of information leads to certain errors in reasoning. In the last section I tried to spell out some plausible ways in which this could occur. But Oyama’s claim ought also to be capable of empirical testing. If she is correct, then there should be an association between using the informational concept of the gene, in which genes are type-identified by the developmental information they contain, and neglect of the role of contextual factors in gene expression. A questionnaire study conducted in Australia on eighty-one post-Ph.D. biologists by Karola C. Stotz and myself in 1999 produced a result that suggests that such an association is worth testing for more carefully (Stotz and Griffiths, in preparation). In that study we found that biologists with training and experience in developmental biology were much less likely to endorse the idea that the gene can be adequately defined as a unit of information than those with backgrounds in biochemistry and pure molecular genetics (fig. 5). Other results from the same study are consistent with the equally widely held view that developmental biologists view dna sequences in the light of contextual factors that affect the expression and processing of gene products. The responses of developmental biologists to questions about whether two dna sequences are ‘‘the same gene’’ were significantly influenced by information about such contextual factors. Molecular biologists without experience in developmental biology tended to neglect these contextual factors, in the sense that their survey responses were not affected by information about them. Putting these two results together suggests that those scientists who are least concerned with contextual effects on gene expression are the happiest to endorse the idea that genes are, fundamentally, carriers of information. This finding does not, of course, allow us to determine whether one of these ideas promotes the other, or whether both ideas are expressions of some deeper conceptual commitments, so the results do not constitute any very strong vindication of my argument above (see ‘‘Genetic Determinism and the Infor-

189

Paul E. Griffiths 60

molecular developmental

% agreement

50 40 30 20 10 0 geneP

struc.

func.

info.

geneD

5 • Results of a survey in which molecular (n) and developmental ( )

biologists were offered a list of ‘‘short definitions’’ of the gene and asked which they would endorse if forced to choose only one: 1. That which makes the difference between two phenotypes [geneP] 2. A nucleic acid sequence with a certain characteristic structure [struc.] 3. A nucleic acid sequence with a certain characteristic function [func.] 4. A carrier of heritable information [info.] 5. A resource for development [geneD]

mational Gene’’). The results are, however, consistent with my theoretically motivated prediction that the informational gene concept should be associated with a neglect of contextual factors in gene expression.

The Research Agenda of Developmental Systems Theory An important part of Kitcher’s critique of Oyama and others who argue that the persistence of genetic determinism has a deeper explanation is his observation that ‘‘neither Lewontin’s ‘dialectical biology’ nor Oyama’s ‘developmental systems theory’ offer anything that aspiring researchers can put to work’’ (Kitcher 2001: 408). Hence, Oyama and Lewontin are calling for a radical new approach to genetic causation when no such approach is available. The result, Kitcher suggests, will be to convince practical scientists that whatever the shortcoming of monocausal genetic explanations, there is no

190

The Fearless Vampire Conservator

practical alternative. In this respect, I suspect, Kitcher has been misled by the heavy emphasis among philosophers of science on evolutionary biology and evolutionary explanations of human behavior. There is a vast philosophical literature on this topic, some of the best of it by Kitcher himself (Kitcher 1985), and in the evolutionary context Kitcher’s complaint has real substance. Russell Gray and I have described the sort of evolutionary research that might be facilitated by a developmental systems perspective (Gray 2001; Griffiths and Gray 2001), and recent work on the evolutionary significance of epigenetic inheritance and niche construction can be regarded as a partial vindication of these claims (Jablonka and Lamb 1995; Odling-Smee, Laland, and Feldman, 1996; Avital and Jablonka 2001; Laland, Odling-Smee, and Feldman 2001). The case for the practical relevance of developmental systems theory is much easier to make, however, in the developmental context. Until recently philosophers of science paid very little attention to developmental biology, and still less to the developmental biology of behavioral traits. But there is a rich experimental tradition in developmental psychobiology dating back several decades, and developmental systems theory is to a large extent an attempt to make explicit and reflect on the core concepts of this research tradition. Developmental psychobiology might perhaps be defined as the experimental elucidation of the effects of genetic and environmental factors and their interactions in the ontogeny of gross behavioral traits. This sort of research was pioneered in the interwar years by American comparative psychologists and continued after World War II in the work of their students and of developmentally oriented workers in the new science of ethology, especially those influenced by Daniel Lehrman (Gottlieb 2001; Johnston 2001). Developmental psychobiology differs from behavioral genetics in its methodological emphasis on experimental intervention in the laboratory, as opposed to the descriptive-statistical study of natural populations, a feature that places it closer to developmental biology. A textbook presentation of this kind of work can be found in George Michel and Celia Moore’s excellent Developmental Psychobiology: An Interdisciplinary Science (Michel and Moore 1995). When I discussed the importance of interactions between genes and environment above, I was referring not so much to the statistical interactions revealed by behavioral genetics at the population level as to the causal interactions revealed by experiments in developmental psychobiology. The call to pay more attention to developmental interactions is not merely an appeal to complexity. It is also an appeal to move beyond using genes as statistical markers for phenotypes and to understand them as biochemical causes of development. As an imperative to researchers, this can mean something

191

Paul E. Griffiths

very practical and not at all unappealing, like ‘‘going molecular’’ and investigating the causes of, for example, mental illness at the level of functional genomics, proteomics, and developmental neurobiology (Schaffner 2001).

Developmental Psychobiology and the Public Understanding of Genetics I suggest that it will be difficult to improve popular understanding of genetics if we continue to rely on what has so far been the main conceptual tool of popularization—the idea that genes are blueprints or programs. As I have argued, this formulation makes a deterministic reading of claims about the role of genes in development almost inevitable. This poses a considerable problem. ‘‘Information talk’’ in molecular biology is not going to disappear in the foreseeable future. There really is a genetic code; there are numerous legitimate applications of technical notions of information in molecular biology; and the informal, quasi-cybernetic notions of ‘‘signaling,’’ ‘‘switching,’’ and ‘‘feedback’’ are the patois of molecular developmental biology. Unfortunately, the mathematical meaning of information is too unintuitive and too far from the usual meaning of the word to become part of popular consciousness, and even terms like signaling irresistibly suggest that what is being signaled is an intentional message. Hence, neither eschewing information talk nor explaining it properly to a wide audience seems to be practicable. The only practical solution to improving the public understanding of what genes do, I suggest, is to popularize other kinds of biological research that can act as a counterweight to popular misinterpretations of information talk. That counterweight ought to be developmental biology and developmental psychobiology. Both disciplines have the advantage that they can discuss gross phenotypic characters that are easily and intuitively grasped by a popular science audience—the carapace of the turtle or the mutual recognition of parent and offspring in ducks, for example. They explain these characters using causal, rather than informational, locutions and by recounting experimental interventions that often involve macrolevel physical processes that are also relatively easy to grasp. Genes often figure in these developmental narratives as things activated by other factors, and those other factors are often environmental. Explanations of development in this idiom thus automatically correct the impression that genes are Godlike ‘‘prime movers themselves unmoved.’’ Developmental psychobiological explanations also have a strong tendency to focus on gene-environment interactions, so they also automati-

192

The Fearless Vampire Conservator

cally stress context dependence. Fortunately, popular writing about this kind of science has started to appear in recent years (Gottlieb 1997; Bateson 1999). David S. Moore’s The Dependent Gene is a particularly successful example of this genre (Moore 2001). It can only be hoped that some stock examples from this tradition will become firmly entrenched in the popular imagination as a counterweight to the dim awareness that ‘‘scientists have discovered’’ genes for this, genes for that, and the genes for the other.

Conclusion Oyama and others have argued that genetic determinism—the view that genetic causes have strong, context-insensitive connections to their phenotypic effects—is kept in existence in the face of contrary evidence by the idea that genes contain information about phenotypes or instructions for development. Kitcher has disputed this diagnosis, arguing that the persistence of genetic determinism is due to the human preference for simple explanations and the difficulty of communicating complex scientific results to a wider audience. I have offered both theoretical and empirical arguments in favor of Oyama’s explanation. At a theoretical level, I have argued that the predominant vernacular conception of information is ‘‘intentional information,’’ and that the concept of intentional information is precisely that of a message whose content is unchanged by contextual effects on the way the message is interpreted. Thus, thinking of genes as containing intentional, imperative messages inevitably leads to the view that the link between genes and phenotypes will be unaffected by changes in other developmental factors. At an empirical level, I have presented a preliminary finding of a statistical association between the informational conception of the gene and a neglect of contextual effects on gene expression, a finding at least consistent with the theoretical picture I have outlined. Kitcher has also suggested that Oyama’s call for a radical new approach to development will be counterproductive because no practical research program exists which will allow scientists to heed that call. This is an important and legitimate concern, but I have argued that Kitcher has overlooked the empirical work in developmental psychobiology that inspires the theoretical work of Oyama and others. Developmental psychobiology is a practical alternative—or supplement—to the kind of work in behavioral genetics that Oyama and Kitcher agree is frequently and illegitimately used to support genetic determinism. Finally, I have suggested that the best way to improve

193

Paul E. Griffiths

the public understanding of genetics is to present popular accounts of experimental findings in developmental psychobiology, and experimental developmental biology more generally, to counteract the effects of simplistic presentations of statistical associations between genes and phenotypes.

Notes 1 Oyama’s influential book The Ontogeny of Information (Oyama 1985), from which the above quotation is drawn, has been reprinted with a new introduction by Richard Lewontin (Oyama 2000b), as have many of her papers (Oyama 2000a). Oyama, Griffiths, and Gray 2001 contains new and classic papers by Oyama and other authors on the developmental systems approach. 2 Kitcher outlines Lewontin’s argument about heritability and genetic causation (p. 399), but unfortunately, his norm of reaction figures do not include one showing nonadditive interaction. 3 The contrasting scientific roles of these two ways of thinking about genes are explored at length in Moss 2001 and Moss 2002. Some exciting potential implications of the first, molecular way of thinking about genes are explored in Neumann-Held 1999 and Neumann-Held 2001. 4 It has been suggested that we should have used the term correlational instead of causal in recognition of the fact that classical information theory makes no reference to causation. Instead, we chose to introduce acausal information channels later as a special case. In the biological contexts we were concerned with it is assumed on all sides that correlations between developmental factors and phenotypes are of interest precisely because they provide evidence of an underlying causal network. 5 The term parity derives from Oyama’s call for ‘‘parity of reasoning’’ in dealing with genetic and environmental causes (Oyama 1985). Parity is the idea that genes and other material causes are on a par. The ‘‘strawman’’ parody of developmentalism says that all developmental causes are of equal importance. The real developmentalist position is that the empirical differences between the role of dna and that of cytoplasmic gradients or host-imprinting events do not justify the metaphysical distinctions currently built upon them. Nucleic acid sequences and phospholipid membranes both have distinctive and essential roles in the chemistry of life, and in both cases there seems no realistic substitute for them. But the facts of development do not justify assigning dna the role of information and control while inherited membrane templates get the role of ‘‘material support’’ for reading dna (Griffiths and Knight 1998: 254; see also Griffiths 2001; Griffiths and Gray 1997; Sterelny, Dickison, and Smith 1996). The obvious candidate for a naturalistic account of intentional information is a ‘‘teleosemantic’’

194

The Fearless Vampire Conservator

account (Millikan 1984; Papineau 1987). According to teleosemantics, a representational state like a thought or a sentence contains information about the state of affairs that it is its function to represent, and something’s function is determined by asking what it was intended to do or designed to do. Many genetic and nongenetic factors in development arguably evolved to exert a certain influence on the phenotype. If so, then it is their function to exert that influence and thus, according to teleosemantics, they contain the instruction to exert that kind of influence. Peter Godfrey-Smith has explored the prospects for a teleosemantic account of developmental information (Godfrey-Smith 1999). 6 Sarkar has proposed that one precondition for regarding a phenotype as ‘‘genetic’’ is that the phenotype itself can be characterized in terms of a specific molecular product or, more usually, its absence (Sarkar 1998). Thus, for example, muscular dystrophy can be defined as the inability to synthesize a key protein. The link between the phenotype and the loss of gene template for that protein is context insensitive because the phenotype more or less is the loss of that template capacity.

References Avital, E., and Jablonka, E. 2001. Animal Traditions: Behavioural Inheritance in Evolution. Cambridge: Cambridge University Press. Bateson, P. P. G. 1999. Design for a Life: How Behavior and Personality Develop. London: Jonathan Cape. Dretske, F. 1981. Knowledge and the Flow of Information. Oxford: Blackwell. Economist Magazine. 1999. Drowning in data. Economist, 26 June, pp. 97–98. Garcia, J., McGowan, B. K., and Green, K. F. 1972. Biological constraints on learning. In: M. Seligman and J. L. Hager (eds.), Biological Boundaries of Learning (pp. 21–43). New York: Appleton Century Crofts. Godfrey-Smith, P. 1989. Misinformation. Can. J. Philos. 19(5): 533–550. Godfrey-Smith, P. 1999. Genes and codes: lessons from the philosophy of mind? In: V. G. Hardcastle (ed.), Biology Meets Psychology: Constraints, Conjectures, Connections (pp. 305–331). Cambridge: mit Press. Gottlieb, G. 1997. Synthesizing Nature-Nurture: Prenatal Roots of Instinctive Behavior. Hillsdale, N.J.: Lawrence Erlbaum. Gottlieb, G. 2001. A developmental psychobiological systems view: early formulation and current status. In: S. Oyama, P. E. Griffiths, and R. D. Gray (eds.), Cycles of Contingency: Developmental Systems and Evolution (pp. 41–54). Cambridge: mit Press. Gray, R. D. 1992. Death of the gene: developmental systems strike back. In: P. Griffiths (ed.), Trees of Life (pp. 165–210). Dordrecht: Kluwer.

195

Paul E. Griffiths

Gray, R. D. 2001. Selfish genes or developmental systems? In: R. S. Singh, C. B. Krimbas, D. B. Paul, and J. Beatty (eds.), Thinking about Evolution: Historical, Philosophical and Political Perspectives (pp. 184–207). Cambridge: Cambridge University Press. Griffiths, P. E. 2001. Genetic information: a metaphor in search of a theory. Philos. Sci. 68(3): 394–412. Griffiths, P. E., and Gray, R. D. 1997. Replicator II: Judgment Day. Biol. Philos. 12(4): 471–492. Griffiths, P. E., and Gray, R. D. 2001. Darwinism and developmental systems. In: S. Oyama, P. E. Griffiths, and R. D. Gray (eds.), Cycles of Contingency: Developmental Systems and Evolution (pp. 195–218). Cambridge: mit Press. Griffiths, P. E., and Knight, R. D. 1998. What is the developmentalist challenge? Philos. Sci. 65(2): 253–258. Harlow, H. F., Dodsworth, R. O., and Harlow, M. K. 1965. Total isolation in monkeys. Proc. Natl. Acad. Sci. USA 54: 90–97. Jablonka, E., and Lamb, M. J. 1995. Epigenetic Inheritance and Evolution: The Lamarkian Dimension. Oxford: Oxford University Press. Johnston, T. D. 2001. Towards a systems view of development: an appraisal of Lehrman’s critique of Lorenz. In: S. Oyama, P. E. Griffiths, and R. D. Gray (eds.), Cycles of Contingency: Developmental Systems and Evolution (pp. 15–23). Cambridge: mit Press. Kay, L. E. 2000. Who Wrote the Book of Life? A History of the Genetic Code. Palo Alto: Stanford University Press. Keller, E. F. 1995. Refiguring Life: Metaphors of Twentieth Century Biology. New York: Columbia University Press. Kitcher, P. 1985. Vaulting Ambition. Cambridge: mit Press. Kitcher, P. 2001. Battling the undead: how (and how not) to resist genetic determinism. In: R. Singh, K. Krimbas, D. Paul, and J. Beatty (eds.), Thinking about Evolution: Historical, Philosophical and Political Perspectives (pp. 396– 414). Cambridge: Cambridge University Press. Laland, K. N., Odling-Smee, F. J., and Feldman, M. W. 2001. Niche construction, ecological inheritance, and cycles of contingency in evolution. In: S. Oyama, P. E. Griffiths, and R. D. Gray (eds.), Cycles of Contingency: Developmental Systems and Evolution (pp. 117–126). Cambridge: mit Press. Lehrman, D. S. 1953. Critique of Konrad Lorenz’s theory of instinctive behavior. Q. Rev. Biol. 28(4): 337–363. Lewontin, R. 1974. The analysis of variance and the analysis of causes. Amer. J. Hum. Genet. 26: 400–411. Lorenz, K. 1965. Evolution and the Modification of Behaviour. U.S. ed. Chicago: University of Chicago Press. Maynard Smith, J. 2000. The concept of information in biology. Philos. Sci. 67(2): 177–194.

196

The Fearless Vampire Conservator

Michel, G. F., and Moore, C. L. 1995. Developmental Psychobiology: An Interdisciplinary Science. Cambridge: mit Press. Millikan, R. G. 1984. Language, Thought and Other Biological Categories. Cambridge: mit Press. Money, J. 1993. The Adam Principle: Genes, Genitals, Hormones and Gender: Selected Readings in Sexology. Buffalo, N.Y.: Prometheus Books. Moore, D. S. 2001. The Dependent Gene: The Fallacy of ‘‘Nature versus Nurture.’’ New York: W. H. Freeman/Times Books. Moss, L. 2001. Deconstructing the gene and reconstructing molecular developmental systems. In: S. Oyama, P. E. Griffiths, and R. D. Gray (eds.), Cycles of Contingency: Developmental Systems and Evolution (pp. 85–97). Cambridge: mit Press. Moss, L. 2002. What Genes Can’t Do. Cambridge: mit Press. Neumann-Held, E. M. 1999. The gene is dead—long live the gene: conceptualising the gene the constructionist way. In: P. Koslowski (ed.), Sociobiology and Bioeconomics: The Theory of Evolution in Biological and Economic Theory (pp. 105–137). Berlin: Springer. Neumann-Held, E. M. 2001. Let’s talk about genes: the process molecular gene concept and its context. In: S. Oyama, P. E. Griffiths, and R. D. Gray (eds.), Cycles of Contingency: Developmental Systems and Evolution (pp. 69–84). Cambridge: mit Press. Odling-Smee, F. J., Laland, K. N., and Feldman, F. W. 1996. Niche construction. Amer. Nat. 147(4): 641–648. Oyama, S. 1985. The Ontogeny of Information: Developmental Systems and Evolution. Cambridge: Cambridge University Press. Oyama, S. 2000a. Evolution’s Eye: A Systems View of the Biology-Culture Divide. Durham: Duke University Press. Oyama, S. 2000b. The Ontogeny of Information: Developmental Systems and Evolution. 2d ed., rev. and exp. Durham: Duke University Press. Oyama, S., Griffiths, P. E., and Gray, R. D. (eds.). 2001. Cycles of Contingency: Developmental Systems and Evolution. Cambridge: mit Press. Papineau, D. 1987. Reality and Representation. New York: Blackwell. Roberts, S. 2002. In the nfl, wretched excess is the way to make the roster. N.Y. Times, 1 August, pp. A21, A23. Sarkar, S. 1996. Biological information: a sceptical look at some central dogmas of molecular biology. In: S. Sarkar (ed.), The Philosophy and History of Molecular Biology: New Perspectives (pp. 187–232). Dordrecht: Kluwer. Sarkar, S. 1998. Genetics and Reductionism. Cambridge: Cambridge University Press. Schaffner, K. 1998. Genes, behavior and developmental emergentism: one process, indivisible? Philos. Sci. 65(2): 209–252. Schaffner, K. F. 2001. Nature and nurture. Curr. Opin. Psychiat. 14(5): 485–490.

197

Paul E. Griffiths

Sterelny, K., Dickison, M., and Smith, K. 1996. The extended replicator. Biol. Philos. 11(3): 377–403. Sterelny, K., and Griffiths, P. E. 1999. Sex and Death: An Introduction to the Philosophy of Biology. Chicago: University of Chicago Press. Stotz, K., and Griffiths, P. E. (In preparation). How scientists conceptualise genes: an empirical study. ms. Tinbergen, N. 1963. On the aims and methods of ethology. Z. Tierpsychol. 20: 410–433. Wolpert, L. 1998. Dolly the sheep. Indep. Internat., 7–13 January, p. 18.

198

8 G E N E T I C S F R O M A N E VO L U T I O N A R Y PROCESS PERSPECTIVE

james griesemer This essay discusses contemporary theoretical perspectives on units of heredity, development, and evolution, including neo-Darwinism, processstructuralism, developmental systems theory, and my own reproducer perspective. Many of these concepts were designed to explore spatial, functional, or processual limits of biological units; for example, whether the units of heredity are restricted to genes, whether adaptations are also units of form, or to what extent the boundaries of developmental systems include environmental resources beyond the skin. In brief, theoretical perspectives in science coordinate models and phenomena by focusing attention in particular respects and committing limited resources to model phenomena in particular ways and degrees. Perspectives specify preferred lines of abstraction from phenomena of interest and also prioritize principles in terms of which models may be constructed to represent phenomena and used to intervene into them (Griesemer 2000a). I argue that a variety of theoretical perspectives on entities, structures, functions, fields, events, and processes is needed in order to achieve a robust account of biology’s theoretical units. I therefore urge a change of philosophical focus, from choice among rival perspectives to comparative analysis of strengths, weaknesses, and complementarities. The essay aims to develop three main points. First, in addition to the standard types of structuralist and functionalist descriptions of hereditary processes there is a third important type: process perspectives. Second, in addition to process-oriented approaches such as process-structuralism and developmental systems theory I propose the ‘‘reproducer’’ approach to amend and reposition the functionalist concept of the replicator within a process perspective on units of reproduction. Third, the availability of a va-

James Griesemer

riety of descriptions or perspectives is not a shortcoming of the present state of theoretical science but an enrichment critical to its success: science works best when it produces empirical results that are robust with regard to theoretical assumptions of its perspectives, as well as of its models. Thus, a multiplicity of perspectives can add to the knowledge base of science by fostering a variety of research programs into different phenomena, and can deepen and secure that knowledge by facilitating robust results. Specifically, I argue that taking a process perspective creates opportunities for robustness analysis of structure- and function-based accounts of evolutionary units, interpreting genetic processes as a special case of reproduction processes. This units-of-reproduction, or ‘‘reproducer,’’ account helps explain the relation between heredity and development, facilitating models of evolutionary transition (Griesemer 2000a, 2000b, 2000c, 2002a, 2002b). Evolutionary transition is the origin of new levels of biological organization (Maynard Smith and Szathmáry 1995). Because the origin of new levels involves the evolution of new developmental processes in addition to the evolution of adaptations, there can be no escaping the need for an account of how heredity and development intertwine to co-produce conditions of evolvability. The addition of the reproducer process perspective to others critical of neo-Darwinism facilitates evaluation of solutions to problems of the origin of biological form that are of interest to process-structuralism (Webster and Goodwin, this volume), various epigenetic theories (Newman and Müller, this volume; Jablonka and Lamb 1995), and developmental systems approaches (Griffiths; Oyama; Neumann-Held: this volume). Thus, the contribution of the reproducer and other process perspectives can be seen not merely as offering new rivals to received views of heredity, development, and evolution, but as contributions to a cooperative effort to enrich biological research. Having a variety of perspectives helps reveal the idealizing assumptions of whole programs of research which, while necessary to effective scientific practice, limit the theoretical and empirical power of any single perspective. A multiplicity of perspectives that differ in guiding assumptions is likely to yield more satisfactory research in biology than any single one could.

200

Evolutionary Process Perspective

Perspectives on Genetics: Structure, Function, Process One advantage of the reproducer perspective is that it brings out relationships among structure and function perspectives which have dominated thinking about genetics and its role in development and evolution.

Structuralism Structure perspectives model phenomena by representing structures. The most common structure perspective on units of evolution is that of a hierarchy of compositional levels of spatial organization: molecules, organelles, cells, tissues, organs, organisms, populations, species. The structural question of units of selection concerns the level(s) of spatial hierarchy at which selection occurs (Lewontin 1970). Elaborating on Ernst Mayr’s classic description of neo-Darwinism (Mayr 1978: 48), the philosopher Robert Brandon describes evolution by natural selection as a three-step process: selection, reproduction, and development (Brandon 1990: 4–5, 81). Selection changes distributions of types within generations, and then reproduction maps type distributions of one generation into the next according to their degrees of heritability. ‘‘But in order to go full circle, in order to get to the stage where selection occurs in the offspring generation, a final step is required. These differing offspring genotypes must develop’’ (Brandon 1990: 5).1 Darwin’s principles of variation, inheritance, and differential reproductive success are thus to be viewed as a description of conditions for the process of evolution by natural selection to occur; that is, for life to go ‘‘full circle’’ (Brandon 1990: 7). These and other ‘‘stage theories’’ of the evolutionary process have spawned several families of models in which life is represented as cycling through a sequence of component steps at specified levels which, in the proper order, constitute a net process of evolution by natural selection. Most of these models are ‘‘genetic’’ in the sense that their elements are genes, gene combinations, genotypic states, or mappings from genotypic states into phenotypes (Brandon 1999). Even if their elements are identified and individuated functionally (as in ‘‘the gene for X’’), the models represent change of unit genetic structures (e.g., allele frequencies). Evolutionary theory has been most fully formalized from genetic perspectives focusing on patterns and mechanisms of heredi-

201

James Griesemer

tary transmission. However, it has been very difficult theoretically to incorporate Brandon’s third, developmental, step into Mayr’s two-step process. Families of models developed under different perspectives can be distinguished by the kinds of state spaces they assume in representations of the component states and processes (selection, inheritance, development) and their combinations.2 A state space for an evolutionary process represents individuals or populations as points whose coordinates are values in each represented dimension. Thus, choice of state space reflects the type of guiding perspective. Processes that cause state transformations are represented as trajectories of state change in the space. For example, population geneticists tend to use allele or genotype state spaces, while quantitative evolutionary geneticists use phenotypic trait state spaces that interpret evolution as change in frequency of genetic units directly (Dobzhansky 1937), change in genotypic or phenotypic states or frequencies (Lewontin 1974), or changes in the mapping relations between genetic and phenotypic units (Wagner and Altenberg 1996; Brandon 1999). The classic evolutionary genetics representation that integrates genotypic and phenotypic state spaces, with arrows representing laws of transformation both within and between spaces (mathematized in some models and not in others), is shown in figure 1 (taken from Lewontin 1974). The evolutionary ‘‘three’’-step process in Lewontin’s diagram is broken down into four sets of transformational laws: ‘‘epigenetic laws’’ of development that map genotypes into phenotypes (T 1 ); laws of mating, migration, and selection that map phenotypes before selection into phenotypes after selection (T 2 ); ‘‘epigenetic relations’’ allowing inferences about distribution of genotypes of the surviving phenotypes (T 3 ); and Mendel’s and Morgan’s ‘‘genetics rules’’ that map genotypes of one generation into the next (T 4 ) (Lewontin 1974: 13). Notice that development is particularly difficult to interpret in this structure perspective. While selection and inheritance transform states within a space and represent processes as change over time, ‘‘development’’ in this representation, in one sense, concerns two sets of logical mappings (T 1 , T 3 ) between genotype and phenotype space, not causal processes that transform state distributions over time. In another sense, the causal process of development is confounded in both causal transformations within spaces: T 2 takes juvenile, preselection phenotypes into adult, postselection phenotypes; T 4 takes parental genotypes (G 2 ) into reorganized offspring genotypes (G 1'). The first complication of development often appears in the biological literature as a note that selection really operates on developmental vectors

202

Evolutionary Process Perspective

1 • The paths of transformation of population genotype from one gen-

eration to the next. G and P are the spaces of genotype and phenotypic description. G 1 , G'1 , G 2 , and G'2 are genotypic descriptions at various points in time within successive generations. P 1 , P'1 , P 2 , and P'2 are phenotypic descriptions. T 1 , T 2 , T 3 , and T 4 are laws of transformation. Details are given in the text. Reprinted from Richard C. Lewontin, The Genetic Basis of Evolutionary Change (New York: Columbia University Press, 1974), p. 14.

rather than on fixed adult trait states. The second complication is rarely acknowledged, but it has been argued that even Mendel’s theory—as distinguished from Mendel’s laws—is a theory of the development of hybrids (Griesemer 2000b, 2002a, forthcoming). Lewontin’s argument concerns ways in which evolutionary genetic models, which tend to be formulated either in phenotype or in genotype state space, inevitably confuse the causal character of their state variables, such as W (mean fitness), since these also depend on mappings between the two spaces. Thus, while genetic structure representa-

203

James Griesemer

tions simplify the task of formalizing selection and inheritance, they complicate the representation of developmental transformation. The converse probably also holds true. The dominant structure perspective on units of selection stems from Lewontin’s generalization of Darwin’s principles of heritable variance in fitness (Lewontin 1970). Lewontin abstracted from Darwin’s reference to organisms to produce a units analysis in terms of individuals that are organized in a compositional hierarchy of parts and wholes: genes inside cells inside organisms inside families and demes inside species and communities (Griesemer 2005). Insofar as this structure hierarchy is compositional, Lewontin’s state space representation for evolution frames genetic structure models in terms of a lower level of genetic units and a higher level of hierarchically organized phenotypic units. Lewontin’s approach reflects a theoretical perspective because the compositional hierarchy is taken as a given, a commitment to hierarchical modeling and the empirical studies it guides; for example, the study of selection at many phenotypic levels: genic selection, gametic selection, group selection, kin selection, and species selection. The existence of the compositional hierarchy itself is not a topic of investigation in these studies. To try to explore its evolutionary origin would be question-begging because the structuralist analysis of units assumes its existence (Griesemer 2000a, 2000c). The perspective is structural in that it takes the generalization of units to be a problem of extending Darwin’s principles beyond the level of organisms. The research programs packaged with this perspective are neo-Darwinian insofar as Darwin’s principles are interpreted in terms of models of evolutionary genetics; for example, population genetics models that represent trait heritability as a consequence of classical gene transmission and expression and which model natural selection as a force altering phenotype or genotype distributions within generations.3

Functionalism Function perspectives model phenomena by representing functions. The most familiar is the replicator-interactor perspective (Hull 1980, 1981, 1988; Dawkins 1983, 1999). The question of units of selection from a function perspective concerns which sorts of entities play the functional role(s) of replicators or interactors. A replicator is any entity in the universe of which copies are made (Dawkins 1983) or which transmits its structure directly or relatively intact in replication (Hull 1988). These definitions specify replicators in terms

204

Evolutionary Process Perspective

E E E E

G

G

A

A

A

A

G

G

G G

G G

2 • A theory of evolution. Figure and legend reprinted from John May-

nard Smith, John Maynard Smith on Evolution (Edinburgh: Edinburgh University Press, 1972), p. 40, with permission from Edinburgh University Press, Edinburgh.

of their functional role in a process—how they work or what they do—irrespective of their particular structural realization in any concrete case. Interactors are entities which interact directly with their external environments in such a way that replication is differential (Hull 1988). That these definitions appear to be circular—replicators defined in terms of replication, interactors in terms of interaction—reveals a fundamental limitation on function perspectives. Functional analyses depend on tacit reference to a process in which a goal is served, but they do not offer an account of any of those processes. A functional perspective that accounted for the processes by which the functions themselves evolved would be question-begging. Dawkins hints at a functionalist state space for models of evolution in his observation that his replicator concept is merely ‘‘extreme Weismannism’’ (1983: 164). Weismannism provides a state space for distinguishing functional roles of replicators (germ) from interactors (soma) in evolution, because selection on somatic interactors in virtue of their phenotypes has the consequence that replication of the associated replicators will be differential, as is shown in figure 2 (Maynard Smith 1972). In this Weismannist perspec-

205

James Griesemer

tive, environments (E) serve as filters on populations of interactors (A); some survive to reproduce and others do not. Replicators (G) that ride inside an interactor share that interactor’s fate. If they are ‘‘active replicators,’’ they influence the phenotype of the interactors in which they ride, so the probability that the replicator is copied (transmitted) is affected by phenotypic selection. If the replicators are also ‘‘germ line,’’ then they are copied (transmitted) to the next generation with differential probabilities. Weismannism separates causal processes of inheritance (transmission across generations, represented by the line of inheritance from G to G to G) from causal processes of development (causal production of A from G). Selection acts directly on A and thus indirectly on G. Because only replicators are transmitted to the next generation, in the form of copies or structure, replicators and not interactors are the units of evolution by means of natural selection.4 From this function perspective, processes of development and inheritance trace to a common cause: the germ/gene/replicator. Selection causes covariance between properties of the soma/phenotype/interactor of the parental generation and properties of the germ/gene/replicator in the next generation. No causal interaction between the processes of development and inheritance can be represented in this state space, however, because distinct functions appear as different vectors in the space. Only the coupling of statistical fates in the common cause suggests a single coherent process moving through time. This difference between hereditary transmission and developmental expression is manifest in the classical distinction between two roles of genes: the autocatalytic function and the heterocatalytic function (Wimsatt 1981). Although the function perspective represents all three causal processes—selection, development, and inheritance—their separation in function state space makes it difficult to represent their causal interaction. In the classical neo-Darwinism of the evolutionary synthesis, genes are inherited, organisms are selected, and populations evolve. The function perspective seeks generalization of units by abstraction from the genotypic and phenotypic functions in evolution: to replicators as the functional entities that serve the inheritance function and to vehicles (Dawkins) or interactors (Hull) that serve the selective interaction function. This kind of functionalism is a theoretical perspective because the division of functional labor in evolutionary processes between genotype and phenotype is taken as given in the theoretical and empirical studies it guides; for example, the extended phenotype and inclusive fitness accounts through which Dawkins argues for a gene’s-eye view of evolution. The functional dichotomy itself is not subject

206

Evolutionary Process Perspective

to evolutionary explanation in these studies. The perspective is functional insofar as it generalizes units beyond Darwin’s focus on organisms in terms of functions that range over all kinds of structural units. That is, whether a structure functions as a unit of selection depends on whether it plays the relevant role of replicator and/or interactor, not where it sits in the structural hierarchy. One motivation for a process perspective stems from a theoretical question similar to that of structuralist and functionalist investigations of units of evolution: How general are Darwin’s principles? Rather than probe the applicability of Darwin’s principles up and down the compositional hierarchy or in and out of the replicator and interactor functions, however, my concern is with the evolution of the hierarchy itself and with the evolution of the genotype/phenotype distinction itself. If these are to be explained evolutionarily, in a non-question-begging account, then both structure and function perspectives on units of evolution, by themselves, are inadequate: they both assume something to be explained. Biologists describe this something as the problem of evolutionary transition—the evolution of new (structural and functional) levels of biological organization (Buss 1987; Maynard Smith and Szathmáry 1995; 1999; Michod 1999). The reason a process perspective is needed is that structuralism and functionalism do not analyze directly the processes by which the ‘‘scaffolding’’ they assume evolved (Bickhard 1992). How did the replicator and interactor functions emerge from entities which did not have them? Indeed, how could they? If evolutionary units are analyzed in terms of the replicator/interactor distinction, then that function perspective cannot answer the question. This becomes evident when analyses of replicators and interactors are probed. Dawkins assumes a process of copying in order to analyze the replicator function: ‘‘a replicator is any entity in the universe of which copies are made.’’ Hull assumes a process of replication: ‘‘replicators are entities that transmit their structure directly in replication.’’ Both appeal to textbook molecular biology to ‘‘explain’’ these processes. However, textbook molecular biology does not seriously address the evolutionary origins of replication, and the mechanisms of xerographic photocopying undermine Dawkins’s analogy, so question-begging about evolutionary transition is avoided only by silence.5 Structuralists fare no better. Entities at a given compositional level of organization are units of selection if they have the properties of heritable variance in fitness at that level. But what is it to be at that level and how did entities get there? The existence of the levels is assumed, not explained, because Darwin’s principles, as many critics of neo-Darwinism have complained, were

207

James Griesemer

designed to account for evolutionary change within levels, not for the evolutionary origin of levels.6 If life evolved from nonlife, then we certainly cannot appeal to the highly evolved separation of roles dna and protein now play in order to explain how genotypes became distinguished from phenotypes. There was no dna at the origin of life or for a long time afterward, and, arguably, there was not any rna either. Likewise, if multicellularity evolved, then we cannot just postulate multicellularity and ‘‘explain’’ it by appeal to its adaptive advantages. The postulation step is barred from research programs that wish to account for origins. The root of these problems is conceptual: the genotype/phenotype distinction and the assumption of an already existing hierarchy of compositional levels of organization make it all too easy to explain origins as ascent up the preexisting developmental ladder (genotypes are ‘‘genetic programs’’ for phenotypes) or the preexisting evolutionary ladder (life evolves ‘‘up’’ the hierarchy).7

Process Perspectives Process perspectives model phenomena by representing processes rather than players in structural or functional terms. Processualists identify units of evolution with processes rather than with objects or functions. Here, I discuss two process approaches to development and evolution: developmental systems theory and process-structuralism. In the next section I discuss my own process perspective, the reproducer.

Developmental Systems Theory Developmental systems theory takes the units of heredity and evolution to be developmental systems and takes development to constitute a collection of interactions among many developmental resources organized in such a way that life cycles—that is, the developmental interactions—recur in subsequent generations (see Oyama, Griffiths, et al. 2001: 2, table 1.1). It denies explanatory privilege to any one kind of developmental resource, such as genes, and recognizes ‘‘control’’ of development to be distributed and decentralized. It takes the explanatory basis of evolutionary change to be historically contingent natural selection for interactions in development, in contrast to the rationalist, universal laws of process-structuralism. Like process-structuralism, it takes the main goal of evolutionary biology to be

208

Evolutionary Process Perspective

the explanation of transgenerational stability and transformation of developmental form. Developmental systems theory’s empirical focus is on characterizing the full array of causally relevant developmental resources and interactions and, more recently, on the subsumption of niche construction as a developmentalist research strategy for modeling ‘‘environmental’’ resources in interaction with ‘‘internal’’ organism resources (Laland, Odling-Smee, et al. 2001; Oyama, Griffiths, et al. 2001). Developmental systems theory has been criticized as too radically holist (Sterelny, Smith, et al. 1996), too willing to ignore developmental and hereditary boundaries between organism and environment (Keller 2001; Sterelny 2001), and too willing to deny distinctions between biological and cultural evolution (Griesemer 2000a). Developmental systems theorists have depicted a process state space in virtue of which they argue that whole developmental systems, rather than organisms-in-environmental-contexts, are replicators (Griffiths and Gray 1994: 285), as illustrated in figure 3. Instead of points, each elementary unit is a line segment indicating a process. Causal interactions among such processes are organized into a developmental system which behaves as a unit life cycle in replication. In this perspective, state components are distinguished according to kind of developmental resource: persistent, collectively generated, parental, developmental, and self-generated. The space has a dynamic order (persistent resources have a different dynamic than self-generated resources), so one could think of the different types of resources as constituting different ‘‘channels’’ in a life cycle.8 However, the characteristics of the causal interactions (arrows) are not well defined. In particular, whether the types of causal interactions that hold developmental systems together are due to flows of energy or matter or perhaps something else entirely is hard to specify. Because of this, the modes of material inheritance and delimitation of offspring life cycles from parental life cycles which are peculiar to the biological realm are hard to represent in this kind of state space (Griesemer 2000a).9

Process Structuralism Process-structuralism emphasizes laws and generic (i.e., nonselective) outcomes of developmental processes as the basis for a rational and mature science of biology, a biology that offers ‘‘exact analyses, which describe necessary and sufficient conditions for a process and so result in well defined solutions’’ (Goodwin 1989: 91).10 This perspective takes field relations that undergo processes of transformation as its objects of study. Because fields rather than functions are taken as the complement to structure, process-

209

James Griesemer a. "Persistent" resources

b. Collectively generated resources c. "Parental" resources

d. Developmental processes

e. Selfgenerated resources

3 • Causal influences in four asexual generations of a lineage of develop-

mental processes. Each arrow represents multiple inputs. Influence of each resource is contingent on the presence of the others. The effects of temporal order of interaction have been overlooked. The broad categories of resources are not intended to be exhaustive, and are made largely for convenience of exposition. Figure and legend reprinted from P. E. Griffiths and R. D. Gray, ‘‘Developmental systems and evolutionary explanation,’’ J. Philos. 91(6) (1994): 277–304, p. 285, with permission from the Journal of Philosophy, New York.

structuralism can draw on the physical sciences in its mathematical models, but at the cost of treating natural selection and its objects as just one more dynamic system. Although there are considerable mathematical virtues in this approach, it puts a substantial burden on explaining how biological processes gain their distinctively biological, emergent character. Rather than a ‘‘historical science,’’ Goodwin notes, with ‘‘species morphology being the result of random variation and natural selection of functionally adapted forms[,] . . . [s]tructuralism assumes that there is a logical order to the biological realm and that organisms are generated according to rational dynamic principles’’ (Goodwin 1989: 91). It offers an account of evolutionary units quite different from functionalist, historical, ‘‘genocentric’’ biology, one in which the morphogenetic field is treated as ‘‘the generative unit in development

210

Evolutionary Process Perspective

and evolution’’ (Goodwin and Trainor 1983: 87). Evolution is understood from this perspective in the same way that a physicist understands it: as the ‘‘time-dependent exploration of [a] set of possibilities under internal (genetic) and external (environmental) parametric variation’’ (Goodwin 1989: 96). The possibilities are given by ‘‘generic states or forms . . . whose distribution in the space of developmental trajectories defines the set of possible forms’’ (Goodwin 1989: 96). Fields can be interpreted as relations among parts of entities (such as organisms). Universal physicochemical laws governing the developmentally invariant fields are taken to be the explanatory basis of biology. The explanatory goal of process-structuralism is to discover the organizing principles of generative processes that are expressed in the field equations (Goodwin and Trainor 1983; Goodwin 1990, 1994; Depew and Weber 1995; Webster and Goodwin 1996; van der Weele 1999). The empirical focus of processstructuralism is modeling and experimentation based on hypothesized field equations as well as simulation and measurement of parameter values, following the example of field theories in physics. Goodwin and Trainor, for example, offer a ‘‘generative model’’ for the pentadactyl limb which involves a Hamiltonian field energy equation. Simulations generate sequences of field solutions that are taken to represent points in a developing limb field where cartilage condensation will occur (Goodwin and Trainor 1983: app. 95 and p. 77). Other simulations model the development of whorl structures on the stem of the unicellular green alga, Acetabularia acetabulum (see Goodwin 1990).11 Proponents of developmental systems theory have taken issue with process-structuralists’ views on rational biology. Griffiths argues that process-structuralism fails to challenge a Darwinian perspective, noting that ‘‘the generic forms that exist in nature may be a tiny subset of the possible generic forms that could have been created by the historical design of alternative developmental systems. . . . The developmental system could have been any one of a number of ways depending on the particulars of evolutionary history’’ (Griffiths 1996: S6). However, we can just as well take Griffiths’s critique constructively, pinpointing a lack of robustness of developmental phenomena to processstructuralism’s idealizing assumption that the possible generic forms are essential, intrinsic properties (Griffiths 1996: S5). If perspectives are viewed as competitors, the critique threatens one of the main goals of processstructuralism: to reorient biological research to include the search for generic forms and generative principles and laws to explain them. If perspectives co-

211

James Griesemer

operate, the critique identifies a problem to attack via robustness analysis and addition of perspectives, rather than criticism and elimination of perspectives. Adding the search for generic laws need not deny historical contingency. For example, Griffiths raises the question: What is the relation between the portion of ‘‘morphospace’’ filled by possible generic forms and the portion specified by natural selection of historical forms? Process-structuralism presumably takes the latter to be a subset of the former. Neo-Darwinism presumably takes the reverse to be the case. The relations between the predictions of process-structuralism and a Darwinian view like Wimsatt’s generative entrenchment model, which retains a selectionist basis, are crucial for establishing a robust account of the evolution of generic forms (Schank and Wimsatt 1988; Wimsatt and Schank 1988; see Griffiths 1996 for discussion; Wimsatt 1999). To evaluate these predictive relations, we need an account of evolution that can explain how evolutionary order is generated as well as how it is maintained and embellished. A more explicit perspective on processes can help identify the contexts in which the evolution of evolvability is creative, productive, and enabling, and not merely a barrier to full selective exploration of the state space of possible forms. Evaluating types of abstraction from process made by structuralists and functionalists for modeling purposes is one means to a heuristic strategy for producing a more robust account of inheritance, development, and evolution. The systematic failure of a given class of models and perspectives, when checked against other models and perspectives making different assumptions and commitments, provides clues to a better understanding of nature. Just as developmental systems theorists have sometimes viewed processstructuralism as a competing perspective, proponents of structuralism and functionalism have often treated their perspectives as competitors for the ‘‘right’’ way to view evolution. For example, Dawkins (1976) and Williams (1966) argue that structuralist accounts of group selection miss the mark because only replicators can play the functional role of important units of selection, and groups of organisms are not replicators (and maybe not even interactors). Since genes and memes (units of cultural transmission) are the only replicators, according to Dawkins, groups cannot be significant units of selection, however much structuralism points to the ‘‘theoretical’’ possibility of a hierarchy of units. In practice, however, function and structure perspectives can cooperate when the assumptions of both are met in particular model-applications: interactor selection among highly evolved organisms in laboratory and field

212

Evolutionary Process Perspective

experiments can be represented in models that take for granted both the existence of the compositional hierarchy and the functional genotype/phenotype distinction. This cooperation has been endorsed by many as a harmless sort of pluralism: whether one views an evolutionary process from the structuralist or the functionalist perspective may just depend on what is convenient for the particular research question at hand.12 Less studied is how the apparent cooperation may lead to a false sense of robustness: rather than complementing one another so as to reveal empirical results that are independent of differing assumptions, the idealizing assumptions shared by the two perspectives may mistakenly be endorsed as plain truths just because they are shared. The genotype/phenotype distinction and the compositional hierarchy, though distinctive foci of the functionalist and structuralist perspectives, may yield pseudorobust results because both are derived from the same grounding in neo-Darwinian principles. In order to detect pseudorobustness—false consilience—a third anchor-point is needed, a perspective from which to ‘‘triangulate’’ results rather than reinforce potentially misplaced trust in dichotomies. Reproduction is a process whose units are both hereditary and developmental, populations of which can evolve by means of natural selection. Thus, an integrated science of genetics, development, and evolution requires a robust account of the units of reproduction. In the next section I describe a process perspective on reproduction which will be used in the final section to formulate questions about the robustness of our genetical understanding of evolution at two levels: robustness to the assumptions of particular process perspectives and robustness to structuralism, functionalism, and processualism.

The Reproducer: A Process Perspective on Development, Inheritance, and Evolution

Evolutionary Transitions and Replicators Theories of evolutionary transition are designed to explain the evolutionary origin of new levels of organization (Buss 1987). They have generally been framed in terms of changes in the mode of replication or transmission of genetic information of biological units. Units that replicated independently prior to a major transition replicate as dependent parts of higher-level, independent replicators afterward (Maynard Smith and Szathmáry 1995, 1999). This kind of analysis captures genetic aspects of the transition process well by

213

James Griesemer

identifying mechanisms that may have played a role in the evolution of chromosomes and multicellularity, such as genetic conflict suppression, contingent irreversibility, duplication-divergence, and developmental bottlenecks. A fundamental problem for this type of theory is to articulate the concept of a replicator. Most theorists have followed Dawkins’s lead and defined replicators as any entities in the universe of which copies are made. Active replicators influence, through their phenotypic effects, the probability that they will be copied. Germ-line replicators are the ancestors of indefinitely long lineages of replicators. Maynard Smith and Szathmáry, for example, adopted Dawkins’s ‘‘active germ-line replicator’’ in their 1995 theory of major evolutionary transitions. The Dawkins replicator concept poses problems for transition theory because, according to Dawkins, there are only two kinds of such replicators: genes and memes. Other candidates fail to satisfy all the necessary properties of longevity, fecundity, and copy-fidelity. The primary downfall of most candidates is that they lack the ‘‘digital’’ properties that give nucleic acid replicators and symbolic memes their copy-fidelity. This is a problem for transition theory because if only genes and memes are replicators, then the number of major transitions is severely limited: once genes became dependent replicators, there could be no further transitions without the evolution of new higher-level independent replicators to become dependent in the next transition. The evolution of replicators is apparently very difficult and, according to Dawkins, has not happened since the evolution of cellular life, except for memes. While this view rightly emphasizes that major transitions are not everyday products of adaptive evolution (longer legs, more efficient enzymes) or mere increases in adaptive complexity, however dramatic (wings, lungs, eyes, homeothermy, internal fertilization), it is nevertheless overly restrictive. According to Maynard Smith and Szathmáry, there have been about eight major transitions in the evolution of life on earth. However, we do not find evidence of seven levels of dependent replicators. Rather, as they later acknowledged (Szathmáry and Maynard Smith 1997), we find evidence of one level of dependent replicators—the genes—and the rest involve ‘‘reproducers’’ up to the transition to human language. This raises the question of the relation between replicators and reproducers.

Reproduction and Multiple Inheritance Systems An amendment to transition theory that revises the concept of replicators addresses this problem (Griesemer 2000c). Suppose evolutionary transition

214

Evolutionary Process Perspective

involves a change in the mode of reproduction, but not necessarily of replication. An evolutionary transition is a process in which entities that were capable of independent reproduction before the transition reproduce only as parts of a larger whole after it (Szathmáry and Maynard Smith 1997; Griesemer 2000c: 79). Mode changes in reproduction result from the evolution of development, so a theory of reproduction must integrate an account of development with that of inheritance. Suppose further that replicators, according to this amendment, are a special class of ‘‘reproducers.’’ Thus, the above description of evolutionary transitions can be understood as a generalization of the Maynard Smith-Szathmáry account, and there is no conflict between Dawkins’s conclusion that origins of replication are rare and Maynard Smith and Szathmáry’s claim that there have been about eight major transitions. The proposed amendment to transition theory requires an analysis of reproduction and an account of how replicators turn out to be a (rare) special case of reproducers. The reproducer analysis offered below suggests a stage model for evolutionary transitions in which inheritance systems evolve at each new level of reproduction. An important implication is that, if the amended theory is correct, there must be a multiplicity of inheritance systems. There is abundant empirical evidence for such a multiplicity beyond the genetic inheritance system. These other systems have recently been classified into genetic, epigenetic, behavioral, and symbolic inheritance systems (Avital and Jablonka 2000; Jablonka 2001). The conceptual basis of inheritance systems remains in doubt, however, because of the dominance of the genetic replicator model for analyzing inheritance processes. Philosophers and scientists from the developmental systems perspective who, like me, want to reintegrate development into Darwinian evolutionary theory have questioned the value of extending replicator talk to other inheritance systems in the search for a more robust theory of complex developmental systems. Developmentalists are suspicious of the ‘‘language of separate ‘inheritance systems’ ’’ (Griffiths 2001: 406), in part because such language has long been grounded in the notion of the replicator as the sole bearer of genetic information. Those working on epigenetic, behavioral, ecological, and cultural inheritance have struggled to overcome this restrictive legacy of explanatory privilege for genes in biological theory. Developmental systems theorists prefer to identify whole developmental systems as replicators to emphasize the complex, nonpartitionable interactions of many developmental resources inside and outside the body that must occur for life to cycle. After I sketch the reproducer analysis I will consider in more detail the variety of process perspectives that might help ‘‘triangulate’’ on the fundamental assumptions of

215

James Griesemer

neo-Darwinian units analysis which may eventually lead to a more robust picture of evolutionary processes.13

Reproducer Analysis The analysis begins with Szathmáry and Maynard Smith’s (1993: 198) account of the units of evolution in terms of three principles, described below: 1. Multiplication. If there is an entity A, then it must give rise to more of the same. [emphasis added]. 2. Heredity. Like begets like: A-type entities produce A-type entities, Btype entities produce B-type entities, and so on. 3. Variability. Heredity is not exact; occasionally A-type objects give rise to A’-type objects (it may even be that A’ = B).

If objects of different types have a hereditary difference in their fecundity and/or survival, the population undergoes evolution by natural selection. Note that unlike the Lewontin analysis mentioned above, the SzathmáryMaynard Smith analysis does not include fitness among the three main principles. It is an analysis of units of evolution. The addition of a fitness principle (fecundity or survival value) analyzes evolution by means of natural selection. The principle of multiplication invokes a principle of development, because multiplication must be of entities of the same kind. Since any two things resemble one another with respect to numerous kinds, biological multiplication must result in entities of the same relevant kind, which it is the task of a theory of development to specify. Development in a minimal sense from an evolutionary point of view is the acquisition of the capacity to reproduce. Acquisition of a particular trait that plays a causal role in a mechanism of development can be analyzed in terms of this general, developmentally acquired capacity to reproduce regardless of the specific ways or degrees in which traits contribute to it. Differently put, traits or parts that develop are means to the evolutionary end of reproductive capacity. Multiplication is the process by which more entities are produced. In that respect the analysis provides a schema for models from this process perspective. This notion of multiplication, however, is too general to serve as an account of biological reproduction. First, biological multiplication is of material objects. Second, the relation between parents and offspring is not merely one of resemblance, but rather is one of material overlap. Offspring are made from physical parts of the parents; they are not merely similar ob-

216

Evolutionary Process Perspective

jects made from wholly distinct materials. Otherwise, copying would be the appropriate concept for biology rather than multiplication (Griesemer 2005, in preparation). The reason biological multiplication involves material overlap of parents and offspring is due to the demands of development (Griesemer 2000a, 2000c). Development is minimally the acquisition of the capacity to reproduce. For multiplication to result in more entities of the same relevant kind, the offspring must be organized so as to have—autonomously from their parents—the capacity to develop. That is, offspring must be born with the capacity to acquire the capacity to reproduce. While it is conceivable that this grade of organization could be transmitted through unorganized bulk matter carrying only the capacity for spontaneous self-organization, it is not probable that such a system of multiplication could compete with biological systems in which highly organized material propagules form the basis for the origin of new individuals. Hence biological reproduction is ‘‘sample-based’’ (Sterelny 2001), and material propagules, not mere informational programs, explain the offspring’s development. Even Dawkins, who generally attributes all significant biological causal powers to replicators (Dawkins 1983: 164), admits that development must be ‘‘bootstrapped’’ by organized developmental propagules in order for replicators to function (Dawkins 1995). The issue raised here is not the empirical question of whether development in life cycles is favored by selection because it leads to complex adaptations (Dawkins 1983: ch. 14). It is rather the observation that Maynard Smith and Szathmáry’s concept of multiplication entails a concept of development. According to this analysis, developmental and reproductive capacities are locked together in a recursive relationship. Reproduction entails development and development entails reproduction. The recursion halts in a condition of ‘‘null development’’ such as chemical autocatalysis. Autocatalytic molecules just have the capacity of autocatalysis; they need not acquire it through ‘‘development.’’ 14 The capacity is, of course, dispositional and requires appropriate triggering conditions (input of ‘‘food,’’ ionic conditions, temperature and pressure range, enzymes, and so forth) in order to be realized. More important, heredity will tend to be exact rather than variationpropagating whenever development is null, because it is very difficult to produce variant autocatalytic cycles that are not ‘‘lethal’’; that is, that tend simply to halt catalysis rather than yield variant cycles (Gánti 2003). Instead, variation and true inheritance must ‘‘emerge’’ in the developmental process of acquiring the capacity to reproduce. Since null developers do not acquire reproductive capacity in development, the opportunity for evolution through

217

James Griesemer

generated variation is lacking. If chemical autocatalysis is the basis for the origin of living systems from nonreproductive chemical systems (Gánti 1979, 1987, 1997, 2003; Dyson 1985), theories of the origin of life must explain the transition from exact to inexact heredity and from autocatalysis with null development to reproduction with development. Copying—the core of Dawkins’s analysis of replicators—in contrast to reproduction does not entail material overlap. The resemblances implied by copying processes are generally insufficient to meet the evolutionary minimum requirement, unless the copying occurs in the context of a more inclusive process of reproduction as outlined above.15 It is possible to make the ‘‘gene’s-eye view’’ work as an analysis of units of replicator evolution only insofar as replication (interpreted as a copying process) takes place in host reproducers serving as interactors. This is so because the process of replication cannot take place without the development of the capacity to replicate, and nothing in the standard analyses of replicators implies their development. If development evolves so that the behavior of material propagules transmitted from the parent involves adaptations and not just physicochemical properties of nonreproductive systems (such as diffusion), then we can call the reproduction process an inheritance process and its units ‘‘inheritors.’’ Further, call the genetic inheritance system a system of replicators insofar as the highly evolved genetic mechanisms that play a role in development have the particular character of a coding system. One possible necessary condition for a coding system is evolved reproducers having the properties of ‘‘digital’’ or ‘‘unlimited heredity’’ in which the number of possible states of the system vastly outnumbers the actual individuals in any reasonable population (Jablonka and Szathmáry 1995; Szathmáry and Maynard Smith 1997; Szathmáry 1999). This combinatorial structure is put to use in evolution by means of certain sorts of processes in development. A coding process is a process that generates a mapping between a code source and a code recipient, for example, between a dna molecule and a protein molecule. Only at this grade of evolution can there be a genotype/phenotype distinction of the modern sort. Note that in keeping with the recursive relation between reproduction and development, genetic coding involves not only the occurrence of molecular events within a generation but also the multiplication of the coding system itself: without the inheritance of ribosomes, synthetase enzymes, and the rest of the decoding apparatus there is no sense in which dna carries a code (Griesemer in preparation). Thus, rather than thinking of ‘‘replicator’’ as a generalization of the gene or the genotype concept, replicators—units of replication—are a special class

218

Evolutionary Process Perspective

of inheritors—units of inheritance—which in turn are a special class of reproducers—units of reproduction—which in turn are a special class of multipliers. Replicators are the most specialized units of a hierarchy of concepts. This is hardly the generalization we need for doing the work of evolutionary units analysis. Just as the evolution of replicators makes them the most specialized and highly evolved units of heredity, the development of modern replicators is deeply embedded in an evolutionary hierarchy of levels of reproductive organization. Replication is a highly context-dependent developmental process because successive evolutionary transitions have altered their mode of transmission and information storage—their mode of development—several times over. Far from being master molecules, genes are prisoners of development locked in the deepest recesses of a hierarchy of prisons. In general terms, one can model evolutionary transition as a progression of developmental ‘‘modes’’ from the general to the specific: from reproducers to inheritors to replicators. Rather than ‘‘point event’’ transitions from independent to dependent replication within larger wholes, evolutionary transitions are extended processes with several distinct stages (Griesemer 2000c). The first stage is the emergence of a new level of reproducers from lower-level processes. The origin of a new level of reproduction does not require a fancy developmental process, as would be entailed by the origination of a new level of replication. Rather, the stabilization and maintenance of a level of reproduction require the evolution of sophisticated (i.e., non-null) developmental mechanisms to block conflict from below and promote cooperation above.16 Replicator-based accounts of transition seem to imply that this evolution of development must coincide with the emergence of a new level of reproduction, which is part of what makes transitions appear improbable. In the second stage, evolution of mechanisms of development that enforce or encourage cooperation—for example, by equalizing genetic interests, by policing rogues, by enforcing centralized control on the means of reproduction, or by some means of blocking outlaw behavior—would help maintain the new level of reproducers (Maynard Smith and Szathmáry 1995; Michod 1999; Sterelny 2001). The evolutionary transition literature, however, with its replicator-based units analysis, does not emphasize the significance of these mechanisms as the evolution of the development of emergent reproducers. The evolution of developmental mechanisms in stage 2 implies that the now-dependent reproducers at the lower level form parts of what I called inheritors (at the higher level). Propagules at the new level must transmit the evolved developmental capacities via the material overlap of developmental

219

James Griesemer

mechanisms at the new level. The contingencies of propagule production and ecology thus place important constraints and conditions on the evolution of mechanisms of development at the new level. These mechanisms may behave as inheritance systems transmissible in parallel with the transmission system of the original reproducers, or they may be propagated through transmission of capacities of lower-level components. To the extent that complex, adaptive evolution of development must occur to stabilize the new level, the second stage of evolutionary transition will result in the evolution of a system of inheritors from the developmental mechanisms that drove the first stage, for example, dna methylation and other epigenetic inheritance systems (reviewed by Jablonka and Lamb 1995). These multiply with material overlap of propagules, conferring the capacity to develop, just as nucleic acid genes do (Griesemer 2002b). The third stage of evolutionary transition occurs when mechanisms of development evolve into a coding system. It is commonly believed that there are only two such coding systems: the nucleic acid system and human language, though cell surface glycoproteins may have enough combinatorial structure in their carbohydrate branching patterns to constitute a potential coding system (Palade 1983; Drickamer and Carver 1992; Sharon and Lis 1993). Nevertheless, suppose the conventional belief is correct. The staging of evolutionary transitions from the reproducer perspective can be framed as a question of heterochrony: why do some transitions appear to go through all three stages whereas others appear to compress or even skip stages? Why did the transitions leading to cells involve all three stages, including the evolution of the nucleic acid coding system, while the transition(s) to multicellularity seems to have involved only the evolution of epigenetic inheritance systems for cell heredity? The evolution of a new level of replicators appears not to have been required for multicellularity. And why did the transition to human language seem to involve replicators while intervening transitions apparently did not? If this amended theory is correct, the transition process requires evolved mechanisms of development to stabilize units at a given level before a new level of autonomous reproducers can emerge. Thus, according to the reproducer analysis, new inheritance systems must evolve at a level before evolutionary transition can occur again. My purpose is not to promote the reproducer perspective as a replacement for either structuralist or functionalist perspectives on neo-Darwinism, or as a rival to other process perspectives such as process-structuralism or developmental systems theory, but rather to highlight shared assumptions of structuralism and functionalism that lend an air of robustness, but which may

220

Evolutionary Process Perspective

not serve the broad project of generalizing Darwinism and may hinder the project of exploring the limits of Darwinism. Since the limitations of shared assumptions cannot be investigated within the confines of the perspectives at issue, I outlined the reproducer perspective in order to have a divergent standpoint from which to evaluate the others (and vice versa). The key problem with both structuralist and functionalist perspectives is that they lack means to represent the process of development in relation to the processes of primary concern to them: heredity and selection. The genotype/phenotype distinction seems to justify treating development as a black box: the only developmental requirement in neo-Darwinism is a mapping relation from genotype to phenotype. The simplest maps in the models of classical biochemical, population, and quantitative genetics were additive. Each gene has a small additive effect on a phenotype, and all the complexities are partitioned according to an analysis of variance into main additive effects and interaction terms (for dominance, epistasis, genotype-environment interaction, and so forth). No matter how complex the mechanistic details get, there will be some mapping relation. Theoretical evolutionary genetics theory can go on independently of developmental biology in the sense that simple maps can be explored without knowing in what respects and degrees they fail to represent actual developmental processes. Empirical results in developmental biology can be incorporated into theoretical evolutionary biology simply by tinkering post hoc with mapping functions. Of course, the evolution of mapping functions themselves—the ‘‘evolvability’’ of the phenotype— cannot be fully explored by neo-Darwinian models because alternative developmental mechanics must be made explicit and that is ruled out by the neo-Darwinian abstraction (see, e.g., Wagner and Altenberg 1996).

Multiplicity of Perspectives, Robustness, and the Enrichment of Science A process perspective in science takes processes to be the primary phenomena to be described (observed, represented, explained, predicted, understood, experimented on, measured, conserved, destroyed, engineered). Structuralism takes objects (entities, things, concrete historical individuals) as its primary subjects, to be described in terms of structure or organizational properties abstracted from functions and processes. Functionalism takes its primary subjects to be functions (roles, purposes, goals, intentions) abstracted from objects that realize them and processes they undergo. Many of

221

James Griesemer

the state spaces pictured above trace in some way to the structuralist tradition in genetics, in which evolution is represented as state transformations of biological objects. Although the transformations themselves can be treated as processes, much of the theoretical apparatus of this kind of research abstracts functions or relations among objects at different times or places from the processes that ‘‘connect’’ them in space-time. Process is pushed into the methodological background, for example, into husbandry practices for rearing and tending the organisms whose transformations in development, heredity, and evolution are tracked and represented (Griesemer 2000b). Since development ‘‘happens’’ in that background, foregrounded relations of heredity have become the subject of a science of genetics which is formally as well as methodologically and empirically separated from embryology. Embryology, in turn, foregrounds development and backgrounds hereditary transmission as the technical means by which developmental ‘‘material’’ is generated for study (Griesemer forthcoming). A theoretical perspective can guide the conduct of modeling (Griesemer 2000a) without much regard for the subjects it pushes into the background. In the units of selection controversy in evolutionary biology, for example, Lewontin’s account of natural selection operating on a hierarchy of individuals is made from a structure perspective in which entities at any level of organization with appropriate structure—that is, they satisfy Darwin’s principles—evolve by means of natural selection (Lewontin 1970; see Griesemer 2005). How these entities develop is of little concern in classical selection theory. Dawkins’s and Hull’s distinctions between replicators and vehicles or interactors identify as units anything which serves or has those identified functions (Hull 1981, 1988; Dawkins 1983). Neither Dawkins nor Hull takes much trouble to analyze the processes through which the functions are served (copying or replication). A perspective on units of evolution like Maynard Smith’s, by contrast, takes the process of organic multiplication as primary, and the structures or functions that instantiate, cause, or explain it as secondary givens (Maynard Smith 1988; see Griesemer 2000c for analysis). To the extent that structure, function, and process together provide an exhaustive taxonomy of phenomena, somewhat as Aristotle’s four causal categories (material, efficient, formal, and final) were supposed to exhaust causality, research packages (Fujimura 1996; Gerson 1998) must at least tacitly make assumptions and commitments in all three respects, though they may take one or another as the primary focus. One or another of these three kinds of perspectives—structure, function, process—may characterize a particular science in a given historical period, distinguish among sciences or subdisci-

222

Evolutionary Process Perspective

plines, or sustain disciplinary rhetoric through transdisciplinary controversies. Their proponents may favor attention for contingent and particular historical reasons to one perspective over the others, even if they recognize that a complete biology would include them all. Perspectives are of ‘‘transcendent’’ significance because they afford means of assessing the robustness of scientific achievements to the idealizing assumptions of modeling strategies and particular scientific research programs. Perspectives do this by specifying what will count as relevant respects and significant degrees in which the models of a family are compared and evaluated, for example, in their predictive successes and failures according to the conventions, procedures, and standards of the research package. Thus, just as we seek scientific results that are robust to the idealizing assumptions of particular models used to represent particular phenomena, we seek scientific achievements—collections of results—that are robust to the assumptions of whole modeling strategies and research packages. This higher-order robustness can be assessed only if a scientific community pursues phenomena from a variety of perspectives with a variety of packages and compares them. It is not enough merely to compete. Process perspectives add a crucial third leg to biological research interpreted in terms of the more familiar structure and function perspectives. In general, consideration of scientific strategies, results, and achievements from only two perspectives tends to promote conceptual dichotomies and the perception that a choice must be made between better and worse or right and wrong interpretations. But perspectives are not the sorts of things that can be right or wrong, better or worse, per se. Rather, they are right or wrong, better or worse for some particular purpose of some particular investigator. Like philosophy itself, theoretical perspectives in science are stances, not claims (Van Fraassen 2002). Since epistemic access to nature by means of scientific research is always limited, partial, contingent, constructed, social, and pragmatic, ‘‘our truth,’’ as Levins so aptly notes, ‘‘is the intersection of independent lies’’ (Levins 1966: 426). Thus, the goal of robustness analysis at all levels is to factor out the idealizations and falsifications that we introduce for the purposes of investigation, not to champion one perspective over all others because it avoids the particular idealizations of the others. False models are our means to truer theories (Wimsatt 1987). A theoretically adequate science of genetics should value variety of theoretical perspectives as much as it values variety of theoretical models. Just as variety of models is essential to robust theories, varying the perspective is a plausible strategy for developing robust concepts of theoretical units such as

223

James Griesemer

the gene, the organism, and the developmental system. I use this argument to promote consideration of ‘‘process perspectives’’ on genetics in addition to traditional structure and function varieties, and to compare several perspectives, including neo-Darwinism, process structuralism, and developmental constructionism, with my own process perspective on reproducers as genetic units. These perspectives collectively yield insights into the nature of genetic units that none by themselves could. Thus, my argument differs from those critics of neo-Darwinism who offer their perspectives as superior alternatives. In my view, it is the theoretical collective, not the individual investigator, that achieves or fails to achieve robust results. Thus variety of models and perspectives is of value in itself for good science. A hallmark of many of the biological perspectives considered here is that they were formulated in reaction to other perspectives, and thus their lists of complaints help constitute the perspective. This rhetoric of opposition, however, makes criticism too easy and cooperation in pursuit of robust scientific understanding needlessly difficult. Perspectives formulated as stances against others drive concepts toward mere contradictories and dichotomies: for or against genes, for or against pan-adaptationism, for or against biological laws, for or against explanation by internal mechanisms. The model assumptions tell a different story: representation of inheritance in terms of genes does not exclude representation of inheritance in terms of genes plus environments or interacting developmental resources. The compatibility or incompatibility of model representations of phenomena depends on much more than the oppositional perspectives from which they are constructed. However, cooperation across perspectives is hindered by identification of research programs with oppositionally defined perspectives. Process-structuralists complain, for example, that neo-Darwinists place too much explanatory weight on natural selection without examining the extent to which dynamic living systems generate and transform phenotypic order ‘‘for free,’’ obviating the need for ‘‘unparsimonious’’ appeals to Darwinian selection on ‘‘underlying’’ genetic sequence variation (Kauffman 1993; Webster and Goodwin 1996). Developmental systems theorists complain that replicator/interactor functionalists, neo-Darwinians, and extended replicator theorists ignore the extent to which processes beyond the skin, such as culture and niche construction, play a strong role in the developmental interactions that structure evolutionary units. These phenomena require identification, they argue, of the replicator function with whole developmental systems rather than with a single resource such as genes (Oyama 1985; Griffiths and Gray 1994, 2001; Oyama, Griffiths, et al. 2001). Extended repli-

224

Evolutionary Process Perspective

cator theorists resist the apparent holism of developmental systems theory on grounds that there is a special and distinct role which genes play but many environmental resources do not (Sterelny, Smith, et al. 1996; Sterelny 2001). Neo-Darwinians express surprise that others are unimpressed by the success of developmental genetics in its reductionist account of evolution in terms of a combination of molecular sequence variation and gene regulation (Coyne 2000). If such critics did understand the progress being made in real science (i.e., molecular developmental genetics), they would, of course, believe. In this essay I have stepped back from these debates and controversies not to assess which perspectives are winning or losing the war for attention and conceptual domination, but rather to reconsider the competitive process by which fundamental concepts and research programs are vetted. In my view, whole research packages—theoretical perspectives and families of models together with collections of research procedures, conventions, and standards —should be evaluated for robustness to idealizing assumptions, just as scientists and philosophers of science have argued that individual models must be so assessed (Levins 1966, 1968; Wimsatt 1980, 1981, 1987).17 However, we cannot easily identify and evaluate packages if our conception of scientific progress is one of simple competitive replacement by the package of the moment. Instead, I urge transperspective cooperation in the service of comparative analysis. Robustness analysis of the sort I advocate requires patient attention to those research packages one does not adopt, for the sake of learning about the prospects and limitations of the one to which one has made a primary commitment. Studies of robustness begin with comparative analysis of assumptions and of results. Provided we have at least three legs to stand on, any one stance serves as a reference point for the detection of similarities and differences in assumptions and empirical findings in the other two. Three-way comparison can reveal the shared assumptions of all three. The hard work of robustness analysis is comparison of the relations between assumptions and results to see if results hold across perspectives despite differences of assumptions. If they do, then results are robust to the particular assumptions of any perspective yielding them. If they do not, then results are not robust, even if they are shared across perspectives. The work of assessing the robustness of results in evolutionary biology to assumptions from different perspectives was advanced considerably in a comparative analysis of theoretical perspectives by Cor van der Weele (1999). She describes three theoretical perspectives on development: a ‘‘switches and

225

James Griesemer

ipl ec aus es

genetic programs

lt mu

int ern al c au ses

fields

constructions contingent causes

4 • Similarities and differences between the three approaches. Details are given in the text. Figure and legend from C. van der Weele, Images of Development: Environmental Causes in Ontogeny (Albany: State University of New York Press, 1999), p. 43.

responses’’ approach which tries to integrate genetics with embryology according to some notion of genetic or epigenetic program (roughly, neoDarwinism as discussed here), a ‘‘fields/structures’’ approach characteristic of concerns with laws of form (roughly, process-structuralism), and a ‘‘networks/constructions’’ approach (roughly, developmental systems theory) in which ‘‘organisms are constructed in contingent historical processes’’ (p. 31). Exemplars of the first include developmental geneticists like Raff and Kaufman as well as epigeneticists like Waddington and Schmalhausen. Goodwin is the prime exemplar of the second. Oyama and Lewontin are exemplars of the third. Van der Weele identifies three dimensions of similarity and difference in a triangle diagram (figure 4) comparing the three perspectives. Perspectives (triangle vertices) united by a similarity of interest or research focus have an edge connecting them. The perspective not connected by that edge differs in that respect in primary interest or focus. Van der Weele’s representation greatly facilitates the formulation of critical, comparative analysis as a project of robustness analysis because what we want to know is which results of evolutionary research are independent of the idealizing assumptions of each of the specified research packages. Results that hold for the perspective of one of these lines of similarity, but not for those perspectives connected by it, are likely to be sensitive to the idealizing assumptions of that perspective. Results that hold across a line of comparison are not likely to depend (directly, straightforwardly) on the assumptions linking the similar perspectives. Van

226

Evolutionary Process Perspective

der Weele’s triangle suggests a general scheme for the initiation of robustness analysis for perspectives. Along these lines, several important similarities and differences can be visualized in a comparison of process-structuralism, developmental systems theory, and the reproducer perspective. The focus on evolution as state transformation of systems over time rather than environmental ‘‘molding’’ of organisms is shared by process-structuralism (ps) and developmental systems theory (ds), though the two perspectives will probably disagree about what parameters of models are appropriate since process-structuralism, like the reproducer perspective (rp), seeks developmental invariants. However, ps seeks developmental invariants as the proper basis for generative principles and process laws, while rp seeks them as clues to the heuristic breakdown of the reductionist research strategy (Griesemer 2000b, 2002a). In contrast to the search for developmental invariants, developmental systems theory focuses on historically contingent recurrences of developmental resources produced by natural selection.18 ds and ps both focus on development as the key to unlocking phenomena left mysterious by neo-Darwinism. rp treats heredity and development as equally relevant and entwined parts of a complex process of reproduction, and thus pursues explanatory parity or symmetry for processes rather than for objects (developmental resources) or functions (replicators and interactors). rp shares with ds the view that historically contingent natural selection is important for explaining the current states of biological systems, but places its significance with respect to units analysis differently. rp follows Maynard Smith and Szathmáry’s rather than Lewontin’s approach to units of evolution, with the addition of a fitness principle to the fundamental three (multiplication, variation, heredity) to define units of selection as a special case of units of evolution. ds appears to follow, in at least some of its statements, the functionalist units analysis of Dawkins or Hull. ds and rp both reject a priori distinctions of internal (organism) and external (environmental) causes, which ps appears to accept in its formulation of what to count as state variables and what as parameters in its developmental models. ds and rp are both amenable to modeling ‘‘environments’’ as parts of developmental systems. rp and ps are amenable to transmission metaphors while ds is not. All three perspectives share an interest in epigenetic inheritance, though they disagree about whether separate inheritance ‘‘channels’’ exist. All three likewise reject Weismannism, though they differ in how and whether to pick up the pieces of that shattered distinction. rp emphasizes origin problems and processes, as do ps and ds, but has yet to guide any serious mathematical model construction of the sort that ps

227

James Griesemer

has. ds also has yet to influence mathematical model construction, in part because it is skeptical of the well-developed modeling strategies of the neoDarwinian research package and endorses the less tractable forms of dynamic coupling of organism and environment characteristic of, for example, niche construction.19 rp treats neo-Darwinian modeling strategies as special cases and looks for ways to relax Weismannist assumptions while retaining them as special cases, thus subsuming traditional genetics instead of rejecting it (Griesemer 2002a). ds subsumes niche construction and draws parallels to the ‘‘dialectical’’ form of interactionism described by Lewontin (see Griffiths and Gray 2001; Oyama, Griffiths, et al. 2001). However, rp is doubtful that the class of mathematical models produced by ps’s physics-based vision of general laws is robust because physics offers a dynamic theory of quantities without a similarly robust account of the chemical objects far from equilibrium that form the material basis for developmental and reproductive processes (Fontana, Wagner, et al. 1994; Fontana and Buss 1996).20 This listing of similarities and differences is intended only to illustrate the kinds of comparisons that might facilitate a robustness analysis among process perspectives. The larger issue is the need for a robust family of process perspectives to complement structuralism and functionalism. Attempting to carry out a realistic robustness analysis would be premature because too few empirical or theoretical results have been generated by any of the new process perspectives to offer a comparative base.21

Conclusion I have argued for the addition of process perspectives in evolutionary biology to the more familiar and conceptually dominant perspectives of structuralism and functionalism. I have also suggested that perspectives—parts of research packages including families of models, practices, standards, and conventions—be viewed as contributors to a cooperative effort to establish a robust understanding of biological phenomena of heredity, development, and evolution. Robustness requires comparative evaluation of the relations between idealizing assumptions and empirical results. This is facilitated when one position is used to judge similarities and differences between at least two others. Thus, a process perspective can be a useful addition to structure and function perspectives, and moreover, a reproducer perspective can be a useful addition to other process perspectives such as process-structuralism and developmental systems theories. I have suggested some of the lines of

228

Evolutionary Process Perspective

investigation that unite and distinguish these three kinds of process perspectives to illustrate the fruitfulness of looking at conceptual evaluation this way. The real work—of articulating and testing models from within several perspectives and then evaluating their empirical results for robustness at both the level of models within families and at the level of perspectives—has barely begun.

Notes I thank Eva Neumann-Held and Christoph Rehmann-Sutter, whose encouragement and insightful comments greatly improved this essay and Mindy Conner for her excellent copyediting. I thank Werner Callebaut and the Konrad Lorenz Institute, Altenberg, for providing time, space, money, and sentiment facilitating completion of the essay. The Dean of the Social Sciences Division, College of Letters and Science, University of California Davis, provided additional financial support. 1 Brandon here refers to genotypes, but in context his account allows ‘‘a more organismic approach’’ in terms of phenotypic change, describing Dawkins’s and Hull’s generalizations of Mayr’s two-step process via a generalization (by abstraction; see Griesemer 2005) of the genotype/phenotype distinction to a replicator/vehicle or replicator/interactor distinction. 2 See Lloyd 1988 for a general review and analysis of the state space approach in biology. 3 Molecular geneticists use allele (haplotype), sequence, or even genomic regulatory state spaces where the elements are cis-regulatory binding sites rather than coding sequences (Davidson 2001). Here, the laws of transformation represent expression processes rather than frequency changes. Some processstructuralists use morphogenetic field representations (Goodwin and Trainor 1983). Others use abstract autonomous Boolean network state spaces (Kauffman 1993). Detailed consideration of genomic regulatory state spaces is beyond the scope of this essay but they are certainly important for process perspectives that attempt to unify evolution and development. 4 On the relation between the replicator concept and Weismannism, see Griesemer and Wimsatt 1989; Griesemer 2005. 5 This is not quite fair. Textbooks often gesture at the problem by mentioning the ‘‘rna world’’ hypothesis to explain the origin of replication and at symbiogenesis to explain the origin of multicellularity. But these mentions are mere gestures, and the evolutionary theory to explain them a mere hand wave. Serious attempts to provide an evolutionary theory adequate to the transition problem run directly into the conceptual issues discussed here. For my critique of Dawkins’s

229

James Griesemer

photocopying analogy, see Griesemer 2005 and Griesemer in preparation. A good presentation on the mechanism of xerographic photocopying can be found at http://www.physics.udel.edu/~watson/scen103/xerox/. Accessed May 4, 2005. 6 A distinction must be made between the evolution of taxonomic levels and the evolution of levels of composition. Darwin did explain the former with his gradualist account of the divergence of character (Darwin 1859: ch. 4). However, this explains only the evolution of genera from species, families from genera, and so on—the evolution of clades. Even if these were compositional levels, Darwin’s theory does not explain the evolution of multicellular organisms from singlecelled organisms or the evolution of life from nonlife—the evolution of grades of organization—as many creationists complain. It leaves the mystery of mysteries untouched. Even Darwinians like Ernst Mayr are discomfited, claiming irony in Darwin’s title since he explained the origin of adaptation, not of species per se. 7 This explanatory problem is familiar. Consider the problem of how to explain the evolution of sex. Sex may confer an advantage on the future prospects of the species over asexual species, but the twofold disadvantage of sex in the short term cannot be explained by appeal to future, teleological benefits because natural selection is ‘‘short-sighted.’’ The explanatory problem with teleology is that knowing the outcome of a process makes it easy to think that any event along the way is part of the mechanism or process ‘‘leading to’’ that outcome. In virtue of knowing the outcome, however, we need never notice that its origin is not accounted for. 8 Griesemer et al. (in press) argue, contra developmental systems theory, that this can be done without the connotations of statistical independence between channels characteristic of theories of inheritance as flow of genetic information. 9 Developmental systems theorists have subsumed niche construction as a strategy for representing environmental inheritance within the developmental constructionist perspective (Oyama, Griffiths, and Gray 2001). Niche construction state space representations adapt Lewontin’s structural state space, insofar as they are defined in terms of ‘‘phenotype pools,’’ so these models imply spatial commitments not specified by the developmental systems perspective (Laland, Odling-Smee, and Feldman 2001). It remains to be seen whether this kind of merger of developmental process and structure perspectives in the formulation of a family of models can clarify and guide fruitful model-building strategies (Griesemer et al. in press). For recent efforts in this direction, see Odling-Smee, Laland, and Feldman 2003. 10 Smith 1992 reviews process structuralism and its avowed connection to preDarwinian rational morphology. 11 A very different research program guided by the process-structuralist perspective is the work of Stuart Kauffman to interpret catalytic sets of interacting proteins and also gene regulation networks in terms of abstract networks of

230

Evolutionary Process Perspective

Boolean ‘‘nk’’ automata: switching networks of elements with on and off states that interact with other network elements to turn them on or off as well (Kauffman 1993, 1995, 2000). Although it would be important to assess Kauffman’s work in a full robustness analysis of model families within process-structuralism, across process perspectives, and across structuralist, functionalist, and process perspectives on evolution, my more limited goal in this essay—to illustrate the need for such an analysis—and want of space and time preclude undertaking that project or discussing Kauffman’s work here. 12 See, for example, Dawkins 1983, ch. 1, on ‘‘Necker Cube pluralism’’; or Sober and Wilson 1998 on the compatibility of many different kinds of models of population structure. 13 Or, to help discover whether or in what respects the picture is pseudorobust. 14 I leave open here the question of whether chains of molecular reactions constitute developmental processes for complex molecules, and thus whether autocatalysis itself could constitute a higher-order reproduction process with recursion down to, say, some quantum mechanical ‘‘null development’’ of individual molecules forming a chain of reactions. 15 This was discovered by von Neumann and reported in his celebrated study of self-reproduction (von Neumann 1966). See also Penrose 1959 and Szathmáry 1994. 16 The levels here are levels of reproduction, not necessarily levels of composition. Because the perspective developed here is a process perspective, no commitment is made directly to spatial, compositional relations among reproducers. I assume, for the sake of illustration, a conflict model of evolutionary transition. Not all models of evolutionary transition are models of conflict from within; they may, for example, result from cooperative evolution of defense against genomic parasites (see Jablonka and Lamb 1995). 17 Elihu Gerson (1998: ch. 10) has a similar, but narrower, notion of research package which does not include perspectives as elements: ‘‘As various aspects of research work [theories, models, procedures, concepts, and descriptions] come to co-specify one another, they tend to form a conceptual and organizational unit. I will call such a unit a package, following a usage from computer science. . . . Co-specification and packages are matters of commitment. That is, when scientists commit to a certain theory (for example), they commit as well to the descriptions and procedures that the theory co-specifies. The parts of a package may co-specify one another in varying degrees. Tightly co-specified packages have few options for commitment among their components; loosely specified packages have many options. For example, the Mendelian-chromosome model as developed by geneticists in the early twentieth century required some sort of controlled breeding program if it was to be tested.’’

231

James Griesemer 18 Griffiths and Gray (1994) actually offer two perspectives: a developmental systems perspective which seeks recurrent developmental resources and a developmental process perspective which seeks recurrent developmental relations. 19 See, for example, the comparison in Griffiths and Gray 2001: 205 between dynamic equations for evolution in neo-Darwinism, Lewontin’s constructionism, and niche construction. 20 ‘‘In a conventional dynamical system all components and all interactions are given explicitly at the outset and the mathematical apparatus is used to compute, for example, equilibrium distributions of the components of the system. In contrast, the components of biological and chemical systems endogenously construct new components upon interaction. The information for this constructive action resides in their structure. The study of the possible organizations of such components hinges on the lawful association between structure and action. A major goal of the AlChemy project is to develop a formalization of the structure/action association called ‘chemistry’ at a level of abstraction that is useful for understanding biological organization, its origin and evolution’’ (Fontana, http://www.santafe.edu/~walter/AlChemy/alchemy.html). Accessed May 4, 2005. 21 The extensive research programs of Kauffman; Wimsatt and Schank; Goodwin, Webster, and colleagues; Newman; and others have offered a number of empirical and theoretical results that are not discussed in detail here. However, all of these contributions combined are dwarfed by those of any one of the neoDarwinian specialties in theoretical or empirical population, quantitative, or evolutionary genetics.

References Avital, E., and Jablonka, E. 2000. Animal Traditions: Behavioural Inheritance in Evolution. Cambridge: Cambridge University Press. Bickhard, M. H. 1992. Scaffolding and self-scaffolding: central aspects of development. In: L. T. Winegar, and J. Valsiner (eds.), Children’s Development Within Social Context: Research and Methodology (pp. 33–52). Hillsdale, N.J.: Lawrence Erlbaum. Brandon, R. 1990. Adaptation and Environment. Princeton: Princeton University Press. Brandon, R. 1999. The units of selection revisited: the modules of selection. Biol. Philos. 14: 167–180. Buss, L. W. 1987. The Evolution of Individuality. Princeton: Princeton University Press. Coyne, J. A. 2000. The gene is dead; long live the gene. Nature 408: 26–27. Darwin, C. 1859. On the Origin of Species. John Murray: London.

232

Evolutionary Process Perspective

Davidson, E. H. 2001. Genomic Regulatory Systems: Development and Evolution. San Diego: Academic Press. Dawkins, R. 1976. The Selfish Gene. New York: Oxford University Press. Dawkins, R. 1983. The Extended Phenotype: The Gene as the Unit of Selection. New York: Oxford University Press. Dawkins, R. 1995. River out of Eden: A Darwinian View of Life. New York: Basic Books. Dawkins, R. 1999. The Selfish Gene. 2d ed. New York: Oxford University Press. Depew, D. J., and Weber, B. H. 1995. Darwinism Evolving: Systems Dynamics and the Genealogy of Natural Selection. Cambridge: mit Press. Dobzhansky, T. 1937. Genetics and the Origin of Species. New York: Columbia University Press. Drickamer, K., and Carver, J. 1992. Carbohydrates and glycoconjugates, editorial overview: upwardly mobile sugars gain status as information-bearing macromolecules. Curr. Opin. Struct. Biol. 2: 653–654. Dyson, F. 1985. Origins of Life. Cambridge: Cambridge University Press. Fontana, W., and Buss, L. W. 1996. The barrier of objects: from dynamical systems to bounded organizations. In: J. Casti and A. Karlqvist (eds.), Boundaries and Barriers (pp. 56–116). Reading, Mass.: Addison-Wesley. Fontana, W., Wagner, G., et al. 1994. Beyond digital naturalism. Artif. Life 1(1–2): 211–227. Fujimura, J. H. 1996. Crafting Science: A Sociohistory of the Quest for the Genetics of Cancer. Cambridge: Harvard University Press. Gánti, T. 1979. A Theory of Biochemical Supersystems and Its Application to Problems of Natural and Artificial Biogenesis. Baltimore: University Park Press. Gánti, T. 1987. The Principle of Life. Budapest: omikk. Gánti, T. 1997. Biogenesis itself. J. Theor. Biol. 187(4): 583–593. Gánti, T. 2003. The Principles of Life. J. R. Griesemer and E. Szathmáry (eds.). Oxford: Oxford University Press. Gerson, E. M. 1998. The American system of research: evolutionary biology, 1890–1950. Ph.D. diss., University of Chicago. Goodwin, B. 1989. Evolution and the generative order. In: B. Goodwin and P. T. Saunders (eds.) Theoretical Biology: Epigenetic and Evolutionary Order from Complex Systems (pp. 89–100). Edinburgh: Edinburgh University Press. Goodwin, B. 1990. Structuralism in biology. Scientif. Progr. 74: 227–244. Goodwin, B. C. 1994. How the Leopard Changed Its Spots: The Evolution of Complexity. New York: C. Scribner’s Sons. Goodwin, B., and Trainor, L. E. H. 1983. The ontogeny and phylogeny of the pentadactyl limb. In: B. Goodwin, N. Holder, and C. C. Wylie (eds.), Development and Evolution (pp. 75–98). Cambridge: Cambridge University Press. Gray, R. 1992. Death of the gene: developmental systems strike back. In: P. E. Griffiths (ed.), Trees of Life (pp. 165–209). Dordrecht: Kluwer.

233

James Griesemer

Griesemer, J. R. 2000a. Development, culture and the units of inheritance. Philos. Sci. 67: S348–S368. Griesemer, J. R. 2000b. Reproduction and the reduction of genetics. In: P. Beurton, R. Falk, and H.-J. Rheinberger (eds.), The Concept of the Gene in Development and Evolution: Historical and Epistemological Perspectives (pp. 240–285). Cambridge: Cambridge University Press. Griesemer, J. R. 2000c. The units of evolutionary transition. Selection 1: 67–80. Griesemer, J. R. 2002a. Limits of reproduction: a reductionistic research strategy in evolutionary biology. In: M. H. V. van Regenmortel and D. L. Hull (eds.), Promises and Limits of Reductionism in the Biomedical Sciences. Chichester: Wiley. Griesemer, J. R. 2002b. What is ‘‘epi’’ about epigenetics? In: G. Vandevijver, L. Vanspeybroeck, and D. Dewaele (eds.), From Epigenesis to Epigenetics: The Genome in Context 981 (pp. 1–14). New York: New York Academy of Sciences. Griesemer, J. R. 2005. The informational gene and the substantial body: On the generalization of evolutionary theory by abstraction. In: M. Jones and N. Cartwright (eds.), Idealization XII: Correcting the Model, Idealization and Abstraction in the Sciences (pp. 59–115). Poznan Studies in the Philosophy of the Sciences and the Humanities, volume 86. Amsterdam: Rodopi. Griesemer, J. R. Forthcoming. Tracking organic processes: representations and research styles in classical embryology and genetics. In: M. D. Laubichler and J. Maienschein (eds.), From Embryology to Evo-Devo: A History of Embryology in the 20th Century. Cambridge: mit Press, Dibner Series. Griesemer, J. R. In preparation. Reproduction and the evolutionary process. Griesemer, J. R., Haber, M., Yamashita, G., and Gannett, L. In press. Critical notice. Cycles of contingency: developmental systems and evolution. Biol. Philos. Griesemer, J. R., and Wimsatt, W. C. 1989. Picturing Weismannism: a case study of conceptual evolution. In: M. Ruse (ed.), What the Philosophy of Biology Is: Essays for David Hull (pp. 75–137). Dordrecht: Kluwer. Griffiths, P. E. 1996. Darwinism, process structuralism, and natural kinds. Philos. Sci. 63(3): S1–S9. Griffiths, P. E. 2001. Genetic information: a metaphor in search of a theory. Philos. Sci. 68(3): 394–412. Griffiths, P. E., and Gray, R. D. 1994. Developmental systems and evolutionary explanation. J. Philos. 91(6): 277–304. Griffiths, P. E., and Gray, R. D. 2001. Darwinism and developmental systems. In: S. Oyama, P. E. Griffiths, and R. D. Gray (eds.), Cycles of Contingency: Developmental Systems and Evolution (pp. 195–218). Cambridge: mit Press. Griffiths, P. E., and Knight, R. D. 1998. What is the developmentalist challenge? Philos. Sci. 65(2): 253–258. Hull, D. L. 1980. Individuality and selection. Annu. Rev. Ecol. Syst. 11: 311–332.

234

Evolutionary Process Perspective

Hull, D. L. 1981. The units of evolution: a metaphysical essay. In: U. Jensen and R. Harré (eds.), The Philosophy of Evolution (pp. 23–44). Brighton: Harvester Press. Hull, D. L. 1988. Science as a Process. Chicago: University of Chicago Press. Jablonka, E. 2001. The systems of inheritance. In: S. Oyama, P. E. Griffiths, and R. D. Gray (eds.), Cycles of Contingency: Developmental Systems and Evolution (pp. 99–116). Cambridge: mit Press. Jablonka, E., and Lamb, M. 1995. Epigenetic Inheritance and Evolution. Oxford: Oxford University Press. Jablonka, E., and Szathmáry, E. 1995. The evolution of information storage and heredity. Trends Ecol. Evol. 10(5): 206–211. Kauffman, S. A. 1993. The Origins of Order: Self-Organization and Selection in Evolution. New York: Oxford University Press. Kauffman, S. 1995. At Home in the Universe: The Search for Laws of Self-Organization and Complexity. New York: Oxford University Press. Kauffman, S. 2000. Investigations. New York: Oxford University Press. Keller, E. F. 2001. Beyond the gene but beneath the skin. In: S. Oyama, P. E. Griffiths, and R. D. Gray (eds.), Cycles of Contingency: Developmental Systems and Evolution (pp. 299–311). Cambridge: mit Press. Laland, K. N., Odling-Smee, F. J., et al. 2001. Niche construction, ecological inheritance, and cycles of contingency in evolution. In: S. Oyama, P. E. Griffiths, and R. D. Gray (eds.), Cycles of Contingency: Developmental Systems and Evolution (pp. 117–126). Cambridge: mit Press. Levins, R. 1966. The strategy of model building in population biology. Amer. Sci. 54: 421–431. Levins, R. 1968. Evolution in Changing Environments: Some Theoretical Explorations. Princeton: Princeton University Press. Lewontin, R. C. 1970. The units of selection. Annu. Rev. Ecol. Syst. 1: 1–17. Lewontin, R. C. 1974. The Genetic Basis of Evolutionary Change. New York: Columbia University Press. Lloyd, E. S. 1988. The Structure and Confirmation of Evolutionary Theory. New York: Greenwood Press. Maynard Smith, J. 1972. John Maynard Smith on Evolution. Edinburgh: Edinburgh University Press. Maynard Smith, J. 1988. Evolutionary progress and the levels of selection. In: M. Nitecki (ed.), Evolutionary Progress (pp. 219–230). Chicago: University of Chicago Press. Maynard Smith, J., and Szathmáry, E. 1995. The Major Transitions in Evolution. Oxford: W. H. Freeman Spektrum. Maynard Smith, J., and Szathmáry, E. 1999. The Origins of Life: From the Birth of Life to the Origin of Language. Oxford: Oxford University Press.

235

James Griesemer

Mayr, E. 1978. Evolution. Sci. Amer. 239(3): 46–55. Michod, R. E. 1999. Darwinian Dynamics: Evolutionary Transitions in Fitness and Individuality. Princeton: Princeton University Press. Odling-Smee, F. J., Laland, K. N., and Feldman, M. W. 2003. Niche Construction, the Neglected Process in Evolution. Princeton: Princeton University Press. Oyama, S. 1985. The Ontogeny of Information: Developmental Systems and Evolution. Cambridge: Cambridge University Press. Oyama, S., Griffiths, P. E., and Gray, R. D. (eds.). 2001. Cycles of Contingency: Developmental Systems and Evolution. Cambridge: mit Press. Oyama, S., Griffiths, P. E., et al. 2001. Introduction: what is developmental systems theory? In: S. Oyama, P. E. Griffiths, and R. D. Gray (eds.), Cycles of Contingency: Developmental Systems and Evolution (pp. 1–11). Cambridge: mit Press. Palade, G. E. 1983. Membrane biogenesis: an overview. In: S. Fleischer and B. Fleischer (eds.), Biomembranes, part J: Membrane Biogenesis: Assembly and Targeting (General Methods, Eukaryotes) (pp. 96: xxix–lv). New York: Academic Press. Penrose, L. S. 1959. Self-reproducing machines. Sci. Amer. 200(6): 105–114. Schank, J. C., and Wimsatt, W. C. 1988. Generative entrenchment and evolution. In: A. Fine and P. K. Machamer (eds.), PSA-1986 (2: 33–60). East Lansing, Mich.: Philosophy of Science Association. Sharon, N., and Lis, H. 1993. Carbohydrates in cell recognition. Sci. Amer. 268(1): 82–89. Smith, K. 1992. Neo-rationalism versus neo-Darwinism: integrating development and evolution. Biol. Philos. 7: 431–452. Sober, E., and Wilson, D. S. 1998. Unto Others: The Evolution and Psychology of Unselfish Behavior. Cambridge: Harvard University Press. Sterelny, K. 2001. Niche construction, developmental systems, and the extended replicator. In: S. Oyama, P. E. Griffiths, and R. D. Gray (eds.), Cycles of Contingency: Developmental Systems and Evolution (pp. 333–349). Cambridge: mit Press. Sterelny, K., Smith, K., and Dickison, M. 1996. The extended replicator. Biol. Philos. 11: 377–403. Szathmáry, E. 1994. Toy models for simple forms of multicellularity, soma and germ. J. Theor. Biol. 169(2): 125–132. Szathmáry, E. 1999. Chemes, genes, memes: a revised classification of replicators. Lect. Math. Life Sci. 26: 1–10. Szathmáry, E., and Maynard Smith, J. 1993. The origin of genetic systems. Abstr. Bot. 17: 197–206. Szathmáry, E., and Maynard Smith, J. 1997. From replicators to reproducers: the first major transitions leading to life. J. Theor. Biol. 187(4): 555–572. Van Fraassen, B. C. 2002. The Empirical Stance. New Haven: Yale University Press.

236

Evolutionary Process Perspective

Von Neumann, J. 1966. Theory of Self-Reproducing Automata. A. W. Burks (ed.). Urbana: University of Illinois Press. Wagner, G. P., and Altenberg, L. 1996. Perspective—complex adaptations and the evolution of evolvability. Evolution 50(3): 967–976. Webster, G., and Goodwin, B. 1996. Form and Transformation, Generative and Relational Principles in Biology. Cambridge: Cambridge University Press. Weele, C. van der. 1999. Images of Development: Environmental Causes in Ontogeny. Albany: State University of New York Press. Williams, G. C. 1966. Adaptation and Natural Selection. Princeton: Princeton University Press. Wimsatt, W. C. 1980. Reductionistic research strategies and their biases in the units of selection controversy. In: T. Nickles (ed.), Scientific Discovery: Historical and Scientific Case Studies (2: 213–259). Dordrecht: D. Reidel. Wimsatt, W. C. 1981. Units of selection and the structure of the multi-level genome (pp. 122–183). PSA 1980, Philosophy of Science Association. Wimsatt, W. C. 1987. False models as means to truer theories. In: M. Nitecki and A. Hoffman (eds.), Neutral Models in Biology (pp. 23–55). London: Oxford University Press. Wimsatt, W. C. 1999. Generativity, entrenchment, evolution, and innateness. In: Biology Meets Psychology: Philosophical Essays (5: 139–179). Hardcastle: mit Press. Wimsatt, W. C., and Schank, J. C. 1988. Two constraints on the evolution of complex adaptations and the means for their avoidance. In: M. Nitecki (ed.), Evolutionary Progress (pp. 231–273). Chicago: University of Chicago Press.

237

9 G E N E S — C AU S E S — C O D E S Deciphering DNA’s Ontological Privilege

eva m. neumann-held If skeptics of the causal privilege of genes are right, how can we explain the intriguing consequences of the manipulation of genes, such as the production of Drosophila flies with numerous wings attached to multiple places on their bodies (Halder et al. 1995a, 1995b)? How can we explain the possibility of predicting phenotypic outcomes by analyzing breeding patterns or dna sequences? Was it not such insight into the causal power of genes that founded new research strategies (and institutes) that are directed at investigating networks of molecular processes in a more general sense? Thus, Thomas Bürglin (this volume) offers a nice example of empirical research on networks of interacting genes (and their products) which allow causal explanations of developmental processes. These approaches should supply some kind of ‘‘explanatory bridging’’ between typical genetic approaches, which analyze aspects of heritability, and developmental biology, which is directed toward understanding multiply regulated biological processes in a developing (and therefore constantly changing, reacting, and adjusting) organism. The ‘‘gene skeptics’’ are consequently opposed by those who believe that we live in a new era of the life sciences and are certain that further development and refinement of current (empirical) successes will lead to new kinds of explanatory and (bio-) technological revolutions, which will allow us to shape not only our biotic environment according to our desires, but sooner or later our bodies and our minds as well. It is beyond the scope of this essay to question this visionary or fantastic prospect. Rather, I am interested in the tension between the scientific and technological successes of the genetic and molecular sciences, on the one hand, and the conceptual and empirical arguments and evidence against the causal primacy of genes, on the other. It seems nonsensical to deny the technical and scientific progress, but at the same time I find the arguments against

Genes—Causes—Codes

gene centrism, as they are laid out in various chapters in this book, very convincing. The question, therefore, is what makes it so difficult to integrate these two aspects of empirical and theoretical research? I believe the difficulty arises when the current successes in molecular approaches are understood as confirming underlying conceptual frameworks and empirical results of genetic approaches that predate the biochemical characterizations of ‘‘genes.’’ Classical and population genetics causally link mechanisms for organismic development, differences in organismic traits, and evolutionary change to some underlying ‘‘action’’ of genes. In a sense the concept of the gene gained a meaning only (Keller 1995: 17) when it could be traced back to the biochemical molecule dna, which is both inherited from generation to generation (of cell lines or organisms) and causally involved in the production of enzymes. Here, the discovery of the ‘‘genetic code’’ is a further building block justifying the transfer of the location of causal mechanisms from their rather abstract site in the genes of classical and population genetics to biochemical-embedded molecular genes. The reason is that the ‘‘genetic code’’ seems to allow a conceptualization according to which the mechanism for transferring a Bauplan (or blueprint) into a developmental process is basically laid down through some kind of preestablished relationship between dna and metabolic processes. So, one might summarize: within the conceptual framework of genetic sciences genes cause developmental and evolutionary outcomes, and the reason they do so can basically be reduced to the causal relationship between the genomic dna and polypeptide synthesis, a relationship that is (pre)arranged and determined by the genetic code. In other words: it is assumed that somehow the ‘‘molecular gene’’ is the biochemical explication of those ‘‘genes’’ which were originally conceptualized in the framework of other genetic approaches such as the classical and the population genetics approaches. The terms basically and somehow point to the vagueness of the assumption of a ‘‘unifying genetic framework.’’ I will argue here that such ‘‘unifying frameworks’’ are too simplistic. They require that it be possible to establish a systematic relationship between the different genetic approaches. I want to show that such a construction of a systematic relationship is necessarily very difficult, if not impossible. Thus, from a systematic and methodological point of view, I will discuss three types of genetic approaches: population (evolutionary) genetics, classical genetics, and molecular genetics. I will show that these approaches differ significantly in their empirical and theoretical methodology, and thus differ also in what is to be explained—the explanandum. The guiding questions for

239

Eva M. Neumann-Held

the first part of my analysis are: What is the explanatory setup in which an (empirical and theoretical) analysis finds the ‘‘cause’’ in ‘‘genes’’? In which ways is the ‘‘gene’’ conceptualized as a ‘‘cause’’? How does the ‘‘gene’’ relate to dna sequences? But I also believe that the answers to these questions alone will not explain the robustness of this ‘‘paradigmatic’’ unified genetic framework. The additional and maybe even more important (because less discussed) problem concerns the assumptions of the ‘‘epistemic nature’’ of the ‘‘genetic code.’’ Thus, I next will discuss the notion of the ‘‘genetic code,’’ arguing for the claim that although it might not exist, it cannot generally be dispensed with. It is my goal to break open the intimate but problematic ‘‘chain of underlying beliefs’’ that link ‘‘genetic causes’’ (in different genetic approaches) and ‘‘genetic code’’ (as a concept used in molecular genetics) to each other. But first, I need to introduce the notions ‘‘cause’’ and ‘‘effect’’ as I want to use them here.1

Cause and Effect ‘‘Causality’’ and causal explanations are, of course, a strongly debated topic in philosophy. For my purpose, however, it should suffice to introduce the terms in a basic way by following Hartmann’s philosophical reconstructions (Hartmann 1993: mainly 79–82), according to which the concept of causality points to a particular consecutive (mostly defined as temporal) relationship in which one incident or state S depends on a particular incident S 1 ; S 1 is a condition of S. We see here already that this concept is closely linked to the concept of explanation. S can be explained by condition S 1 under the effect of a particular rule.2 Often, however, the explanation of S requires not just condition S 1 (and a particular rule) but a set of conditions S 1 , . . . , S n . The incidents which together result in the presence of the conditions S 1 , . . . , S n , can be called the cause of the occurrence of S, whereas S is the effect of the occurrence of S 1 , . . . , S n .3 A single Si from {S 1 , . . . , S n } can be called a component of the cause. The introduction of the terms cause and effect allows us to formulate analytically the law of causality: ‘‘Same causes have same effects’’ (Hartmann 1993: 81).4 The use of the notions ‘‘cause,’’ ‘‘effect,’’ ‘‘condition,’’ and ‘‘component’’ of a cause allows us to describe and distinguish between explanatory goals and empirical strategies in genetic sciences.

240

Genes—Causes—Codes

The ‘‘Gene’’ Beyond DNA—Evolutionary Genes In population genetics, mathematical and empirical approaches are used to model evolution as the change in frequencies of alleles in a population. The concept of the evolutionary gene has been spelled out quite extensively by G. C. Williams (1966) and more recently by Richard Dawkins (1982, 1989), on whose writings the following analysis is based.5 According to these concepts, an evolutionary gene can be identified only in a scenario of selection among different varieties of traits of the same type. It is (1) the cause of the difference between the traits, that is, that condition which when changed (mutated) leads to the expression of a change in the corresponding trait; and (2) by supplying the cause of these different expressions of the trait it also supplies a difference on which selection can act, thus causing a shift in the distribution of traits (and the underlying cause) in the population. The relationship can be expressed as follows. First, we need to describe how the change in a particular condition from S i to S i* leads to a change in the trait S to S*: (1) If S i (in set of conditions S 1 , S 2 , . . . , S n ), then trait S (2) If S i* (in set of conditions S 1 , S 2 , . . . , S n ), then trait S*

The evolutionary gene is then an abstract notion to indicate differences in replication rates, that is, the (shifting) ratio of S i and S i* which (due to selection) results in the shifting ratio of S and S* in a population. For my purposes we can leave out the shifting aspect. Then the concept of the evolutionary gene states that the ratio between two conditions is the cause of the ratio between two traits, because only the difference between these conditions causes a difference between the respective traits: (3) S i /S i* has the effect (is the cause of ) S/S*

The important point here is that (3) may not (in a lawlike manner) be equated with any of the following formulations: (4) S i has the effect (is the cause) of S (5) S i* has the effect (is the cause) of S* (6) S i /S i* has the effect (is the cause) of S (7) S i /S i* has the effect (is the cause) of S*

That means that there is no logical reason to conclude that what causes a difference between traits in a selective scenario (the evolutionary gene) is the

241

Eva M. Neumann-Held

same as the complete cause of the trait (and its mutation) in question! Rather, it might very well be that S i and S i* are just components of the cause of S and S*, respectively. Therefore, to speak in this context of ‘‘a gene for a trait’’ (in the sense of an evolutionary gene for a trait) must be seen as sloppy shorthand for ‘‘that which causes a difference in the traits upon investigation.’’ 6 The next point to consider is whether evolutionary genes necessarily consist of dna. If we think of an evolutionary gene as the difference between two alleles of the same dna segment, then it follows that these dna sequences do not even have to be directly involved in any polypeptide synthesis processes; that is, an evolutionary gene could very well be localized between ‘‘reading frames’’ (Dawkins 1982: 87f.) and could be a very local event—even a single base pair.7 Thus, it is clear that the concept of the evolutionary gene differs from the molecular concept, in which a gene is defined as the sequence that codes for a polypeptide (see below). But what is the empirical evidence for a systematic relation between different changes in traits and changes in corresponding dna sequences? Interestingly, such a systematic relationship must be ruled out. First, different combinations of dna sequences, for example, can bring about the same changes in phenotypic expression. That means that there is no selection among these different kinds of combinations of dna sequences (Wagner 1988, 1990, 1996). As a consequence, evolutionary genes do not have to correspond to specific stretches of dna. Second, on the basis of the insight that a trait is caused by a multitude of conditions that encompass more than just dna, it can be shown that evolutionary change can be traced back—in principle—to any change in frequencies of any of the conditions necessary for developing a trait (that is, the developmental resources) which result in stable inheritance of differential expression of that trait (Griffiths and Gray 1994). In summary: evolutionary gene denotes that set of conditions—among others which are necessary for the expression of some trait—which causes a heritable difference in the expression of that trait in an evolutionary scenario. Thus, in a particular instance, it is an empirical question as to which developmental resource or which stable combination of developmental resources plays a role as ‘‘evolutionary gene.’’ A concept according to which ‘‘genes’’ in evolutionary modeling have to be particular dna sequences (although they might be), or are nothing more than dna, is not defensible. Before I end this section I want to discuss briefly a different understanding of gene in the context of supplying ‘‘evolutionary explanations’’ of traits, particularly in sociobiology and so-called evolutionary psychology. Here,

242

Genes—Causes—Codes

a reference to ‘‘genes’’ and ‘‘our genetic inheritance’’ points to an implicit (and often explicit) understanding according to which traits are causally determined by ‘‘genes,’’ with the more or less distinct implication that these (often behavioral) traits are hard to change.8 These approaches have numerous theoretical and empirical problems which have been criticized by numerous researchers of different disciplines on empirical and theoretical grounds.9 I do not want to repeat them here, but—in accordance with the goal of this essay—instead to analyze the meaning of the concept ‘‘gene’’ in such contexts in view of its relationship to the ‘‘gene’’ as made up of dna sequences. In sociobiological accounts traits are thought of as evolutionary, that is, as the historic results of selection processes. Selection acted on conditions that are among the causes of the trait. The cause of the trait in question is thus the sum of conditions for the trait, and the conditions have been shaped by selection processes. These conditions can be called ‘‘evolutionary genes*.’’ The major distinction between evolutionary genes as discussed above and ‘‘evolutionary genes*’’ is that the former refers to ‘‘bits and pieces’’ which cause a difference between traits of the same type, whereas the latter would be the sum of components of the cause of a particular trait itself. That means an ‘‘evolutionary gene*’’ is a further abstraction from a complicated etiological account of a number of ‘‘evolutionary gene incidents’’ linked to the evolutionary shaping of a trait in the framework of some particular ecological setting. Thus, ‘‘gene talk’’ combines here a functional evolutionary analysis, which in turn depends significantly on the underlying evolutionary theory and on assumed ecological conditions, with an assumption about an underlying causal story about the mechanism, the form and interaction of those (evolutionary shaped) conditions that produce the trait in question. It is very important to note here that the causal mechanisms themselves have to be treated (correctly) as a black box. Likewise—and therefore!—it remains unclear what those ‘‘genes’’ are made of. However, the use of the notion ‘‘gene’’ apparently seduces us into speculating whether ‘‘evolutionary genes*’’ consist of nothing but dna. I think a belief that this can be taken for granted can lead to profound confusion, and therefore this ‘‘speculation’’ should be treated carefully. Here, the same arguments hold true that were applied to the treatment of ‘‘evolutionary genes.’’ If we concentrate on the idea that an ‘‘evolutionary gene*’’ is the set of all conditions that cause a trait, then it is most plausible to assume that an ‘‘evolutionary gene*’’ is not reducible to dna sequences, because the production of a trait necessarily depends on more factors than dna alone. (I return to this point below in ‘‘Molecular Genes and Developmental Outcomes.’’) Furthermore, if we cannot find a sys-

243

Eva M. Neumann-Held

tematic one-to-one relationship between traits and dna, that is, if a trait— even if it is favored by selection—can be produced by different means, then selection might act on the trait but not on single specific underlying causal mechanisms. Therefore, I conclude that both ‘‘evolutionary genes’’ and ‘‘evolutionary genes*’’ are hypothetical constructs. The latter concept refers to a set of conditions for causing a trait, whose existence is functionally explained by reference to an evolutionary story of the shaping and integration of its conditions. The conditions and their (causal) interaction remain a black box, which may be opened only by investigating the developmental processes that lead to the trait in question. In contrast to the evolutionary gene, which might consist only of dna (that, as we have seen, is an empirical question), an ‘‘evolutionary gene*’’ always encompasses more than dna if it is to refer to the complete cause (with all its components) for the presence of a trait. Thus, a reconstruction of the (evolutionary) history of the genesis of a trait is not sufficient for the conclusion that the trait in question is ‘‘preestablished’’ (in some ‘‘genes’’) and unchangeable in an individual life history.

Classical Genetics and the Use of ‘‘Molecular Tools’’ The availability of experimental tools for molecular research has changed the range of explanatory approaches in genetic disciplines. As I want to show in this and the next section, there is an important difference between approaches which are rooted in classical genetics and might use the tools of molecular research (this section), and typical molecular approaches as employed in developmental genetics (next section). Thus, the claim is: the employment of the same tools does not guarantee the same explananda. One might call breeding technologies the basic methodological root of classical genetics (Gutmann and Janich 1997). The development and application of breeding practices in agriculture and animal husbandry go back a long way in the history of humankind, and such practices have been and still are used to enhance particular traits in cattle, plants, and, more recently, bacteria and phages. The method allows sexual reproduction only between those exemplars of a type which show a distinctive desired trait, and only under highly selective conditions, so that only those exemplars exhibiting a desired trait will survive.10 In classical genetics, breeding is used to analyze patterns of heritability of

244

Genes—Causes—Codes

morphological or physiological (sometimes behavioral) traits, thereby defining a gene (I speak of the classical gene) phenomenologically as an underlying ‘‘factor’’ for a trait (Gutmann and Janich 1997: 19).11 So far, a gene is again a kind of hypothetical unit (Johannsen 1911). Pragmatically and empirically, it is not easy to find and define appropriate traits which permit reference to underlying ‘‘factors’’ and for which rules of transgenerational distribution could be formulated. This requires not only standardization of environmental conditions but also standardization of the organismic constitution. That is, the required organisms have to be ‘‘produced’’ so as to allow for successful analysis of patterns of inheritance in breeding experiments. This need to produce suitable strains for breeding experiments was acknowledged by early researchers in genetics such as Mendel and Johannsen (1911).12 On the basis of our present knowledge let us now ask again how those classical genes could relate to dna sequences. And again we see that the classical gene is basically a black box. Breeding under highly selective conditions might produce a particular trait—a particular phenotype—but this outcome reveals nothing about the mechanisms involved in producing this trait. Resistance to some antibiotic could be caused by a metabolic change in some enzymatic pathway traceable to some mutation in dna (so that would be, e.g., S 1 among Sn for S) or by changes in the organization of membranes or cell walls (e.g., S 2 among Sn for S).13 New methods and the use of new organisms made it possible to narrow down the analysis to more specific—and less complex!—types of conditions S i for S. In addition to crossbreeding technologies, techniques to produce chromosomal mutations through X-ray or chemical treatment and to analyze chromosomes cytologically became available.14 Application of these methods to suitable (i.e., standardized) strains of organisms (particularly Drosophila) and phenomenological analysis of the organisms allowed chromosomal and genetic maps to be constructed through which mutational effects in traits could be linked or ‘‘mapped’’ to loci. These approaches thus focused on changes in the patterns of traits caused by changes on the chromosomal level, because the experimental setup itself allowed a focus on only a subset of conditions, chromosomal loci, producing phenotypic changes. Here at the chromosomal loci the decisive variations were introduced, whereas all other conditions were kept as stable as possible through the elegant breeding approach of the laboratory strains. Further experimental conditions, particularly the choice of another type of organism, allowed the complexity of condition-cause-effect relationships to be reduced even more. In bacteria and phage genetics, feeding experi-

245

Eva M. Neumann-Held

ments and mutants were used to reconstruct the steps in biosynthetic pathways. Obviously, the steps depended on the presence of certain polypeptides (enzymes), as could be concluded from mutant strains which depended on proper nutritional supplementation of intermediate products to complete the synthesis of the final product. That means that the capacity to synthesize particular biochemical substances could be viewed as a ‘‘physiological’’ trait, and by phenomenological analysis genetic (chromosomal) changes could be related to the production of polypeptides. Basically, these approaches led to the so-called one gene-one enzyme hypothesis (I speak of the second classical gene concept). It is important to note here that phage genetics, too, deals with phenomenological genes, because strictly speaking we still have here no explanation of how ‘‘a gene’’ could produce an enzyme. We only know that variations in particular loci on chromosomes are one condition (S i ) that affects the expression of trait S, the enzyme. From a theoretical perspective what we have seen so far is that the narrowing down of the condition-cause-effect relationship, by using appropriate organisms and thereby appropriate phenotypes (reduction of complexity) in a highly standardized environment, also allows a new conceptualization of the underlying gene concept. The classical gene concept encompasses more conditions S i for expressing trait S than the second classical gene concept, if only because of the differing complexities of the trait under investigation. Empirical evidence for these considerations can be found, for example, in the fact that the phenomena of classical genetics—such as characterizations of patterns of inheritance as dominant or recessive or so-called position effects —cannot be reduced to one kind of molecular mechanism (Gilbert 2000: 164f.). In a few cases, however, a particular pattern of inheritance can be traced to a single polypeptide, that is, to one single gene as a particular dna sequence. Here I want to discuss classical genetics analysis by dna technology, which allows us to link chromosomal loci to (biochemically defined) dna sequences. In such cases, patterns of inheritance can often be linked not only to chromosomal maps but also, by application of molecular techniques, to the level of dna sequences. What exactly is being done here? By techniques of classical genetics we find on the phenomenological level that the cause of differences in the expression of a particular organismic trait can be linked to differences in the respective chromosomal and genetic maps. Molecular approaches—dna sequencing—add here a biochemical analysis. We know now that in many cases the difference at the chromosomal loci consists of differences in the particular structure of the dna sequence. We might also be able

246

Genes—Causes—Codes

to contextualize this locus on the dna so that we can identify the functional contribution of this dna sequence in the cell’s biochemical pathways; for example, in the production of a particular polypeptide. We know then that a particular change in S i results in a change in S. However, we still have not explained how S is made, we still do not leave the explanatory framework which was directed toward explaining the cause of the difference. Why should that be important? Theoretically, it is important because the ‘‘gene’’—here a dna sequence—is only one condition, S i , among others that produce S on the phenotypic level. Therefore, we can predict a change in S i only if all other conditions are held constant, which is guaranteed only by strictly regulated and standardized experimental setups. As soon as these restrictions are loosened, however, prediction is possible only—at best—in a statistical manner. The observations then become much more difficult to interpret, as in human (and medical) genetics. Let us look at this in more detail. Genetics in humans (fortunately) does not use breeding experiments; the classical method is based on family tree analysis, which traces patterns of the inheritance of particular phenotypic traits in families.15 In addition, molecular technologies allow dna sequences to be screened for variations that are linked to particular phenotypic traits such as (inherited) diseases.16 I noted above that in addition to transgenerational transmission of dna there are several other channels of inheritance available,17 but for the sake of the argument let us consider here only cases in which heritable differences in traits have been traced to differences in dna sequences. Quite often, what we find empirically is that the presence of such mutations does not permit the prediction of the trait; that is, we cannot predict the outbreak of the disease in individual cases. In other words, the same dna constellations do not always result in the same patterns of traits, and particular traits can come about through a combination of different (including genetic) developmental resources (e.g., Wolf 1995, 1997, 2002). Even if a tendency to aggressive behavior in a family can be traced back to a mutation in monoamine oxidase A, the presence of the mutation does not allow ‘‘abnormal aggressive’’ behavior to be predicted (see the claims by Brunner et al. [1993a, 1993b], discussed in Neumann-Held 1997). To summarize: The important point here is that the application of molecular tools in the context of classical (and human) genetics approaches is important but still of limited use. Although many inherited diseases can be traced back to particular dna constellations (changes—that is, mutations—in the sequence in comparison with sequences of healthy people), human geneticists themselves stress that known correlations between genes and phenotypic outcomes are only of a statistical nature. That means that individual

247

Eva M. Neumann-Held

predictions are often not possible, and the kind and degree of so-called inherited diseases vary to an enormous extent because the effects (phenotypic outcomes) of gene activation depend on interactions with other causally relevant entities. Variation in such entities (conditions) can change the phenotypic outcome dramatically, leaving a mutation in a gene almost ineffective on the phenotypic level—even in cases in which there is good statistical evidence for a close correlation with genetic constitution and phenotypic outcome.18 So far we have seen that neither evolutionary genes nor classical genes easily resolve into dna sequences, although they may do so in single cases. Furthermore, we have seen that the experimental and conceptual setup of both approaches is suited only to seeking the causes of differences between phenotypes and—eventually—developmental pathways.

Biochemical Accessibility of DNA and the Genetic Code If we now turn to ‘‘molecular genes,’’ the situation changes in two respects. The experimental accessibility of dna and its participation in biochemical reactions seems, first, to supply a material substrate for ‘‘genes,’’ and, second, to allow for opening the black box with respect to how a gene affects a trait. In other words, it seems possible now to investigate directly the activity of genes (now equated with dna), and thereby to understand how developmental processes can be understood causally on the molecular level. However, in the transition from classical genetic sciences to molecular genetics, and thereby in the transition from phenomenological to biochemical analysis, some conceptual confusion crept in. Apparently, researchers were evaluating the results of these new techniques within the framework of the former genetic approaches, and thus interpreted them as confirming deep-rooted assumptions about the causal power of genes by supplying causal-mechanistic explanations. Before we look at these points of confusion in more detail, it might be helpful to review briefly the explanatory expectations with which the genetic sciences started and how they seemed to fulfill those expectations over time. The upsurge of classical genetics in the 1930s began with Thomas Hunt Morgan’s explicit decision to move from research on developmental (embryological) processes to genetics because developmental research had reached its methodological limits (see Sarkar, this volume). With this move he prepared the path for the development and employment of genetics as

248

Genes—Causes—Codes

a new research strategy. The guiding question of developmental biology— How does an egg develop into a complex multicellular organism?—was now turned into a genetic question: How do genes achieve their effects? The explanatory goal changed, and with it the respective empirical tools, and conversely one may also say that new empirical tools became available which enabled new explanatory goals to be formulated. Thus, this move in research strategies can be characterized and defended as an instrumental one. However, and this is an important point, it seems that the move from developmental research to genetic research was already leading to a conceptual confusion (Keller 1995: ch. 1) because geneticists began to claim to be able to resolve the developmental problems through genetic approaches.19 Obviously, the switch in perspective and in experimental approach led to some kind of underlying ontological belief system, according to which genes are the most important (if not the sole) causative power for developmental processes and outcomes. As long as classical genetics did not explain how genes affect a trait or what actually constitutes a gene materially, this ‘‘belief system’’ could be viewed as a guiding hypothesis. This changed with the rise of molecular approaches. The advent of molecular genetics might be dated to Oswald Avery’s discovery in 1944 that the biochemical substrate of genetic loci consisted of dna. This opened the path to further investigation of dna and its involvement in metabolic (in a wider sense, developmental) processes. At this point the question of classical genetics—how to explain (inheritable) differences in phenotypes—could be reduced to the question of how differences in dna cause differences in polypeptides. Beadle and Tatum (1941) took the next empirical and conceptually important step by providing, as Keller explains it, ‘‘a particular kind of answer to the question of how a gene produces its effects—namely, it catalyzes a specific chemical reaction . . . it makes an enzyme. At last the mysterious notion of gene action seemed to have a real content’’ (Keller 1995: 17f.). The biochemical approach allowed a turn from the question What causes a difference? to the question What causes a trait—in particular, a polypeptide? At this point confusion entered the (molecular) approach unnoticed, but with important consequences. Morgan and his followers had transformed developmental questions into genetic ones. This approach had led to the successful biochemical explanation of gene action. Due to the deeply rooted ontological assumption that genes control development, it now seemed possible that the genetic approach, turned around, could explain development. The problem was—and is—that locating the cause of phenotypic differences

249

Eva M. Neumann-Held

in differences in gene sequences (one of several putative difference-generating developmental resources) turned into locating the causes for traits exclusively in the genes, thereby assuming that the genes are the only causal source of development. In other words, the switch in explanatory goals and the way this proceeded, with a single focus on the ‘‘activity of genes,’’ reflected an assumption that whatever causes differences in traits also explains the development of those traits. This assumption is not a logical one; it needs—at best—empirical support (see next section). And although this assumption may have been successful in the early days of molecular genetic research (Kay 2000: 326), it has hindered the development of newer models for contemporary research problems. In earlier days mainstream research concentrated on testing the theoretical and empirical consequences of this assumption, to the extent that the—instrumentally useful—assumption turned into some kind of certainty. One anchoring point that stabilizes this confusion seems to be the understanding (1) of the relationship between dna and the synthesis of polypeptides, and (2) of the ‘‘(epistemic) nature’’ of the genetic code. The interpretation of the work of Beadle and Tatum described above paved the way for Watson and Crick, who introduced a representation of dna coupled with a semantic move that had major consequences. ‘‘In a long molecule,’’ they wrote, ‘‘many different permutations are possible, and it therefore seems likely that the precise sequence of the bases is the code which carries the genetical information’’ (Watson and Crick 1953: 967). In this 1953 paper Watson and Crick introduced two notions that became determinants in future interpretations: ‘‘information’’ and ‘‘code.’’ As Keller stresses: ‘‘Watson and Crick have gotten a lot of credit for their work and deservedly so, but one contribution has, I fear, been overlooked: their introduction of the information metaphor to the repertoire of biological discourse was a stroke of genius. The story of this metaphor—its uses and implications—is immensely rich and has been extensively explored by others’’ (Keller 1995: 18). And although, according to Kay (2000: 150), ‘‘the code which carries the genetical information’’ was mentioned only in passing by Watson and Crick, others—in particular Gamov, Rich, and Ycas—used the ‘‘information idiom’’ deliberately and ‘‘with all its connections to mathematics, logic, cryptoanalysis, linguistics, computers, operations research, and weapons systems . . . as a means of framing the ‘coding problem.’ The trope of information served to integrate mechanisms of molecular specificity, structural considerations, mathematical relations, and linguistic attributes within a single explanatory framework. It reshaped the problem of genetic specificity through a discourse that resonated with the technosciences of command and control’’

250

Genes—Causes—Codes

(Kay 2000: 150). In other words, during the period 1953–1961, ‘‘the scriptural representation of the genetic code as text, reading, alphabet, and words was introduced; they served as the conceptual framework and as analytical tools for establishing the correlations between nucleotide triplets and amino acids. . . . [T]hese communication tropes were instantiated as the discursive framework of molecular biology’’ (Kay 2000: 329). These references to ‘‘codes’’ and ‘‘information’’ were not unusual and not specific to genetic topics at the time. On the contrary, as Keller (1995: 18) notes, the introduction of these two notions came at the beginning of the 1950s, when the mathematical ‘‘information theory’’ proposed by Claude Shannon was a popular topic in the world of communication systems and seemed an appropriate tool for the analysis of all kinds of complex systems. And although, as Keller points out, it was also clear by 1952 that the technical terms information and code, as used in mathematical information theory, were not applicable to biological problems,20 the ‘‘metaphorical’’ (Keller 1995: 19) use of these terms, taken from the discursive framework of information theory, lent plausibility to and thereby affirmed the explanatory framework of molecular genetics, according to which the causal power has to be located in the genes (that is, in dna). However, in the light of the successful story of molecular biology, one should not forget that irrespective of whether these concepts were clear-cut or fuzzy, the framework thus set allowed the framing of the problem at hand, namely how the correlation between nucleic acids and proteins is established (see also below). The answer to this question was given when empirical (not formal!) research established an alignment between three successive nucleic acid base pairs (triplets or codons) on the dna corresponding specifically to particular amino acids.21 Today any textbook on molecular biology contains a formal table relating triplets to specific amino acids. And so today as in the mid-1960s, the ‘‘view of dna as a universal language is ubiquitous both in biology and the culture at large’’ (Kay 2000: 330). After these introductory remarks on the conceptual framework of molecular genetics I will now turn to the claims of the ‘‘causal power’’ of molecular genes. These claims are based on two points: • Genes, that is dna segments, determine phenotypes (at least developmental outcomes) by causing specifically the synthesis of corresponding polypeptides. • The determination of the sequence of polypeptides through dna is grounded in the ‘‘coding’’ function of dna. The sequence of polypeptides is laid down— in coded form—in the dna and is decoded by biochemical mechanisms.

251

Eva M. Neumann-Held

Molecular Genes and Developmental Outcomes The descriptions of dna as containing information through coding mechanisms ‘‘authorized the expectation . . . that biological information does not increase in the course of development: it is already fully contained in the genome’’ (Keller 1995: 19). Thus, the new molecular findings seemed to confirm the framework, traceable to the beginning of genetics, which posits ‘‘genes’’ to be the only active power in the organism.22 Although developmental genetics has certainly provided some intriguing and surprising research results, the question arises as to whether its underlying research strategy is too ‘‘gene focused’’ for a comprehensive explanation of developmental processes. Here we can return to the approach chosen in the former treatments of evolutionary and classical gene concepts and ask for the explanandum. The important point is that if developmental processes or developmental outcomes (specific organismic traits) are to be explained, then the focus is no longer on the effect of variations of single causes S i to S, but on the contribution of all S n to S. In a framework in which genes are believed to be the only driving causal power in developmental processes, it might seem obvious also to believe that S n consists only of genes (or genetic networks) or that at least those genetic causes determine traits (effects) more forcefully than do other, nongenetic influences. The primacy of genes is thus defended by the informational content of genes (and vice versa). Meanwhile, however, there is enough empirical evidence and conceptual reframing available to show that these assumptions are losing their charm and that a privileged role (Gray 2001) cannot be attributed to the genes. GodfreySmith (1999) summarizes: ‘‘We can regard the environment as a background condition against which genes carry information about phenotypic traits. But . . . we can also view genetic conditions as background conditions or part of the ‘channel.’ Against such a genetic background, we can see environmental conditions as carrying information about phenotypes. And we can also see phenotypes as carrying information about environmental conditions. If any of these attributions of informational properties is acceptable, then all of them are’’ (1999: 312). And therefore, Gray (2001) concludes: ‘‘We argue that assigning a gene an informational role and relegating the environment to secondary background support cannot be justified using either a mathematical conception of information theory derived from communication theory . . . or a semantic view of information’’ (2001: 190).

252

Genes—Causes—Codes

This argument of arbitrariness can be extended to the relationship between dna and polypeptides. Usually this is described as a causal relationship: the sequence of base pairs of a specific dna segment determines the sequence of amino acids in a polypeptide. The corresponding dna segment is called a ‘‘gene,’’ and according to the usual wording a gene thus ‘‘codes for’’ a polypeptide as each triplet in the gene codes for an amino acid. According to this description it seems again that all the information for synthesizing a polypeptide from single amino acids is embedded in the sequence of a dna segment (a ‘‘gene’’): the gene is viewed as the single cause (or at least the most important cause) of a polypeptide.23 And again, it can be shown that this description is at least misleading. It is known that the mrna can be ‘‘worked on’’ by several mechanisms in the cell. For example, through alternative mrna splicing, a particular dna sequence might give rise to different (alternatively spliced) mrna sequences. Through mrna editing, the sequence of the mrna can be changed by insertion, deletion, or conversion of the bases, which then gives rise to a new mrna that in no way can be matched to the original dna sequence. In both cases the regulation of the mechanisms in a single developing organism is time and space dependent, and therefore the mechanisms and their regulation involve numerous causal factors. That means that the amino acid sequence is not preestablished in the dna; there are numerous conditions, the dna sequence being only one of them, which taken together form the cause of a polypeptide. I have suggested elsewhere (for an extended treatment, see Neumann-Held 1999, 2001) that based on these descriptions of biochemical interactions it is important to distinguish between dna sequences, on the one hand, and a ‘‘gene’’ as whatever is the cause of a polypeptide, on the other. These ‘‘genes’’ include but are not restricted to dna sequences. Such a gene, which I call the process molecular gene (pmg), encompasses all factors which are necessary for the entire time- and space-regulated process that results in the synthesis of specific polypeptides. Thus defined, a pmg might be a frequent event in any individual of a population (one might think of so-called housekeeping genes), or it might be a rare or even a single event in the life history of an individual. This uncertainty leads Moss to consider the pmg concept to be of ‘‘negligible biological value’’ (Moss 2001: 91). I do not share these doubts and would ask: biological value for what kind of research interest? If we are interested in understanding the mechanisms of the differential use of dna sequences in ontogenetic (organismic) processes, then we need to look for precisely those kinds of mechanisms that pmg highlights. pmg allows us to

253

Eva M. Neumann-Held

focus on individual processes and hence is better understood as a conceptual framework from which particular research interests could be derived. Thinking of even linear polypeptides in any individual as the products of an interaction between multiple causal factors opens new perspectives for designing different strategies, if the manipulation of these polypeptide expression patterns is a goal.24 If, however, the goal is to conceptualize some kind of entities that are transgenerationally, stably transmitted so that evolutionary change can be modeled, then pmg is probably not suitable—although it might remind researchers that any theory-guided conceptualization of stable units (modules) for modeling evolution should consider the (empirically approachable) complexity of individual developmental processes. So I conclude that pmg is a valuable gene concept for particular explanatory goals. Independent of that, the evidence presented indicates that dna is not even causally privileged with regard to the synthesis of a linear polypeptide chain. This claim, however, seems to be contradicted by the discovery of the ‘‘genetic code.’’ Is the sequence of amino acids in polypeptides not laid down—in a coded form—in the dna, to be decoded by biochemical mechanisms?

To Code or Not to Code? In her historical analysis ‘‘Who wrote the book of life?’’ Lily Kay states: ‘‘Had dna and rna been of genetic interest in the 1930s, the problem of their correlations with amino acids would have probably assumed central importance. But then the terminology and modes of reasoning would have not been informational and scriptural, since the information discourse had not yet come into being’’ (Kay 2000: 328). Following Kay we might conclude that in this case scientists would not have described the biochemical relations between dna and polypeptide synthesis in terms of a ‘‘coding relation’’; the ‘‘genetic code’’ would never have been invented. And indeed, when following the arguments presented in the last section according to which dna segments are only one condition causing the sequence of an amino acid, one might wonder whether saying that a dna sequence codes for an amino acid sequence of a polypeptide is still justified. Some philosophers continue to defend the existence of the genetic code even in the face of strong arguments to the contrary. This claim, however, must be handled with care. Paul Griffiths (this volume), for example, explains that ‘‘the facts of molecular developmental biology do not correspond

254

Genes—Causes—Codes

to the popular idea that the genetic code is a language in which the genome contains instructions about phenotypes’’ and cites Kenneth Schaffner (1998), who makes the point ‘‘by saying that there are no tiny ‘traitunculi’ living in the genome.’’ But although one has to be careful to avoid slipping ‘‘from a code for protein structure to a language for specifying phenotypes,’’ one nevertheless has to concede that ‘‘there really is a genetic code’’ (Griffiths, this volume). Likewise, Godfrey-Smith points out explicitly: ‘‘All the genes can code for, if they code for anything, is the primary structure (amino acid sequence) of a protein’’ (Godfrey-Smith 1999: 328). Furthermore, he specifies that the claim that there really is a genetic code—which clearly is an ontological statement —does not necessarily claim causal privilege for genes: The idea that genes code because of their developmental role may or may not be coupled with a claim about the preeminent causal importance of genes. One might hold that genes code without holding that ‘‘genes are destiny.’’ The point of the concept of coding, on a developmental role analysis, is to pick out one particular causal role among many. Within developmental and metabolic processes there are raw materials (like amino acids), cutters and joiners (enzymes), stores of energy (like atp), readers and assemblers (ribosomes)—and there are coded instructions as well (the genes). Raw materials and stores of energy might be just as important as messengers, but they are different kinds of causal players. (Godfrey-Smith 1999: 314)

Characterizing the differences between causal inputs to developmental outcomes is certainly a valuable goal, and a necessary one for conceptual and empirical research in developmental biology. There are different ways of doing this. One might follow Godfrey-Smith and focus on biochemical molecules. One could then characterize the different kinds of metabolic interactions in which different types of biochemical molecules are involved. But Godfrey-Smith goes a step further and claims that genes exist not just as another biochemical molecule, dna, but as ‘‘coded instructions’’ (Godfrey-Smith 1999: 312).25 It is but a short step to the claim that the genes exhibit some kind of ‘‘information’’ independent of all the other causal inputs that are necessary to synthesize a polypeptide. Can one not then conclude that the existence of ‘‘coded instructions’’ is independent of time and space in the developmental process—even if the reading of those instructions depends on the context of the organism? If one concedes that the genes are really coding for linear polypeptides, it seems that one should also admit that genes contain at least the information of the polypeptide setup of a cell, that genes are rather

2 55

Eva M. Neumann-Held

‘‘essence’’ and ‘‘form’’ than ‘‘accident’’ and ‘‘contingency’’ (see Oyama 2000: ch. 6, 89). On the other hand, the fascinating possibility of aligning dna sequences and polypeptide sequences as they are exemplified in any molecular textbook makes it almost unimaginable not to speak of a coding relationship here. I am not sure how far Godfrey-Smith and Griffiths would go, but I believe that the talk of ‘‘coded instructions’’ allows genetic determinism (in some meager form) to reenter the explanatory framework of biological processes. That seems reason enough to investigate carefully the arguments given for this ontological concession to the ‘‘genetic code.’’ Godfrey-Smith offers three arguments in defense of the genetic code. The first two are as follows: ‘‘Coding is (i) specific to the relationship between genes and phenotypes, and (ii) asymmetric, as genes code for phenotypes but not vice versa.’’ 26 He adds that this holds true at least for ‘‘the coding relationships discussed in molecular biology’’ (Godfrey-Smith 1999: 312). While it is true that the coding relationships in molecular biology are discussed in this way, that is not sufficient evidence to claim that the relationship between gene (dna) and phenotype (linear polypeptide sequence) is really a code because only this relationship could be described as such. I want to argue against this claim and show that the description of biochemical interactions as ‘‘coding’’ does not have to be restricted to the interaction between dna and polypeptide. Let us compare, for example, the metabolic reactions leading to the synthesis of a linear polypeptide chain with a general pattern of enzymatic chain reactions. To simplify the argument, I think it is safe to begin the analysis of the polypeptide synthesis process with the processed mrna transcript, which then enters the translation processes. After all, in the face of mechanisms like mrna editing, it is hard to maintain that the sequence of the dna itself is ‘‘translated’’ unchanged into a polypeptide. So, if we want to defend the application of the ‘‘coding’’ notion only to the processes of polypeptide synthesis, the argument is strongest when we concentrate on processed mrnas.27 Transfer rnas (trnas) play a crucial role in translation processes. These are molecules with a cloverleaf-like structure and two binding sites, one of which binds specifically to triplets of mrna through base pairing of the corresponding nucleic acid bases. There is a second specificity involved, because the trna triplet that is suited to bind to a triplet on the mrna corresponds to a binding site of the same trna for a specific amino acid. Transfer rnas are thus loaded with amino acids according to their triplet structure at the site at which they bind to the mrna. Therefore, two trna molecules binding at adjacent sites on a mrna bring two amino acids in close proximity to

256

Genes—Causes—Codes

each other, allowing a chemical bond to be formed between them by further enzymatic processes.28 Figure 1 illustrates a formal, typical alignment of mrna triplets with polypeptide amino acids, which is meant to express this ‘‘unparalleled’’ coding relationship. Triplets are denoted with one capital letter and the corresponding amino acids with a small letter, reflecting the correspondence guaranteed by the mediation of the two specific binding sites of trna molecules. Now let us think of some other kind of metabolic reaction by which some products—for example, amino acids—are synthesized in a cell from precursors with the support of enzymes and cofactors. Usually, such reactions are described as shown in figure 2. Some complex molecule is synthesized in a cell from precursors. Each step in the synthesis is catalyzed by a specific enzyme and with the support of certain cofactors (like atp or nadph) which enable the reaction between precursor and enzyme. Let us assume that these enzymes are all incorporated in a membrane. The enzymes used in this synthesis process are referred to by capital letters, and the components which are added by each enzymatic reaction by small letters.The drawing in figure 2 focuses on the growing chain of the product and views the necessary precursors, enzymes, and cofactors as inputs to each single step. The point I want to make is that both types of reaction can be drawn in both ways, although usually they are not. Thus, in the first case we could equally well focus on the growing chain of the polypeptide and view the single triplets together with the enzymatic functions of cofactors and precursors (single amino acids ‘‘activated’’ by binding to a trna) as inputs to each step in the synthesis reaction (see figure 3). Likewise, the relationship between the enzymes in the membrane and the synthesis of some complex cell component can be aligned as a ‘‘coding relation,’’ as shown in figure 4. I have formalized both cases, and, of course, biochemical (synthesis) reactions differ from one another in many respects. There are differences between the uses and kinds of cofactors, between enzymes and enzyme location, and between the numbers of components. For example, I have assumed here that the enzymes in my ‘‘formal’’ example are bound in a membrane to imitate the situation of triplets in which the alignment through chemical bonding guarantees the order in which the triplets are ‘‘used’’ by the respective trnas. One might argue that although such membrane-bound complexes are known, many other enzymatic (anabolic) reactions in the cell are not membrane bound. However, using these enzymes in the right order to synthesize the corresponding products is decisive here, too, and the details

257

1 • Typical alignment of mRNA and polypeptide chain, which is meant

to express a coding relationship.

enzyme B co-factors B

enzyme C co-factors C

enzyme B co-factors

a

a

…

enzyme C co-factors

a

b

b

etc.

…

etc.

…

c

2 • Typical illustration of a biosynthetic pathway.

t–RNAB co-factors triplet B

t–RNAC co-factors triplet C

t–RNAB co-factors

a

a

b

…

t–RNAC co-factors

a

b

c

3 • Illustration of the relationship between mRNA and a polypeptide chain as a biosynthetic pathway.

membrane enzymes:

a

b

c

a

b

c

d

products: d

4 • Alignment of components of a biosynthetic pathway, which is meant to express a coding relationship.

Genes—Causes—Codes

of the mechanisms by which the availability of these enzymes is guaranteed are still an open question in cell biology. In any case, the highly specific reactions which are needed for a proper synthesis of products are comparable, and although I would not deny the differences between the types of reactions, I do not see any difference major enough (in qualitative and quantative respects) to restrict the description of ‘‘coding’’ only to the relationship between dna and linear polypeptide synthesis. Likewise, the asymmetric relation that Godfrey-Smith describes (‘‘coding is asymmetric, as genes code for phenotypes but not vice versa’’; see above) holds for other metabolic reactions as well. Amino acids or fatty acids, as products of reaction chains, do not produce the enzymes which are part of the steps in the chain. But the amino acids will themselves be used in the production of enzymes, and the fatty acids might be used in the reproduction of membranes in which the enzyme complexes can become embedded. Likewise, enzymes (in whose production dna is always involved) are certainly also part of the conditions for the reproduction of dna (and all the subsequent steps of transcription and translation processes). And in a much broader framework that takes into account the never-ending sequence of metabolic steps in living organisms (or even beyond, in the transgenerational line of living cells) one might very well argue that polypeptides produce dna and that amino acids are necessary reactants for the production of enzymes.29 The point is that the sequence of what is producing what depends more on the starting point of the description than on some underlying ‘‘natural order.’’ Therefore, the notion of ‘‘coding’’ depends on the description applied rather than on a ‘‘real’’ property of dna. Now we can proceed to Godfrey-Smith’s third argument for assigning a ‘‘coding function’’ only to the relationship between genes and polypeptides. He writes: ‘‘The peculiar characteristic of dna that justifies its being treated as a code lies in the fact that its sequence is physically read by the cell during the construction of proteins. The cell first creates an mrna molecule whose sequence corresponds to the sequence of bases in the dna, and then part of the cell’s machinery physically moves along the mrna molecule, at each step interacting with the base sequence, producing with each step a chain of amino acids whose linear structure corresponds, by a standard rule, to the linear structure of the mrna’’ (Godfrey-Smith 1999: 314). As Godfrey-Smith himself admits, the concept of ‘‘reading’’ which is used to defend the concept of ‘‘coding’’ can be applied only in an extended sense— which might be necessary for the concept of coding, too. But this concession could reasonably be regarded as a minor one by a biologist; it is consistent

259

Eva M. Neumann-Held

with the claim that the processes in which dna is involved have a remarkable and theoretically important similarity to ordinary processes of reading and interpretation. And that is what many biologists might regard as the important point; it does not matter exactly how the genetic sense of ‘‘reading’’ and ‘‘interpretation’’ are connected to everyday usage (Godfrey-Smith 1999: 315). So, the claim is that the ‘‘reality’’ of the genetic code is based on the fact that dna is physically ‘‘read,’’ at least in an ‘‘extended sense.’’ It seems to me that the application of the concept of ‘‘reading’’ is understandable only after the description of ‘‘coding’’ has been applied. Therefore, the allusion to ‘‘reading properties’’ adds nothing to the argument about the ‘‘coding properties,’’ which was rejected above.

Do We Need the ‘‘Genetic Code’’? Some Concluding Remarks In the last section I tried to convince the reader that there really is no code in the dna. Of course, that does not mean that the relationship between dna/mrna and polypeptide synthesis is no longer interesting. It remains extremely interesting and important—as are the energy-supplying pathways of glycolysis and the Krebs cycle and the biosynthetic pathways of, for example, hormones and amino acids. Moreover, all the pathways that have been described biochemically are dependent on each other. But does this mean the whole idea of ‘‘coding’’ is futile? Does the concept of the ‘‘code’’ lack any explanatory weight (Sarkar 1996)? Is it nothing but a metaphor, once helpful for heuristic reasons but now gone astray? I do not think so, and I will make two brief points to support my opinion. First, looking at the history of molecular biology, I agree with Godfrey-Smith (2000) that the introduction of the notion of a ‘‘code’’ into molecular genetic research helped to resolve a specific problem which Godfrey-Smith (2000: 31) put in the following question: ‘‘How can a gene control the exact sequence of a long chain of amino acids strung together by the cell?’’ What exactly was done here may be methodically reconstructed like this: the concept of ‘‘code’’—as a system of rules which relates the signs of two alphabets to each other—was used not as a metaphor but as a model for the relationship between dna and polypeptide chain. Using this model allowed the question to be framed and a line of research to be set up with the result that the alignment of triplets to amino acids could be demonstrated empirically (!)—in a very simplified in vitro system—before the detailed mechanisms were known (see above).

260

Genes—Causes—Codes

An epistemological problem arose when the methodical order was reversed. While the ‘‘code’’ in a technical and/or everyday life-world context was used as a model for turning some problem into a biological research program, this order was reversed when—under certain conditions—this approach turned out to be successful. The code as a model for investigating molecular interactions turned into a model of molecular interactions.30 This might be called a case of misplaced concreteness, an ontological shortcut, with the result that ‘‘suddenly’’ coding properties became attributed to the dna. But this, as we have seen, is hard to defend, and not only on methodical and methodological grounds, once the molecular details are taken seriously. Nevertheless, under different explanatory and/or pragmatic approaches, notions that have been refuted under particular descriptions can become useful under others. The ‘‘genetic code,’’ for example, can be a useful description—a pragmatic tool—when the purpose is to manipulate particular organisms so that they produce some particular biochemical compound. One might think here of the goal of making some bacteria strain produce a polypeptide with a slightly altered amino acid composition. In that case one might genetically modify the corresponding dna sequence, based on the ‘‘genetic code,’’ to correspond to the amino acid sequence of the desired polypeptide and then transform this modified dna into the bacteria strain. It is important to be precise about what gene technologists actually do here. They treat developmental, interacting processes in which many (actually an almost unlimited number of ) causal components are involved as linear production chains with a limited number of reactants, defined inputs, and, most important, clearly defined outputs. This works pretty well if—and this is of utmost significance—the whole modified gene construction, its implementation in the cell and the conditions in which the organisms are kept, is controlled and adjusted constantly with the goal of optimizing the developmental processes of the organisms toward the production of the product of interest. Nevertheless, living beings are difficult to control completely, and companies complain that it is almost impossible to make living beings produce genetic products unchanged for an extended period of time and over generations.31 In addition to the ‘‘code,’’ another kind of well-known but previously refuted description is useful under the conditions of gene technology, that is, with the goal of producing particular compounds. I refer to the description under which a particular dna sequence would be the important causal input for ‘‘producing’’ a particular polypeptide. Think of a typical problem in the context of gene technology, in which one tries to understand why one ‘‘gene construction,’’ transferred into some bacteria strain, produces the desired

261

Eva M. Neumann-Held

product perfectly, whereas an alternative construction leads to no useful production rates. The question is which construction of dna in which genetic environment supplies the best conditions for production of the protein.32 Acknowledging the efforts which are necessary to implement such appropriate ‘‘dna constructs,’’ the argument that ‘‘genes’’ constructed technically by humans deserve patenting might be considered plausible. A side note here: when dna is viewed conceptually and pragmatically from such perspectives it might seem difficult to apply a notion of the ‘‘gene’’ as suggested in pmg. However, the difficulties of maintaining standardized production in genetically modified organisms that have been mentioned show that ‘‘genes’’ in organisms—even when they are designed only to produce some product—are nevertheless embedded in a developmental context. Thus, it might be easier to identify the causes for such unreliability in a framework which contextualizes those dna sequences as, for example, in pmg. I have shown here that the explanatory goals and the pragmatic approach taken determine whether ‘‘genes’’—as hypothetical constructs, indicating the work of unknown causal factors, or as dna sequences—can rightly be described as the decisive causal power in producing some particular effect. I want to stress again that such attribution is not possible if we want to understand developmental processes and the causes of developmental outcomes. I will conclude by making two points. First, it is, of course, possible to describe the interactions of genes in developmental processes, and this results in scientific knowledge. However, this means focusing on only one causal input in the complex interactions of developmental processes. A next step, I suggest, would call for a reconceptualizing of networks as consisting of more than only dna sequences. However, it will not be a trivial task to select those conditions that should be considered to explain some particular aspect of developmental processes. Second, clearly we learn a lot when we investigate some aspects by modeling them as processes of production. But is production the same as development? I do not think so, and at this point it becomes clear that what we need but—as far as I know—still lack in biology (and elsewhere) is a clear definition of the concept of development. Until we have such a conceptualization it will be difficult to predict whether we will ever achieve a complete understanding of developmental processes and outcomes or to determine whether this problem is too holistic for scientific approaches— which of course does not deny the very existence of development, as we know it from our life experience.

262

Genes—Causes—Codes

Notes Part of this work was funded by the foundation Mensch-Gesellschaft-Umwelt at the University of Basel (grant 42/95). For stimulating discussions and helpful criticism I am grateful to Susan Oyama, Christoph Rehmann-Sutter, and Michael Weingarten. My special thanks go to Jackie Leach Scully and, in particular, to Louise Röska-Hardy for the insightful English revision. 1 Certainly there are other ways to distinguish, describe, and evaluate genetic approaches. For example, in a sense I am considering one aspect of the problem whether all gene concepts can be reduced to a molecular gene concept—that is, to molecular details. A thorough analysis of the problem of reductionism, however, cannot be carried out in the framework of this essay (for recent discussion, see Schaffner 1998). Such an analysis is presented by Sarkar (1998, 2001), who rejects the possibility convincingly. This is not to deny the possibility of a novel unified gene concept, which could subsume a molecular understanding. A wellreasoned suggestion has been put forward by Beurton (1998, 2000) and has been discussed by Neumann-Held and Rehmann-Sutter (1999). 2 This introduction of explanation is of course close to the so-called HempelOppenheim scheme (Hempel 1965), except that I use the term rule here and not natural law. The latter already seems to imply important ontological statements, which I would prefer to avoid (for a criticism, see, e.g., Janich 1997: in particular pp. 59ff.). 3 Note that according to this suggestion causes and effects are occurrences (and groups of occurrences) and not, for example, things, features, or states (‘‘Dinge, Eigenschaften oder Zustände,’’ Hartmann 1993: 80). 4 Furthermore, Hartmann (1993) argues that the so-called principle of causality, namely the claim that each incident has a cause, does not follow from his formulation of the law of causality. This is particularly important when trying to reduce reasons to causes. 5 For a more thorough analysis of Dawkins’s concept of the ‘‘evolutionary gene,’’ see Neumann-Held 1998; and Griffiths and Neumann-Held 1999. 6 It should be noted that what causes the difference among traits is neither necessarily a cause for selection at all (think of Kimura’s neutral mutations; Kimura 1994) nor a putative single cause for selection. After all, selection between traits depends mainly on the ecological embeddedness and the general (morphological, physiological, metabolic) organization of the phenotype. Phenotypic differences are thus only one (necessary but not sufficient) condition for selection. However, these problems concern general principles for modeling evolution; here, I am concerned with the question of how a gene concept, developed in the framework of population genetics, relates to others, in particular to molecular gene concepts.

263

Eva M. Neumann-Held 7 Dawkins claims, for reasons that I cannot follow in his own framework, that an evolutionary gene must occupy more space than only a single base pair (Dawkins 1982: 91). For a broader discussion and more detail, see Neumann-Held 1998; and Griffiths and Neumann-Held 1999. 8 Sociobiological explanations are offered for a broad range of phenomena of individual and social life in animals and humans. Regarding the latter, one might think of the claims for ‘‘biological foundations’’ of ‘‘marriage rules’’ (Thornhill 1992) or of morality (Alexander 1987; Wuketits 1990). I would include here ‘‘linguistic nativism,’’ which suggests nativistic aspects of language structure (Pinker 1994). For a thorough criticism of the latter, see Tomasello 1995. 9 There are numerous critical analyses of the methods of ‘‘sociobiology’’ and its followers; unfortunately, none of the well-argued and reasonable criticisms seems to have impressed the movement. The structure of its arguments—although questionable and hard to verify empirically—has basically remained unchanged. Two classical treatments are Rose, Lewontin, and Kamin 1984; and Kitcher 1990. 10 In bacteria and phages the exchange of genetic material is of course mostly or always asexual. 11 See here also Moss 2001: 91. Moss calls this gene ‘‘Gene-P,’’ which he distinguishes from the molecular Gene-D. Moss characterizes Gene-P: ‘‘A Gene-P allows one to speak predictively about phenotypes, but only . . . in a limited number of cases and within some contextually circumscribed range of probabilities.’’ 12 For references to Mendel and for a methodological analysis of Mendel’s experiments, see Gutmann and Hanekamp 1996. In contemporary molecular research one might think here of the numerous laboratory strains of E. coli which are defective in their dna repair system, which makes them suitable for recombination manipulations. 13 See Jablonka and Szathmáry 1995 for the treatment of different channels of inheritance. 14 Aside from T. H. Morgan, the most important name in the context of developing and refining this method is undoubtedly H. J. Muller (see also Sarkar, this volume). 15 Of course, other methods are applied as well, for example, the cytological testing of chromosomal structures. I can omit these here because they neither add to nor subtract from my argument. 16 I leave aside here the grave normative and ethical problem attached to the question of who determines which traits are desirable and which are not (Henn 1998, 2001). I believe, though, that there are some heritable diseases for which one may assume a general preference to prevent them if a suitable therapy is available. Since therapeutic measures are often not available, the evaluation of the explanatory value of the diagnostic methods is even more important. This is the point of my current contribution here.

264

Genes—Causes—Codes 17 Examples are dna methylation patterns (Jablonka and Szathmáry 1995), environmental conditions during developmental (embryonic) processes (Barker 1998, 2001), genomic imprinting (for a case example, see Heutink et al. 1992; for a general introduction, see Strachan and Read 1996: 201f.; for a discussion, see also Neumann-Held 2002). 18 This is not to deny that there are cases in which it is hard to imagine— and there is no empirical evidence so far to confirm—that some contextual setup could compensate for a mutational event on the dna. However, I do not see a strict theoretical (preempirical) way of predicting whether compensations are possible or not. 19 In this line of thought it is very interesting that T. H. Morgan himself cautioned against an interpretation according to which the genes would always ‘‘behave’’ in the same way. But clearly this warning had little effect on the interpretational path of genetics and was not even followed by Morgan (Keller 1995: 13f.). 20 Keller gives one argument here and writes: ‘‘But as early as 1952, geneticists recognized that the technical definition of information simply could not serve for biological information (because it would assign the same amount of information to the dna of a functioning organism as to a mutant form, however disabling that mutation was). Thus the notion of genetical information that Watson and Crick invoked was not literal but metaphoric’’ (Keller 1995: 19). 21 This is quite an interesting point. As is demonstrated in Kay’s historical analysis (2000), all formal approaches to the problem of the ‘‘genetic code’’ remained unsuccessful. In the end, it took a particular empirical approach, a highly artificial in vitro setup, to relate biochemically nucleic acid triplets to amino acids. One could interpret this as indicating that the ‘‘coding relationship’’ is a much more complicated issue in living beings (cells) than it is in an in vitro system, where it was established. But it also seems that the time then was not ripe, and more empirical evidence and more conceptual problems were required to break up the framework of ‘‘gene power.’’ I return to the ‘‘coding problem’’ below. 22 In this sense Keller refers to the Human Genome Project and writes: ‘‘Twenty years later, the progression from Watson and Crick to the Human Genome Initiative, as Watson himself has so often reminded us, appeared straightforward and logical. If all development is merely an unfolding of preexisting instructions encoded in the nucleotide sequences of dna—if our genes make us what we are—it makes perfect sense to set the identification of these sequences as the primary and, indeed ultimate, goal of biology’’ (Keller 1995: 21). 23 I am aware that the notion of ‘‘information’’ is not very clear here. The important point and new insight of dsa (Developmental System Approach) is that ‘‘information’’ is not located in the genes but is rather diffusely distributed through the developing system; that is ‘‘a difference that makes a difference’’ (Oyama 2000: 3, 161f., 199f.). Gray (2001), Godfrey-Smith (1999, 2000), and

265

Eva M. Neumann-Held

Griffiths (this volume) have tried to characterize this notion of ‘‘information’’ further, and they distinguish it strictly from any semantic view of information as conceptualized in some approaches in philosophy of mind (e.g., Millikan 1984). However, one could also criticize the use of the notion of information in genetics in general (e.g., Sarkar 1996) and from a methodological point of view (see, e.g., Janich 1998, 1999). I will leave this aspect aside here and concentrate on the notions of ‘‘cause’’ and ‘‘coding.’’ In any case I think that for any criticism of even a metaphorical use of the notion of ‘‘information’’ in genetics the seemingly obvious existence of the ‘‘genetic code’’ is a challenge. 24 pmg denotes only those processes which result in a linear polypeptide. Moss (2001: 91) finds this ‘‘unacceptably arbitrary’’ because ‘‘biologically there is no temporal boundary between polypeptide backbone synthesis and polypeptide folding and modification.’’ I counter this criticism by pointing out that I am limiting this concept to a linear polypeptide not because I think there is a temporal boundary but because there is a conceptual boundary—for the purpose of understanding particular aspects of producing polypeptides in vivo or in vitro. Setting such boundaries is a general scientific strategy and often reflects research interests more than ‘‘real’’ temporal or spatial (!) boundaries. 25 I take it that we should read ‘‘gene’’ here to mean dna sequences which are used in the processes of polypeptide synthesis. Furthermore, I take it that Godfrey-Smith refers to the effects of dna and ‘‘genes’’ only insofar as they are contextualized in an organism. 26 As quoted above, Godfrey-Smith specified that the coding relationship exists, if at all, only between dna and the linear polypeptide product, and I assume that this is what he means in this quote, although he refers to ‘‘phenotypes’’ here. 27 In a sense molecular geneticists acknowledge pragmatically that processed mrnas ‘‘make more sense’’ in analyzing ‘‘active’’ genomic dna or protein products when they use cdnas (made from mrnas through reverse transcriptase) and not genomic dna to (1) identify genomic sequences involved in polypeptide synthesis, or (2) identify polypeptide synthesis by cloning the cdna segments in expression vectors. For further details, see any standard molecular biology textbook, e.g., Knippers 1997: 262ff.). 28 I am neglecting here the involvement of ribosomes, molecules which stabilize and enzymatically support the reactions between mrnas, trnas, and amino acids. However, no one has claimed so far—and I do not see how one could— that the presence of ribosomes argues for the existence of ‘‘the code.’’ 29 Thus, developmental biologist Scott Gilbert argues that from a physiological point of view there is no beginning of an individual life but only a differentiation process of cells from a (fertilized) egg cell to an adult life form, which produces germ cells that differentiate (Gilbert 2002). 30 For an elaborate treatment of the (necessary) use of models in scientific

266

Genes—Causes—Codes

practice and of the difference between ‘‘model for’’ and ‘‘model of,’’ see Gutmann 1995. In a similar way Susan Oyama criticizes the application of computer models and the program metaphor to living beings: ‘‘Order in a process is perceived and formulated as descriptive rules. From these, prescriptive rules are derived and imposed on a mechanical medium to allow simulation of the original process. The prescriptive rules are then projected back into the original process as cognitive agents, programs, accounting for the original order in terms of the simulated order. The working of the original is then said to be ‘‘like’’ that of the imitation, and therefore due to the same kind of intentional control that created that imitation. To say it another way, order is abstracted from one system and imposed on a second, then the imposed order-as-program is abstracted from the second and projected into the first’’ (Oyama 2000: 72). 31 This can have grave consequences because the quality and the side effects of products synthesized through genetically modified organisms cannot be guaranteed. For a critical review, see Bahnsen 2002. 32 What I am referring to here is the necessity of controlling as far as possible the construction in terms of choosing the right vector, a good promoter, maybe some specific binding sites that enhance transcription or translation, and, if available and needed, some sequences which allow the implementation of the construct in the genome to be controlled.

References Alexander, R. D. 1987. The Biology of Moral Systems. New York: Aldine. Bahnsen, U. 2002. Kopierfehler im Bioreaktor. Die Zeit, no. 51, 12 Dec., pp. 35f. Barker, D. J. P. 1998. Mother, Babies and Health in Later Life. Edinburgh: Churchill Livingstone. Barker, D. J. P. 2001. A new model for the origins of chronic disease. Med. Health Care Philos. 4: 31–35. Beadle, G. W., and Tatum, E. L. 1941. Genetic control of biochemical reactions in Neurospora. Proc. Natl. Acad. Sci. USA 27: 499–506. Beurton, P. J. 1998. Was sind Gene heute? Theory Biosci. 117: 90–99. Beurton, P. J. 2000. A unified view of the gene, or how to overcome reductionism. In: P. Beurton, R. Falk, and H.-J Rheinberger (eds.), The Concept of the Gene in Development and Evolution: Historical and Epistemological Perspectives (pp. 286–314). Cambridge: Cambridge University Press. Brunner, H. G., et al. 1993a. X-linked borderline mental retardation with prominent behavioral disturbance. Phenotype, genetic localization, and evidence for disturbed monoamine metabolism. Amer. J. Hum. Genet. 52: 1032–1039. Brunner, H. G., et al. 1993b. Abnormal behavior associated with a point mutation in the structural gene for monoamine oxidase A. Science 262: 578–580.

267

Eva M. Neumann-Held

Dawkins, R. 1982. The Extended Phenotype. Oxford: W. H. Freeman. Dawkins, R. 1989. The Selfish Gene. New ed. Oxford: Oxford University Press. Gilbert, S. 2000. Developmental Biology. 6th ed. Sunderland, Mass.: Sinauer. Gilbert, S. 2002. Genetic determinism: the battle between scientific data and social image in contemporary developmental biology. In: A. Grunwald, M. Gutmann, and E. M. Neumann-Held (eds.), On Human Nature: Anthropological, Biological, and Philosophical Foundations (pp. 121–140). Berlin: Springer. Godfrey-Smith, P. 1999. Genes and codes: lessons from the philosophy of mind? In: V. G. Hardcastle (ed.), Where Biology Meets Psychology: Philosophical Essays (pp. 305–331). Cambridge: mit Press. Godfrey-Smith, P. 2000. On the theoretical role of ‘‘genetic coding.’’ Philos. Sci. 67: 26–44. Gray, R.D. 2001. Selfish genes or developmental systems? In: R. S. Singh et al. (eds.), Thinking about Evolution: Historical, Philosophical, and Political Perspectives (pp. 184–207). Cambridge: Cambridge University Press. Griffiths, P. E., and Gray, R. D. 1994. Developmental systems and evolutionary explanations. J. Philos. 91: 277–304. Griffiths, P. E., and Neumann-Held, E. M. 1999. The many faces of the gene. BioScience 49: 656–662. Gutmann, M. 1995. Modelle als Mittel wissenschaftlicher Begriffsbildung: Systematische Vorschläge zum Verständnis von Funktion und Struktur. In: W. F. Gutmann and M. Weingarten (eds.), Die Konstruktion der Organismen, II. Struktur und Funktion. Aufsätze und Reden Senck. Naturf. Ges. 43: 15–38. Gutmann, M., and Hanekamp, G. 1996. Abstraktion und Ideation—Zur Semantik chemischer und biologischer Grundbegriffe. J. Gen. Philos. Sci. 27: 29–53. Gutmann, M., and Janich, P. 1997. Zur Wissenschaftstheorie der Genetik. Materialien zum Genbegriff. In: Europäische Akademie zur Erforschung von Folgen wissenschaftlich-technischer Entwicklungen Bad Neuenahr-Ahrweiler GmbH (ed.), Graue Reihe No. 5. Halder, G., Callaerts, P., and Gehring, W. J. 1995a. Induction of ectopic eyes by targeted expression of the eyeless gene in Drosophila. Science 267: 1788–1792. Halder, G., Callaerts, P., and Gehring, W. J. 1995b. New perspectives on eye evolution. Curr. Opin. Genet. Dev. 5: 602–609. Hartmann, D. 1993. Naturwissenschaftliche Theorien. Wissenschafts-theoretische Grundlagen am Beispiel der Psychologie. Mannheim: B. I. Wissenschaftsverlag. Hempel, C. G. 1965. Aspects of Scientific Explanation and Other Essays in the Philosophy of Science. New York: Free Press. Henn, W. 1998. Der dna-Chip—Schlüsseltechnologie für ethisch problematische neue Formen genetischen Screenings? Ethik Med. 10: 128–137. Henn, W. 2001. Sind wir alle erbkrank? Zur Normalität des genetisch Abnormen. Universitas 657: 266–274.

268

Genes—Causes—Codes

Heutink, P., et al. 1992. A gene subject to genomic imprinting and responsible for hereditary paragangliomas maps to chromosome (11q23-qter). Hum. Mol. Genet. 1: 7–10. Jablonka, E., and Szathmáry, E. 1995. The evolution of information storage and heredity. Trends Ecol. Evol. 10: 206–211. Janich, P. 1997. Kleine Philosophie der Naturwissenschaften. Munich: Beck. Janich, P. 1998. Informationsbegriff und methodisch-kulturalistische Philosophie. Ethik Sozialwiss. 9: 169–182. Janich, P. 1999. Kritik des Informationsbegriffs in der Genetik. Theory Biosci. 118: 66–84. Johannsen, W. 1911. The genotype conception of heredity. Amer. Nat. 45: 129–159. Kay, L. E. 2000. Who Wrote the Book of Life? A History of the Genetic Code. Stanford: Stanford University Press. Keller, E. F. 1995. Refiguring Life: Metaphors of Twentieth-Century Biology. New York: Columbia University Press. Kimura, M. 1994. Population Genetics, Molecular Evolution, and the Neutral Theory. Chicago: Chicago University Press. Kitcher, P. 1990. Vaulting Ambition: Sociobiology and the Quest for Human Nature. 3d ed. Cambridge: mit Press. Knippers, R. 1997. Molekulare Genetik. Stuttgart: Thieme Verlag. Millikan, R. G. 1984. Language, Thought, and Other Biological Categories. Cambridge: mit Press. Moss, L. 2001. Deconstructing the gene and reconstructing molecular developmental systems. In: S. Oyama, P. E. Griffiths, and R. D. Gray (eds.), Cycles of Contingency: Developmental Systems and Evolution (pp. 85–97). Cambridge: mit Press. Neumann-Held, E. M. 1997. ‘‘Gene’’ können nicht alles erklären. Universitas 52: 469–479. Neumann-Held, E. M. 1998. Jenseits des ‘‘genetischen Weltbildes.’’ In: E.-M. Engels, T. Junker, and M. Weingarten (eds.), Ethik in den Biowissenschaften (pp. 261–280). Berlin: Verlag für Wissenschaft und Bildung. Neumann-Held, E. M. 1999. The gene is dead—long live the gene: conceptualizing genes the constructionist way. In: P. Koslowski (ed.), Sociobiology and Economics: The Theory of Evolution in Biology and Economic Thinking (pp. 105– 137). Berlin: Springer. Neumann-Held, E. M. 2001. Let’s talk about genes: the process molecular gene concept and its context. In: S. Oyama, P. E. Griffiths, and R. D. Gray (eds.), Cycles of Contingencies: Developmental Systems and Evolution (pp. 69–84). Cambridge: mit Press. Neumann-Held, E. M. 2002. Can we find human nature in the human genome? In: A. Grunwald, M. Gutmann, and E. M. Neumann-Held (eds.), On Human

269

Eva M. Neumann-Held

Nature: Anthropological, Biological, and Philosophical Foundations (pp. 141– 161). Berlin: Springer. Neumann-Held, E. M., and Rehmann-Sutter, C. 1999. Individuation and the reality of genes. A comment to Peter J. Beurton’s article ‘‘Was sind Gene heute?’’ Theory Biosci. 118: 85–95. Oyama, S. 2000. The Ontogeny of Information: Developmental Systems and Evolution. 2d ed., rev. and exp. Durham: Duke University Press. Pinker, S. 1994. The Language Instinct: How the Mind Creates Language. New York: William Morrow. Rose, S., Lewontin, R. C., and Kamin, L. J. 1984. Not in Our Genes: Biology, Ideology and Human Nature. Harmondsworth: Penguin. Sarkar, S. 1996. Decoding ‘‘coding’’—information and dna. BioScience 46: 857– 864. Sarkar, S. 1998. Genetics and Reductionism. Cambridge: Cambridge University Press. Sarkar, S. 2001. Reductionism in genetics and the human genome project. In: R. S. Singh et al. (eds.), Thinking about Evolution: Historical, Philosophical, and Political Perspectives (pp. 235–254). Cambridge: Cambridge University Press. Schaffner, K. 1998. Genes, behavior and developmental emergentism: one process, indivisible? Philos. Sci. 65: 209–252. Strachan, T., and Read, A. P. 1996. Molekulare Humangenetik. Heidelberg: Spektrum Verlag. Thornhill, N. W. 1992. Evolutionsbiologie und historische Wissenschaften. In: E. Voland (ed.), Fortpflanzung: Natur und Kultur im Wechselspiel. Frankfurt: Suhrkamp. Tomasello, M. 1995. Language is not an instinct. Cognit. Develop. 10: 131–156. Wagner, G. 1988. The vexing role of replicators in evolutionary change. Biol. Philos. 3: 232–236. Wagner, G. 1990. What has survived of Darwin’s theory?—the domestication of replicators: a neo-Darwinian commentary on the concept of replicator selection. Evol. Trends Plants 4: 71–73. Wagner, G. 1996. Homologues, natural kinds and the evolution of modularity. Amer. Zool. 36: 36–43. Watson, J. D., and Crick, F. H. C. 1953. Genetical implications of the structure of deoxyribonucleic acid. Nature 171: 964–967. Williams, G. C. 1966. Adaptation and Natural Selection. Princeton: Princeton University Press. Wolf, U. 1995. The genetic contribution to the phenotype. Hum. Genet. 95: 127– 148. Wolf, U. 1997. Identical mutations and phenotypic variation. Hum. Genet. 100: 305–321.

270

Genes—Causes—Codes

Wolf, U. 2002. Genotype and phenotype: genetic and epigenetic aspects. In: A. Grunwald, M. Gutmann, and E. M. Neumann-Held (eds.), On Human Nature: Anthropological, Biological, and Philosophical Foundations (pp. 111–119). Berlin: Springer. Wuketits, F. M. 1990. Gene, Kultur und Moral. Soziobiologie—Pro und Contra. Darmstadt: Wissenschaftliche Buchgesellschaft.

271

10 B O U N DA R I E S A N D ( C O N S T R U C T I V E ) I N T E R AC T I O N We have to look inside ourselves as well as out. . . . we cheat . . . and diminish ourselves by suggesting that we are made only of reactions to the world around us, as if we bring no inner self into our lives.

susan oyama These words come from Deborah Blum’s (1997: 261) popular treatment of biological sex differences in humans. I begin with her admonition to remember our biological interiors neither to embrace it as scientifically ratified wisdom nor to hold it up as an example of unscientific journalism. Rather, I quote Blum because she captures in her book a quite standard scientific and popular way of dealing with the boundary between our insides and our outsides. There is an enormous amount of historical and conceptual complexity wrapped up within the apparently straightforward boundary line of the skin, and I would like to examine just a few aspects of it. As my title suggests, my treatment of such boundaries is related to a certain view of biological interactions. In describing this view, however, I will touch on more general characteristics of the developmental systems approach (dsa; sometimes known as dst or dsp, for ‘‘theory’’ or ‘‘perspective’’ in place of ‘‘approach’’) with which, along with several other contributors to this volume, I am associated. I begin with an introduction to constructivist interactionism, sketching the view of the organism that emerges from this approach. Next comes a discussion of the role of biological boundaries in delimiting not just bodies or other concrete entities but classes of causal influences as well. I then make some observations on the question of individualism and offer some recent examples of theoretical boundary work in developmental biology.

Boundaries and Interaction

Interactions and Construction In recent years there has been a change in the terms I use to refer to interactions in the living world. This is less a result of some basic alteration in my thinking than a capitulation to the power of existing practice: the usages of interaction against which I have long argued have proven to be impressively stubbornly entrenched. In the mid-1980s, I noted in my book The Ontogeny of Information (2000c, originally published 1985), that some scholars whose work I found conceptually congenial avoided the label interactionist because they thought it implied separate, already defined entities that might affect each other in an interaction but that existed as such apart from those interactions.1 While acknowledging this problem, I chose not to coin a new term; there were already too many isms about. These misgivings about terminological overload linger, but in the meantime I have been misunderstood often enough that I have had to become even more careful to dissociate myself from that kind of conventional ‘‘interactionism’’ in which individuals and their environments are characterizable independently of each other (Oyama 2000a, 2000b, 2001). I call my own position constructivist interactionism to distinguish it from such standard views. Like any other term, of course, it carries perils of its own, among them confusion with antibiological, environmentaldeterminist positions in a variety of disciplines. Believing that magic words are as rare as magic bullets, however, I place less store in choosing precisely the right ones than in using whatever terms I do choose in useful and persuasive ways; this essay is part of that effort. First of all, the construction involved in developmental systems implies no constructor. I do not say ‘‘of course’’ here, because although I assume it is clear that I am not invoking a constructing deity, neither is there an autonomous constructor person (or ‘‘culture’’), as in some caricatures of social constructionism, or, if we are talking about entities at a smaller scale, a constructor molecule, nucleus, or cell. When the focus is on development, most of the moves I try to head off involve agency ‘‘from the inside-out,’’ for this is where one most often encounters the notion of internally driven, goaldirected forces, typically thought of as emanating from the dna. When we turn to evolution, we find the opposite: the constructor is ‘‘the environment,’’ or ‘‘natural selection,’’ or even ‘‘mother nature.’’ Although it may seem unnecessary, or even whimsical, to say there is no constructor, then, it is not the practice of personification that I am contesting, but rather the notion of formgiver separate and independent from that which is formed. Because dsa does

273

Susan Oyama

not make the usual move of casting development and evolution as contrasting sorts of processes, one controlled from the inside (‘‘programmed,’’ perhaps) and the other from the outside (‘‘contingent,’’ ‘‘historical’’), the sense of construction is similar for both: the appearance over time of an interactive product, in this sense emergent rather than prefigured or predefined. The emphasis here is on the concrete processes of coming into being. The precise characteristics of these processes and their outcomes are a matter of the particular entities and conditions in place rather than of preexisting necessities. By preexisting necessities I mean internal natures, representations, instructions, or codes (in the case of development), or already defined external environmental problems and demands (in the case of evolution). In a developmental system, there are no repositories of form, but rather complexes of interactants and interactions that can be very similar or stable across organisms and through time. Although processes and outcomes can be exceedingly predictable, these constructive interactions are not to be thought of as ‘‘expressions’’ or ‘‘impositions’’ of separately existing forms or requirements. They are a function of the state of the whole complex. The organism and current conditions in turn bear the marks of their histories. Stability or reliability, then, are themselves contingent. Systemic interdependence does not imply automatic predictability or general resistance to perturbation, any more than interaction implies malleability or variability (Oyama 2001).

Organisms and Outlines: Containing Causation An organism’s skin is a theoretically and empirically salient boundary. There are other important borderlines, as we shall see later, including the one drawn around the cell. They are often used to distinguish classes of developmental causes, which are then given different roles in explaining ontogeny. Similar inside/outside polarities are implemented in discussions of evolution as well. Thus we find the familiar contrasts between internal and external causes: genes and environments, biology and culture. Relations of mutual constitution, however, are incompatible with treatments that rely on prespecified environments, whether they are the ontogenetically relevant environments in a developmental story or the selectively relevant ones in an evolutionary one. Accounts of entities that carry their ‘‘natures,’’ their possibilities and propensities—their personalities, if you will—strictly within them, independent of their settings (albeit only revealed in those settings), are no better (see Rehmann-Sutter 2002 on ‘‘the metaphysical genome’’).

2 74

Boundaries and Interaction

In the quotation with which I began, for instance, our ‘‘inner selves’’ are being distinguished from our ‘‘reactions’’ to the outer world, as though the self or one’s true nature were separable from traffic with the outside. It is precisely such metaphors of containment that are at issue here: an entity’s futures are imagined somehow to be enclosed or prefigured in it. We ask whether an acquaintance ‘‘has it in her’’ when we are not sure whether she will be able to accomplish some task or successfully face some challenge. We thus transform uncertainty about a person’s future into a query about her interior, making an interaction into a context-free capacity. Such figures of speech are often harmless. They are so familiar and deeply embedded in our everyday talk, however, that when they appear in the scientific literature their metaphoric colors are invisible. Insofar as they reflect, and feed, a set of problematic oppositions between insides and outsides, they are not as innocent as they seem. It is one thing, for example, to define a norm of reaction as the range of phenotypes found when organisms with a given genotype develop in a variety of environments, and quite another to explain those phenotypes as manifestations of the possibilities already specified by, contained by, the genome. Given the prevalence of this explanatory style, it is no surprise that the unwashed public thinks that our genes control our futures. Such ‘‘vulgar’’ misunderstandings are actually quite accurate adoptions of the scientific habit of identifying stability with insides and considering outsides important mainly as sources of variation (for other examples, see van der Weele 1999: ch. 2). By parity of reasoning, we can ask, Is the range of phenotypes that can develop in a given environment specified or contained by that environment? The emphasis in the developmental systems approach is on mutually constituting entities and their surrounds. This, along with the associated skepticism about the usual distinctions between internal and external causes, is sometimes taken to mean that scientific analysis must grind to a paralyzed halt because no distinctions can be made.2 On the contrary, however, the framework provides for a multitude of questions and analytical techniques, with the advantage that they are clearly distinguished rather than being run together willy-nilly, and therefore are capable of generating well-specified, properly circumscribed hypotheses (on the pragmatics of research, see van der Weele 1999; Neumann-Held 2002). What is at issue is the possibility of identifying, if we are speaking at the organismic level, a biological nature apart from a lifetime’s worth of (constructive) interactions with a lifetime’s worth of multileveled environments. Perhaps not that many people would disagree with this last statement if

2 75

Susan Oyama

presented with it point-blank; obviously, the precise wording would be important. Yet taking it seriously would mean that one could never again speak of an ‘‘inner self ’’ different from ‘‘reactions to the world,’’ or of a person’s biological nature ‘‘interacting’’ with a culture, or of a child’s phenotype being inconsistent with, or even misrepresenting, its genotype (see Sterelny and Griffiths 1999: 104, on intentionality and misrepresentation in discussions of biological information). Organisms, including humans, are entirely biological—they are alive, and their physiologies involve metabolism, the selective production of proteins and other biochemicals, and so on. The genetic processes entailed in all this are possible only by virtue of the immediate environments of the dna, as well as the larger contexts within which it becomes possible to understand why this stretch of dna is transcribed rather than another, and why in this cell and not another, and at this time, not earlier or later (Nijhout 1990, 2001; Strohman 1997). Actual attention to the functioning of genetic material, then, ‘‘naturally’’ leads us outward as well as inward, and there is no obvious place to stop. The stopping point—that is, the extent of the research context—depends on the nature and goal of the inquiry, and every scientist knows that one possible fruit of investigation is guidance on whether, and how, to expand or contract the scope of the search. Although there are many discriminations to be made about the consequences of some biochemical, morphological, or ecological difference for an organism, its offspring, its prey, or its more general surround, or about patterns of features in a population, none of these is illuminated by locating fundamental causes on one side of the skin and mere conditions and materials on the other.

Insides, Outsides, and Individualism Given all these indefinite outlines and interpenetrating, mutually constituting entities and surrounds, it is perhaps surprising that colleagues occasionally tell me that the developmental systems approach is ‘‘individualistic.’’ 3 The question has not, to my knowledge, been raised in print. Nor has it always been clear what meaning(s) of individualism these colleagues may have had in mind, or what prompted the claim in the first place. In one case the charge stemmed partly from a mistaken conviction that social and cultural factors are excluded from the ‘‘environment’’ in developmental systems writing. There are, however, numerous explicit statements to the contrary in the developmental systems literature, in which examples at the social and

2 76

Boundaries and Interaction

physiological levels are often juxtaposed precisely to make the point that the social environment is part of the developmental environment. Other interpretations of individualism would seem similarly inappropriate: the emphasis on multiple scales and system dynamics in dsa militates against any idea that social processes are best understood as aggregations of individual actions or in terms of egoistic motivation, for instance, or that evolution must be approached only at the organismic level. In any case, individualism is hydra-headed and unruly, conceptually, politically, and ethically. Touching as it does on issues of identity and responsibility, it enters into many debates about public policy, where it typically indexes a general emphasis on people’s characters or personalities rather than their surroundings (including, of course, their social surroundings), especially with respect to initiative, effort, achievement, and moral or social responsibility. Those who are guided by such an emphasis on individual choices and actions are apt to be charged with ‘‘blaming the victim,’’ and they may then countercharge that arguments about social conditions lead to moral chaos by absolving persons of responsibility for their own behavior. Such debates depend on precisely the conceptual framework that the developmental systems approach is meant to demolish. After all, a salient aspect of that framework is the way it cleaves people (or ‘‘selves,’’ as in the earlier discussion) from their worlds. It is true that a focus on individuals could result from a limited investigative strategy in a particular research project. This would amount to a wager that focusing on individuals will be more immediately productive than starting with groups, say, or institutions (see Godfrey-Smith 1996 on internalist and externalist explanations of mind as empirical bets). As the previous paragraph suggests, however, a concentration on individuals is not usually so narrowly and pragmatically specific, but is more likely to be covertly or overtly linked to broader ontological allegiances: beliefs about what is ‘‘real,’’ ‘‘primary,’’ or ‘‘truly explanatory.’’ Often individualism involves a general privileging of the internal over the external. In a more abstract vein, it can indicate a belief that individuals are somehow prior to collectivities, the latter being simply aggregations of more or less free-standing entities with their qualities already given. The disciplines of psychology and sociology, in fact, have historically been defined largely by just such allegiances, and by their associated methodological choices. It is worth noting that opting for externalism does not necessarily remove the focus on individuals; taking ‘‘the mind’’ as one’s object of study usually suggests a psychological, and thus to some extent individualist, bias in itself. The contrast class changes as we shift levels, from internal versus external

277

Susan Oyama

explanations of individual minds to individual versus social explanations of behavior or larger-scale patterns. I cannot analyze these matters in detail here. Because the worry that the notion of a developmental system somehow implies any of these varieties of individualism is so terribly ill-founded, however, at least a few remarks are in order.4 First I will say a bit about levels of analysis; next comes a brief look at the delineation of entities, like individual organisms; and then I will consider the causal complexes that produce these entities.

Choosing Units of Analysis The possibility of shifting scales, of both time and magnitude, is part and parcel of the developmental systems approach. Because dsa does not rely on the usual conceptual apparatus of internal and external causes, it does not force a choice between explaining behavior, for instance, by the characteristics of persons or by their social settings. As noted above, such unsavory dilemmas have often figured in the competition between psychologists and sociologists to define the proper unit of analysis. The idea that the individual is the correct object of study, or the best unit with which to explain collective phenomena, can be related to the distinctly non-system-theoretic habit of explaining wholes by their independently existing, separately describable parts. In the public and academic arguments about whether persons (that is, particular characteristics of individuals, typically understood as being internal to them and therefore the proper object of psychology) or societies (often seen as external to persons) are to blame for various misfortunes and misdeeds, one can readily see how the disciplinary politics of individualism merges with ethics and politics in a larger sense. Whether such arguments are conducted in scholarly journals or in the popular press, they are usually thought to bear on a whole host of social issues, from welfare to mental health, from criminal justice to education, from feminism to taxation. In each of these areas, attributing behavior primarily to inside or outside forces leads to differing prescriptions for the personal and societal dilemmas that confront us all. There is no place in the developmental systems framework for the kinds of conceptual moves that support such polemics. dsa, that is, does not look to already defined, individual constituents that can simply be aggregated to arrive at the properties of the whole; but neither does it allow the style of explanation that treats people as mere results or products of the autonomous

2 78

Boundaries and Interaction

operation of collectivities, institutions, or social forces. This last point is crucial, for it blocks the conclusions that will tempt some: that dsa forsakes the individual for the social, or internalism for externalism. Ironically enough, in light of the previous discussion, this would in some cases amount to fleeing one or another variety of individualism to embrace its mirror image.

Delineating Entities If proper units or levels of analysis do not unequivocally announce themselves to us, apart from our preoccupations and projects, their qualities conveniently ‘‘preregistered,’’ as Hendriks-Jansen puts it (1996: 55; see also Varela, Thompson, and Rosch 1991: 148 on a ‘‘pregiven world’’), then neither do cleanly bounded entities identify themselves. One should not think, then, that because I do not explicitly define organism I consider its identification to be unproblematic. Enough hard cases come readily to mind to make it obvious that it is not a simple thing to draw an outline around an individual organism (for a richly provocative treatment, see Turner 2000). Think of clones, colonial organisms, and symbionts—especially endosymbionts, some of which have become intimately integrated into host life cycles, and even host cell nuclei (Margulis 1981), to say nothing of conjoined twins and multiple personalities in humans. In fact, when pondering this matter I have been reminded of the notion of the infinitely long coastlines described by mathematicians who point out that shifting scale—from bays and peninsulas to inlets and promontories, and then to smaller features such as stones and pebbles—inexorably increases the distance one must trace between two locations (Mandelbrot 1983: ch. 5). If we follow our skin to its transition to the mucous membrane of the mouth and throat and beyond, in a kind of topological analogy to the coastline, we can ask whether our gut symbionts are inside or outside us. A particle of food could be considered inside once it has been absorbed into the bloodstream or into a cell, but one of our cells similarly resolves, if we look closely enough, to a maze of structures, channels, and pores, constantly changing their configurations and traversed by frantic traffic. The problems of defining what individuals are go far beyond distinguishing inside from outside. This difficulty generally does not prevent us from discussing cells and organisms or groups in interesting and productive ways. We can alter or refine our definitions and strategies as the need arises, and it is useful to keep in mind that this is what we are doing—adopting locally useful criteria and rules of thumb for a reason. But recall that one way to interpret

2 79

Susan Oyama

individualism is as a belief in the primacy—ontological, analytic, or some other kind—of discrete individuals, and insofar as this requires a commitment to independently specifiable and clearly and unequivocally bounded organisms, or to one level of analysis over others, the developmental systems approach does not supply the necessary raw materials.

Delineating Causal Complexes The boundaries of a developmental system, at whatever scale we may be working, are no more fixed, once and for all, than those of the entities themselves, organismic or otherwise. They change over time and, again, with the type of investigation. The extent of the set of causally relevant factors will differ according to the project and the stage of the investigation, but this again poses no barrier to research. In fact, explicitly acknowledging, and then evaluating, our choices should reduce the risk of missing important influences because of unexamined assumptions and analytic traditions (Griffiths and Gray 1994; van der Weele 1999; Oyama 2000a). Whether we are speaking of dna segments, cells, organisms, or groups, however they are individuated, we must, for coherence, consistency, and comprehensiveness, include the context in the explanatory complex, and not only as a container or a causally secondary set of modulators or materials, but as constitutive of the processes and products in question. Hence, any kind of individualism that requires explaining organisms by invoking only generative mechanisms within those organisms is also blocked by constructivist interactionism.

Boundary Work: Cellular Programs and Hybrid Disciplines Units of analysis, biological entities, and causal complexes do not simply report for duty, uniformed and ready to march; considerable work is sometimes required to prepare them for deployment. Some recent writings on developmental biology by Evelyn Fox Keller and Scott Gilbert provide examples of this kind of boundary work—work that is at once concrete and conceptual, strategic and disciplinary. Juxtaposing these two authors’ explanatory strategies with my own should help clarify the issues and throw into relief the similarities and differences in our positions.

280

Boundaries and Interaction

Cellular Developmental Programs? In her contribution to Cycles of Contingency, the collection I edited with Paul Griffiths and Russell Gray, Evelyn Fox Keller (2001: 302; see also this volume) takes me to task for some skeptical comments I once made about the claim that cells explain development (Oyama 2000b: ch. 1). Keller starts with the single-cell stage in complex organisms, when the boundaries of cell and organism coincide. She worries that the developmental systems approach erases that boundary, along with cells’ special qualities; for her, the cell is ‘‘a unit of development.’’ Now, I have no argument with viewing cells as important to the study of development, but there is a problem. Here is the context of our exchange: I had quoted Sidney Brenner, known for his work on the worm Caenorhabditis elegans, who once said: ‘‘The total explanation of all organisms resides within them, and you feel there has to be a grammar in it somewhere. Ultimately, the organism must be explicable in terms of its genes, simply because evolution has come about through alterations in dna’’ (cited in Lewin 1984: 1327). This reasoning locks in a developmental internalism that I do indeed dislike, and although the quotation about development and evolution comes from several decades back, the reasoning is still current. Interestingly enough, the passage that troubled Keller was one in which I objected to the practice of investing genes with special agency— something she herself has criticized with great skill (Keller 1985). My complaint was not about taking bounded entities as objects of inquiry (‘‘a unit of development,’’ say) but about the way they are described and placed in the larger story of development. Brenner’s self-professed failure to ‘‘find’’ a program in the genes had forced him up to the cellular level, but he had not abandoned his wish for the unit of development. There is an important difference between a and the here, especially when the definite article is justified by the genes’ ability, in Brenner’s words, to ‘‘get hold of the cell.’’ I mistrust this search for ‘‘the’’ unit. The possibility of treating the cell as an analytic unit is embedded in the developmental systems framework as firmly as the cell is embedded in its developmental contexts. At the same time, the implausibility of employing the cell as the explanatory unit is demonstrated by the lower- and higher-level interactions needed for them to function as they do. What is crucial is the multiple embeddings in a developmental system, the multiple levels that both allow provisional focus on particular entities and prevent any from being elevated to the kind of prime movers that students of development have so often sought. Also crucial is the distinction

281

Susan Oyama

between the line defining a cell or an organism, however one draws it, and the one required to encompass their formative influences. Developmentalists are perennially whipsawed between their allegiance to insides and their need to take appropriate account of outsides, and the play has often been in exactly what it means to take appropriate account of each. Without a consistently theorized distinction between the boundary of the target entity and the boundary of its developmental system, workers have resorted to rather arbitrary causal privileging: acknowledging outside influences but making them secondary to the formative (informative, active, fundamental) ones inside. More causal evenhandedness—the parity of reasoning that is basic to developmental systems thinking—does not require us to ignore the ability of the cell to function in a variety of contexts, a robustness Keller emphasizes but that we should not exaggerate. A commitment to parity does oblige us to accept the importance of the surroundings in making such resilience possible, and so to accept the causal role of what is constant in environments, not just what is variable: what is predictable about them as well as what is not. I sympathize with one of Keller’s aims in making her case for the bounded cell: to remember the body. (One of her concerns is the human, especially the female, body, and she cites Barbara Duden in this regard. One should note, though, that this way of pursuing a concern with bodies raises questions about the claim, discussed earlier, that it is dsa that is individualist. See Rehmann-Sutter 2002 for remarks on developmental systems, essentialism, and embodiment.) Keller rightly observes that bodies are in constant danger of disappearing when genes dominate discourse. Taking the moment when the limits of the body and the cell are the same (hence her rhetorical movement between the skin of a body and the membrane of a cell), she suggests reversing the convention of the genetic program. Exploiting the relative interchangeability of data and program, Keller proposes that we view ‘‘the fertilized egg as a massively parallel processor in which ‘programs’ (or networks) are distributed throughout the cell’’ (2001: 307; see also Keller 2000: 99). dna then serves as their data. Saying she ‘‘accept[s] the metaphor of program, warts and all’’ (2001: 302), Keller criticizes only its identification with the genes: for her, the cell is programmed. In The Ontogeny of Information (2000c: 79) I used the reversibility of program and data for a different purpose: to undermine the program as an explanatory device in biology and psychology. One chooses one’s battles where one will, and Keller and I agree on a great deal. Still, when she adds

282

Boundaries and Interaction

in a note that this ‘‘developmental program . . . is everywhere to be found’’ (2001: 307; see also Keller 2000: 162), I’m itching to ask, everywhere? Or only, as it seems, everywhere in the cell, whose functional autonomy she attributes to its special boundary? And if only there, why? Anticipating the question, Keller charges developmental systems theorists with treating the body—including the cellular ‘‘body’’—as mere environment for the genes. But again the ability to move among levels, or scales of magnitude, changes the picture. The intracellular milieu is immediate environment to the dna, but of course it is not just environment, not just container or locale; it is constitutively part of genetic processes. The cell is also composed of a host of molecular ‘‘bodies’’ with shapes, sizes, internal relations, and locations. And that cell is a bounded entity among others. Its membrane does indeed keep things together, as Keller observes, even as it ‘‘regulates . . . traffic between inside and out’’ (2001: 301). I would argue that it is exactly such spatial and temporal relationships that are captured by the idea of a developmental system (which includes, after all, all sorts of things bumping up against each other). Much of the appeal of the language of programs and information, on the other hand, lies in its formal air: its detachment from matters like mere matter. For many infophiles, what is exhilarating is precisely the indifference to particular material that such language encourages. The ‘‘substrate neutrality’’ that theorists like Daniel Dennett (1995) celebrate means that programs can be instantiated in brains, silicon, or tin cans. Privileging the formal over the material, the abstract over the concrete, infotalk too often obscures the physicality and temporality of developmental interactions.5 When time, shape, and spatial propinquity are forgotten, the felt need arises for something else—some agent to bring those molecules together, and back we go on the hunt for designer genes. In like manner, the multicellular organismic body, however tricky it is to delineate, is at once environment to its internal parts and itself a more or less bounded entity in a more or less bounded surround. In a developmental system the organism’s body is not, strictly speaking, environment to the genes, as Keller would have it; the immediate intracellular environment is, while the dna is in turn ‘‘environment’’ to other cellular constituents, just as one bit of dna is part of the context to other bits. It would seem that there are boundary issues here that are yet to be resolved.

283

Susan Oyama

Eco-Devo, Evo-Devo Recall that in the developmental systems perspective stability is attributed not to enclosed interiors but to an extended, heterogeneous system. Scott Gilbert’s recent article on eco-devo (the intersection of ecology with developmental biology) catalogs many nice examples of environmental participation in early development, something Cor van der Weele (1999) has ably done as well. Gilbert begins with a standard description of ontogeny as differential gene expression (2001: 1), but then continues, ‘‘However, the regulators of gene expression need not all reside within the embryo.’’ He speaks of ‘‘tertiary induction’’ (p. 4) which, unlike standard embryonic induction by agents internal to the embryo, involves inducers from the outside. I join Gilbert in wishing for more attention to the substantive (not just supportive, modulatory, or disruptive) role of environmental factors in development, as I do in looking forward to increased exchange between ecological and developmental studies (Oyama 2000c: afterword). I wonder, though, whether the (perhaps vestigial) developmental internalism in his introduction to eco-devo does justice to a full synthesis of the internal and external worlds. Does his reference to the reaction norm as ‘‘a property of the genome’’ (2001: 3), for instance, suggest the most integral ecological developmental biology one could envision? Or could it be an instance of the historical tendency that he rightly complains of on the same page, the tendency to make developmental variation a property of the genes? Is placing the norm of reaction in the dna, that is, another of those strategies of conceptual containment? And does Gilbert’s emphasis on phenotypic plasticity reflect the traditional association of environmental factors with developmental variation? Cor van der Weele (1999), who challenged Gilbert to attend more seriously to environmental factors in ontogeny, included in her book on the subject what is largely absent from Gilbert’s treatment: the oft-invisible role of the surround in normal or invariant developmental systems—that is, when it is not a source of variation. The ‘‘ecology of development’’ that she envisions (p. 199) would thus seem more inclusive than one still focused on variation. I would, in addition, want to include the role of organisms in creating, maintaining, and altering those ecological surroundings. The organism’s ontogenetic story requires its developmentally relevant environment to complete the developmental system. Just as the cell membrane keeps things together without benefit of a program, so do an organism’s interactions with its surround, and the same can be said of larger-scale processes, many of which are quite regular; Yrjö Haila (1999), in fact, writes of ecology’s

284

Boundaries and Interaction

seasonal cycles as ‘‘developmental.’’ Colonies, ranges, niches, and territories are spatially bounded areas within which crucial interactions occur, and outside which they become less likely. Special techniques may be required to make those boundaries visible, just as they are for cell and nuclear boundaries. The ecological entities are not just orderly to varying degrees; they also give spatial and temporal order to the organisms of which they are partially composed, participating in the network of relations that enable cells and other units to be so well behaved (when they are well behaved). Cellular processes in turn affect these larger ones, sometimes very indirectly, so that, as I noted above, we can eventually ask how an organism’s developmental interactions aid the very reconstruction of the seasonal changes that support its continuation in its ecological surroundings. This would be part of the research program of eco-devo, which Gilbert rightly notes would embrace organisms outside the small set that have traditionally been studied. In large part these have been selected for study, he adds (p. 3), precisely because they live and work so happily under impoverished laboratory conditions—that is, because they are not apt to show the kinds of developmental context-dependencies he outlines in his paper. Such robustness, of course, then supports internalist explanations of the developmental courses that are observed.6 Gilbert has also been involved in another scientific hybrid: evolutionary developmental biology, now referred to as evo-devo. He, Opitz, and Raff (1996) seek to rehabilitate embryology’s concept of the morphogenetic field. The cell, they say, is the ‘‘unit of organic structure and function,’’ but morphogenetic fields rather than cells or genes are ‘‘a major unit of ontogeny whose changes bring about changes in evolution’’ (p. 357). These fields are defined as ‘‘discrete units of embryonic development . . . produced by the interactions of genes and gene products within specific bounded domains. They are therefore defined in terms of information that becomes translated into spatial entities’’ (p. 366). Apparently locating the generation of form squarely within the organismic skin, the authors describe these fields as ‘‘based on genetically defined interactions among cells’’ (p. 367). What happens if we place these two interdisciplinary initiatives, evo-devo and eco-devo, in confrontation with each other? From some points of view (like mine) they seem to be in some tension. The paper on evo-devo expresses developmental biology’s informational, genetic, and internalist loyalties, while eco-devo would seem to point outward, toward the surrounding world. Gilbert’s (2001: 4) piece on ecology and development, in fact, does both: as I noted earlier, it has some quite traditional developmentalist ten-

285

Susan Oyama

dencies, but it also includes that reference to induction from the outside. Once factors outside the organism’s skin are given the serious consideration Gilbert argues for in eco-devo, it is natural to ask whether evo-devo’s (or, indeed, any developmentalists’) explanatory units must be wholly contained within that skin. What, for instance, should we make of Gilbert, Opitz, and Raff ’s (1996: 369) statement that ‘‘genes make the cells and the cells form the body’’? Asked whether even morphogenetic fields (traditionally conceived in quite internalist terms) could include extraorganismic factors, Gilbert (pers. comm., 15 October 2001) gave examples in which gravity and temperature played crucial roles in morphogenesis. Though I did not rely on the concept of the morphogenetic field in my 1985 book, I did cite both gravity and temperature as examples of extraorganismic factors whose contributions to development made them part of the developmental system. Indeed, it would be silly to exaggerate my differences with either Gilbert or Keller. On the full spectrum of views, we would be virtually indistinguishable. Although it is too early to know what will come of the encounters between dsa and the two hybrid developmental biologies, there are surely ample affinities among them.7

Conclusion dsa’s somewhat heterodox notion of interaction as developmental construction replaces the traditional schemes of internal and external causes with the notion of a developmental complex that extends far beyond the organism’s skin. Because it treats organisms and their environments as interdefining and interdependent (in what Levins and Lewontin 1985 call organism-environment interpenetration; see also Lewontin, Rose, and Kamin 1984), it would seem to be inhospitable to most kinds of individualism. First, it denies that the organism (or for that matter the gene or the cell) is somehow the natural or basic unit of analysis, instead emphasizing dependencies among levels. Its generally pragmatic stance also means that the delineation of individual organisms is apt to be neither obvious nor unequivocal. Finally, its conceptualization of causal complexes eliminates many of the moves traditionally used by individualist approaches and by many of their (ineffective) critics as well. All this opens up the possibility of giving appropriate weight (not necessarily ‘‘equal’’ weight or importance, as some readers worry—the metric and weighting themselves depend on the inquiry) to the many kinds of factors that can contribute to the continuous formation and re-formation of

286

Boundaries and Interaction

organisms, but without making the organism-environment divide primary in characterizing the development of the organism itself.

Notes Earlier versions of parts of this chapter were delivered to the Boston Colloquium for Philosophy of Science and to the Indiana Seminar on Science, Language and Culture. My thanks go to those who were willing to exchange ideas with me before, during, and after those occasions. I am also grateful to the editors for their thoughtful and penetrating comments. 1 See The Ontogeny of Information for an extensive introduction to the ‘‘developmental systems approach’’; a brief overview is also given by Griesemer, this volume. 2 . . . and no research grants won! 3 Gerson, in a personal conversation, 19 February 2002. 4 A possible exception is the pragmatically circumscribed and provisional research strategy mentioned earlier, but as I suggested, it is unlikely to exist in a conceptual vacuum. 5 I need hardly add that these are also crucial characteristics of the bodies Keller wishes to keep firmly before us. 6 Anne Fausto-Sterling (2000: 358–359) gives a nice account of the interaction between theory and animal models. 7 See Robert, Hall, and Olson 2001 for a somewhat skeptical view; see Robert (2003) for more enthusiasm.

References Blum, D. 1997. Sex on the Brain: The Biological Differences Between Men and Women. New York: Viking. Dennett, D. C. 1995. Darwin’s Dangerous Idea. New York: Simon and Schuster. Fausto-Sterling, A. 2000. Sexing the Body: Gender Politics and the Construction of Sexuality. New York: Basic Books. Gilbert, S. F. 2001. Ecological developmental biology: developmental biology meets the real world. Dev. Biol. 233: 1–12. Gilbert, S. F., Opitz, J. M., and Raff, R. A. 1996. Resynthesizing evolutionary and developmental biology. Dev. Biol. 173: 357–372. Godfrey-Smith, P. 1996. Complexity and the Function of Mind in Nature. Cambridge: Cambridge University Press.

287

Susan Oyama

Griffiths, P. E., and Gray, R. D. 1994. Developmental systems and evolutionary explanation. J. Philos. 91: 277–304. Haila, Y. 1999. ‘‘Biodiversity’’ and the nature/culture divide: conflicting tendencies. Biodivers. Conserv. 8: 165–181. Hendriks-Jansen, H. 1996. Catching Ourselves in the Act. Cambridge: mit Press. Keller, E. F. 1985. Reflections on Gender and Science. New Haven: Yale University Press. Keller, E. F. 2000. The Century of the Gene. Cambridge: Harvard University Press. Keller, E. F. 2001. Beyond the gene but beneath the skin. In: S. Oyama, P. E. Griffiths, and R. D. Gray (eds.), Cycles of Contingency: Developmental Systems and Evolution (pp. 299–312). Cambridge: mit Press. Levins, R., and Lewontin, R. 1985. The Dialectical Biologist. Cambridge: Harvard University Press. Lewin, R. 1984. Why is development so illogical? Science 224: 1327–1329. Lewontin, R. C., Rose, S., and Kamin, L. J. 1984. Not in Our Genes. New York: Pantheon. Mandelbrot, B. 1983. The Fractal Geometry of Nature. Updated and augmented. San Francisco: W. H. Freeman. Margulis, L. 1981. Symbiosis in Cell Evolution: Life and Its Environment on the Early Earth. San Francisco: W. H. Freeman. Neumann-Held, E. M. 2002. Can we find human nature in the human genome? In: A. Grunwald, M. Gutmann, and E. M. Neumann-Held (eds.), On Human Nature: Anthropological, Biological, and Philosophical Foundations (pp. 141–161). Wissenschaftsethik und Technikfolgenbeurteilung, Band 15. Berlin: Springer Verlag. Nijhout, H. F. 1990. Metaphors and the role of genes in development. BioEssays 12: 441–446. Nijhout, H. F. 2001. The ontogeny of phenotypes. In S. Oyama, P. E. Griffiths, and R. D. Gray (eds.), Cycles of Contingency: Developmental Systems and Evolution (pp. 129–140). Cambridge: mit Press. Oyama, S. 2000a. Causal democracy and causal contributions in dst. Philos. Sci. 67 (Proceedings): 5332–347. Oyama, S. 2000b. Evolution’s Eye: A Systems View of the Biology-Culture Divide. Durham: Duke University Press. Oyama, S. 2000c. The Ontogeny of Information: Developmental Systems and Evolution. 2d ed., rev. and exp. Durham: Duke University Press. Oyama, S. 2001. Terms in tension: what do you do when all the good words are taken? In: S. Oyama, P. E. Griffiths, and R. D. Gray (eds.), Cycles of Contingency: Developmental Systems and Evolution (pp. 177–193). Cambridge: mit Press. Rehmann-Sutter, C. 2002. Genetics, embodiment and identity. In: A. Grunwald, M. Gutmann, and E. M. Neumann-Held (eds.), On Human Nature: Anthropo-

288

Boundaries and Interaction

logical, Biological, and Philosophical Foundations (pp. 23–50). Wissenschaftsethik und Technikfolgenbeurteilung, Band 15. Berlin: Springer Verlag. Robert, J. S. 2003. Developmental systems and animal behaviour. Review of Evolution’s Eye: A Systems View of the Biology-Culture Divide, by Susan Oyama. Biol. Philos. 18 (2003): 477–489. Robert, J. S., Hall, B. K., and Olson, W. M. 2001. Bridging the gap between developmental systems theory and evolutionary developmental biology. BioEssays 23: 1–9. Sterelny, K., and Griffiths, P. 1999. Sex and Death: An Introduction to Philosophy of Biology. Chicago: University of Chicago Press. Strohman, R. C. 1997. The coming Kuhnian revolution in biology. Nat. Biotechnol. 15 (March): 194–200. Turner, J. S. 2000. The Extended Organism: The Physiology of Animal-Built Structures. Cambridge: Harvard University Press. Varela, F. J., Thompson, E., and Rosch, E. 1991. The Embodied Mind. Cambridge: mit Press. Weele, C. van der. 1999. Images of Development: Environmental Causes in Ontogeny. Albany: State University of New York Press.

289

11 B E YO N D T H E G E N E B U T B E N E AT H T H E S K I N

evelyn fox keller The end of the twentieth century witnessed an efflorescence of critical commentary deploring the excessively genocentric focus of contemporary molecular and evolutionary biology. Among philosophers, the best known work may well be that of developmental systems theorists (dst) Oyama, Griffiths, and Gray (and sometimes including Lewontin and Moss), but related critiques have also emerged from a variety of other quarters (see Oyama et al. 2001). Expressing diverse intellectual and philosophical preoccupations, and motivated by a variety of scientific and political concerns, these analyses have converged on a number of common themes and sometimes even on strikingly similar formulations (see, e.g., Keller 1995, 1999; Griffiths and Neumann-Held 1999). Common themes include conceptual problems with the attribution of causal primacy (or even causal efficacy) to genes (e.g., Lewontin 1992; Moss 1992; Keller 1995; Strohmann 1997); disarray in contemporary uses of the very term gene; and confusion and misapprehension generated by use of the locution genetic program. Much of the impetus behind these critiques issues from long-standing concerns, and indeed, many of the critical observations could have been (and in some cases were) made long ago. Why, then, their particular visibility today? An obvious answer lies close at hand: Critiques of genocentrism have found powerful support in many of the recent findings of molecular biologists. Indeed, I would argue that it is from these empirical findings that the major impetus for a reformulation of genetic phenomena now comes. Three developments (or findings) are of particular importance here: (1) the need for elaborate mechanisms for editing and repair of dna to ensure sequence stability and fidelity of replication, (2) the importance of complex (and nonlinear) networks of epigenetic interactions in the regulation of transcription, and (3) the extent to which the ‘‘sense’’ of the messenger transcript

Beyond the Gene but Beneath the Skin

depends on highly regulated mechanisms for editing and splicing. The implication of the first is that the structure of the gene (or sequence of dna) may be the (or, from the perspective of dst, ‘‘a’’) locus of heredity constancy, but it can no longer be supposed to be its source: particular genes (or sequences) persist as stable entities only as long as the machinery responsible for that stability persists. The dependence of gene function on complex epigenetic networks challenges (or at least seriously complicates) the attribution of causal agency to individual genes. Finally, the third finding radically undermines the assumption that proteins are simply and directly encoded in the dna (indeed, it undermines the very notion of the gene as a functional unit residing on the chromosome). On this much one finds a certain general agreement. But differences— deriving in part from the different intellectual, scientific, and political perspectives of their authors—can also be found. Sometimes these are matters of emphasis, sometimes of focus, and sometimes of more substantive import. In this essay I will focus on what I believe to be a substantive issue distinguishing my own perspective from that which tends to dominate the dst literature. That issue can be put in the form of a question: Is there a place on our biological map for the material body of the organism, for that which lies beyond the gene yet beneath the skin? And if so, where is that place?

The Body in Question I share with proponents of dst the conviction that the oppositional terms in which the nature-nurture debate has historically been framed are both artificial and counterproductive. But the particular question I pose here reflects an additional source of unease, and that is over the tacit elision of the body not only implied by the framing of the classical controversies but at least partially continued in the solutions that have thus far been put forth. The most conspicuous roots of my question are doubtless to be found in the history of genetics and neo-Darwinian evolutionary theory. With the emergence of genetics in the early part of the twentieth century, debates over the relative force of nature and nurture (first framed as such by Francis Galton in 1874) were recast, initially, in terms of heredity and environment (see, e.g., Barrington and Pearson 1909; Morgan 1911; Conklin 1915), and soon after, in terms of genes and environment (see, e.g., analyses of the relative importance of heredity and environment in Fisher 1918; Wright 1920). This second reframing may have been an inevitable consequence of the terminological

291

Evelyn Fox Keller

shift in the biological literature of the 1920s and later, in which the term heredity came to be replaced by the newer term genetics.1 But the consequence of this shift was more than terminological: it amounted to a conceptual reduction of ‘‘nature’’ to ‘‘genes,’’ and with that reduction allowed only one of two possible statuses for the extragenetic body: either its complete elision or its relegation to the category of ‘‘nurture’’ or ‘‘environment.’’ Genetics further contributed to this relegation with its recasting of another and far older controversy, namely that concerning the relations between form (generally construed as active) and matter (construed as passive). That ancient discussion could now be (and was) reconceptualized in terms of genes as the agents of ‘‘action’’ (later, as sources ‘‘information’’), and of a cellular or extracellular environment that is simultaneously acted on and informed serving as passive material substrate for the development (or unfolding) of the organism (for further discussion, see Keller 1995; Griesemer 2005). But where many of the discussions of heredity and environment among geneticists focused on their relative force in individual development, elsewhere such debates more commonly focused on their relative force in shaping the course of evolution. Here, the neo-Darwinian synthesis was of particular importance. In identifying genetic continuity and change as the sole fundament of evolution, it contributed powerfully to the polarization of debates over the relative force of genes and environment in such highly charged arenas as eugenics and the ‘‘heritability’’ of intelligence and other behavioral attributes; it also helped pave the way for the mid-twentieth-century recasting of the nature-nurture debate in one of its crudest forms: as a battle between advocates of ‘‘Darwinian’’ and ‘‘Lamarckian’’ evolution (see, e.g., Keller 1991; Jablonka and Lamb 1995). To the extent that such debates imply a logical disjunction (form or matter, nature or nurture, genes or environment), they are clearly counterproductive. But my particular argument here is that replacing an implied disjunction with an explicit conjunction (nature and nurture, genes and environment) does little to ameliorate the problem of the role of the body that resides beyond the gene yet beneath the skin. To be sure, the disjunctive framing absolutely denies informational function to material dynamics, whereas the conjunctive framing advocated by dst clearly does permit such a function.2 But both exhibit a discernible tendency to figure the organismic body qua environment (and qua nurture), and hence, as having no more specific formative agency than, say, gravity or temperature. This leaves what is distinctively informing about the organism’s ‘‘internal environment’’ and the role it might play in development and evolution still concealed from view.

292

Beyond the Gene but Beneath the Skin

Which Body, Which Skin? And Why Stop at the Skin? Should the organism’s body be singled out as having particular biological significance? And if so, which body, and which skin? Biology recognizes many bodies, corresponding to many skins: in higher organisms, there is the multicellular body contained within an outer integument; in all organisms, cellular bodies are contained by cell membranes; and in eukaryotic organisms, nuclear bodies are contained by nuclear membranes. To avoid some of this ambiguity, I choose to focus on that moment in the life cycle of higher (metazoan) organisms in which the outer integument is the cell membrane, and the organism’s body is the cellular body—that is, the moment when the body in question is the fertilized egg or zygote. But why stop at the skin? Certainly, no biological integument provides an absolute divide between interior and exterior, and the cell membrane of a fertilized egg is, of necessity, more porous than most. Furthermore, because it regulates so much of the traffic between inside and out, the cell membrane is itself an active agent in shaping the body it contains—and indeed, in determining the very meaning of interiority. These facts constitute a warning against conceptualizing the organism as an autonomous individual sealed off from an exterior world by a static or preexisting boundary. Yet even so, the cell membrane, dynamic and permeable though it may be, defines a boundary which evolution has not only crafted into a cornerstone of biological organization but has endowed with vital significance. And given the dire effect the physical erasure of this boundary would have on the survival of the organism or cell, it scarcely seems necessary to elaborate on the inappropriateness of its conceptual erasure. In other words, the immediate and most obvious reason for taking this boundary seriously is grounded in its manifest indispensability for viability. But this being said, we are still no closer to understanding why it is so important. By way of addressing this last question I would like to suggest that the primary function of the cell membrane (as of any other biological skin) is simply to hold things together—more specifically, it keeps in proximity the many large molecules and subcellular structures required for growth and development. Proximity is crucial because it enables a degree of interconnectivity and interactive parallelism that would otherwise not be possible, but that is required for what I take to be the fundamental feature of the kind of developmental system we find in a fertilized egg, namely, its robustness.

293

Evelyn Fox Keller

Prior to all its other remarkable properties—in fact, a precondition of these —is the capacity of a developmentally competent zygote to maintain its functional specificity in the face of all the vicissitudes it inevitably encounters. This paradigmatic body may not be autonomous, but as embryologists have always known, it is far more tolerant of changes in its external environment than in its internal milieu. Indeed, were it not for the robustness of early embryos—that is, their ability to tolerate being moved from one environment to another—embryonic manipulation would not be possible, and much of what we know of as experimental embryology would never have emerged. All of this may seem too obvious to need saying, but there are times when the obvious is what most needs saying. We have learned that no elision is innocent. Nor, for that matter, is any reminder of elision. Politics are everywhere. Just as there are important political dimensions to the history of debates over genes and environment, so too there are political dimensions to the elision of the body in genetics discourse, as there also are, inevitably, to my insistence here on the boundary of the skin.3 In fact, the title of this essay contains its own elision: it reminds us of another title (Barbara Duden’s The Woman Beneath the Skin) while at the same time suppressing the subject of that other title. There is, of course, a reason for my choice. This is not an essay in feminist theory; nor is it about women. The ‘‘woman’’ in my title is signified only by its absence—intended, by that absence, to evoke nothing more than a recognition of the trace of the woman beneath the skin that still lurks, if not in the body more generally, then surely in the reproductive body of the fertilized egg. Because in sexual reproduction the cytoplasm derives almost entirely from the unfertilized egg, it is no mere figure of speech to refer to it as the maternal contribution. Furthermore, the representation of that body as ‘‘genetic environment,’’ as nothing more than a source of nurture for the developing organism, is a bit too reminiscent of conventional maternal discourse for at least this author’s comfort. My title, in short, is deliberate in its allusiveness: I want to indicate the possibility that gender politics has been implicated in the historic elision of the body in question without at the same time reinscribing the woman in that or any other body. The primary aim of this essay is, finally, a biological one, by which I mean that it is to reclaim the possibility of finding biological significance and agency in that no-man’sland beyond the gene but beneath the skin. Contra Oyama (1992),4 I want to argue that taking the cell rather than the gene as a unit of development does make a difference: not only does it yield a significant conceptual gain in the attempt to understand development, it also permits better conformation to the facts of development as we know them.

294

Beyond the Gene but Beneath the Skin

Is There a Program for Development? And If So, Where Is It? Many authors have taken issue with the concept of a program for development, noting its teleological implications, its metaphoric reliance on computer science, its implication of a unidirectional flow of information (see, e.g., Stent 1985; Newman 1988; Oyama 1989; Moss 1992; de Chardarevian 1994). But my concern here is not with the concept of program per se: rather, it is with the more specific notion of a genetic program, especially in contradistinction to its companion notion, that of a developmental program. In other words, for my particular purposes here, I accept the metaphor of program, warts and all, and focus my critical attention instead on the implications of attaching to that metaphor the modifier genetic. And I ask two questions: First, what is the meaning of a ‘‘genetic program’’? Second, how did this concept come to be so widely accepted as an ‘‘explanation’’ of biological development? 5 Taken as a composite, the meaning of the term genetic program simultaneously depends on and underwrites the particular presumption that a ‘‘plan of procedure’’ for development is itself written in the sequence of nucleotide bases. Is this presumption correct? Certainly it is almost universally taken for granted; but I want to argue that, at best, it must be said to be misleading, and at worst, it is simply false. To the extent that we may speak at all of a developmental program, or of a set of instructions for development, in contradistinction to the data or resources for such a program, current research obliges us to acknowledge that these ‘‘instructions’’ are not written into the dna itself (or at least are not all written in the dna), but rather are distributed throughout the fertilized egg. Furthermore, it is not only the material resources of the cell that can serve as data, but nucleotide sequences as well. Indeed, if the distinction between program and data is to have any meaning in biology, it has become abundantly clear that it does not align (as had earlier been assumed) either with a distinction between ‘‘genetic’’ and ‘‘epigenetic’’ or with the precursor distinction between nucleus and cytoplasm. To be sure, the informational content of the dna is essential—without it development (life itself ) cannot proceed. But for many developmental processes, it is far more appropriate to refer to this informational content as data than as program (Atlan and Koppel 1990). Indeed, I want to suggest that the notion of genetic program both depends on and sustains a fundamental category error in which two independent distinctions, one between genetic and epigenetic, and the

295

Evelyn Fox Keller

other between program and data, are pulled into mistaken alignment. The net effect of such alignment is to reinforce two outmoded associations: on the one hand, between genetic and active, and, on the other, between epigenetic and passive. Development results from the temporally and spatially specific activation of particular genes, which in turn depends on a vastly complex network of interacting components including not only the ‘‘hereditary codescript’’ of the dna, but also a densely interconnected cellular machinery made up of proteins and rna molecules. Necessarily, each of these systems functions in relation to the others alternatively as data and as program. If development cannot proceed without the ‘‘blueprint’’ of genetic memory, neither can it proceed without the ‘‘machinery’’ embodied in cellular structures. To be sure, the elements of these structures are fixed by genetic memory, but their assembly is dictated by cellular memory.6 Furthermore, one must remember that more than genes are passed from parent to offspring. To forget this is to be guilty of what Richard Lewontin calls an ‘‘error of vulgar biology.’’ As he reminds us, ‘‘An egg, before fertilization, contains a complete apparatus of production deposited there in the course of its cellular development. We inherit not only genes made of dna but an intricate structure of cellular machinery made up of proteins’’ (Lewontin 1992: 33). Assuming one is not misled by Lewontin’s colloquial use of the term inherit to refer to transmission over a single generation (as distinct from multigenerational transmission), none of this is either controversial or news, nor does it depend on the extraordinary techniques now available for molecular analysis. Yet, however surprising this may be, it is only within the last decade or two that the developmental and evolutionary implications of socalled maternal effects have begun to be appreciated.7 Current research now provides us with an understanding of the mechanisms involved in the processing of genetic data that make the errors of what Lewontin calls ‘‘vulgar biology’’ manifest. Yet, even when elaborated by the kind of detail we now have available, such facts are still not sufficient to dislodge the confidence that many distinguished biologists continue to have in both the meaning and the explanatory force of the genetic program. Why not? What grants the ‘‘genetic program’’ its apparent explanatory force even in the face of such obvious caveats as those above? To look for answers I will turn to history, more specifically, to the history of the term itself.

296

Beyond the Gene but Beneath the Skin

‘‘Programs’’ in the Biological Literature of the 1960s The ‘‘program’’ metaphor, borrowed directly from computer science, entered the biological literature in the 1960s not once but several times, and in at least two distinctly different registers. In its first introduction, simultaneously by Mayr (1961) and by Jacob and Monod (1961), the locus of the program was explicitly identified as the genome, but over the course of that decade another notion of ‘‘program,’’ a ‘‘developmental program,’’ also surfaced, and repeatedly so. This program was not located in the genome, but instead was distributed throughout the fertilized egg (see, e.g., Apter 1966). By the 1970s, however, the ‘‘program’’ for development had effectively collapsed into a ‘‘genetic program,’’ with the alternative, distributed, sense of a ‘‘developmental program’’ all but forgotten. François Jacob, one of the earliest to use the genetic program concept, contributed crucially to its popularization. In The Logic of Life, first published in 1970 (in English 1973, 1976), Jacob describes the organism as ‘‘the realization of a programme prescribed by its heredity,’’ claiming that ‘‘when heredity is described as a coded programme in a sequence of chemical radicals, the paradox [of development] disappears’’ (Jacob 1976: 2, 5). Jacob views the genetic program, written in the alphabet of nucleotides, as the agent responsible for the apparent purposiveness of biological development; it and it alone gives rise to ‘‘the order of biological order’’ (p. 8). He refers to the oft-quoted characterization of teleology as a ‘‘mistress’’ whom biologists ‘‘could not do without, but did not dare to be seen with in public,’’ and writes, ‘‘The concept of programme has made an honest woman of teleology’’ (pp. 8–9). Although Jacob does not exactly define the term, he notes that ‘‘[t]he programme is a model borrowed from electronic computers. It equates the genetic material of an egg with the magnetic tape of a computer’’ (p. 9). Equating the genetic material of an egg with the magnetic tape of a computer does not, however, imply that that material encodes a ‘‘program.’’ It might just as well be thought of as encoding ‘‘data’’ to be processed by a cellular ‘‘program,’’ or by a program residing in the machinery of transcription and translation complexes, or by extranucleic chromatin structures in the nucleus. Computers have provided a rich source of metaphors for molecular biology, but they cannot by themselves be held responsible for the notion of genetic program. Indeed, as already indicated, other, quite different, uses of the program metaphor for biological development were already in use. One

297

Evelyn Fox Keller

such use was in the notion of a developmental program—a term that surfaced repeatedly through the 1960s and stood in notable contrast to that of a genetic program. Let me give an example of this alternative use. In 1965, Michael Apter, a young graduate student steeped in information theory and cybernetics, teamed up with the developmental biologist Lewis Wolpert to argue for a direct analogy not between computer programs and the genome, but between computer programs and the egg: ‘‘If the genes are analogous with the subroutine, by specifying how particular proteins are to be made . . . , then the cytoplasm might be analogous to the main programme specifying the nature and sequence of operations, combined with the numbers specifying the particular form in which these events are to manifest themselves. . . . In this kind of system, instructions do not exist at particular localized sites, but the system acts as a dynamic whole’’ (Apter and Wolpert 1965: 257). Indeed, throughout the 1960s a number of developmental biologists attempted to employ ideas from cybernetics to illuminate development, and almost all shared Apter’s starting assumptions (for examples, see Keller 1995: ch. 3)—that is, they located the program (or ‘‘instructions’’) for development in the cell as a whole. The difference in where the program is said to be located is crucial, for it bears precisely on the controversy that had been raging among biologists since the beginning of the century over the adequacy of genes to account for development. By the beginning of the 1960s this debate had subsided, largely as a result of the eclipse of embryology as a discipline during the 1940s and 1950s. Genetics had triumphed, and after the identification of dna as the genetic material, the successes of molecular biology had vastly consolidated that triumph. Yet the problems of development, still unresolved, lay dormant. Molecular biology had revealed a stunningly simple mechanism for the transmission and translation of genetic information, but, at least until 1960, had been able to offer no account of developmental regulation. James Bonner, a professor of biology at California Institute of Technology, in an early attempt to bring molecular biology to bear on development, puts the problem well. Granting that ‘‘the picture of life given to us by molecular biology . . . applies to cells of all creatures,’’ he writes, this picture ‘‘. . . is a description of the manner in which all cells are similar. But higher creatures, such as people and pea plants, possess different kinds of cell. The time has come for us to find out what molecular biology can tell us about why different cells in the same body are different from one another, and how such differences arise’’ (Bonner 1965: v).

298

Beyond the Gene but Beneath the Skin

Bonner’s own work was on the biochemistry and physiology of regulation in plants in an institution well known for its importance in the birth of molecular biology (see, e.g., Kay 1992). In the work under discussion, published in 1965, like Apter and a number of others writing in the mid-1960s Bonner employs the conceptual apparatus of automata theory to deal with the problem of developmental regulation. But unlike them, he does not locate the ‘‘program’’ in the cell as a whole, but rather in the chromosomes, and more specifically in the genome. Indeed, his paper begins with the by then standard credo of molecular biology, asserting, ‘‘We know that . . . the directions for all cell life [are] written in the dna of their chromosomes’’ (Bonner 1965: v). Why? An obvious answer is suggested by his location. Unlike Apter and unlike other developmental biologists of the time, Bonner was situated at a major thoroughfare for molecular biologists, and it is hard to imagine that he was uninfluenced by the enthusiasm of his colleagues at Cal Tech. In any case, Bonner’s struggle to reconcile the conceptual demands posed by the problems of developmental regulation with the received wisdom among molecular biologists is at the very least instructive, especially given its location in time, and I suggest it is worth examining in some detail for the insight it has to offer on our question of how the presumption of a genetic program came—in fact, over the course of that very decade—to seem self-evident. In short, I want to take Bonner as representative of a generation of careful thinkers about an extremely difficult problem who opted for this (in retrospect, inadequate) conceptual shortcut.

Explanatory Logic of the Genetic Program From molecular biology Bonner inherited a language encoding a number of critical if tacit presuppositions. That language shaped his efforts in decisive ways. Summarizing the then current understanding of transcription and translation, he writes: ‘‘Enzyme synthesis is therefore an informationrequiring task and . . . the essential information-containing component is the long punched tape which contains, in coded form, the instructions concerning which amino acid molecule to put next to which in order to produce a particular enzyme’’ (Bonner 1965: 3). At the same time, he clearly recognized that only the composition of the protein had thus been accounted for, and not the regulation of its production required for the formation of specialized cells, that is, cell differentiation remained unexplained. ‘‘Each kind of specialized cell of the higher organ-

299

Evelyn Fox Keller

ism contains its characteristic enzymes,’’ he writes, ‘‘but each produces only a portion of all the enzymes for which its genomal dna contains information’’ (p. 6). But he continues: ‘‘Clearly then, the nucleus contains some further mechanism which determines in which cells and at which times during development each gene is to be active and produce its characteristic messenger rna, and in which cells each gene is to be inactive, to be repressed’’ (p. 6). Two important moves have been made here. Bonner has argued that something other than the information for protein synthesis encoded in the dna is required to explain cell differentiation (and this is his main point), but on the way to making this point he places this ‘‘further mechanism’’ in the nucleus, with nothing more by way of argument or evidence than his ‘‘Clearly then.’’ Why does such an inference follow? And why does it follow ‘‘clearly’’? Perhaps the next paragraph will help: ‘‘The egg is activated by fertilization. . . . As division proceeds cells begin to differ from one another and to acquire the characteristics of specialized cells of the adult creature. There is then within the nucleus some kind of programme which determines the property [sic] sequenced repression and derepression of genes and which brings about orderly development’’ (p. 6). Here, the required ‘‘further mechanism’’ is explicitly called a program, and once again it is located in the nucleus. But this time around, a clue to the reasoning behind the inference appears in the first sentence, ‘‘The egg is activated by fertilization.’’ This is how I believe the (largely tacit) reasoning goes: If the egg is ‘‘activated by fertilization,’’ the implication is that it is entirely inactive prior to fertilization. What does fertilization provide? The entrance of the sperm, of course; and unlike the egg, the sperm has almost no cytoplasm: it can be thought of as pure nucleus. Ergo, the active component must reside in the nucleus and not in the cytoplasm. Today, the supposition of an inactive cytoplasm would be challenged, but in Bonner’s time it would have been taken for granted as a carryover from what I have called ‘‘the discourse of gene action’’ of classical genetics (Keller 1995). And even then it might have been challenged had it been made explicit, but as an implicit assumption encoded in the language of ‘‘activation,’’ it was likely to go unnoticed both by Bonner’s readers and by Bonner himself. Bonner then goes on to ask the obvious questions: ‘‘What is the mechanism of gene repression and derepression which makes possible development? Of what does the programme consist and where does it live?’’ (Bonner 1965: 6). And he answers them as best he can: ‘‘We can say that the programme which sequences gene activity must itself be a part of the genetic information since

300

Beyond the Gene but Beneath the Skin

the course of development and the final form are heritable. Further than this we cannot go by classical approaches to differentiation’’ (p. 6). In these few sentences Bonner has completed the line of argument leading him to the conclusion that the program must be part of the genetic information, that is, to the ‘‘genetic program.’’ Once again we can try to unpack his reasoning. Why does the heritability of the course of development and the final form imply that the program must be part of the genetic information? Because—and only because—of the unspoken assumption that it is only the genetic material that is inherited. The obvious fact that the reproductive process passes on (or transmits) not only the genes but also the cytoplasm (the latter through the egg for sexually reproducing organisms) is not mentioned. But even if it were, this fact would almost certainly be regarded as irrelevant, simply because of the prior assumption that the cytoplasm contains no active components. The conviction that the cytoplasm could neither carry nor transmit effective traces of intergenerational memory had been a mainstay of genetics for so long that it had become part of the ‘‘memory’’ of that discipline, working silently but effectively to shape the very logic of inference employed by geneticists. Yet another ellipsis becomes evident (now, even to Bonner himself ) as Bonner attempts to integrate his own work on the role of histones in genetic regulation. Not all copies of a gene (or a genome) are in fact the same: Because of the presence of proteins in the nucleus capable of binding to the dna, ‘‘in the higher creature, if it is to be a proper higher creature, one and the same gene must possess different attributes, different attitudes, in different cells’’ (1965: 102). The difference is a function of the histones. How to reconcile this fact with the notion of a ‘‘genetic program’’? There is one simple way, and Bonner takes it—namely, to elide the distinction between genome and chromosome. The ‘‘genetic program’’ is saved (for this discussion) by just a slight shift in reference: now it refers to a program built into the chromosomal structure— that is, into the complex of genes and histones, where that complex is itself here referred to as the ‘‘genome.’’ But the most conspicuous inadequacy of the location of the developmental program in the genetic information becomes evident in the final chapter, in which Bonner attempts to sketch out an actual computer program for development. Here, the author tries to reframe what is known about the induction of developmental pathways in terms of a ‘‘master program,’’ proposing to ‘‘consider the concept of the life cycle as made up of a master programme constituted in turn of a set of subprogrammes or subroutines’’ (p. 134). Each

301

Evelyn Fox Keller

subroutine specifies a specific task to be performed—for a plant, his list includes cell life, embryonic development, how to be a seed, bud development, leaf development, stem development, root development, and reproductive development. Within each of these subroutines is a list of cellular instructions or commands, such as, ‘‘divide tangentially with growth,’’ ‘‘divide transversely with growth,’’ ‘‘grow without dividing,’’ ‘‘test for size or cell number,’’ and so on (p. 137). He then asks: ‘‘[H]ow might these subroutines be related to one another? Exactly how are they to be wired together to constitute a whole programme?’’ (p. 135). Conveniently, this question is never answered. If it had been, the answer would have necessarily undermined Bonner’s core assumption. To see this, two points emerging from his discussion need to be underscored: First, the list of subroutines, although laid out in a linear sequence (as if following from an initial ‘‘master program’’), actually constitute a circle, as indeed they must if they are to describe a life cycle. Bonner’s own ‘‘master program’’ is in fact nothing but this composite set of programs, wired together in a structure exhibiting the characteristic cybernetic logic of ‘‘circular causality.’’ The second point bears on Bonner’s earlier question, ‘‘Of what does the programme consist and where does it live?’’ The first physical structures that were built to embody the logic of computer programs were built out of electrical networks (hence the term ‘‘switching networks’’), and this was Bonner’s frame of reference. ‘‘That the logic of development is based upon [a developmental switching] network,’’ he maintains, ‘‘there can be no doubt’’ (p. 148). But what would serve as the biological analogue of an electric (or electronic) switching network? How are the instructions specified in the subroutines that constitute the life cycle actually embodied? Given the dependence of development on the regulating activation of particular genes, Bonner reasonably enough calls the developmental switching network a ‘‘genetic switching network.’’ But this does not quite answer our question; instead it obfuscates it. The clear implication is that such a network is constituted of nothing but genes, whereas in fact, many other kinds of entities also figure in this network, all playing critical roles in the control of genetic activity. Bonner himself writes of the roles played by histones, hormones, and rna molecules; today, the list has expanded considerably to include enzymatic networks, metabolic networks, transcription complexes, signal transduction pathways, and so on, with many of these additional factors embodying their own ‘‘switches.’’ We could, of course, still refer to this extraordinarily complex set of interacting controlling factors as a ‘‘genetic switching network’’—insofar, that is, as the regulation of gene activation remains central to development—but only if we

302

Beyond the Gene but Beneath the Skin

avoid the implication (an implication tantamount to a category error) that that network is embodied in and by the genes themselves. Indeed, it is this ‘‘category error’’ which confounds the very notion of a ‘‘genetic program.’’ If we were now to ask Bonner’s question, ‘‘Of what does the programme consist and where does it live?’’ we would have to answer, just as Apter did long ago, that it consists not of particular gene entities, and lives not in the genome itself, but is of and in the cellular machinery integrated into a dynamic whole. As Garcia-Bellida writes, ‘‘Development results from local effects, and there is no brain or mysterious entity governing the whole: there are local computations and they explain the specificity of something that is historically defined’’ (1998: 113). Thus, if we wish to preserve the computer metaphor, it would seem more reasonable to describe the fertilized egg as a massively parallel processor in which ‘‘programs’’ (or networks) are distributed throughout the cell.8 The roles of ‘‘data’’ and ‘‘program’’ here are relative, for what counts as ‘‘data’’ for one ‘‘program’’ is often the output of a second ‘‘program,’’ and the output of the first is ‘‘data’’ for yet another ‘‘program,’’ or even for the very ‘‘program’’ that provided its own initial ‘‘data.’’ Thus, for some developmental stages, the dna might be seen as encoding ‘‘programs’’ or switches which process the data provided by gradients of transcription activators; or, alternatively, one might say that dna sequences provide data for the machinery of transcription activation (some of which is acquired directly from the cytoplasm of the egg). In later developmental stages, the products of transcription serve as data for splicing machines, translation machines, and so on. In turn, the output of these processes makes up the very machinery or programs needed to process the data in the first place. Sometimes, this exchange of data and programs can be represented sequentially, and sometimes as occurring simultaneously.

Into the Present In the mid-1960s, when Bonner, Apter, and others were attempting to represent development in the language of computer programs, automata theory was in its infancy and cybernetics was at the height of its popularity. During the 1970s and 1980s these efforts lay forgotten: cybernetics had lost its appeal to computer scientists and biologists alike, and molecular biologists found they had no need of such models. The mere notion of a genetic program sufficed by itself to guide their research. Today, however, provoked in large part by the construction of hard-wired parallel processors, the project to simulate

303

Evelyn Fox Keller

biological development on the computer has returned in full force, and in some places has become a flourishing industry. It goes by various names— Artificial Life, adaptive complexity, and genetic algorithms among them. But what is a genetic algorithm? Like Bonner’s subroutines, it is ‘‘a sequence of computational operations needed to solve a problem’’ (see, e.g., Emmeche 1994). And once again we need to ask, why ‘‘genetic’’? Furthermore, not only are the individual algorithms referred to as ‘‘genetic,’’ but ‘‘in the fields of genetic algorithms and artificial evolution, the [full] representation scheme is often called a ‘genome’ or ‘genotype’’’ (Fleischer 1995: 1). In Complexity, an account of the science written for the lay reader, Mitchell Waldrop quotes Chris Langton, the founder of Artificial Life, as saying: ‘‘[Y]ou can think of the genotype as a collection of little computer programs executing in parallel, one program per gene. When activated, each of these programs enters into the logical fray by competing and cooperating with all the other active programs. And collectively, these interacting programs carry out an overall computation that is the phenotype: the structure that unfolds during an organism’s development’’ (Waldrop 1992: 194). Like their counterparts in molecular genetics, workers in Artificial Life are not confused. They well understand, and when pressed readily acknowledge, that the biological analogues of these computer programs are not in fact ‘‘genes’’ (at least as the term is used in biology), but complex biochemical structures or networks comprising proteins, rna molecules, and metabolites which often, although certainly not always, execute their tasks in interaction with particular stretches of dna.9 Artificial Life’s ‘‘genome’’ typically consists of instructions such as ‘‘reproduce,’’ ‘‘edit,’’ ‘‘transport,’’ or ‘‘metabolize,’’ and the biological instantiation of these algorithms is found not in the nucleotide sequences of dna but in specific kinds of cellular machinery such as transcription complexes, spliceosomes, and metabolic networks. Why, then, are they called ‘‘genetic,’’ and why is the full representation called a ‘‘genome’’? Is it not simply because it so readily follows from the usage the term genetic program had already acquired in genetics? Words have a history, and their usage and meaning depend on this history. History does not fix the meaning of words; rather, it builds into them a kind of memory. In the field of genetic programming, ‘‘genes’’ have come to be considered not as particular sequences of dna, but as the computer programs required to execute particular tasks (as Langton puts it, ‘‘one program per gene’’); yet, at the same time, the history of the term ensures that the word gene, even as adapted by computer scientists, continues to carry its original meaning. And perhaps most important, that earlier meaning remains avail-

304

Beyond the Gene but Beneath the Skin

able for deployment whenever it is convenient to use it. Much the same can be said for the use of the terms gene and genetic programs by geneticists.

A Recapitulation I have taken some time in examining Bonner’s argument for genetic programs, not because his book played a major role in establishing the centrality of this notion in biological discourse, but rather because of the critical moment in time at which it was written and because of the relative accessibility of the kind of slippage on which his argument depends. The very first use of the term program that I have been able to find in the molecular biology literature had appeared only four years earlier.10 In 1961, Jacob and Monod published a review of their immensely influential work on the operon model, a genetic mechanism for enzymatic adaptation in Escherichia coli. The term program is introduced in their concluding sentence: ‘‘The discovery of regulator and operator genes, and of repressive regulation of the activity of structural genes, reveals that the genome contains not only a series of blue-prints, but a coordinated program of protein synthesis and the means of controlling its execution’’ (Jacob and Monod 1961: 354). In a paper written three decades later, Sydney Brenner refers to the belief ‘‘that all development could be reduced to [the operon] paradigm,’’ that ‘‘it was simply a matter of turning on the right genes in the right places at the right times,’’ in rather scathing terms. As he puts it, ‘‘Of course, while absolutely true this is also absolutely vacuous. The paradigm does not tell us how to make a mouse but only how to make a switch’’ (Brenner et al. 1990: 485).11 And even in the first flush of enthusiasm, not everyone was persuaded of the adequacy of this particular regulatory mechanism to explain development.12 Lewis Wolpert was one of the early skeptics. In the late 1960s he seemed certain that an understanding of development required a focus not simply on genetic information, but also on cellular mechanisms.13 But by the mid-1970s even Wolpert had been converted to the notion of a ‘‘genetic program’’ (see, e.g., Wolpert and Lewis 1975). What carried the day? Certainly it was not more information about actual developmental processes. Far more than most histories of scientific terms, the history of genetic program bears the conspicuous marks of a history of discourse and power. Initially founded on a simple category error, in which the role of genes as subjects (or agents) of development was unwittingly conflated with their role as objects of developmental dynamics, the remarkable popularity of this term in molecular genetics over the last three decades cries

305

Evelyn Fox Keller

out for an accounting. Certainly it provided a convenient gloss, an easy way to talk that rarely if ever trips up scientists in their daily laboratory work. But it does trip them up in their efforts to explain development; indeed, the term has proven remarkably effective in obscuring enduring gaps in our understanding of developmental logic. Arguably, it has also contributed to the endurance of such gaps. So why did it prevail? If its popularity cannot be accounted for in strictly scientific or cognitive terms, we must look elsewhere. I suggest we look to the consonance of this formulation with the prior history of genetic discourse, particularly with the discourse of ‘‘gene action’’ that earlier prevailed. Fortifying the ‘‘genetic program’’ in the postwar era, with its easy and continuing elision of the cytoplasmic body, were an entirely new set of resources. Primary among these were the new science of computers, the imprimatur of Schroedinger, and the phenomenal success of the new molecular biology. Jacob’s Logic of Life was of key importance in the popularization of the concept of ‘‘genetic program.’’ Invoking the approval of both Schroedinger and Wiener, Jacob endowed the transition from past to future metaphors with the stamp of authority.14 Elsewhere (Keller 2000), I have argued that the notion of a genetic program has now begun to give way to that of a developmental program, particularly in the literature on cloning. And clearly, I consider this a step forward. The use of the term genetic to describe developmental instructions (or programs) encourages the belief even in the most careful readers (as well as writers) that it is only the dna that matters; it helps all of us to lose sight of the fact that, if that term is to have any applicability at all, it is primarily to refer to the entities on which instructions directly or indirectly act and not of which these instructions are constituted. The necessary dependency of genes on their cellular context, not simply as nutrient but as embodying causal agency, is all too easily forgotten. It is forgotten in laboratory practice, in medical counseling, and perhaps above all, in popular culture. For some authors (e.g., Oyama, this volume), however, the shift from genetic to developmental represents only marginal progress. For them, the word program is at least as problematic as genetic; indeed, because of the widespread assumption that only genes can encode programs, the use of one implies the other. To be sure, the concept of program has changed considerably since the 1960s, but it has not lost its facile assimilation with information, or, more generally, its disembodied aura. They are right. Perhaps we should employ a different word. Alternatively, we can work to change the meaning, and

306

Beyond the Gene but Beneath the Skin

the usage, of the original term. After all, the word program has been around a lot longer than have computers. But there is also another, and possibly more serious, problem, and that has to do with boundaries. Why privilege the cell as a unit of development? Or even the organism itself ? As Oyama notes, ‘‘it is not a simple thing to draw an outline around an individual organism’’ (this volume). Boundaries are both porous and fluid. Nor is it possible to ‘‘absolutely’’ demarcate inside from outside. Boundaries are relative—that is, they are porous and fluid to some degree—and so are distinctions between inside and outside. Everything depends on degree, and on scale. The development of biological organisms depends on many different kinds of boundaries, none of which is closed, and all of which permit (indeed, help to structure) different levels of communication. Cells, for example, need to communicate with each other through intercellular signaling. But because large molecules cannot get through cell membranes, such intercellular signals must be mediated by smaller molecules, synapses, and so on. Similarly, prenatal communication between mother and fetus is mediated both by intercellular signaling and by the flow of materials through the placenta. Given that the placenta does not survive birth, and that intercellular signaling of the kind that works internally (i.e., with respect to the outer skin of the organism) is not effective between organisms, postnatal communication must be mediated in other ways—for example, by metabolic and respiratory exchange, by sensory perception, by perceptionmotor coordination, and eventually by language. None of this, however, leads me to disavow concepts of ‘‘inside’’ and ‘‘outside,’’ ‘‘internal’’ and ‘‘external’’—even if it does serve to remind me how critically these categories depend on the focus of one’s concern. For embryonic development, we need to attend particularly to boundaries composed of biological membranes and skin: nuclear membranes, cellular membranes, and the skins of both fetus and mother—not because other boundaries (e.g., familiar, sociocultural, economic) do not matter, but because their relevance is less immediate, and of a lesser degree. Nature does not respect principles of parity. For embryogenesis, I privilege the boundary of the fertilized cell, at the stage in which that boundary converges with that of the growing organism. Whether I am right is not so much a matter of principle as it is of the magnitude of relative effects—perhaps one could say it is a matter of degree, and of scale.

307

Evelyn Fox Keller

Notes This essay is reprinted from S. Oyama, P. E. Griffith, and R. D. Gray (eds.), Cycles of Contingency: Developmental Systems and Evolution (2001); thanks to mit Press for permission. 1 Earlier, in the late nineteenth century, the term heredity had commonly been used far more inclusively, encompassing the study of both genetics and embryology (see, e.g., Sapp 1987). Furthermore, in the 1920s, the term genetics was largely understood to refer solely to transmission genetics. 2 Indeed, much of this literature in essence argues for symmetry between the role of genes and other developmental resources. Thus, for example, in arguing against the conventional view that genes code for traits, Griffiths and Gray suggest that ‘‘we can talk with equal legitimacy of cytoplasmic or landscape features coding for traits in standard genic backgrounds’’ (1994, 1998: 122). 3 See, for example, Kevles 1986 and Paul 1995, 1998 for discussions of the politics of eugenics debates, and Sapp 1987 for a discussion of the impact of Lysenko’s antigenetics crusade in the Soviet Union just before and during the cold war. See my discussion of the ‘‘discourse of gene action’’ in Keller 1995: ch. 1. 4 I refer in particular to Oyama’s discussion of Brenner’s abandonment of the concept of a ‘‘genetic program’’ and his emerging conviction that the proper ‘‘unit of development is the cell.’’ She writes, ‘‘Having given up genetic programs, [Brenner] now speaks of internal representations and descriptions. In doing so he is like many workers who have been faced with the contradictions and inadequacies of traditional notions of genetic forms and have tried to resolve them, not by seriously altering their concepts, but by making the forms in the genome more abstract: not noses in the genes, but instructions for noses, or potential for noses, or symbolic descriptions of them. This solves nothing’’ (Oyama 1992: 55). 5 The remainder of this paper is adapted from Keller 1999. 6 A vivid demonstration of this interdependency was provided in the 1950s and 1960s with the development of techniques for interspecific nuclear transplantation. Such hybrids almost always fail to develop past gastrulation, and in the rare cases when they do, the resultant embryo exhibits characteristics intermediate between the two parental species. This dependency of genomic function on cytoplasmic structure follows as well from the asymmetric outcomes of reciprocal crosses demonstrated in earlier studies of interspecific hybrids (Markert and Ursprung 1971: 135–137). 7 ‘‘Maternal (or cytoplasmic) effects’’ refers only to the effective agency of maternal (or cytoplasmic) contributions (such as, e.g., gradients). Because such effects need not be (and usually are not) associated with the existence of permanent structures that are transmitted through the generations, they should not be confused with ‘‘maternal inheritance.’’

308

Beyond the Gene but Beneath the Skin 8 Supplementing Lenny Moss’s observation that a genetic program is ‘‘an object nowhere to be found’’ (Moss 1992: 335), I would propose the developmental program as an entity that is everywhere to be found. 9 Executing a task means processing data provided both by the dna and by the products of other programs; that is, by information given in nucleotide sequences, chromosomal structure, gradients of proteins and rna molecules, the structure of protein complexes, etc. 10 Simultaneously, and probably independently, Ernst Mayr introduced the notion of ‘‘program’’ in his 1961 article ‘‘Cause and effect in biology,’’ adapted from a lecture given at mit on 1 February 1961. There he wrote, ‘‘The complete individualistic and yet also species-specific dna code of every zygote (fertilized egg cell), which controls the development of the central and peripheral nervous system . . . is the program for the behavior computer of this individual’’ (Mayr, 1961: 1504). 11 As Soraya de Chadarevian points out (1994), Brenner had criticized the use of the operon model for development as early as 1974 (see his comments in Brenner 1974). 12 Or even of the appropriateness of the nomenclature. Waddington, for example, noted not only that it ‘‘seems too early to decide whether all systems controlling gene-action systems have as their last link an influence which impinges on the gene itself,’’ but also redescribed this system as ‘‘genotropic’’ rather than ‘‘genetic’’ in order ‘‘to indicate the site of action of the substances they are interested in’’ (Waddington 1962: 23). 13 For example, Wolpert wrote in 1969: ‘‘Dealing as it does with intracellular regulatory phenomena, it is not directly relevant to problems where the cellular bases of the phenomena are far from clear’’ (Wolpert 1969: 2–3). 14 He writes: ‘‘According to Norbert Wiener, there is no obstacle to using a metaphor ‘in which the organism is seen as a message’’’ (Jacob 1976: 251–252). And two pages later, ‘‘According to Schrödinger, the chromosomes ‘contain in some kind of code-script the entire pattern of the individual’s future development and of its functioning in the mature state. . . . The chromosome structures are at the same time instrumental in bringing about the development they foreshadow. They are law-code and executive power or, to use another simile, they are architect’s plan and builder’s craft all in one’ ’’ (p. 254).

References Apter, M. J. 1966. Cybernetics and Development. Oxford: Pergamon Press. Apter, M. J., and Wolpert, L. 1965. Cybernetics and development. J. Theor. Biol. 8: 244–257.

309

Evelyn Fox Keller

Atlan, H., and Koppel, M. 1990. The cellular computer dna: program or data. Bull. Math. Biol. 52 (3): 335–348. Barrington, A., and Pearson, K. 1909. A First Study of the Inheritance of Vision and of the Relative Influence of Heredity and Environment on Sight. Cambridge: Cambridge University Press. Bonner, J. 1965. The Molecular Biology of Development. Oxford: Oxford University Press. Brenner, S. 1974. New directions in molecular biology. Nature 248: 785–787. Brenner, S., et al. 1990. Genes and development: molecular and logical themes. Genetics 126: 479–486. Chardarevian, S. de. 1994. Development, programs and computers: work on the worm (1963–1988). Paper presented at the Summer Academy, Berlin, Germany. Conklin, E. G. 1915. Heredity and Environment in the Development of Men. Princeton: Princeton University Press. Duden, B. 1991. The Woman Beneath the Skin. Cambridge: Harvard University Press. Emmeche, C. 1994. The Garden in the Machine. Princeton: Princeton University Press. Fisher, R. A. 1918. The correlation between relatives on the supposition of Mendelian inheritance. Trans. Roy. Soc. (Edinburgh) 52: 399–433. Fleischer, K. 1995. A multiple-mechanism developmental model for defining selforganizing geometric structures. Ph.D. diss., California Institute of Technology, Pasadena. Galton, F. 1970. English Men of Science: Their Nature and Nurture. 1874. Reprint. London: Cass. Garcia-Bellida, A. 1998. Discussion. In: The Limits of Reductionism. Novartis Foundation Symposium 213. Chichester: Wiley. Griesemer, J. R. 2005. The informational gene and the substantial body: on the generalization of evolutionary theory by abstraction. In press, Biology and Philosophy. Griffiths, P., and Gray, R. (1994) 1998. Developmental systems and evolutionary explanation. Reprinted in: D. Hull and M. Ruse (eds.), The Philosophy of Biology (pp. 117–145). Oxford: Oxford University Press. Griffiths, P. E., and Neumann-Held, E. M. 1999. The many faces of the gene. BioScience 49: 656–662. Hull, D., and Ruse, M. 1998. The Philosophy of Biology. Oxford: Oxford University Press. Jablonka, E., and Lamb, M. 1995. Epigenetic Inheritance and Evolution. New York: Oxford University Press. Jacob, F. 1976. The Logic of Life. New York: Vanguard. Originally published in French by Editions Gallimard, 1970.

310

Beyond the Gene but Beneath the Skin

Jacob, F., and Monod, J. 1961. Genetic regulatory mechanisms in the synthesis of proteins. J. Mol. Biol. 3: 318–356. Keller, E. F. 1991. Between language and science: the question of directed mutation in molecular genetics. Perspect. Biol. Med. (Winter): 292–307. Keller, E. F. 1995. Refiguring Life. New York: Columbia University Press. Keller, E. F. 1999. Decoding the genetic program. In: P. Beurton and R. Falk (eds.), Genes. Cambridge: Cambridge University Press. Keller, E. F. 2000. The Century of the Gene. Cambridge: Harvard University Press. Kevles, D. J. 1986. In the Name of Eugenics: Genetics and the Uses of Human Heredity. Berkeley: University of California Press. Lewontin, R. 1992. The dream of the human genome. N.Y. Rev. Books, 28 May: 31–40. Markert, C. L., and Ursprung, H. 1971. Developmental Genetics. Englewood Cliffs, N.J.: Prentice-Hall. Mayr, E. 1961. Cause and effect in biology. Science 134: 1501–1506. Morgan, T. H. 1911. The influence of heredity and of environment in determining the coat colors in mice. Ann. N.Y. Acad. Sci. 21: 87–118. Moss, L. 1992. A kernel of truth? On the reality of the genetic program. Philos. Sci. Assoc. 1: 335–348. Newman, S. A. 1988. Idealist biology. Perspect. Biol. Med. 31(3): 353–368. Oyama, S. 1989. Ontogeny and the central dogma: do we need the concept of genetic programming in order to have an evolutionary perspective? In: M. R. Gunnar and E. Thelen (eds.), Systems and Development (pp. 1–34). Hillsdale, N.J.: Lawrence Erlbaum. Oyama, S. 1992. Transmission and construction: levels and the problem of heredity. In: G. Greenberg and E. Tobach (eds.), Levels of Social Behavior: Evolutionary and Genetic Aspects (pp. 51–60). Wichita, Kans.: T. C. Schneirla Research Fund. Reprinted in Oyama 2000. Oyama, S. 2000. Evolution’s Eye: A Systems View of the Biology-Culture Divide. Durham: Duke University Press. Oyama, S., Griffiths, P. E., and Gray, R. D. (eds.). 2001. Cycles of Contingency: Developmental Systems and Evolution. Cambridge: mit Press. Paul, D. 1995. Controlling Human Heredity. Atlantic Highlands, N.J.: Humanities Press International. Paul, D. 1998. The Politics of Heredity. New York: New York University Press. Sapp, J. 1987. Beyond the Gene. Oxford: Oxford University Press. Stent, G. S. 1985. Hermeneutics and the analysis of complex biological systems. In: D. J. Depew and B. Weber (eds.), Evolution at a Crossroads (pp. 209–225). Cambridge: mit Press. Strohmann, R. C. 1997. The coming Kuhnian revolution in biology. Nat. Biotechnol. 15: 194–200.

311

Evelyn Fox Keller

Waddington, C. H. 1962. New Patterns in Genetics and Development. New York: Columbia University Press. Waldrop, J. M. 1992. Complexity. New York: Simon and Schuster. Wolpert, L. 1969. Positional information and the spatial pattern of cellular differentiation. J. Theor. Biol. 25 (1): 1–48. Wolpert, L., and Lewis, J. H. 1975. Towards a theory of development. Fed. Proc. 34 (1): 14–20. Wright, S. 1920. The relative importance of heredity and environment in determining the piebald pattern of guinea-pigs. Proc. Natl. Acad. Sci. USA 6: 320– 332.

312

12 POIESIS AND PRAXIS Two Modes of Understanding Development

christoph rehmann-sutter Aristotle’s well-known distinction between poiesis and praxis as two distinct kinds of activities focuses on the relation between process and goal. A praxis is an activity whose process is itself its goal, whereas in poiesis the process is instrumental in reaching a different goal, external to the process itself. Aristotle thought that this distinction also applied to processes in living organisms. In this essay I will explore what it means to understand development as a kind of organic praxis. What, if anything, does that change in our descriptions of life processes in development; in our understanding of what biology tells us today about the molecular, genetic, cellular, organismic, and social (and perhaps other) processes that make development happen and that, in the end, make us the beings we are? There is a set of issues to be sorted out. The most obvious concerns the relation between activities that we can plan, whose meanings we can understand, and biological processes which are not intentional and which we describe and analyze in mechanistic terms. Categories that make sense for sorting activities do not necessarily make sense in organic development. We cannot intentionally grow legs or arms or heads, however hard we try. It is not clear, therefore, whether categories such as poiesis or praxis are applicable to growth processes, let alone to sequences of cellular movements in nonhuman organisms. A second issue concerns the two sides that are to be brought into coherence. We need to clarify which parts of our activities the categories of poiesis and praxis emphasize, and on the other side, which elements constitute the basis of our ideas about development. I start by giving an outline of the sequence of the arguments. In the first paragraph (1), the precise meaning of the poiesis/praxis distinction is first discussed as it appears in Aristotle’s practical philosophy. (2) There is textual evidence that Aristotle used the same distinction in his philosophy of biology

Christoph Rehmann-Sutter

as well. He gave it a very prominent status as the base of his definition of the soul and used it in his explanation of what distinguishes living from nonliving entities. (3) I next describe the effect of applying the poiesis/praxis distinction to developmental processes. A ‘‘poietical’’ understanding can be distinguished from a ‘‘practical’’ one. The former holds developmental processes as instrumentally important for bringing about functional results or for constructing the organism, while the latter holds that developmental processes are parts of the being’s continuous presence in the world, steps of its organic existence. If we were to adopt a ‘‘practical’’ understanding of developmental processes we would also believe that each process, in addition to being functional or structural, should be considered meaningful in itself, just by virtue of being a lived process. (4) Why does this picture conflict with the standard interpretation of developmental biology? I can discuss only a few of the several reasons for this here. One is that biology is deeply rooted in mechanistic lines. The functional aspects of parts and processes are considered to be the point of the explanatory endeavor. This functionalist (or mechanist) paradigm was preferred over competing paradigms because of its seemingly low metaphysical content and pragmatic advantages for doing experimental science. (5) Another reason is the idea of a ‘‘genetic program’’ that has been vindicated by genetics. If we think that the dynamic structure and developmental capacities of an organism are essentially stored in the sequence of its dna, there is little room for speculations about any inherent sense of the developmental steps; they appear to be nothing but the executions of instructions. (6) There may be a basic insecurity about whether to apply hermeneutic terms like describing or interpreting to biology. The most convincing answer to this insecurity is the fact that we are living organisms ourselves: we do biology essentially from the perspective of participant observers. This causes some epistemological troubles but also gives us an advantage; at the same time it makes a critical hermeneutic reflection indispensable. Otherwise we would be in danger of being unaware of ‘‘what we make of ourselves’’ when adopting a particular descriptive regime. (7) The final issue to be tackled is the question of why the adoption of another descriptive regime would be preferable. What kind of choice is this? I will end by suggesting that it is a choice among the kinds of relationships we want to live in. Describing (and thereby necessarily interpreting) development means positioning ourselves in a morally relevant way with regard to the living entities we describe.

314

Poiesis and Praxis

Aristotle on Activities Among the processes in which our own lived bodies are involved, there is a category of processes that can be accompanied by intention, planning, criticism, and ethics: our own activities. Aristotle divides activities into two types and calls them poiesis and praxis. The distinction is explained in a twofold way, using the terms (1) end (telos), and (2) work (ergon). 1. Poiesis has its end outside itself, but praxis does not, because good praxis (eupraxia) is an activity that is itself the end of what is being done (Aristotle Nicomachean Ethics VI.1140b.6–7). According to this definition, poiesis is an activity that is undertaken for the sake of something other than the activity itself, while praxis is an activity that is undertaken for the sake of itself. Aristotle’s examples of poiesis are taken from the realm of craft. Building a house means doing a variety of things toward the end of establishing a house. The building procedure, as complex as it may be, is not an end in itself. As soon as a better, more efficient technique for achieving one of its steps is available, the procedure will be changed. Poietic steps are thus replaceable. Whether their replacement is recommendable depends, among other things, on their efficacy. An exception might be a builder of a house who identifies so much with the building procedure that he likes the process of building the house as much as the completed structure. Examples of praxis, in Aristotle’s account, are seeing, thinking, and living (Metaphysics IX.1050a.36f.). All of these can be undertaken for the sake of themselves. Of course, a praxis like seeing also has poietic aspects: producing a sharp image of things, for example, or reducing the light intensity by narrowing the pupil. There, the processes will be judged by their efficacy in achieving the end (successful vision). But the act of seeing, which is performed by executing all those functions, is done for its own sake. When I watch a person, seeing may be useful in order to coordinate my body movement with hers, but seeing her is also a way of being in a visual relationship with her, which is itself the end of looking at her. In the passage quoted from the Nicomachean Ethics, Aristotle connects the term praxis with the moral term good. Eupraxia, the good praxis, is conceptualized as an activity worth doing for the sake of itself. This phrase contains in a nutshell the whole great idea of Aristotle’s ethics: the virtuous human, for Aristotle, is not primarily the person who restricts her or his aims (by obeying moral rules), but the person who refines her or his aims to the highest level. The truly virtuous person is the truly happy person. Therefore, virtu-

315

Christoph Rehmann-Sutter

ous activities—just, sensible, gentle, generous, caring actions, and so on— are those which are truly worth doing for their own sake (cf. Urmson 1988). They are practices. The differentiation between poiesis and praxis on the basis of ends focuses on the teleological structure of human behavior. But in addition to the end there is another central term in the teleology of activities. The same division can be made by considering the work an activity represents. 2. Some activities involve work (ergon) that is different from the movement and use (chresis) of the body or its organs; others involve nothing besides this movement or use. In the latter, the activity is itself the achievement (Aristotle Metaphysics IX.050a.23–27). The connection between these two parallel distinctions, (1) and (2), can easily be seen: the work is the end of the activity. In seeing, to take the example given above, the real work lies in the act of seeing itself. There is no work besides sight that is brought into being by the act of seeing, whereas in, say, the craft of weaving, the real work lies in what is woven. According to this distinction between poiesis and praxis, there are two kinds of ends and equally two kinds of achievements. The first ones are the activities themselves, and the second are the works beside them. The ends of poietical activities are the separable works, although very often those activities themselves can also be regarded as works. A person may experience the act of weaving as more important than what she weaves. You cannot weave without weaving something, she might say.1 This means that the same activity can sometimes be regarded at once as both a poiesis and a praxis, depending on the point of view. Writing a chapter for a book (which I am presently doing) has a poietic aspect, in that my intention is to produce an understandable argument and also to learn from having done so. But the same procedure has practical aspects too, both in my enjoyment of the very process of formulating the words on paper, and also in that I will hold myself (and will be held by my readers) accountable for doing it mindfully. Thinking, for Aristotle, is a praxis whose real work occurs in the thinking person (Metaphysics IX.1050a.36). Breathing, to take another example, has a poietic aspect in that it transports oxygen to our lungs and carbon dioxide out of them. But it has also a practical aspect, in that it is a way of relating ourselves to the environment. Breathing in, I sense the fresh morning air and find myself at this point in the world, afresh in a new moment of my life.

316

Poiesis and Praxis

In the Mode of Entelecheia One of the central theoretical terms in Aristotle’s ontology of living beings is entelecheia. He says that psuche is the entelecheia of an organized body (De Anima 412a.9f. and 21f ). By psuche, as he explains in De Anima, he means the eidos of living beings, and eidos is (see Metaphysics VII) the principle that makes a thing, a being, what it is: to ti en einai. However, at least in the writings that have survived, he never clearly defines what he means by the neologism entelecheia. The term is composed from en, telos, and echein, but the etymology does not give an unequivocal explanation of its meaning, because it is not clear what telos refers to. Its use in philosophy after Aristotle is correspondingly very diverse (a brief overview is given in Zubiria 1999). The most significant use in our contemporary intellectual discourse in biology is undoubtedly the vitalist one. Hans Driesch believed that there is a peculiar life substance capable of controlling the development of organisms in a goal-directed fashion, and focused on entelechy as precisely this entity or ontological principle. Entelechy, in his view, makes final causes dominate efficient causes and brings the telos into the organism (Driesch 1909: 283ff.). Historically, vitalism has been an intellectual countermovement to physicalism—the claim that the whole world, including life, can in principle be fully explained by physical mechanisms. Vitalists claim that the telos of development (the adult form, or at least the sequence of morphological changes that takes place during the life of an individual organism) is already present at the start of the developmental processes in the zygote or the embryo. Mechanist thinking was inspired by a concern that scientific explanation in biology, an explanatory mode that has to work without unprovable ‘‘metaphysical’’ assumptions, might be in danger if vitalism were adopted. The rejection of the existence of ‘‘entelechies’’ has for some time been a ritual element of the strong Darwinian tradition. Today, this battle has been won; criticizing vitalism no longer appears necessary (Mahner and Bunge 2000: 135f.). Modern thinking is considered to have no place for entelechies (Zubiria 1999). I agree, insofar as this vitalist meaning of entelecheia can be established unequivocally. A careful exegesis of Aristotle’s metaphysical writings, however, does not support this interpretation. The vitalist use can easily be unmasked as an illegitimate appropriation of the Aristotelian term. Furthermore, the philosophical discipline of metaphysics is also profoundly misunderstood if identified tel quel with a dogmatic belief system. Metaphysics, as I wish to use the

317

Christoph Rehmann-Sutter

term, is the discipline of philosophical reflection about modes of being and the possibilities of understanding them, including cultural criticism of what being-in-the-world—for instance as living beings carrying dna genomes in their cells—is held to mean.2 Aristotle can be defended against the accusation that he shared such a variant of finalism (for a nuanced treatment, see Mayr 1991: ch. 3; and Moya 2000). The idea that final causes can be represented in living organisms in such a way as to act as efficient causes, and thus to be capable of steering development from within, has been rightly criticized as obscurely metaphysical. Entelecheia should be understood in other ways. The translation I prefer is parallel to praxis. Praxis has the structure of being an end in itself. Both terms, entelecheia and praxis, signify the inherence of an end. The key idea would be to take the definition of the teleological structure of praxis as a grid for translating the Aristotelian entelecheia. My suggestion, therefore, is the following: praxis means a coincidence with the goal in the realm of activities, whereas entelecheia means the same coincidence in the realm of the life of a living being. If an act of being consists of being an end in itself, it is an entelecheia. If an activity consists of being an end in itself, it is a praxis. I admit that this usage deviates from the standard interpretations of entelecheia. Three other suggestions have been prominent in the literature: (1) a state of being at the goal, (2) a state of carrying the goal of the developmental process already in itself, and (3) actuality.3 The first proposal, that entelecheia means full or complete reality or completion (reaching the goal), suggests a process which has come to a halt, to a state that is the end. This presents a difficulty for developmental biology, because here we almost never have processes coming to a stop. But if the full reality means the ongoing process itself, in all its parts and states, this proposal coincides with mine that the process is itself an end. The second proposal holds that entelecheia contains a teleological assertion about a goal-directed process. It is hardly conceivable, however, that Aristotle would have neglected those species, like trees, which do not have a discernible constant adult form but are in a process of constant growth and change. This interpretation also causes the logical problem that a goal in this sense is necessarily a future state, and it is difficult to explain how it could be active already in the present, leading the steps in the right direction. This thesis implies, somewhat implausibly, that Aristotle has not noticed his own distinction between four different types of causes and has conflated efficient and final causes. Today, some believe that the idea of a genetic program has

318

Poiesis and Praxis

solved this problem, containing the double idea that programs have evolved by the mechanisms of Darwinian natural selection and that those programs are now stored as the information content of dna (see Mayr 1991: 81–86). But, as the other chapters of this book establish, the belief in the existence of a developmental genetic program that is stored in the dna sequence has come under considerable criticism within developmental genetics and molecular biology itself. The third possibility is unattractive because it cannot reflect Aristotle’s obvious distinction between entelecheia and energeia. In Metaphysics (IX. 1050a.21ff.) he states clearly that energeia refers to, or is directed at, entelecheia —which excludes synonymity between both terms. It would also have a strangely trivializing effect on the psychology of De Anima: The soul would be nothing but ‘‘the actuality’’ of an organism. The explanation of entelecheia as parallel to praxis, as a dynamic mode of being whose process is itself the end, leads to substantial theoretical consequences in the reading of Aristotle’s metaphysics. Their discussion cannot be the point of this essay. Here, I want to defend a claim for an alternative understanding of development: if living processes could be understood in the mode of entelecheia, this would mean that the change in composition, dynamic structure, and complex interactive form that characterizes its existence is a process with the ontological quality of being itself an end that is performed. Living processes have in this sense an inherent teleological structure, but in a totally different way from what has hitherto been claimed as ‘‘teleology.’’ Development, organic movements in general, would not therefore be fully explained in terms of productions, effects, or mechanisms, because these terms all carry the structure of poiesis.

Morphomes and Organic Practice I want first to restate my point in more modern terminology, while keeping the reference to its Greek origins, and call this model ‘‘organic practice.’’ Applied to developmental processes, a ‘‘practical’’ understanding would hold that these processes—in all their molecular details—are not simply instrumental in bringing about certain results, results that play functional roles in the construction of the organism. Developmental processes as practices are more than that. How can we express this ‘‘more’’? We can say that their significance is not only instrumental but also intrinsic. Developmental processes

319

Christoph Rehmann-Sutter

do not simply lead to the next steps or the next intermediate states, but are themselves significant. They are bits (steps) of a being’s continuous presence in the world. In other words, developmental processes are constituents of the existence of the living being, and existence can be seen as a process of staying present in an environment. The identity of the organism is therefore (among other things) the continuity of its presence. It is evident that this ‘‘identity’’ cannot be adequately understood when it is described by the composition, the construction, or the function of the organism. For convenience I introduce a term, morphome, that is constructed in parallel with genome, the total of all genes of an organism (analogous neologisms include proteome and rnome; Riddihough 2002). Morphome is an abstract term covering all morphological and structured states and processes a given organism is capable of performing during its lifetime. We can formulate a basic theorem of the organic practice model: Living beings develop along the course of their particular morphome, in such a way that in each moment of their developmental processes they remain present as the intrinsic ends of those processes. What are the implications of this? I outline five main points here. 1. According to the view of organic practice, there is no element of the morphome that is ontologically prior to or higher than all the others. There is no one privileged state or phase like the ‘‘imago’’ or the ‘‘adult’’ state, making the states before its occurrence a production ( poiesis), and those after its realization a decay. The development of an organism is nothing less than a continuous and changing presence in the world. Every stage of the processes is important per se. Its stages may differ considerably in a multitude of respects. And relative to certain perspectives (which can be explicitly stated), some stages may be especially important. The reproductive age, for example, may be of particular importance with reference to an evolutionary perspective, and the age of greatest bodily strength may be of special importance relative to competition. But aside from such particular perspectives, all phases are intrinsically important. The organism may have different inner and outer shapes, capabilities, needs, and so on, that mark distinctions between the developmental stages. But these differences are not differences in terms of ontological relevance for this being. 2. The organic practice view implies a certain form of subjectivity. In order to make sense, the model must take into account a certain subjective dimension of the organisms themselves; organisms cannot be described solely, objectively as biochemical machines. To be meaningful, the concept of praxis requires a dimension of experience, perception, communication, action, or the

320

Poiesis and Praxis

like. Otherwise, it would not be possible to expect a process to be performed for its own sake. The processes would simply happen and that would be all. Life, experienced through the subjective perspectives of different organisms, must have highly different qualities. In the subjective perception of a caterpillar the environment must be dramatically different from that of the butterfly into which it develops. Metamorphosis is not only a structural change but also possibly a qualitative change in subjectivity, changing its perceptive and active radius and its perceptive qualities. These subjective qualities (essentially perception and behavior) are accessible only from the perspective of the being itself, and this is not reducible, at least not fully, to our scientific understanding, to empathy, or to philosophical imagination (Nagel 1974). We say that it makes sense to start from the assumption that there might be some subjective perspective and a genuinely active mode for each organism, as simple or as bizarre as it may appear if somebody could compare it with our perspective as humans. The continuous and changing ‘‘presence’’ in the world, as we described organic practice above, requires a subject that can be aware and present in some way. When used as a model for describing living organisms, the organic practice view opens a space for subjectivity and for the subjective sense of the living being. Accepting this affects our relationships: we can approach organisms as a kind of concrete other in the realistic sense of the word: as a being for which we can be responsible.4 The success of molecular biology, especially developmental genetics, is therefore not the strongest argument against the hypothesis of subjectivity in life. Rather, development can be understood in all its molecular aspects and complexities as the particular process of necessarily subjective presence in the world. This general thesis also includes a plurality clause: subjectivity may have as many variations as the diverse forms of objective bodies. Our human experience of subjectivity is by no means exclusive or representative; it is only one example of living subjectively. It is what we know—because we happen to be humans. 3. The end of the developmental process leading through the different stages of the morphome is the process itself: the continuous presence in the world; maintenance of the presence. And to this continuous organic practice belongs the ability of the organism to organize itself in such a way as to endure the next moment physically, to survive, to cope with difficulties, to procreate, and to give way to subsequent generations. Practice, therefore, also has an eminently productive side. Each step is capable of producing the next, in a historical (Stent 1981) sequence. ‘‘Historical’’ means that no preestablished

321

Christoph Rehmann-Sutter

plan is needed to establish the order of the sequence of steps. The regularity of subsequent developmental steps emerges because each step, and the whole situation of the organism after this step, in a complex way causes the next. However, life history as understood in the mode of organic practice not only has the sequential aspect that each dynamic element of the morphome has a causal explanatory significance for the steps to follow within a given environmental and/or social context. There is also a genuinely presentistic aspect: all the steps constitute part of the existence of the organism in the sense of its active presence in the world, not only as a biochemical machine which functions, but also as entities that can meaningfully be described as being in the presence of a world. Organisms are in this sense subjects of a life. 4. If we allow development to have this practical side (the reasons for doing this have an ethical quality and will be discussed below under ‘‘Choice’’), we have the philosophical preconditions for appreciating the existence of living beings as more than their molecular-mechanistic functioning, without losing interest in what science finds out about those mechanisms. Each molecular mechanism is another detail in the fascinating complexity of organic practices. Each molecular explanation does not eliminate questions but deepens the wonder at the ability of organisms to maintain and develop their presence within interacting ecological and social networks. Hence, developmental biology together with a thorough philosophical interpretation is in its essence a theory of existence, an account of what it means to be in the world. Within a model of organic practices, it becomes possible to debate which account of organic identity a particular biological theory should have, which explanations of ‘‘being an organism’’ are better and which worse. The organic practice model should therefore protect us from rash metaphysical conclusions. It is a basis to challenge the alleged nonexistence of any deeper meaning in organic life beyond chance and survival. 5. But there is still another dimension to the importance of considering the ‘‘sense’’ life possibly has to the developmental systems themselves, in the mode of practice: they may be performing and celebrating their presence in the world without us being aware of it. Our world may be populated by myriad other living things who are also, like us, occupied with performing and celebrating their existence. Perhaps—I must speak once more in anthropomorphic terms—they have considerably less need than we do to overcome meaninglessness, absurdity, and triviality. The terms performing and, a fortiori, celebrating, are metaphors. They are taken from our human realm of experiences. Organic practice, if taken seriously and if carefully considered, also means to allow for distinct kinds of

322

Poiesis and Praxis

sense of those organic, developmental works which are not reducible to what we as human interpreters can anticipate. This granted, there is still a considerable difference depending on whether we have reason to expect a space of intrinsic sense which is connected to the term of organic practice at all— or not. Let me summarize: each step in a developing sequence has a double significance. On the one hand, the step is a necessary intermediate for the success of the sequence or a prerequisite for the next steps, or a necessary but not sufficient condition for other developmental cascades taking place at another location in the same organism. This is the poietic, productive aspect. But on the other hand, the same sequence of steps is a constituent of the organism’s life, and matters, step by step, in itself. Each process matters in being performed, because the ‘‘life’’ of the organism means essentially the act of performing. Any one step cannot be singled out. Each refers necessarily to all the other steps of the morphome. Each step makes reference to the whole of the life of an organism and to its contexts.

Functionalism The modern scientific perception of living organisms—its analysis of their molecular structure and its explanation of their functions—prima facie seems to be closed to an organic practice understanding. All attempts to see ‘‘more’’ in natural processes appear as pure projections, wishful thinking. The modern scientific understanding of nature has been deeply antiteleological. It rejects both poiesis and praxis as models for natural processes, along with the whole attempt to connect natural processes with the realm of subjective human understanding. A process in the strict sense is nothing but a sequence of events; it has its efficient causes, and questions about ends, achievements, or final causes make no sense within this approach.5 Because processes are nothing but causal sequences, this criticism culminates in stating that natural processes cannot and should not be understood (as actions can be understood), but rather should be explained with reference to natural laws, physical forces, and the structure and function of the parts of the entities involved. My observation is that real biology has always practiced a little heresy. It never has been so consequently descriptivist that references to the poiesis model were omitted. What has really been left out of biology is the praxis model. The first evidence for this is the concept of functionalism, or mechanism.

323

Christoph Rehmann-Sutter

The biological understanding of biomolecules (such as dna or proteins), of cellular components (such as vesicles, membranes or microfilaments), or of organs (such as the brain, the heart, or the uterus) is inspired by an interest in their function. The function of x can be defined as that y which x serves, is built for, or is used for.6 The function of dna is to provide information for the synthesis of proteins. The function of insulin is to control the chemical degradation of sugar. If we compare the structure of such functional descriptions with poiesis and praxis, a parallel to poiesis is evident: y (in what ‘‘the function’’ of x consists) is not x itself but something different from it. The function of the synthesis of insulin in the pancreas is not seen in the process of insulin synthesis itself (this would not ‘‘explain’’ the function of synthesis in the context of whole system of the human body) but in the production of insulin molecules, which are metabolically important for sugar degradation. This overwhelming interest in function may be a particular epistemological characteristic of the biosciences. Alexander Rosenberg notes that ‘‘explanations in physical science do not assign functions to the phenomena they explain. They assign causes. But the molecular biologist’s explanation of the difference between rna and dna proceeds by citing the effects of the difference and not its causes. This is in fact what a function seems to be, a certain kind of effect’’ (Rosenberg 1985: 41). The assignment of function means to see in organic processes a poietic structure. The effects (use, service) of a part or process x in a dynamic system in biology explain the evolutionary advantage of the existence of x in this system. ‘‘Without the restrictions in the way in which a goal is attained that are provided by successively more detailed accounts of the operative internal mechanisms,’’ Rosenberg writes, ‘‘no improvement in the predictive powers of biological theory is forthcoming’’ (pp. 65f.). I want to distinguish between the use of the function concept as a tool in research, and functionalism as a philosophical extrapolation. In my view the former is legitimate, its success the best proof of its legitimacy. The question of the function of, for example, a gene has undoubtedly been very fruitful for contemporary research. I cannot imagine how genetics could proceed without searching for functions. But when this is converted into a philosophy of life, claiming that living organisms are essentially, or nothing but functional systems, that this is their very mode of existence, their principle of being, functionalism turns to metaphysics and as a distinct metaphysical theory excludes organic practice as a philosophical view on biology.

324

Poiesis and Praxis

Beyond Genetic Programs The mechanistic approach to organic systems was the first site in modern biological thought where the model of poiesis was used as a mode of understanding. The second site has been the predominant mode of genetic explanation of development in the twentieth century. The question was: How can a zygote produce a new organism? One answer is: by a genetic program that is somehow encoded in the sequence of dna. The term genetic program seems to have appeared for the first time in a notebook of Jacques Monod in 1959 (see Kay 2000: 221), but the idea itself goes back at least to August Weismann’s concept of the Ids, the dark bits of chromosome he could see under the microscope. In his view, it was these which provided the Architektur for the new organism (1892: 91, 112). The idea that an invisible formative principle is responsible for instructing the development of the organism from within is itself Aristotelian (see Kupiec 1999; Kupiec and Sonigo 2000: 8). But the spectrum of possible understandings of Artistotle’s text has not been considered. According to the dominant interpretation, the original concept was of a productive or fabricative principle. The idea of the genetic program sees the relation between form and process as a productive one. John Maynard Smith and Eörs Szathmáry describe the ‘‘basic picture’’ of development informing modern biology in the following way: ‘‘The basic picture, then, is that the development of complex organisms depends on the existence of genetic information, which can be copied by template reproduction. Evolution depends on random changes in that genetic information, and the natural selection of those sets of instructions that specify the most successful organisms . . . . What is transmitted from generation to generation is not the adult structure, but a list of instructions for making that structure’’ (Maynard Smith and Szathmáry 1999: 2; see the discussion in Rehmann-Sutter 2000). Several alternative approaches to understanding development, all rejecting the idea of a genetic program, have been proposed (see the chapters by Griesemer, Griffiths, Hoffmeyer, Oyama, Webster, and Goodwin in this volume; Morange 2001). Alternative views on the genome-organism relation rediscover the organism as an ‘‘informational network’’ (Keller 1995: 118). They have a better chance of explaining the complexities of real genes, whose nucleotide sequences are frequently edited before being used in polypeptide synthesis (the sequence is changed after transcription), chemically modified (methylated in imprinting), multiply spliced (different mature mrnas pro-

325

Christoph Rehmann-Sutter

duced from one and the same gene), or whose gene products have totally different functions in different cellular contexts (Fogle 2000; Neumann-Held 2001). Alternatives such as the developmental systems approach have the considerable theoretical advantage that they are—in contrast to the program approach to dna—hospitable to an interpretation of developmental processes in the practical mode. Susan Oyama starts from the idea that dna segments obtain their causal significance only within the process of their actual involvement in developmental steps. The macromolecular interactions in which dna gets involved take place in the context of a dynamic cellular microarchitecture, an organismic body, and very often involve environmental relations as well. These interactions provide a given dna sequence with its real information value. It does not make sense to say that genes carry information before or outside these contexts and interactions (Oyama 1985, 2000, this volume; Oyama, Griffiths, and Gray 2001). dna sequences are not genes by themselves. They need first to be involved in the molecular mechanisms of polypeptide synthesis. The question of how a sequence becomes a gene, how it is ‘‘genetized,’’ so to speak, is crucial (Neumann-Held 2001: 73). Eva Neumann-Held concludes that we would do better to call this interactive process that actually leads to a polypeptide ‘‘the gene’’ rather than the stretch of dna that is involved. This process encompasses much more than a bit of dna, and the dna itself cannot signify what the cellular processes can do with it. This is the basic idea behind her fascinating ‘‘process molecular gene concept’’ (Griffiths and Neumann-Held 1999; Neumann-Held 1999). Given that each developmental step is the precondition and, together with the dna and the milieu, the sufficient causal condition for performing the next step, and given that we have theoretical models to conceptualize the role of the dna in development along these lines, the idea of organic practice can more easily be integrated. A systemic approach explains the significance of dna, both as the whole genome and as particular sequence stretches in the developing organism, by the interrelations between parts and processes in the morphologically structured ‘‘system’’ (which also includes features of the environment). In a systemic theory of the genome there is no genotypic hinterland behind or within the organism destined to be a producing agent. The phenotypic organism, through its internal and external networks of relationships, develops in its own ‘‘historic’’ logic. Step by step, it integrates the dynamic features of its organic structure into a new moment of presence. The organism is not a dna-driven machine for producing more dna; rather, the organism is the author of the genetic information that is necessary to in-

326

Poiesis and Praxis

form each new developmental step (cf. Rehmann-Sutter 2002). If we discard the genetic program theory, developmental genetics leaves space for such a re-reading. But why bother with the effort? What is the legitimacy of interpreting development in such a weighty way? Why not just listen to what biology tells us about all those fascinating genetic mechanisms? I have demonstrated that science does not contradict the theorem of organic practice. But I have not yet given philosophical reasons for why the model should be attractive. Therefore, I now propose to consider an additional point of view: that of our own embodiment.

Body Knowledge ‘‘I am witness of my environment,’’ Maurice Merleau-Ponty notes in a draft for a lecture on the idea of nature at the Collège de France in 1959–1960 (Merleau-Ponty 2000: 295). Merleau-Ponty has argued that our participant perspective on organic life and organic development, the perspective which we hold and cannot leave (even when we forget it or resist taking it into consideration), offers us a key epistemological advantage. Being ourselves organisms, we cannot confine ourselves to describing organisms as objects. We are witnesses in nature, not trapped in an objectivizing observer perspective. We are ourselves specimens of those organisms that we describe, and we are this not only accidentally. Our status as witnesses includes our being in our environment as one of them. This has an interesting consequence. Our body is not just one object among the myriad other objects in the world, one that differs from the others in the bewildering fact that it is accompanied by a conscience and the others are not. Being the being does not muddy the view by including subjective, emotional biases in the mix; it enables us to see. The body ‘‘is closer to me than the things,’’ Merleau-Ponty continues, because only by being a body can we perceive other things (2000: 295). When we take this phenomenological and explicitly subjective stance, we do not merely transfer the image of the body from the objective sciences— that is, we do not judge ourselves as having ‘‘in our own possession’’ the objective body of the natural sciences just by attaching our consciousness to it like a tag saying ‘‘this is me.’’ The body we live is somehow a different thing. It is not an object at all, but rather something which is a prerequisite for all our capacities to see, perceive, think, investigate, and objectify. And it can

327

Christoph Rehmann-Sutter

itself be regarded as an object, as a functional system. This ability concerns a special reciprocal activity. The object-body that is investigated by biomedical research (see La Mettrie’s Homme Machine) is the result of and not the prerequisite for that activity. When we consider this epistemological reflection as fundamental for our perspective on the world as well, we begin to suspect that other living things who share the environment with us are not just objects either. Our objectivizing gaze ‘‘makes’’ them objects. In order to see them in a ‘‘correct’’ perspective the observer must be active, not just receptive. The scientific observer must actively detach and must standardize a selection procedure for perceptive phenomena. One rule in this procedure is the following: only those perceptions that can be repeated by anybody who masters the experimental procedures should be seen as valid empirical evidence. Relatedness of our self and other subjective facts and features are not among these. The lived phenomenal body is a ‘‘macro-phenomenon’’ says MerleauPonty, because its phenomenality does not end at the skin, at the boundaries of our physical body. But objects, and also the body in the objectivist perspective, are microphenomena (2000: 295), and hence, the lived phenomenal body offers a different perspective on the whole environment. Merleau-Ponty’s terminology can help clarify the implications of what we discussed above: that the being we are ourselves and the other beings in our environment are not objects in themselves. We can, and perhaps should, consider that they might also be seen as beings with whom we share presence in a common space. The model of organic practice reminds us that living beings are present in their developmental processes. These processes are intrinsically (as praxis) ‘‘what organisms are up to.’’ Being in the world is not a triviality; it is the basic accomplishment of life. This includes being present to other things, including ourselves. There is a further convergence between the organic practice model and body phenomenology. So far we have discussed two kinds of knowledge of the body: the body within an objectivist perspective and the body of our own presence in the world. Both belong to us, but in different kinds of relationships: the first is the body we have, the second the body we are. Recent anthropological research in medicine—a field where these two perspectives on embodiment are particularly close together—has shown that there is a third dimension to embodiment: the body we do. The body we do is something we practice (Mol and Law 2004). Annemarie Mol’s (2002) study of atherosclerotic patients provides evidence that the body and its atherosclerosis are different realities when ‘‘enacted,’’ for example, through X-ray angi-

328

Poiesis and Praxis

ography or through ultrasound. Different practices multiply the body, and the different enacted realities of the body hang together in an intricate way. The model of organic practice is related to this idea that an organism becomes what it is through practices. Doing, in a very basic sense, is primary for being an organism. Organisms practice what they are. The Aristotelian term of practice provides us with an idea simple and basic enough not to involve anthropomorphic projections. We could not uncontroversially claim for all organisms and cells that they ‘‘enact’’ their bodies, as is perhaps possible for self-conscious humans. But it may make sense, even against the background of molecular biology, to say that the actual ‘‘meaning’’ (information/effect) of the genome arises from its processes of interaction within the organism. The genetic information is not there before development starts, as prescriptive inscription on the chromosomes that needs only to be realized, transformed from a one-dimensional sequence to a three- (or four-) dimensional being. What genes ‘‘do’’ (bring about) depends to a great extent on the context of the particular cell and its place within the developing body. ‘‘Meaning’’ is not provided by a static or eternal being, as in classical ontology; it develops through practices. The notion of intrinsic sense that I have used throughout this essay depends not on an ontology of being, but rather refers to an ontology of practices whose senses arise in their operation. Developmental biology has a particular significance among the biological subdisciplines because it attempts to understand ontogeny, the coming into being of living things, including ourselves. Investigating the development of a living being is therefore, in Merleau-Ponty’s words, ‘‘the best way of understanding its being’’ (2000: 310). My argument about organic practice makes use of the epistemological advantage that results from our witness status in development. Taking into account the fact that we are witnesses of organic life opens a reflective space around scientific insights into the mechanisms and the genetics of development. There is room for developing organic beings to be present in their own intrinsic logic of organic practice; in other words, in their own spaces of sense.

Choice The term space of sense indicates that microcosmos of sense that is implied by the practice view of organic processes and movements. These are more than what they produce. This ‘‘more’’ is the sense of the processes or movements themselves. In order to see a living being in such a way that its developmen-

329

Christoph Rehmann-Sutter

tal processes have an intrinsic sense, we have to assume that this being, by its very processes of development, constitutes a realm of sense of its own. The theoretical choice between two fundamental ways of thinking about the role of the genome within development—the ‘‘program’’ view and the ‘‘systemic’’ view—therefore opens or closes the space needed to conceive living things as beings in the true sense, developing their lives as a practice within their own spheres of sense. So far we have established that there is such a choice. But we still need to establish the truth of the theorem of organic practice itself. Why should we adopt this option? The organic practice theorem is not an empirical hypothesis. We cannot falsify it through experiments or measurements. It is, rather, a hermeneutic hypothesis which says that it makes sense to see developmental processes in two modes, one parallel to poiesis, the other parallel to praxis. But still, there must be reasons to find the organic practice theorem trustworthy, or, conversely, obscure and futile. The relevant reasons come to mind when we focus on the contribution of biology to our understanding of relationships: the relationship to nonhuman life—that is, the human-nature relationship—and the relationship to our own embodiment. Developmental genetics is about formative processes which take place in organisms. It describes these processes in a certain way, and in doing so constructs a model of them. This model is located within human minds, and from there it influences how scientists and scientifically informed people see those beings that the model describes. The model qualifies perception, and by doing so structures the interpretive relationship. In this relational context one question takes on particular significance: What difference does the choice between the ‘‘poietic’’ and ‘‘practical’’ models make? The latter model leads to an appreciative perception of developing organisms, to a mode of perception that expects a sphere of intrinsic sense of developmental processes superseding functional explanations and establishing those processes as parts of the being’s presence. This intrinsic sense closely accompanies all the complex mechanisms of physical development. What is claimed here is the possibility that the developing being has its own perspective. This is not the same as the existence of a metaphysical force, life essence, or soul substance. Within the framework of the poietic model, on the other hand, there is no possibility of such a qualified relationship. There are three metaphysical options: creationism, preformationism, or nihilism. Creationism claims that there is an intentional plan that made the biosphere (God, Nature, or Evolution—with a capital N and a capital E). Preformationism claims that dna

330

Poiesis and Praxis

codes for genetic blueprints of organisms and that the genome controls development. But if we have no convincing reason to assume the truth of either creationism or preformationism and decide to stay within the limits of the poietic model, we necessarily end up in a nihilistic existentialism with a negative metaphysics (‘‘Ni dieu ni gène’’; Kupiec and Sonigo 2000). The praxis model is comprehensive enough that the poietic (productive, functional, mechanistic, materialistic) aspect can remain present, whereas the poietic model would exclude the praxis aspect. In this respect the two modes of understanding are not symmetrical. The first choice, therefore, is not between theories but between relationships: a morally qualified relationship is established between describers and the beings they describe. This relationship is essentially configured by the describers’ ethical attitude. Description is not merely a representational activity that can succeed or fail, be correct or misleading. Description is itself a form of relational practice, a form of being in the world, a practice of being present to those other beings we describe. Like other practices it has its own virtue. The virtue for this practice of description or its ethical criterion is contained in the following question: Is the practice capable of coming into the presence of other living beings? 7 If the suggested interpretation of developmental biology in the mode of organic practice makes any sense at all, it should contribute to the richness and depth of our commitment as witnesses of life. If there is any necessity in this choice, it is an ethical necessity.

Notes Part of this work was funded by the foundation Mensch-Gesellschaft-Umwelt (mgu) at the University of Basel (grant 42/95). I am grateful to my project partner, Eva M. Neumann-Held, for inspiring discussions, and to Jackie Leach Scully, Monica Buckland Hofstetter, and Mindy Conner for help in clarifying the English text. 1 Problems of this distinction in the theory of action are discussed in Charles 1986; Kraut 1989; Broadie 1991; and Sherman 1991; see also Rehmann-Sutter 1996: ch. 5.1. 2 Mauron 2002 identifies ‘‘genomic metaphysics’’ with ‘‘the belief that the genome is the ontological hardcore of an organism, defining its distinctive traits, its individuality, as well as underpinning its membership in a particular species.’’ Accordingly, alongside the decline in this belief, a decline in the metaphysics of the genome would be expected. I think we should not hope to discard the meta-

331

Christoph Rehmann-Sutter

physics of the genome when we hope to discard a false belief about its role; for this we should not define metaphysics as a particular (false) belief but as a space for critical reflection on the ontological significance of the genome. 3 I have discussed these extensively elsewhere; see Rehmann-Sutter 1996: 139ff. 4 I use the term concrete other as it was introduced by Seyla Benhabib (1987). 5 Stephen Jay Gould notes: The ‘‘aspect of good organic design does express a final cause in adaptation, but any evolutionary changes must still be crafted by an efficient cause—and Darwinian natural selection generally acts as the efficient cause we seek for our explanations’’ (Gould 2002: 1208). 6 There is a considerable literature on the concept of function; see Cummins 1975; Mahner and Bunge 2000: ch. 4. 7 The expression ‘‘coming into presence of ’’ is from James Kellenberger’s relationship ethics (1995: ch. 6).

References Aristotle. 1957. De Anima. T. E. Page (ed.). Loeb Classical Library. London: William Heinemann; Cambridge: Harvard University Press. Aristotle. 1961. Metaphysics. T. E. Page (ed.). Loeb Classical Library. London: William Heinemann; Cambridge: Harvard University Press. Aristotle. 1962. Nicomachean Ethics. T. E. Page (ed.). Loeb Classical Library. London: William Heinemann; Cambridge: Harvard University Press. Benhabib, S. 1987. The generalized and the concrete other. In: E. F. Kittay and D. T. Meyers (eds.), Women and Moral Theory (pp. 154–177). Savage, Md.: Rowman and Littlefield. Broadie, S. 1991. Ethics with Aristotle. New York: Oxford University Press. Charles, D. 1986. Aristotle: ontology and moral reasoning. Oxford Stud. Ancient Philos. 4: 119–144. Cummins, R. 1975. Functional analysis. J. Philos. 72: 741–765. Driesch, H. 1909. Philosophie des Organischen. 4th ed. Leipzig: Quelle and Meyer, 1928. Fogle, T. 2000. The dissolution of protein coding genes in molecular biology. In: P. Beurton, R. Falk, and H.-J. Rheinberger (eds.), The Concept of the Gene in Development and Evolution. Historical and Epistemological Perspectives (pp. 3– 25). Cambridge: Cambridge University Press. Gould, S. J. 2002. The Structure of Evolutionary Theory. Cambridge: Belknap Press. Griffiths, P. E., and Neumann-Held, E. M. 1999. The many faces of the gene. BioScience 49: 656–662. Kay, L. E. 2000. Who Wrote the Book of Life? A History of the Genetic Code. Stanford: Stanford University Press.

332

Poiesis and Praxis

Kellenberger, J. 1995. Relationship Morality. University Park: Pennsylvania State University Press. Keller, E. F. 1995. Refiguring Life: Metaphors of Twentieth Century Biology. New York: Columbia University Press. Kraut, R. 1989. Aristotle on the Human Good. Princeton: Princeton University Press. Kupiec, J.-J. 1999. L’influence de la philosophie d’Aristote sur l’élaboration de la théorie de l’évolution et sur la génétique. Rev. Eur. Sci. Social. 37/115: 89–116. Kupiec, J.-J., and Sonigo, P. 2000. Ni Dieu ni gène: pour une autre théorie de l’hérédité. Paris: Seuil. Mahner, M., and Bunge, M. 2000. Philosophische Grundlagen der Biologie. Berlin: Springer. Mauron, A. 2002. Genomic metaphysics. J. Mol. Biol. 319: 957–962. Mayr, E. 1991. Eine neue Philosophie der Biologie. Munich: Piper. Merleau-Ponty, M. 2000. Die Natur: Vorlesungen am Collège de France 1956–1960. Munich: Fink. Mol, A. 2002. The Body Multiple: Ontology in Medical Practice. Durham: Duke University Press. Mol, A., and Law, J. 2004. Embodied action, enacted bodies. The example of hypoglycaemia. Body Sci. 10 (2–3): 43–62. Morange, M. 2001. The Misunderstood Gene. Cambridge: Harvard University Press. Moya, F. 2000. Epistemology of living organisms in Aristotle’s philosophy. Theor. Biosci. 119: 318–333. Nagel, T. 1974. What is it like to be a bat? Philos. Rev. 83: 435–450. Neumann-Held, E. M. 1999. The gene is dead—long live the gene! Conceptualizing genes the constructionist way. In: P. Koslowski (ed.), Sociobiology and Bioeconomics (pp. 105–137). Berlin: Springer. Neumann-Held, E. M. 2001. Let’s talk about genes: the process molecular gene concept and its context. In: S. Oyama, P. E. Griffiths, and R. D. Gray (eds.), Cycles of Contingency: Developmental Systems and Evolution (pp. 69–84). Cambridge: mit Press. Oyama, S. 1985. The Ontogeny of Information. Cambridge: Cambridge University Press. Oyama, S. 2000. Evolution’s Eye: A Systems View of the Biology-Culture Divide. Durham: Duke University Press. Oyama, S., Griffiths, P. E., and Gray, R. D. 2001. Cycles of Contingency: Developmental Systems and Evolution. Cambridge: mit Press. Rehmann-Sutter, C. 1996. Leben beschreiben. Über Handlungszusammenhänge in der Biologie. Würzburg: Königshausen and Neumann. Rehmann-Sutter, C. 2000. Review of The Origins of Life, by John Maynard Smith and Eörs Szathmáry (1999). J. Bioecon. 2: 183–188.

333

Christoph Rehmann-Sutter

Rehmann-Sutter, C. 2002. Genetics, embodiment and identity. In: A. Grunwald, M. Gutmann, and E. M. Neumann-Held (eds.), On Human Nature (pp. 23–50). Berlin: Springer. Riddihough, G. 2002. The other rna world. Science 296: 1259. Rosenberg, A. 1985. The Structure of Biological Science. Cambridge: Cambridge University Press. Sherman, N. 1991. The Fabric of Character: Aristotle’s Theory of Virtue. Oxford: Clarendon. Smith, J. M., and Szathmáry, E. 1999. The Origins of Life: From the Birth of Life to the Origin of Language. Oxford: Oxford University Press. Stent, G. 1981. Strength and weakness of the genetic approach to the development of the nervous system. Annu. Rev. Neurosci. 4: 163–194. Urmson, J. O. 1988. Aristotle’s Ethics. Oxford: Oxford University Press. Weismann, A. 1892. Das Keimplasma: Eine Theorie der Vererbung. Jena: G. Fischer. Zubiria, M. 1999. Entelechie. In: H.-J. Sandkühler (ed.), Enzyklopädie Philosophie (pp. 327f ). Hamburg: Meiner.

334

13 D E V E L O P M E N TA L E M E R G E N C E , GENES, AND RESPONSIBLE SCIENCE

brian c. goodwin Organisms can be described as dynamic wholes that embody specific constraints, but that is not saying much. This description applies to almost any natural process we wish to consider, from a snowflake to a spiral galaxy. Each of these has a particular form and behavior that express its dynamic organization as a whole while embodying specific constraints that give the phenomenon its distinctive properties. However, biologists are not accustomed to describing organisms in these terms, which seem too simple and physical to capture the properties of a living system. An organism is complex, the result of a historical process. But so is a galaxy or a snowflake. The specific pattern of water crystals in a snowflake is a reflection of inherent organizational constraints resulting in a characteristic hexagonal structure which, together with the particular circumstances acting during crystallization, make it different in detailed form from all other snowflakes. An organism, however, is this and more. It carries its history with it in its genes, and reproduces these as an inherent aspect of a distinctive dynamic process expressed in a unique life cycle. Reproduction is one of the properties that make an organism different from snowflakes and galaxies. And this is a major reason why biologists no longer emphasize the obvious fact that organisms are dynamic wholes that embody specific constraints (including genes). They prefer to focus on genes as constraints described as information, or as algorithms, rules for making an organism of a particular type. This is fine, as far as it goes. However, bits of information or rules have no meaning except within a context of possibilities, among which they specify alternatives. That context is the organism as life cycle. So we come back, willy-nilly, to the organized dynamic whole within which genes act as part of the organizational constraints that give the life cycle its specific form. The problem now in biology is to describe dynamically the nature of this self-

Brian C. Goodwin

reproducing whole and how it differs from other complex processes. This can be seen as a major task of postgenomic biology. Now that it has become clear once again that the information in genes cannot explain how an organism is made, particularly with the recognition that there is not enough difference in numbers or natures of genes to account for the morphological and behavioral differences between us and chimpanzees and asparagus plants, we can get back to the job that was somewhat eclipsed during the 1980s and 1990s while the spotlight played on the genome project. The job to be done in the fields now variously called postgenomics, proteomics, and biocomplexity has to do with the problem of organized complexity at different levels of the organism, from genes and molecules to the distinctive pattern of morphology and behavior that characterizes the life cycle of a species (cf. Keller 2000). Two related areas of research have been maturing since the late 1970s that can make significant contributions to this job: nonlinear dynamics and complexity theory. Both are highly dependent on sophisticated computing tools, and their results make clear why science has reached a new frontier that requires significant change in our relationship to complex systems.

Emergent Properties of Complex Systems Complex systems are described as entities that are made up of many units or agents in interaction that together constitute a dynamic and structural whole. This is a very general definition with a diversity of applications throughout the living and nonliving worlds, and that is what makes it possible to compare animate and inanimate systems within the same conceptual framework. What distinguishes living from physical systems is often the type of unit that constitutes the system and the kind of agency it can express. Multicellular organisms are complex systems whose constituents are molecules at one level, cells at another, tissues and organs at yet another. Insect societies are organizations of interacting individuals. Ecosystems are organized systems made up of individuals belonging to diverse species. The distinctive properties that characterize these systems can be understood as a consequence of the way the constituents are organized dynamically. As an example, the cellular slime mold is made up of several thousand separate cells that come together (aggregate) and undergo a distinctive pattern of spatially ordered change leading to a distinctive reproductive form, the fruiting body (fig. 1). The sequence of initial spatial changes can be modeled as a process in which the interacting cells have the properties of what is

338

Developmental Emergence

1 • The life cycle of the cellular slime mold, Dictyostelium discoideum. See B. C. Goodwin, How the Leopard Changed Its Spots: The Evolution of Complexity (London: Weidenfeld and Nicolson, 1994), p. 46.

called an ‘‘excitable medium.’’ An excitable medium has dynamic behavior of the type observed in developing slime molds and so provides an explanation of their distinctive type of agency. An example is the characteristic spiral patterns cells generate by their collective movements during the aggregation process (fig. 2a–c). Vasiev, Siegert, and Weijer (1997) used a model of the aggregating field of interacting amoebae as an excitable medium to simulate slime mold development and observed that the whole sequence of pattern changes, from separate cells to multicellular form, occurred without the need to change any of the parameters of the model. That is, the sequence occurs spontaneously as a result of the dynamic order in the system. The authors conclude that ‘‘this complicated process could take place without the necessity for transcription of new genes. In reality, however, it is known that many different genes are transcribed during aggregation and mound formation.’’

339

Brian C. Goodwin

2 • The gradual breakup of the original lawn of amoebae into separate aggregates centered on initiating centers. Each aggregate forms an organism of several thousand cells which then develops as shown in figure 1. See B. C. Goodwin, How the Leopard Changed Its Spots: The Evolution of Complexity (London: Weidenfeld and Nicolson, 1994), p. 54.

These observations lead to two questions. First, what are the genes doing during this process? And, second, what constitutes an explanation of the developmental process? It seems that in the aspect of slime mold development described by Vasiev, Siegert, and Weijer the genes are reinforcing and stabilizing the transitions from one form to another. For example, as the cells aggregate, their initially uniform spatial pattern breaks into a set of converging streams (fig. 2d–f ). After this process has begun, genes become active, which makes the cells stickier, in turn enhancing stream formation by causing cells to adhere more strongly to one another. Stream formation begins as an expression of intrinsic dynamic instability of the uniform pattern (Höfer et al. 1965), and genes then stabilize the process. However, there are other aspects of development in which genes initiate instability, as when two types of cell arise and segregate in the process of making the fruiting body. So genes can play different roles in development: they can enhance and stabilize patternforming processes that occur spontaneously as a result of the dynamic order expressed by the complex system; and they can initiate dynamic changes that result in morphogenesis of particular types. In neither case, however, can we say that gene activity explains the change of form observed. The actual morphology of the organism is explained by a model that describes the full dynamic of the process, of which genes are a part, and the particular form generated. This is all obvious and has been known for some time. Unfortunately, the language of developmental genetics often appears to imply that genes are the cause of development. The cause of development is in fact the dynamic

340

Developmental Emergence

process, organized in space and time, that is expressed within the developing organism; genes are a part of this. What is now needed in biology is a theory of organization of the living state of the type explored by Maturana and Varela (1986) and by Kauffman (2000) that could answer the question: How do living systems reproduce not only their constituent parts but the organizational constraints that produce the distinctive order of the whole?

Wholes, Parts, and Unpredictability Are the emergent biological phenomena described by nonlinear dynamic models used in the study of complex systems reducible to the properties of a level below that of the phenomenon studied? That actually depends on the form of the explanation achieved by a model, and on scientific consensus. In a study of rhythmic patterns observed in the activity of ants (Leptothorax) tending the queen and the young in the colony’s brood chamber, Cole demonstrated that individual ants or ants at low density display deterministic chaos in their activity-inactivity pattern, while at higher densities such as those in the brood chamber the behavior becomes rhythmic with a well-defined periodicity (Cole 1991). A model that reproduced these observations involved individuals with intrinsically chaotic behavior interacting by excitation, as observed experimentally. Rhythmic activity patterns from chaotic individuals interacting by excitation are not something that could have been predicted, so the phenomenon is regarded as an emergent property of the system (see Solé and Goodwin 2000). Such properties of the whole (the colony in the brood chamber) are not reducible to the properties of the individual ants together with their interactions, even though the model consistently produces this result. And they are certainly not reducible to the activity of genes. Someone may develop a theorem which describes the conditions under which chaotic individuals in interaction give rise to periodic behavior, but none exists at the moment, so far as I am aware. If and when it does, the higher-level phenomenon will probably be regarded as predictable from the lower-level behavior, and in this sense reducible. However, unexpected emergent properties that could not have been predicted from lowerlevel properties are being continually discovered in nature and in models of complex systems. Goldstein notes that there seems to be no end to the emergence of emergents. Therefore, the unpredictability of emergents will always stay one step ahead of the ground won by prediction. As a result, it seems that emergence is now here to stay (Goldstein 1999).

341

Brian C. Goodwin

It is important to recognize that even if a particular phenomenon has been explained as a property of a complex system, and in some sense reduced to the dynamic organization of lower-level components, any change in the nature of the components or their interactions can result in unpredicted behavior of the whole. For instance, the introduction of a new species into an ecosystem or a new gene into the genome of an organism will in general have unpredictable consequences. Ecological management and biotechnology are intrinsically unpredictable interventions into natural systems. Another way of putting this is to acknowledge that nature is, in general, unpredictably creative. Those aspects of the natural world that have turned out to be largely predictable and controllable, which constitute the basis of our major technologies (lights, computers, cars, tv, hydroelectric generators, etc.), belong largely to the linear realm of cause-effect relationships and occupy a very small fraction of natural processes. Most of nature is nonlinear and complex, and hence unpredictable in its response to disturbance. We see this in the climatic consequences of global warming that we are now being forced to recognize, in the increasing incidence of epidemics of new as well as old pathogens, and in the dramatic extinction of species that we have unwittingly unleashed through habitat destruction and irresponsible farming practices. We live in a world that is well beyond our control, for reasons that have become clear from scientific study particularly with the development of chaos theory and complexity theory. Nevertheless, we must interact with this creative, unpredictable world; our lives depend on it. If we cannot predict, manipulate, and control complex systems, which include organisms, ecosystems (agricultural and natural), communities, organizations, and economies, what is the appropriate form of behavior in our relationship with them? Science itself has taken us to a new frontier of understanding, but it is not clear what form of praxis goes with it. We have lost the innocence of believing that we can always fix things with new technology, but the alternative way of being in the world is only slowly emerging into general consciousness. What can we say about this?

From Control to Responsible Participation The properties of complex systems are revealed not only through measurable quantities, such as the diversity of species in a habitat, but also through their qualities, which we perceive as the health or beauty of a landscape, be it farmland or a natural ecosystem. These are indicators of the condition of

342

Developmental Emergence

the whole, perceived directly through the complex relationships of the parts that underlie the organization of the whole. In the assessment of the state of health or disease of an individual, a practitioner pays attention not only to quantities such as pulse rate, blood pressure, and body temperature, but also to reports of pain, to the person’s complexion, posture, tone of voice, and other indicators of overall condition, which are qualities. Anyone involved with domestic or farm animals is aware of evaluations of their condition, often reported in terms such as distressed, nervous, uneasy, or lively, attentive, playful, and so on. Our descriptions of nature are generally more in terms of qualities than quantities. However, for perfectly understandable historical reasons, Western science has chosen to regard measurable quantities as the basis of reliable, objective knowledge of the world, while qualities are considered to be subjective, personally idiosyncratic evaluations that have no validity as indicators of real, objective conditions. There is, however, a nonquantitative aspect of scientific activity which is acknowledged to be an important element in discovering how observed parts constitute coherent wholes: the use of intuition, or noninferential knowing. This usually involves the sudden realization that the component aspects of natural processes cohere into self-consistent wholes by virtue of relationships, often expressed mathematically in science, that describe and explain the regularities of behavior observed in the system. Arriving at such insight is recognized to be a creative act that requires considerable preparation. Such groundwork, which includes paying close attention to the phenomena under investigation, extensive acquaintance with current ways of understanding them, and openness to new integrative relationships, comes through scientific apprenticeship. It is arrived at through participation in scientific procedures and direct experience of failure and success, as judged consensually by the scientific community. Intuition is cultivated through experience, not from books. The same can be said of the evaluation of any quality of a natural process: capacity to evaluate the health of a person or a habitat or an organization comes from direct experience of participating in the processes occurring in the person, habitat, or organization. Measurements may help this evaluation, but they cannot substitute for the direct perception of the quality of the whole that comes through attentive observation and participation. In simple systems, regularities expressed through mathematical relationships based on direct causal connections allow for prediction, and hence control. In such cases the observer can largely detach from the process and manipulate it. But in complex systems the causal connections are such that the results of

343

Brian C. Goodwin

interaction or manipulation are largely unpredictable. Discovering the consequences of interaction then requires continuous evaluation of the properties of the whole, seen through both quantitative and qualitative change, the latter often providing the more sensitive indicator. This involves both participation and observation, as described systematically in cooperative inquiry (Reason 1998; Reason and Goodwin 1999). The result is that the evaluation of qualities, which was not an explicit aspect of the Western scientific tradition but actually lived in the shadow as intuitive insight into the creative process of grasping the consistency of natural process, now comes into the light as a necessary, vital feature of living with complex systems. Our science of quantities is then extended to a science of qualities, forced upon us by the dialectics of science in its discovery of the intrinsically unpredictable properties of complex systems and by the limitations of prediction and control to provide appropriate practical procedures for relating to the natural systems on which the quality of our lives depend. Responsible behavior requires that we transform our scientific praxis from control to participation, which requires an explicit extension of scientific method from quantitative measurement to include qualitative evaluation.

A Science of Qualities and Intrinsic Values If we are to speak of a science of qualities, it is necessary to show that there are scientific procedures whereby qualitative evaluation of the holistic properties of complex systems can be repeatably and reliably carried out. The basis of scientific inquiry is a method of investigation practiced by a community which can lead to consensus about particular properties of that which is being examined. In Western science, the emphasis is on methods of measurement which result in agreed quantitative aspects of natural systems together with descriptions of logical relations between these quantities that capture the regularities of behavior observed (often cast in mathematical form). The result sought for is a description of a natural process (a falling body, transmission of electromagnetic waves, circulation of the blood) that is taken to provide insight into the real, intrinsic nature of the process being studied. Is there an equivalent procedure that allows for consensus to be reached about the qualities of natural processes, such as the experience of an animal in particular circumstances, the condition of health or well-being of an individual, or the quality of particular types of food or drink?

344

Developmental Emergence

As far as the last example is concerned, there are procedures regularly practiced by food and drink tasters that are taken to provide reliable indicators of quality. These can involve quite sophisticated analytical methods for determining the degree of consensus between different tasters. Wemelsfelder and colleagues applied an adaptation of this procedure to assess the degree of consensus between observers asked to evaluate the quality of experience of different pigs under similar circumstances: in a familiar large pen containing straw and a human with whom the pig was free to interact (Wemelsfelder et al. 2000). The method of ‘‘free choice profiling’’ allowed the observers to choose their own terms to express how the pig was feeling, using descriptors such as ‘‘playful,’’ ‘‘aggressive,’’ ‘‘nervous,’’ ‘‘laid-back,’’ and so on. The crucial question posed by the researchers was whether different observers evaluating the pigs independently would produce consensus of qualitative assessment. In their analysis they used a generally accepted mathematical procedure that looks for clustering (agreement) and then allows the terms in the cluster to be ordered (principal coordinate analysis) and examined for semantic consistency. The result was a striking degree of consensus between independent observers about the quality of experience of the different pigs. This demonstration of consensus provides us with the type of evidence required to claim that what is observed as the pigs’ quality of experience is a property of the pigs themselves, not an idiosyncratic projection of the observers’ feelings onto the pigs (Wemelsfelder 2001). In a similar manner, the claim for the ‘‘objective reality’’ of quantities in nature is arrived at by consensus between subjects using an agreed methodology. The criterion for the reality of qualities is the consensus in the evaluations of the observers. A foundation is thus established to recover qualities as reliable indicators of natural process, which is what common sense in any case asserts. We can indeed perceive the intrinsic quality of life experienced by other beings, which we can value for its own sake rather than how these beings serve us. The ability of living beings to experience the quality of their lives and our capacity to evaluate it result in the recognition that their lives have intrinsic value for them. This opens the door to a much more systematic use of intuitive insight, or direct knowing, in the assessment of conditions that affect quality of life, not just for humans but for all life on this planet. It also supports forms of practical action that depend on the Precautionary Principle in making decisions about the development or use of technologies. It is not necessary to have quantitative scientific evidence of risk in order to impose a moratorium on the use of a technology. It is sufficient to recognize a legiti-

345

Brian C. Goodwin

mate consensus on its potential danger to the quality of life of other beings and their complex interactions in ecosystems and the planet as a whole, and to require more extensive investigation into its consequences, both quantitative and qualitative. The most prominent current example of lively debate on these issues relates to an aspect of biotechnology: the use of transgenic animals and crops in agriculture and the marketing of food from genetically modified organisms (gmos). Public opinion in the United States, Canada, Europe, Japan, and many other countries favors labeling these foods so that consumers can make a choice, not only because it is a basic right but also because there is a strong intuitive sense that the technology may be dangerous to quality of life and health. The science of complexity supports this in making clear that transferring a gene or genes from one plant species to another using genetically designed vectors has unpredictable consequences concerning the composition of the food and the impact of the transgenic variety on the ecosystem into which it is released. Since there is no benefit to the consumer from this technology, and there is a risk of danger to health of humans and animals fed gm food, as well as to the health of ecosystems, the judgment of many is that this technology should not be used until much more is known about its consequences from control trials. There are many others who believe that there is already enough evidence of danger to call for a complete stop to biotechnology based on this particular form of genetic manipulation. There are alternatives that can be used to increase food production while maintaining or enhancing food quality. It is not my intention to argue the pros and cons of this debate, but simply to point to it as evidence of a new dialogue that has arisen from developments in science and technology that deal with complex living systems: organisms and ecosystems. People are using scientific principles that transcend the purely quantitative in judging these, and I have argued that these principles have a good foundation that connects with new developments within science itself that are forcing an extension of legitimate, reliable knowledge to the qualities of natural systems and their intrinsic values. Since knowledge constitutes a basis for appropriate action, the immediate consequence of this is to recognize that technologies and their consequences should not violate our knowledge of the intrinsic values of these natural systems. The result is that values reenter an extended science at a basic level. Responsible science necessarily involves examining the possible consequences of the applications of scientific knowledge, both quantitative and qualitative, and taking action (or refraining from action) in ways appropriate to this knowledge. Although

346

Developmental Emergence

cultivation of qualitative judgment should become a much more prominent aspect of scientific training, qualitative evaluations are not the prerogative of scientists but are shared by everyone affected by scientific decisions. Civil society should therefore have a much more active role to play in decisionmaking, as is being demanded by the public in a range of activities from genetically modified food to the World Trade Organization. People are necessarily involved in the process of responsible science. In conclusion, it is evident that the encounter between science and the living condition that was a major feature of the twentieth century is producing an enrichment of the content and the practice of science by the inclusion of qualities and the exercise of responsibility, both consequences of the study of complex systems and their emergent properties. This entails basic transformations in science education as well as in decision-making procedures about research and its applications. The century of the gene has taught us humility about the limitations of what we thought was the ‘‘secret of life’’ (Keller 2000). Discovering the further, deeper secrets of life could take us into a relationship with the other inhabitants of our planet, forming a society that celebrates qualities and intrinsic values and is based on responsible participation.

References Cole, B. J. 1991. Is animal behavior chaotic? Evidence from the activity of ants. Proc. Roy. Soc. Lond. B 244: 253–259. Goldstein, J. 1999. Emergence as a construct: history and issues. Emergence 1: 49– 72. Goodwin, B. C. 1994, 2001. How the Leopard Changed Its Spots: The Evolution of Complexity. London: Weidenfeld and Nicolson; Princeton: Princeton University Press. Höfer, T., Sherratt, J. A., and Maini, P. K. 1995. Cellular pattern formation during Dictyostelium aggregation. Physica D 85: 425–444. Kauffman, S. A. 2000. Investigations. Oxford: Oxford University Press. Keller, E. F. 2000. The Century of the Gene. Cambridge: Harvard University Press. Maturana, H. R., and Varela, F. J. 1998. The Tree of Knowledge: The Biological Roots of Human Understanding. Boston: Shambala Publications. Reason, P. 1998. Co-operative inquiry as a discipline of professional practice. J. Profess. Care 12: 419–436. Reason, P., and Goodwin, B. C. 1999. Toward a science of qualities in organisations: lessons from complexity theory and postmodern biology. Concepts Transform. 4: 281–317.

347

Brian C. Goodwin

Solé, R. V., and Goodwin, B. C. 2000. Signs of Life: How Complexity Pervades Biology. New York: Basic Books. Vasiev, B., Siegert, F., and Weijer, C. J. 1999. A hydrodynamic model for Dictyostelium discoideum mound formation. J. Theor. Biol. 194: 441–450. Wemelsfelder, F. 2001. The inside and outside aspects of consciousness: complementary approaches to the study of animal emotion. Anim. Welfare 10: S129– S139. Wemelsfelder, F., Hunter, E. A., Mendl, M. T., and Lawrence, A. B. 2000. The spontaneous qualitative assessment of behavioural expressions in pigs: first observation of a novel animal welfare measurement. Appl. Anim. Behav. Sci. 67: 193–215.

348

14 N OT H I N G L I K E A G E N E

jackie leach scully In this essay I will look at the ethical consequences of choosing models to describe gene action in the development and maintenance of an organism. What happens ethically, that is, when such models diffuse out of the scientific domains in which they were originally devised into public hands, and are then manipulated, used and misused, subverted, and otherwise dressed up in clothes more appropriate to the new tasks they are required to perform? I will argue that models of gene action can influence contemporary cultural attitudes to human identity, relationships, and social responsibilities as rarified conceptions about genes become integrated into everyday modes of thought; and that in doing so, these concepts inevitably affect our ethical behavior. ‘‘Nothing like a gene’’ is an ambiguous phrase, chosen deliberately to underline the elusiveness of the subject. If I say there is nothing like a gene, I might mean it in the sense that genes are unique, that there is absolutely nothing else like them. For both the practitioners of biology and the ordinary public, I think this is more or less how genes are currently understood. Whether in the mythopoeic phrases that are used repeatedly to describe the genome—the blueprint for life, the vision of the Grail (e.g., Gilbert 1992), the ultimate arbiter (Bains 1987: 7), the Book of Life (Bodmer and McKie 1995)— or implicitly in the sheer concentration of time, effort, and funding devoted to them, the message is clear that there is nothing comparable to genes in importance for understanding human behavior, and even human destiny. By saying there is nothing like a gene, I might also be claiming that there is, literally, nothing like a gene. That is, the reality of gene action in development is quite other than the picture of it we have come to hold. As other chapters in this book describe, a growing body of biological data suggests that genes are not what we thought they were—or as we might sometimes wish they are.

Jackie Leach Scully

Modeling Gene Action: Implications for Identity The discussions in other chapters of this book show that a reductionist molecular strategy—working, at least methodologically, from the premise that ‘‘life, at base, is the end result of the properties of the materials from which it is made, and that we can understand it by treating it as a collection of complicated molecules acting together rather like miniature clockwork’’ (Bains 1987: 6)—has been extraordinarily effective in the advance of modern biology. The methodological success of reductionism has concomitantly enhanced the credibility of a reductionist view of gene action as an inflexible, determining genetic program. In recent years this picture of how genes work has come under increasing criticism for its biological inadequacy (see other chapters in this volume), as well as for more ideological reasons (see, e.g., Hubbard and Ward 1983; Rose, Lewontin, and Kamin 1984; Lewontin 1991), and in parallel other ideas have been elaborated that can broadly be described as constructivist, systems, or process models (Griffiths and Neumann-Held 1999; Neumann-Held 1999). Dichotomizing the types of model like this is, of course, an oversimplification, but nevertheless it enables us to see how the overall properties of the two views will be reflected in the different constructs of human being they are used to produce. Deciding on the most valid or appropriate model of gene action is more than an arcane problem for molecular and developmental biologists. Scientific models are never neutral aids to thinking: they are heuristic tools that provide frameworks to make some things easier to think about and other things harder. Each model also carries its own load of unspoken associations, which are important because it is these associations as much as the frameworks themselves that inform the thinking of society beyond laboratories and science journals. It is because of this that the role we attribute to the chemical dna in the life history of an organism, and how we envisage the process of getting from genotype to phenotype, matter ethically as well as scientifically. Models of gene action are crucial not just for interpreting the biological chain of events that connects genome to organism, but also for cultural constructions of identity. The first use is about causal biological mechanism: if my elaboration as an embodied being with a specific set of characteristics derives in some way from the action of my genome, then developmental biologists will be interested in tracking the mechanisms involved in that process. In addition to an explanation of mechanism, however, there is also the use of genes in ontological analogy. Most people are not biologists, and when

350

Nothing Like a Gene

they use the discourse of genes they are likely to be doing something other than explaining biological phenomena. One of the most important activities human beings engage in is describing themselves and others, to themselves and to others. In this context, the narrative of genes will be used to help articulate people’s qualities or behaviors in terms of fantasies of where those identities originated and how they are modified by experience. Of course, these two uses of gene discourse, as description of a mechanism and as ontological analogy, are not as neatly separated as I have made out: biologists (must) use analogy and metaphor to explore their models of mechanism, while a grasp of the biologically credible mechanism makes some analogies and metaphors more culturally persuasive than others. The point I want to discuss further is that both mechanistic and ontological usages of different models have ethical (as well as social and political) consequences. The contemporary intellectual authority of science means that scientific narratives profoundly influence our cultural narratives of identity and selfhood. In this discussion, by identity I mean the sense each of us has of the self as a coherent, enduring, reasonably consistent collection of embodied characteristics, with a subjective sense of agency and the ability to identify with any number of externally defined categories (woman, working class, Asian, Labour voter, dog lover, etc.). I am not here primarily concerned with sociopolitical allegiance or with psychodynamic processes of identification, but with cultural notions about where people’s characteristics come from and how people show agency in the use they make of the characteristics they have. The crucial point for ethics is that the stories told about identity shape our concepts of what constitutes moral behavior between agents, or indeed what constitutes a moral agent at all. In other words, exactly how genes are implicated as a developmental resource will influence ideas of how simple or complex a phenomenon ‘‘identity’’ is considered to be, whether identity is amenable to reductionist analysis, what parts of human identity are altered in parallel with the manipulation of genetically associated characteristics, and whether a specific identity can ever be exactly reproduced. Up to now, ethicists have shown most interest in genetics at the point where the science is put into practice: and so the issues which have been extensively analyzed in recent years have tended to be ones that arise downstream of implementation, such as the (re)appearance of eugenic ideas, the problem of insurance discrimination, or justice in health resource allocation. But points of ethical, social, and political relevance appear at an earlier stage than this, when models are first constructed as aids to comprehending biological processes.

351

Jackie Leach Scully

The purest version of the deterministic model of gene action sees human characteristics as resulting in a linear and directly causal fashion from the presence in the genome of particular dna sequences. Identity (as an individual’s set of characteristics) is entirely or almost entirely generated in isolation from nongenetic factors, and neither communicates nor interacts with them. Identity is formed and can be expressed in one direction only—from the gene(s), which are or contain the necessary program of information, to the individual embodiment. The original version of the Watson-Crick central dogma—dna makes rna makes protein—is extended to something like dna makes rna makes protein makes person. In the classic molecular model, the genes are isolated from any influence of the organismic embodiment they have elaborated, or any effects of the embodiment’s interaction with her environment, because the flow of information is one-way. As with the original statement of the central dogma— which is now acknowledged to be biologically inadequate—nothing goes in the opposite direction: person does not make protein or nucleotide. The flow of information between genome and identity is therefore also one-way. Not only is the latter shaped (entirely or predominantly) by the genetic constitution, the identity which results can have no impact on the genes that are its point of origin. Because of this, identity is also predictable and reproducible irrespective of environment or context. Rather than arising from an interaction between various genetic and body-internal and body-external environmental resources, including relationship with other embodied agents, identity is a discrete component that can in some way be stored in these little nuggets of dna. It is perceivable by us, Dawkins’s ‘‘lumbering robots,’’ only when writ large in the form of our own or others’ physical embodiment, but it is ‘‘really’’ present long before that, in the genome. This is why some molecular biologists can look to the day when each person’s genome will be recorded on a cd-rom, and concur with Gilbert’s now notorious claim that one will be able to pull a cd out of one’s pocket and say: ‘‘Here’s a human being: it’s me!’’ (Gilbert 1992: 96; my italics). If we adhere to a strict deterministic program model of the action of genes in human development, in which the procedure by which individuals are made is a kind of ‘‘molecular Meccano’’ (Bains 1987: 7), or where genes ‘‘spell out a recipe . . . for building, growing and running a living human body,’’ 1 then our ideas about human identity are likely to tend toward certain conclusions. One is that, owing to the direct causal relationship between the two, by changing our genes we can change our identity. Moreover, since all

352

Nothing Like a Gene

other factors—internal or external environmental influences, for example, or individual decisions—are secondary to the genetic ones, then changing our genes is also the most effective way of changing our identity. Faith in this strong one-to-one mapping between genes and identity is what lies behind the reactions that have been aroused by the idea of human cloning. The argument that it is a violation of human dignity to attempt to reproduce an individual identity may be a legitimate one, but it works only if by replicating a specific genome we are in fact replicating a specific identity as well. In a similar way, the reservation that engineering of the human genome will irrevocably alter the essence of being human (whatever that might be) requires the essence, or some element which is absolutely needed for its expression, to be contained entirely or predominantly within the genome. Otherwise, genomic alterations would have correspondingly less impact. Evidence has been around for some time that the one genome = one end product model fails to explain all the behavior of genes and gene products. Nevertheless, the model remains explicit or implicit in much of the professional and, perhaps more important, popular biological literature. Even writers who agree that genes alone do not determine human disease or human behavior, and allow that environmental factors and free will play a role as well, may use unwittingly revealing language to convey the opposite message. Characteristics are said to be determined by genes but influenced by nongenetic factors. ‘‘The general colour of your skin is genetically determined . . . many characteristics may be influenced by the environment as well as by genes . . . altering environmental influences . . . can markedly influence the course of development’’ (Bains 1987: 13); and, ‘‘Each one of us has a special genetic recipe that makes us unique and which dictates many of the characteristics we take for granted.’’ 2

Analogy and Metaphor The diffusion of ideas about genetics from the laboratory into the public domain started in the 1970s and accelerated throughout the 1990s in parallel with an avalanche of published results in the science. A literature search of the quality United Kingdom newspapers produced only a handful of references to genes or genetics prior to 1994 that were not in scientific articles.3 By 1996 there had been an explosion in the use of genetic allegories and metaphors in other settings. A sample includes a theater critic asking if transposing the settings of the classics is the artistic equivalent of genetic engineering, a tele-

353

Jackie Leach Scully

vision review describing characters as ‘‘totally deficient in the gene which gives the English their celebrated reserve,’’ politicians grounding policy decisions in the fact that ‘‘there has always been a receptive [sic] gene in the American character that is isolationist’’ while in retaliation a political commentator could say that most politicians lack the compassion gene or carry a gene that makes them self-serving liars, and a fashion page asking the readers to ‘‘think of the designers as you would those mad genetic engineers mating the cardigan and the jacket.’’ 4 Popular science writing plays an interpretive role in this process, translating genetic language into forms that the general public finds palatable and accessible. A second step of interpretation is also involved as the readers of popular science books and reports interpret what they read or hear; and finally on top of this comes a level of interpretative subversion, as people use the gene motif for their own ends. As suggested earlier, everyone has a stake in making sense of their lives, in self-interpretation, and it is for this work that the gene concept is currently proving most useful. ‘‘Gene talk has entered the vernacular as an explanation for human behavior,’’ notes S. Lindee. ‘‘In supermarket tabloids and soap operas, in television programming, in women’s magazines and parenting advice books, genes appear to explain obesity, criminality, shyness, directional ability, intelligence, political leanings, and even preferred styles of dressing’’ (Lindee 1997). In many cases the references are ironic ones or are intended to be funny. The ‘‘gene for’’ locution is an economical and vivid way of talking about a defining characteristic: English reserve, isolationism, compassion. The writer ‘‘knows’’ that there is no gene for reserve as a national characteristic, and expects the readers to know this as well. But far from reducing the significance of its incorporation into popular writing, an ironic usage only underlines how firmly embedded in the cultural consciousness the gene idea now is. Its semiserious use relies on the reader having a sophisticated grasp of which bits are directly transferable and which bits are intended to provoke. And we have known since Freud, if not before, that joking plays an important role in containing unease and defending against things that are felt to be dangerous (Freud 1905: 236). As an explanation of human being, the gene idea is simple to grasp, all-embracing, and still profoundly weird: the use of jokes here only underlines the psychic disturbance of the irruption of a new source of beliefs about human identity into everyday thinking. These examples from the media also illustrate that the deterministic program model of gene action is currently the only one available for popular discourse on human nature. Although the transfer from professional science

354

Nothing Like a Gene

to popular discourse resembles what was done in the twentieth century with ideas from evolutionary theory, endocrinology, and psychoanalysis, the deterministic program model has crossed over with an ease that, to my mind, has not yet been satisfactorily accounted for. There has been no comparable migration by any version of the system or process models. In part this is undoubtedly because these models were developed much more recently and have not yet had time to follow the interpretation/incorporation path taken by the program model. An additional factor is that they lack the genetic program model’s conceptual simplicity; ‘‘one gene = one characteristic’’ taxes nobody’s logic, even if it does remain vague about what a gene actually is. I suggest also that the program model’s associations with linguistic metaphors of coding and text (dna as a code, the genome as the Book of Life, the analogies in popular science between sequence variation and typographical errors) have played a major part in its attraction. Our text-based culture makes Westerners very comfortable with the textual metaphor. Furthermore, most people use text as an instrument, not an object of reverence or even of academic study but a tool to be used, whether reading an airport paperback or the instructions on a packaged ready meal. On an everyday level, questions of reflexivity, interpretation, or authorship—all of which are more likely to be invoked by developmental systems or process models—are not to the fore. The textual metaphor, finally, also carries a reassuring message about the whole project of elucidating the meaning of the genome. The postEnlightenment understanding of the written word is as a means of illumination, not of obfuscation. If we think about the genome as a book, and of gene action as the reading of it, we can legitimately (in both a moral and a practical sense) expect the genome to yield its secrets, because that is what books do.

Agents Normal and Abnormal Ethical discourse depends on notions of identity. Traditional ethical analysis has given preference to universalizability (what is right for one person must be right for everyone else in relevantly similar circumstances) and impartiality as fundamental requirements for a just morality. Perhaps more questionably, ethics has also tended to assume that the best way to ensure both universalizability and impartiality is to eliminate the sources of particularity and bias that lurk in the specifics of identity—a kind of ethical reductionism, in which the identifying characteristics of agents A and B are pared down

3 55

Jackie Leach Scully

to the bare minimum. Equating universalizability and impartiality with anonymity in this way requires the agent to be reduced to her essential core, the moral equivalent of the idealized perfectly round and perfectly smooth balls used to teach elementary mechanics. It is this kind of agent that MacIntyre describes as an ‘‘unencumbered self,’’ transcending the snares of personal affinities or community loyalties (MacIntyre 1981). But in the interactions between agent A and agent B, it matters who agent A and B are; and who they are depends on the specifics of their lives, including their biographies and the conditions in which their interactions play out. Zygmunt Bauman notes that the agent who is completely anonymous and without identifying features cannot be an object of knowledge in any meaningful sense, because there is nothing ‘‘there’’ to know (Bauman 1993). In the absence of such knowledge, it is questionable whether any kind of relationship—including any moral relationship—can exist. What kind of rights or obligations can be claimed by a featureless ball? The response might be that the agent is not totally featureless, but in that case we have to ask which features are retained and who decides which features are morally relevant. Criticism along these lines has encouraged an increasing number of contemporary ethicists to advocate a turn toward taking particularity seriously as part of meaningful ethical description. Feminist ethics in particular has charged the allegedly minimalist moral agent of traditional ethics with being, in reality, a distinctly featured one. Rather than being truly featureless, the features it has are taken for granted as normative for persons, and therefore need not be made explicit: rationality, autonomy, and by the feminist analysis, gender. More generally, this analysis says that having credibility as a moral agent depends strongly on having the features which are taken to be normative for persons. In line with medicine’s vastly expanded technical capacities and biomedicine’s authority in the modern world as a provider of explanations of human development and health, biomedicine and genetics are increasingly seen as providing normative accounts of human ontology. Thus, as the number of gene loci associated with particular diseases or abnormalities (more or less convincingly) has expanded, the authority of nongenetic accounts of abnormality has correspondingly shrunk. This has profound consequences for the ways that normality and abnormality are conceptualized. The methodology that developmental biology uses to characterize normal development is an indirect one, in which abnormalities of development are first identified and then related to associated mutations at gene loci. For each locus, multiple mutations may be found. In this way an increasingly refined

356

Nothing Like a Gene

picture of the abnormal is assembled, but assembled in order to delineate the normal, by collecting instances in which it is absent. Similarly, the characterization of alleles associated with abnormal states termed disease or disability is a process which defines not abnormality (disease) but normality (health) in absentia. On reductionist accounts, this is the best (possibly the only) way of analyzing within a realistic time frame the numerous molecular components that make up a multicellular organism. Whether we agree with this methodological claim or not, the indirect approach to the analysis of normality skirts the possibly insurmountable problem of giving an unambiguous definition of what the normal, wild-type, or healthy organism might be. Instead, normality is recognized as the ‘‘speaking silence’’ of the abnormality discourse, defined by all the abnormalities that it is not. The model of identity derived from deterministic gene processes goes: genome a = identity a.

Analogously, genome n = identity n,

where n stands for normal. A number of consequences might follow from these statements. We might conclude, for example, that there is only one or a very few normal identities, because there is only one or a very few normal genomes—assuming the latter from references in popular and professional science to ‘‘the’’ human genome. (There is no reason why there should be only one or a very few of either normal genomes or normal identities. But the locution of ‘‘the human genome’’ reflects what seems to be a firmly rooted belief. ‘‘Normality’’ rarely occurs in the plural in genetic discourse.) From this we might then say that there is a definable set of abnormal states of being, which can be unambiguously identified from their corresponding genomes. If genome a = identity a, then we can transform identity a into identity b by changing the corresponding genome, genome a, into genome b, resulting in identity b. And finally, it also follows that if we wish to normalize an abnormal identity (identity ab) we can do this by transforming the abnormal genome ab into normal genome n. Irrespective of whether abnormal can ‘‘really’’ be transformed into normal through genetic manipulation like this, splitting genomes or identities into antagonistic categories comes at an ethical price. The division of phenomena into two analytical categories (normal versus abnormal, legitimate versus illegitimate, healthy versus sick, able versus disabled) in practice seems to lead inexorably to the preference of one over the other. The dichotomization

357

Jackie Leach Scully

of genomes and identities into normal and abnormal, together with the generally accepted premise that normality is more desirable than abnormality, entails the preference of the normal genotype and phenotype above the alternatives. So, although the concern has frequently been expressed that the practice of genetic screening—a technology which permits the selection, either pre- or postnatally, of ‘‘normal’’ genomes over ‘‘abnormal’’ ones—is morally wrong because it is a covert form of eugenics and may increase discrimination against disabled or otherwise anomalous people (Paul 1995; Rock 1996; Shakespeare 1998), there is a more subtle objection (Kaplan 1993; Shakespeare 1999) that arises long before any eugenic practices come into play. The model in which a person’s identity is generated in a deterministic fashion from a genetic program serves to reinforce the claim that normality is something that can be defined, in some way, for human beings, and moreover can be distinguished clearly and unambiguously from abnormality: because ‘‘the’’ normal genome can be distinguished clearly and unambiguously from abnormal ones, and because the model says that the genome gives rise in a programmatic way to a set of characteristics that constitute an identity. Normality no longer has to be assessed subjectively in terms of an individual’s ability to function in society, contribute to production, or form relationships with others. Normality and its opposite become ontological categories for which there exist known biological markers that can be run on a gel. I think it is significant for this discussion that of all categories with deep ontological meaning (gender, race, class, sexual orientation, and age), disability is the one distinguished by the sloppiness of its boundaries. It is not possible to change age, sexual preference, or even class overnight, and although there are now surgical and medical techniques for changing anatomical sex, complete reassignment of gender requires physical and psychosocial modifications that are lengthy and often less than satisfactory. But it is possible to be transported out of the category of normality and into the category of ‘‘disabled’’ in the time it takes to be hit by a car. I might add that gender reassignment is also usually voluntary, while the transition to a disabled embodiment is not often undertaken by choice.5 I have argued at greater length elsewhere (Scully 1998) that people with impairments pose a unique problem for the social order unlike that posed by other marginalized groups, not only because they occupy a liminal position between categories of illness and health but because they literally embody the instability of the line between ‘‘us’’ and ‘‘them.’’ For a disabled person to be accepted as normal means acknowledging that it is normal for at least

358

Nothing Like a Gene

some people to be disabled, and that would mean acknowledging that any ‘‘normal’’ person might experience disability. Sander L. Gilman has written that ‘‘it is the fear of collapse, the sense of dissolution, which contaminates the western image of all diseases. . . . [W]e project this fear onto the world in order to localize it and, indeed, to domesticate it. For once we locate it, the fear of our own dissolution is removed. Then it is not we who totter on the brink of collapse, but the other’’ (Gilman 1988:1). One way of dealing with this fear is to reinforce the reality of the boundary by naming a group called ‘‘the disabled.’’ I am not arguing here that doing so is wrong—there are genuine political and administrative advantages to grouping people (or people grouping themselves) according to some shared characteristic—but rather that the creation of any group, including the one called ‘‘the disabled,’’ is a device of representation rather than discovery. Much of the attraction of the deterministic genetic program lies in the opportunity it offers to concretize the frustratingly woolly concepts of normal and abnormal, and to replace their subjective evaluation with an apparently more objective approach that appears grounded in an authoritative biological reality. Arguably, the genetic program’s reinforcement of a normal/abnormal dichotomy has at least as much potential for undermining the social position of disabled people as any practice of genetic screening.

Nonprogram Models The deterministic program model of gene action regards the characteristics of an organism as the products of the genetic program. Note, however, that it makes no especially strong claim for the universality of purely genetic causation in the development of human health or behavior. The possibility that it applies only to some characteristics, with others the result of nongenetic or mixed influences, remains open. In practice, adherence to the program model tends to go along with a conviction that it applies to all or the majority of human qualities. If I turn to a model that considers human characteristics to be derivatives of an overwhelmingly powerful genetic program, then I am likely to think of my identity as predetermined by the invariant action of the program in very early developmental processes and, at least after developmental maturity, as fixed and stable. Systems or process theories of gene action see the organism less as a passive product of instructions encoded in dna than as actively and interactively engaged in its own development. The interaction is between parts and pro-

359

Jackie Leach Scully

cesses of the organism and parts and processes of the environment in which it finds itself. The aetiological significance of the genome alone is reduced, so that it becomes only one of a ‘‘complex of factors . . . related to the cellular morphology, to the dynamics of biochemical transitions, to external circumstances, to previous developmental history of the system or to dna sequences available’’ (Rehmann-Sutter 2002: 46).6 Models that relate gene expression to the organism as a whole, to the processes of organismic development and maintenance, and to the specific environment in which development takes place are needed to explain some pieces of biological evidence. Beyond this description of mechanism, when used as analogies for ontology as described earlier they offer an image of a more radically contingent identity emerging from a number of potential pathways and histories of interaction, and fully interpretable only with a knowledge of the conditions and relationships that were involved (Rehmann-Sutter 1999). The picture of the individual that is extracted from this differs significantly from the one produced by the genetic program metaphor. For one thing, it is relational: an individual self does not simply fall out of the dna sequences; it is generated (and keeps on being generated) in the relationship between genes and other elements and processes, psychosocial and political ones as well as intra- and intercellular. The analogy allows for these factors to be variably effective under different circumstances. It is possible to say, for example, that this agent’s selfhood emerges predominantly from her embodied experience of her biological materiality, while for another the political world in which he exists has been more significant in shaping his sense of self. It allows for flexibility and change over time. An example might be the apparently heterosexual man who leaves wife and family for another man at the age of fifty. Whatever the ‘‘real’’ physiological, emotional, or social factors involved, drawing on a systems or process model to think about ontology allows this man’s biography to be understood as a narrative of the flexibility inherent in the dynamic interaction between biological, psychic, and sociohistorical givens. It is difficult to make use of a program metaphor that places identity, including sexual identity, as stably set early in development without also having to interpret this man’s trajectory as a developmental error: either he was always ‘‘really’’ gay but didn’t realize it, or else he is ‘‘really’’ straight and is making a big mistake now. Thus program or systems models are each respectively more conducive to certain ways of thinking about human identity. If identity is perceived as a fixed state, determined by very early developmental processes and with the end result of these processes clearly classifiable into normal and abnormal,

360

Nothing Like a Gene

events and experiences will have a different significance than if identity is seen as a flexible, interactive, and dynamic process. To take a different example, understanding heritable forms of deafness solely under their description as medical conditions resulting from abnormalities in the genome places them in a different moral context than if deafness, whatever the aetiology of the hearing impairment, is understood as locating a person within a linguistic minority identity for reasons to do with culture and history as much as dna sequence. Looked at that way, the phenomenon ‘‘not being able to hear’’ has a very different ontological significance, especially in issues of illness and cure (Scully 2003). But in addition to this, each model favors a certain type of inference about concepts like personal responsibility, self-determination, and the malleability of human nature. I do not intend to go into these possible inferences here, not least because much of the ethical debate over the new genetic techniques has been about just these consequences of genetic knowledge (usually understood in terms of the program paradigm) for cherished ideas about free will or the value of social engineering. It is worth noting in passing, however, that the genetic program model slips rather comfortably into political frameworks that reconstruct social problems as nominative pathologies (e.g., the social phenomenon of drug abuse exists because individuals have genes for becoming addicted to psychoactive substances) that are most effectively addressed at the level of an individual biology (the problem of drug abuse will ultimately be solved by biomedical interventions that alter metabolism or receptor function).

Choosing a View I have argued that the conceptual framework explicitly offered by different developmental models, and the implicit assumptions that accompany them, are readily adopted by popular discourses about human beings or human nature, even though the job they do in the cultural domain is ontological rather than mechanistic. In this area, models are adopted not on the basis of their power to describe biological phenomena but on their ability to provide accessible, collectively recognized tools of self-description. Whether we find a theory of human selfhood grounded in genetic reductionism more acceptable than, say, Marxism therefore says something about how we (want to) understand ourselves. Opting for a particular vocabulary to describe how I came to be and how I continue to live is an act of moral

361

Jackie Leach Scully

self-representation with necessary effects on the contours of my ethical beliefs about human behavior, and also on the acting out of these beliefs in reallife situations of personal, political, and social consequence. This important moral decision is neither entirely conscious nor free, because it remains contingent on the models that are culturally available. In this light, expanding the explanatory repertoire becomes a moral responsibility placed on those who have the authority or ability to influence the passage of scientific concepts into the public domain. How many different models are needed, and the problems of pluralism and relativism that might emerge, are questions beyond the scope of this essay. Within the community of scientists and science communicators there tends to be a high degree of pessimism about the public understanding of science and—for those who find deterministic gene program models inadequate— concern about the apparent enthusiasm of the public for oversimplified genetic explanations. The key deficiency at the moment is the lack of any popular form of a process or systems model, in contrast to the ready availability of program ones. Given that systems or process models are proving necessary to biological explanation, at least in certain disciplines, it is conceivable that the dynamic, processual faces of nature will become a strong new paradigm in science. In that case they will eventually make their way in some form into the public consciousness. Driven by the need to provide meaningful conceptualizations of identity and behavior rather than to explain biology, mutant forms of these developmental models are likely to emerge. From a biologist’s perspective, they might be as little like ‘‘real’’ genes as the model of a genetic program; but their power lies in providing a conceptual framework that makes interaction, responsiveness, and contingency central to the discourse of identity, and ultimately to moral agency and evaluation.

Notes 1 Taken from the brochure of the Centre for Life, Newcastle, United Kingdom, 1999. 2 Brochure of the Centre for Life, Newcastle, 1999, p. 8. 3 At that time these were the Guardian, Independent, Times, and Daily Telegraph. 4 With regard to ‘‘receptive,’’ the writer may have been misusing the phrase ‘‘recessive gene.’’ The fashion analogy is from ‘‘Prepare to split your suits,’’ Sunday Independent, 28 February 1999, p. 8.

362

Nothing Like a Gene

The latter is not invariably true; see Elliott 2000 on elective amputees. Rehmann-Sutter (2002: 46) describes ‘‘the body [as] the author of its genetic information.’’ To my mind this ascribes too much purpose to what is going on here (authors have a goal in mind when they write things). I would like to suggest, without further elaboration here, that the analogy of musical improvisation captures rather better the collaborative, responsive, and relational nature of the interaction between organism and developmental resources, including the genome. 5

6

References Bains, W. 1987. Genetic Engineering for Almost Everybody. Harmondsworth: Penguin. Bauman, Z. 1993. Postmodern Ethics. Oxford: Blackwell. Bodmer, W., and McKie, R. 1995. The Book of Man: The Human Genome Project and the Quest to Discover Our Genetic Heritage. Oxford: Oxford University Press. Elliott, C. 2000. A new way to be mad. Atlantic Monthly 286: 72–84. Freud, S. 1905. Jokes and Their Relation to the Unconscious. Standard Edition of the Complete Psychological Works, trans. and ed. J. Strachey et al. London: Hogarth Press, 1966. Gilbert, W. 1992. A vision of the Grail. In: Kevles and Hood (eds.), The Code of Codes. Cambridge: Harvard University Press. Gilman, S. L. 1988. Disease and Representation: Images of Illness from Madness to aids. New York: Cornell University Press. Griffiths, P. E., and Neumann-Held, E. M. 1999. The many faces of the gene. BioScience 49: 656–662. Hubbard, R., and Ward, E. 1993. Exploding the Gene Myth. Boston: Beacon Press. Kaplan, D. 1993. Prenatal screening and its impact on persons with disabilities. Clin. Obstet. Gynaecol. 36: 605–612. Lewontin, R. C. 1991. The Doctrine of dna. Harmondsworth: Penguin. Lindee, S. 1997. The cultural powers of the gene—identity, destiny and the social meaning of heredity. In: J. Wirz and E. T. Lammerts van Bueren (eds.), The Future of dna (pp. 24–34). Dordrecht: Kluwer. MacIntyre, A. 1981. After Virtue: A Study in Moral Theory. London: Duckworth. Neumann-Held, E. M. 1999. The gene is dead—long live the gene: conceptualizing genes the constructionist way. In: P. Koslowski (ed.), Sociobiology and Bioeconomics: The Theory of Evolution in Biological and Economic Theory (pp. 105–137). Berlin: Springer-Verlag. Paul, D. E. 1995. Eugenic origins of medical genetics. In: The Politics of Heredity:

363

Jackie Leach Scully

Essays on Eugenics, Biomedicine and the Nature-Nurture Debate. Albany: State University of New York Press. Rehmann-Sutter, C. 1999. Contextual bioethics. Perspekt. Philos. 25: 315–338. Rehmann-Sutter, C. 2002. Genetics, embodiment and identity. In: A. Grunwald, M. Gutmann, and E. M. Neumann-Held (eds.), On Human Nature. Berlin: Springer. Rock, P. J. 1996. Eugenics and euthanasia: a cause for concern for disabled people, particularly disabled women. Disabil. Society 15: 121–128. Rose, S., Lewontin, R. C., and Kamin, L. J. 1984. Not in Our Genes. New York: Pantheon Books. Scully, J. L. 1998. When embodiment isn’t good. Theol. Sexual. 9: 10–28. Scully, J. L. 2003. Disabled embodiment and an ethic of care. In: D. Diniz (ed.), Bioética e Genética. Brasilia/São Pãolo: LettresLivres/Loyola. Shakespeare, T. 1998. Choices and rights: eugenics, genetics and disability equality. Disabil. Society 13: 665–681. Shakespeare, T. 1999. Losing the plot? Medical and activist discourse of contemporary genetics and disability. Soc. Health Illness 21: 669–688.

364

CONTRIBUTORS

thomas r. bürglin, ph.d. Professor Bürglin studied biology and obtained his Ph.D. at the University of Basel. He held postdoctoral fellowships at Harvard Medical School and Massachusetts General Hospital, Boston; was a start Fellow at the Biozentrum, University of Basel; and is currently an associate professor at the Karolinska Institutet, Stockholm, Sweden. His research interests include development of Caenorhabditis elegans, Homeobox genes, and Hedgehog-related genes. brian c. goodwin, ph.d. Professor Goodwin was born in 1931 in Canada. He studied biology in Canada, then took a mathematics degree at Oxford and a Ph.D. involving biology and mathematics at Edinburgh University. He has held research and teaching positions at mit, the University of Sussex, and the Open University, UK, where he was a professor of biology. He was associated with the Santa Fe Institute for several years. He now teaches holistic science at Schumacher College in England. His research interests include the science of complexity and development of a science of qualities that can address issues of health and quality of life in diverse areas, contributing to sustainable lifestyles. His publications include How the Leopard Changed Its Spots: The Evolution of Complexity (1994). james griesemer, ph.d. Professor Griesemer has been a fellow of the Wissenschaftskolleg zu Berlin, a guest of the Rektor at Collegium Budapest, and a guest of the Max Planck Institut für Wissenschaftsgeschichte in Berlin and the Konrad Lorenz Institute in Altenberg. Currently he is a professor and chair of philosophy, a member of Science and Technology Studies, and in the Center for Population Biology at the University of California, Davis. In addition to occasional empirical research in population biology, he studies genetic, developmental, ecological, and evolutionary theories and their interrelations; visual representation and the use of diagrams in formal problem-solving in biology; and the history of research organization in biological laboratories and natural history museums. Professor Griesemer is currently completing a book to be titled Reproduction and the Evolutionary Process.

paul e. griffiths, ph.d. Previously director of the Unit for History and Philosophy of Science at the University of Sydney, Australia, Professor Griffiths is now a professor of history and philosophy of science at the University of Pittsburgh. His publications include Sex and Death: An Introduction to Philosophy of Biology (coauthored with Kim Sterelny, 1999) and Cycles of Contingency: Developmental Systems and Evolution (coedited with Susan Oyama and Russell D. Gray, 2001). jesper hoffmeyer, ph.d. Professor Hoffmeyer studied genetic and biochemical regulation of purine nucleoside biosynthesis during the 1970s. He began

Contributors work on the history of science and technology and gradually moved on to theoretical biology, establishing the Biosemiotics Group at the Institute of Molecular Biology in 1988. He is currently an associate professor in the Department of Biological Chemistry, University of Copenhagen. He is a member of the board for the Centre for Ethics and Law, University of Copenhagen, and a member of the International Committee of Jakob von Uexküll Centre, Tartu University, Estonia. In 2000 he was named a Thomas A. Sebeok Fellow by the Semiotic Society of America. His most recent publication is ‘‘Baldwin and biosemiotics: what intelligence is for’’ (coauthored with Kalevi Kull, in: B. Weber and D. Depew [eds.], Evolution and Learning. The Baldwin Effect Reconsidered, 2003).

evelyn fox keller, received her Ph.D. in theoretical physics at Harvard University, worked for a number of years at the interface of physics and biology, and is now a professor of history and philosophy of science in the Program in Science, Technology, and Society at mit. She is the author of many articles and books, including A Feeling for the Organism: The Life and Work of Barbara McClintock; Reflections on Gender and Science; Secrets of Life, Secrets of Death: Essays on Language, Gender and Science; Refiguring Life: Metaphors of Twentieth Century Biology; The Century of the Gene; and Making Sense of Life: Explaining Biological Development with Models, Metaphors, and Machines. gerd b. müller, m.d., ph.d., is a professor in the Department of Theoretical Biology at the University of Vienna and chair of the Konrad Lorenz Institute for Evolution and Cognition Research at Altenberg, Austria. His primary scientific interest is the relationship between development and evolution in the generation of organismal form. This work includes theoretical and experimental studies of the genetic and epigenetic determinants of structural organization and involves the development of computational tools for the three-dimensional representation and analysis of gene expression during embryonic development. Recent coedited books include Konrad Lorenz und seine verhaltensbiologischen Konzepte (2001), Origination of Organismal Form (2003), and Environment, Development, and Evolution (2003). eva m. neumann-held, dr. rer. nat., dipl. biol., received a Ph.D. in biology, followed by four years of postdoctoral training and research as molecular biologist in Denmark and the United States. In Germany, she studied philosophy and participated in several biophilosophical projects, among them ‘‘Chances and Limitations of an ‘Evolutionary Ethics’’’ and ‘‘Cognition, Brain, and Neuronal Networks,’’ both at the Ruhr-Universität Bochum; ‘‘Genome and Organisms: Philosophical Interpretations of Development Biology,’’ sponsored by the University of Basel; and ‘‘Xenotransplantation of Cells, Tissues, or Organs’’ at the Europäische Akademie Bad Neuenahr-Ahrweiler GmbH. As a fellow at the Kulturwissenschaftliches Institut Essen (kwi, Germany) she was a member of the study group ‘‘What Makes a Life Form Human? On the Normative and Empirical Challenges of Cultural Sciences through Biological Naturalism.’’ Currently, she is the coordinator of the research group ‘‘What Is a Human Being? Culture—Language—Nature,’’ a joint project of the University of Dortmund and the kwi. In this group she participates in the projects ‘‘Determinism and Plasticity—How Free Are We in Our Personal De-

366

Contributors velopment?’’ and ‘‘Bio-social Inheritance.’’ Her publications include ‘‘‘Organisms’ —Historical and Philosophical Issues’’ (with M. Gutmann and C. Rehmann-Sutter, guest editors of Theory in Bioscience, vol. 119, 2000), ‘‘Let’s Talk about Genes’’ (in: Cycles of Contingencies, ed. S. Oyama, P. Griffiths, and R. Gray, 2001), and On Human Nature (coedited with A. Grunwald and M. Gutmann, 2002).

stuart a. newman, ph.d. Professor Newman has been an inserm Fellow at the Pasteur Institute, Paris; a Fogarty Senior International Fellow at Monash University, Australia; and a visiting scientist at the University of Paris-Sud, the French Atomic Energy Center-Saclay, the Indian Institute of Science in Bangalore, and the University of Tokyo. He is a professor of cell biology and anatomy at New York Medical College, Valhalla, New York, and a fellow of the Institute on Biotechnology and the Human Future (Chicago). His research interests include cell differentiation, theory of biochemical networks and cell pattern formation, protein folding and assembly, mechanisms of morphological evolution, and the cultural background and social implications of biological research. He is coeditor (with Gerd B. Müller) of Origination of Organismal Form: Beyond the Gene in Developmental and Evolutionary Biology (2003) and coauthor (with Gabor Forgacs) of Biological Physics of the Developing Embryo (2005). susan oyama, ph.d. Professor Oyama trained at Harvard University’s Social Relations Department. She has written widely on the nature/nurture opposition and on the concepts of development, evolution, and genetic information. Her publications include her essay collection Evolution’s Eye: A Systems View of the BiologyCulture Divide (2000); an expanded edition of The Ontogeny of Information (2000), considered by many to be the foundational text of the ‘‘developmental systems’’ perspective; and Cycles of Contingency, a volume on developmental systems by scholars from many fields (coedited with P. Griffiths and R. Gray, 2001). Currently, she is professor emerita at the John Jay College of Criminal Justice and at the Graduate School and University Center of the City University of New York. christoph rehmann-sutter. After receiving academic training in molecular biology, he studied philosophy and sociology in Basel, Freiburg, and Darmstadt. In 1997–1998 he was a research fellow at the University of California, Berkeley. Currently he is assistant professor and head of the Unit of Ethics in Biosciences, University of Basel, Switzerland. Since 2001 he has been president of the Swiss National Advisory Commission on Biomedical Ethics. His research interests include bioethics, natural philosophy, philosophical and ethical issues of gene therapy and genetic testing. His books include Ethik und Gentherapie (coedited with Hansjakob Müller, 2d enl. ed., 2003), Leben beschreiben (1996), Partizipative Risikopolitik (with Adrian Vatter and Hansjörg Seiler, 1998), Zwischen den Molekülen: Beiträge zur Philosophie der Genetik (2005), and Bioethics in Cultural Contexts: Reflections on Methods and Finitude (coedited with Marcus Düwell and Dietmar Mieth, forthcoming). sahotra sarkar, ph.d. Professor Sarkar received his Ph.D. in philosophy from the University of Chicago in 1989 and is currently a professor in the Department of Philosophy and Section of Integrative Biology at the University of Texas,

367

Contributors Austin. His research interests include philosophy of science, conservation biology, history of science, evolution, and ecology. His publications include Genetics and Reductionism (1998), J. B. S. Haldane—A Scientific Biography (forthcoming from Oxford), the six-volume Science and Philosophy in the Twentieth Century: Basic Works of Logical Empiricism (ed., 1996), and the two-volume The Philosophy of Science: An Encyclopedia (ed., 2003).

jackie leach scully, ph.d., was trained as a molecular biologist and has worked in the fields of cancer and neurobiology research, the public understanding of science, and bioethics. She is currently at the Office of Ethics in the Biosciences, Institute for the History and Epistemology of Medicine, University of Basel. Her research interests include empirical approaches to ethical evaluation, embodiment and disability, and feminist ethics. Recent publications include Quaker Approaches to Moral Issues in Genetics (2002), ‘‘A postmodern disorder: moral encounters with molecular models of disability’’ (in: Disability/Postmodernity: Embodying Disability Theory, 2002), and ‘‘Disability: stigma and discrimination’’ (in: Encylopaedia of the Human Genome, 2003). gerry webster, ph.d., b.sc., studied at the University of London. He was a lecturer (1966–2000) and the director of the Human Sciences Degree Programme (1981–2000) at the University of Sussex, and is now retired. Dr. Webster’s research interests include philosophical problems in the biological and human sciences. His most significant recent publication is Form and Transformation: Generative and Relational Principles in Biology (coauthored with Brian Goodwin, 1996).

ulrich wolf, professor em. dr. rer. nat. Professor Wolf studied biology and physical anthropology at the Universities of Tübingen and Munich (1952–1961); did postgraduate work at the University of Wisconsin, Madison; was a lecturer at the University of Freiburg (1969); and then was professor and director of the Institute of Human Genetics and Anthropology, University of Freiburg (1972– 2001). In 1975 he was president of the European Society of Human Genetics. He is a member of the Deutsche Akademie der Naturforscher Leopoldina, Oesterreichische Akademie der Wissenschaften, and Accademia di Science e Lettere Istituto Lombardo Milan. His main research areas are experimental and clinical cytogenetics, genetic mechanisms of vertebrate evolution, genetics of sexual development, genotype-phenotype relationship, and the organism as an autopoetic system.

368

INDEX

Abnormal, as term, 357 Abstraction, and digital codes, 168–169 Acquired behaviors, vs. innate, 178 Adhesivity, 50–51, 55–57 Agassiz, L., 105 Alberch, P., 140 Alberts, B., 143 Analog code, vs. digital code, 168 Analogy, in science, 353–355. See also Metaphors Analysis, units of, 278–279 Ancient metazoans, 40, 47–48 Approaches, types of, 239–240. See also specific types of approaches Apter, Michael, 298–299, 303 Aristotle, 222, 313, 315–317, 319 Artificial Life, 304 Asakara, S., 119 Atomism, 100, 109, 118 Avery, Oswald, 78, 249 Bateson, Gregory, 168 Bateson, William, 78, 127 Beadle, G. W., 249–250 Behaviors, innate and acquired, 178 Bernard, Claude, 105 Biosemiotics, 8, 157–160. See also Semiosis Blum, Deborah, 272 Bodies, 290–307; boundaries of, 8 (see also Membranes); discussing objectively, 327–329. See also Form; Organisms Body plans, 43–53 Body segmentation, 48–49

Bonner, James, 298–304 Borges, J. L., 110 Boundaries, biological, 272–287; and causation, 274–276; delineating, 281– 283; of developmental systems, 280. See also Membranes Brady, Ron, 132 Brandon, Robert, 201–202 Brenner, Sydney, 4, 15, 281, 305 Bürglin, Thomas, 8, 238 Buss, Leo, 92 Caenorhabditis elegans: cell types in, 17–18; dna sequences in, 21–22; gene disruptions in, 28–29; genome of, 18–21, 88; homeobox gene in, 16, 29–32; life cycle of, 16; methods for studying, 26–28; as a model system, 15–36; molecular genetics of development of, 8; mutations in, 26; nervous system of, 18, 29–30, 34; neurons on, 31–32; rna in, 19, 28 Cambrian explosion, 51–52 Campbell, Philip, 1 Carroll, S. B., 143 Cassirer, E., 8, 108, 132, 135, 138–139, 145–146 Causality: attributing to genes, 8, 238– 262; and biological boundaries, 274–276; concept of, 105; determinants of, 39, 280; Neo-Darwinian interpretation of, 41; as term, 240; ways of thinking about, 177 Cell adhesion, 44–46 Cell division, 22–23, 48

Index Cell model, sequestered modular template (smt) of, 89–90 Cell polarity, and lumen formation, 46–48 Cells: differentiation of, 22–23; subpopulations of, 50–51; as units of development, 294 Cellular programs: developmental, 281–283; and hybrid disciplines, 280–283 Cellular slime mold, 338–339 ‘‘Central directing agency,’’ 104–107, 113, 118, 122 Chance, concept of, 152–153 Chaos dynamics, 154 Chetverikov, S., 81 Choice, concept of, 329–331 Classical genetics, 239–240, 244–248, 250 Code-duality, 166–169 Codes and coding: and enzymes, 257– 259; historical references to, 251; as term, 2, 251; uses of, 260–261; vs. reading, 259–260. See also Genetic code Cognitive science, language in, 153–154 Cole, B. J., 341 Compartment formation, and differential adhesion, 44–46 Complex systems, 338–344 Complexity theory, 338 Complexity (Waldrop), 304 Computer metaphors, 297–305 Computer modeling, 33–34, 303–305 Conceptual system: derived from Darwinism, 100–101; ‘‘functional’’ vs. ‘‘structural,’’ 102–103 (see also Functionalism; Structuralism) Constructivist interactionism, 138, 272–274 Content, in natural theology, 108–109 Contingency, and inherency, 42 Control, and responsible science, 342–344

370

Copying, 218 Crick, F., 78–79, 250 Critique of Judgment (Kant), 105, 136 Cuvier, G., 101–103, 106, 109 Darwin, C., 104, 107–108 Darwinism, 107–111; form in, 107– 110; inheritance in, 111. See also Neo-Darwinian interpretation Dawkins, R., 205–206, 212, 214, 217, 222, 227, 241, 352 Deacon, Terrance, 153 De Anima (Aristotle), 317 Dennett, Daniel, 283 Dennis, Carina, 1 Dependent Gene, The (Moore), 193 Determinism, genetic, 175–195; defined, 177–182; and developmental psychobiology, 193–194; and intentional information, 187–189; model of, 352– 353, 354–355; reasons for persistence of, 175–177, 181–182, 193–194 Development: and ‘‘central directing agency,’’ 104–107; and embryology, 83; evocators of, 79–83; genes role in, 340; historical notes on, 77–92; how form is brought about in, 145–148; mechanisms by which genes control, 4–5; molecular genetics of, 8; perspectives on, 213–221, 225–226, 313–331; ‘‘responsible science’’ of, 9; role of dna in, 1; and semiosis, 152– 170; units of, 294. See also specific approaches to development Developmental, as term, 306–307 Developmental biology: alternative interpretation of, 333; ethics in, 355–359; and genome analysis, 15–36 Developmental constraint, 142 Developmental control genes, 23–24. See also Homeobox genes Developmental emergence, 337–347 Developmental genetics, history of, 85–86

Index Developmental integration, 59–61 Developmental mechanisms, evolution of, 38–63, 219–220 Developmental outcomes, 252–256 Developmental plasticity, 40 Developmental process: applying poiesis/praxis distinction to, 314; different interpretative approaches to genes in, 8–9; genes as causal power in, 251–253; and morphomes, 321–323; study of, 3 Developmental psychobiology, 191–194 Developmental Psychobiology (Michel and Moore), 191 Developmental systems approach (dsa), 179–181; compared with other perspectives, 226–228; constraints of, 278–279; emphasis in, 275; and individualism, 286–287; information metaphor in, 8; philosophical advantage of, 326–327; on processstructuralists’ view, 211–212; research agenda of, 190–192; as theory, 208– 209, 282–283. See also Oyama, Susan Developmentalists, vs. gene centrists, 1–2 Differential adhesion, and compartment formation, 44–46 Differentiation, and induction interactions, 53–54 Diffusion, and formation of spatial gradients, 43–44 Digital codes: and abstraction, 168–169; system of, 167–168 dna: biochemical accessibility of, 248–252; decipherment of structure of, 78–79; heredity as aspect of, 1; information as term in, 154– 155; as resource, 77–92; role of, 1–3; significance of, 1 dna sequences: and classical genes, 245–248; and genes, distinguished, 253–254 Dobzhansky, T., 62

Driesch, Hans, 102, 118, 124–126, 132 Drosophila melanogaster: alleles in, 83; differential adhesion in, 46; genome of, 88; historical research on, 77–78; homeobox genes in, 26; mutations in, 23, 77–78, 81; segment formation in, 49 Dunn, L. C., 115 Eckstein, H., 42 Eco-devo (ecology and developmental biology), 284–286 Ehrlich, Paul, 79 Embryology, and genetics, 3–4 Embryonic induction, 53 Embryonic organizer, 82–83, 106 Embryos, developmental mechanisms in, 53–54 Emmeche, Claus, 166, 167 Empiricist positivist approach, 109–110 Entelecheia, 317–319 Environmental influences, 35–36, 39–40 Enzymes, and coding, 257–259 Epigenetic, as term, 41 Epigenetic approach, vs. organismic approach, 8 Epigenetic factors, in evolution, 42 Epigenetic integration, 61 Epigenetic interactions, 52–59 Epigenetic mechanisms, 41–42 ‘‘Epistemic cut position,’’ 156 Epithelial tissues, 46–47, 54–58 Erkenntnisproblem in der Philosophie und Wissenschaft der neueren Zeit, Das (Cassirer), 136 Escherichia coli, 16 Ethics, 349–362. See also Responsible science Evo-devo (evolutionary developmental biology), 284–286 Evolution: account of units of, 216–217; epigenetic factors in, 42; inherency in, 38–63; problem of form in, 130; semiotic interactions in, 159–160; as

371

Index Evolution (continued ) three-step process, 201–202. See also specific perspectives Evolutionary developmental biology (evo-devo), 284–286 Evolutionary genes, 241–244 Evolutionary paradigm: critique of, 117–124; defined, 114–115; and molecular biology, 116–117; and Weismann, 111–117 Evolutionary systems, 152–153 Evolutionary transitions, 213–214 Excitability, and segmentation, 48–50 ‘‘Excitable medium,’’ described, 339 Expressivity, as term, 82 Fankhauser, G., 144 Form: assumed by metazoans, 63; concept of, 107; correlation with genotype, 38–39; in Darwinism, 108–110; how brought about in development, 145–146; in natural theology, 108–109; and Newtonian mechanics, 128; in structuralism, 130; ‘‘system’’ approach to, 102; and theory of evolution, 130; Weismann on, 113. See also Bodies; Organisms; Segmentation; Structures Formalization, and realization, 152–154 Form-theoretical considerations, 145– 148 Fox Keller, Evelyn, 8, 145, 249, 251, 280–283, 286 Functionalism, 204–208, 323–324; as preferred paradigm, 314, 323–324; shared assumptions with structuralism, 221; subjects in, 221–222. See also Structuralism Galileo, 127 Galis, F., 60 Gamov, G., 84, 250 Garcia-Bellida, A., 303 Gene: extension of concept of, 82;

372

informational, 175–195; understandings of the term, 62, 141, 242–243, 305; as unit of development, 294 Gene action, modeling, 350–353 Gene centric approach, 1–2, 7–9, 137, 290 GeneChips, 32–33 Gene function, studying, 28–29 Genes: causality attributed to, 5–6, 8, 177, 238–262; as determinants, 77– 92; and dna sequences, 245–248, 253–254; evolutionary, 241–244; how researchers study, 15–36; as intentional, 187–188; models to describe, 349–362; regulatory, 4; role in development, 340; and semiosis, 152–170; structural, 4 Genetic, as term, 306–307 Genetic approaches, 239–240 Genetic code: and biochemical accessibility of dna, 248–252; existence of, 254–262; interpretations of, 8; notion of, 240; system of, 167. See also Codes and coding Genetic determinism. See Determinism, genetic Genetic information, demystifying, 169–170 Genetic processes, and the origin of body plans, 43–52 Genetic programs, 38; beyond, 325–327; and computer metaphors, 299– 305; explanatory logic of, 299–303; functioning of, 4; history of, 305– 306; idea of, 314; introduction of, 4; meaning of, 295–296; reason for persistence of, 306; as term, 325–326 Genetic screening, 357–358 Genetically modified food, debate over, 346–347 Genetics: cultural meanings of, 7; development and, as research strategy, 4; and embryology, 3–4;

Index from evolutionary process perspective, 199–229; historical notes on, 77–92; and identity, 7, 9; perspectives on, 201–208; public’s understanding of, 6–7, 188–189, 192–193; as term, 305 Gennes, Pierre-Gilles de, 39 Genome: as blueprint, 84–85; monistic conceptualization of, 129; relationship to organism, 325–326; in structuralism, 129 Genome analysis, and developmental biology, 15–36 Genotype, 38–39, 141 Geoffroy (Étienne Geoffroy St. Hilaire), 102–103 Germ plasm, 112–113, 115–116, 121 Gilbert, Scott, 280, 286 Gilbert, Walter, 3, 87, 91 Godfrey-Smith, Peter, 183, 252, 255– 256, 260–261 Goldstein, J., 341 Golomb, Solomon W., 84 Goodwin, Brian, 8, 9, 140, 165–166, 226 Gould, Stephen J., 146 Gray, Russell, 191, 281, 290 Griesemer, James, 8 Griffiths. P. E., 211, 256, 281, 290 Gutmann, W., 8 Habit formation, and semiosis, 158 Hagen, Margaret, 132 Haila, Yrjö, 285 Haldane, J. B. S., 85 Hartmann, D., 240 Hataro, S., 119 He, Opitz, Raff, 226, 285 Hendriks-Jansen, H., 279 Heredity: as aspect of dna, 1; descriptions of, 199–200; in developmental systems approach, 208–209 History, and structure, 129–130 Hoffmeyer, Jesper, 8 Holism, 100

Homeobox genes: ceh-14, 29–32; as developmental control genes, 23–24; discovery of, 85; in Drosophila, 26; function of, 16, 24–26; of the Hox cluster, 24–26 Homology, and developmental integration, 59–61 Horvitz, Bob, 15, 32 Hox genes, 24–26, 85, 143 Hull, David, 132, 222, 227 Human Genome Project, 86–89 Human genome sequence, properties of, 88–89 Hume, D., 107, 109 Identity: and ethics, 355–359; and genetics, 7, 9; and modeling gene action, 350–353; and program models, 360–361 Individualism, 272; and developmental systems approach, 286–287; and outlines, 276–278 Individuals, defining, 279–280 Induction interactions, 53–54 Information: cellular interpretation of, 154–156; historical references to, 251; intentional, 184–185, 187–189; and semiosis, 154–156; as term, 2, 155, 182–186, 251 Information, metaphors of, 8, 169–170 Information talk, influence of, 189–190 Informational gene, 175–195 Ingber, D. E., 147 Inheritance: and contingency, 42; in Darwinism, 111; in the evolution of developmental mechanisms, 38–63; functionalism on, 206; process perspective on, 213–221; as term, 296; Weismann on, 111–112, 206 Innate behaviors, vs. acquired, 178 Instruction, as term, 2 Interactions, and constructions, 272– 287 Intrinsic values, 344–347

373

Index Introduction to Modern Genetics (Waddington), 82 Jacob, François, 4, 84–85, 100, 116, 297, 305–306 Johannsen, Wilhelm, 80, 141, 245 Kaneko, K., 50 Kant, I., 103, 105–106, 131, 136 Kasai, H., 119 Kauffman, Stuart, 153, 226, 341 Kay, L. E., 250, 254 Kirschner, M., 144 Kitcher, Philip, 175–195 Knight, Robin, 186 Kuhn, Thomas, 114 Lander, Eric, 136–137 Langton, Chris, 304 Lehrman, Daniel, 191 Lévi-Strauss, C., 131 Lewis, J., 121–122 Lewontin, Richard, 137–138, 176, 179, 202–204, 222, 226–227, 296 Logic of Life, The (Jacob), 297, 306 Lorenz, Konrad, 178 Lumen formation, and cell polarity, 46–48 MacLeod, Colin M., 78 Malthus, T., 107 Mangold, H., 82 Master genes, 4, 5 Master molecule, age of, 84–86 Maturana, H. R., 341 Maynard Smith, John, 184, 186, 215, 216, 222, 227, 325 Mayr, Ernst, 85, 201–202, 297 McCarty, Maclyn, 78 Membranes, 163–166. See also Boundaries, biological Mendel, G., 202, 203, 245 Mendelism, 77–78; atomism of, 100– 101, 115; molecular basis of, 141; physical interpretation of, 78

3 74

Mepham, J., 124 Merleau-Ponty, Maurice, 327, 328, 329 Mesenchymes, and epithelia, 54–58 Mesoderm induction, 53–54 Metaphors: and analogies, 353–355; computer, 297–305; information, 8, 169–170 Metaphysics (Aristotle), 319 Metazoans: basic body plans of, 52–53; development of, 59–61; differentiation in, 50–51; forms assumed by, 47–48, 63; plasticity of ancient vs. modern, 40 Methodical culturalist position, 6 Metz, J. A., 60 Michel, George, 191 Model system, C. elegans as, 15–36 Models, 349–362; choosing, 361–362; and identity, 360–361; nonprogram, 359–361. See also specific models Modern organisms, segment formation in, 49–50 Mol, Annemarie, 328–329 Molecular analysis: of the gene, 141– 142; proteins in, 143–144; range and limits of, 139–144 Molecular approach, 137–139; in analyzing morphogenetic processes, 142–144; as gene centric, 137; and genome of C. elegans, 18–21; range and limits of, 139–144; vs. organismic approach, 135–148 Molecular biology: computers as metaphor for, 297–298; delimiting domain of, 85; and evolutionary paradigm, 116–117; problems of morphogenesis in, 145; social impact of, 2 Molecular genes, 251–254 Molecular genetics, 239–240, 249–250 Molecular model: genes in, 352; readings of, 7–9; research strategy of, 3–4; as term, 3

Index Molecular tools, 244–248 Monod, Jacques, 4, 84–85, 100, 116, 118–120, 297, 305, 325 Moore, Celia, 191 Moore, David S., 193 Morgan, Thomas Hunt, 3, 77, 83, 141, 202, 248–250 Mori, Ikue, 31 Morphogenesis: and cell adhesion, 44–46; Driesch on, 125–126; mechanical stress as determinant of, 58; molecular approach in analyzing, 142–144; and molecular biology, 145; Monod on, 118–120; nongenetic causal determinants of, 39; Wolpert on, 121–122 Morphological innovation, sources of, 41–42 Morphological novelties, 58–59 Morphological plasticity, influences on, 39–40 Morphology: pre-Darwinian, 101–102; rational, 101–104, 107 Morphomes: and developmental process, 321–323; and organic practice, 319–323 Moss, L., 253 Müller, Gerd, 8 Muller, H. J., 77–79 Multicellular organisms: evolution of genes in, 35; physical processes mobilized in, 57–58 Multiple inheritance systems, 214–216 Mutations: in C. elegans, 19–21; in Drosophila melanogaster, 23; in organismic approach, 138 Natural theology, 107–108 Nature and nature, reframing issue of, 291–292 Nematodes, 16. See also Caenorhabditis elegans Neo-Darwinian interpretation: and evolving generating mechanisms,

62; on phenotype, 39, 41. See also Darwinism Neumann-Held, Eva M., 8, 326 New Patterns in Genetics and Development (Waddington), 85 Newman, Stuart, 8 Newton, I., 127–128 Nicomachean Ethics (Artistotle), 315 Nilsson-Ehle, Herman, 80 Nonlinear dynamics, 338, 341 Nonprogram models, 359–361 Normal, as term, 357 Norms of reaction, 80–81, 178–179 Novelties, morphological, 58–59 Objectivity, and digital codes, 167–168 Ontogeny of Information, The (Oyama), 273, 282 Ontological analogy, 350–351 Oosawa, F., 119 Operon model, 85 Organic practice model, 321–323, 330– 333; and lived phenomenal body, 327–329; and morphomes, 319–323; and praxis, 329 Organismic approach, 137–139; mutation in, 138; selection in, 138; vs. epigenetic approach, 8; vs. molecular approach, 135–148 Organisms: body of, 290–307; as coherent constructions, 146–147; conception of, 106; defining internal environment of, 292–294; delineating, 279–280; describing, 337; environment for, 164; and outlines, 274–276; relationship to genome, 325–326; as structures, 126–129; as wholes, 103–104, 106, 341–342. See also Bodies; Form Organization, concept of, 107 Organizer, embryonic, 82–83, 106 Origination, and development of body plans, 50–52 Origins, questions of, 104

3 75

Index Owen, Richard, 102 Oyama, Susan, 8, 138, 169, 175–176, 179–182, 189, 226, 290, 294, 307, 326; Kitcher’s critique of, 190–191; on persistence of genetic determinism, 181–182, 193–194 Pattee, Howard, 155–156, 167 Peirce, Charles Sanders, 157, 158, 160– 163 Perspectives, theoretical: characterizing periods of science, 222–223; compared, 225–227; competition among, 212–213; function of, 199; multiplicity of, 221–228. See also specific perspectives Phenotypes, neo-Darwinian interpretation of, 39, 41 Piaget, J., 125–126, 131 Plasticity, 80 pmg (process molecular gene), 8, 253–254, 262 Poiesis, 9, 313–331; defined, 313, 315–316; model of, 331–332 Polanyi, Michael, 153 Popper, Karl, 152 Population (evolutionary) genetics, 239–240, 241–244 Praxis, 9, 313–331; defined, 313, 315–316; and entelecheia, 318; and the organic practice model, 329 Predictable complexity, 81 Prigogine, Ilya, 159 Problem of Knowledge in Philosophy and Science of Modern Times, The (Cassirer), 136 Process models, 361–362. See also specific models Process molecular gene (pmg), 8, 253–254, 262 Process perspectives, 208–213; on development, 213–221; on evolution, 213–221; on inheritance, 213–221;

3 76

multiplicity of, 221–228. See also specific perspectives Process structuralism, 8, 209–213, 226–228. See also Structuralism Program, as term, 305. See also Genetic programs Program metaphor, 297–299 Protein-coding gene, 19–20 Proteins: functions of, 21; in molecular analysis, 143–144; as transcription factors, 24 Public discourse: analogy and metaphors in, 353–355; impact on discussion of genetics, 6–7; over genetically modified food, 346–347 ‘‘Reading,’’ vs. ‘‘coding,’’ 259–260 Reaktionsnorm, 80 Realization, and formalization, 152–154 Reductionism, and modern biology, 350 Rehmann-Sutter, Christoph, 8–9 Reichert, Karl, 102 Replicators, 199–200, 213–214, 217, 219 Reporter constructs, 27–28 Reproducer perspective, 199–201, 216–221, 226–228 Reproducers, and replicators, 213–214 Reproduction, and multiple inheritance systems, 214–216 Responsible science, 9, 337–347 Rich, Alexander, 250 Riddle, D.L., 15 rna: in C. elegans, 19, 28; genes producing, 21; splicing of ‘‘genes,’’ 4–5; and translation process, 256–257 Robustness, 221–229 Romashoff, D. D., 81 Rosenberg, Alexander, 324 Russell, E., 108 Salthe, Stanley, 162 Santaella, Lucia, 160 Sarkar, Sahotra, 8, 170, 183

Index Saussure, Ferdinand de, 157 Schaffner, Kenneth, 255 Schelling, F. W. J., 104 Schmalhausen, I. I., 226 Schrödinger, Erwin, 78–79, 84, 306 ‘‘Science of qualities,’’ 344–347 Scully, Jackie Leah, 9 Searle, John, 157 Segmentation: and excitability, 48– 50; in modern organisms, 49–50; regulatory processes for, 48–49; sequential, 49; simultaneous, 49 Self-organizing processes, 144–145 Self-regulation, in structuralism, 126– 129 Semiosis, 152–170. See also Biosemiotics Sequestered modular template (smt) model, 89–92 Shannon, Claude, 251 Siegert, F., 339, 340 Signification, and biosemiotics, 157– 158, 160–163 Skin, biological significance of, 293– 294. See also Boundaries, biological; Membranes smt (sequestered modular template) model, 89–92 Sonneborn, T., 119 Soviet Union, genetics research in, 80–82 ‘‘Space of sense,’’ as term, 329–330 Spatial gradients, formation of, 43–44 Spemann, H., 82, 106 Stengers, Isabelle, 159 Stent, G. S., 137 Stephens, C., 87 Sterelny, Kim, 182–183 Stotz, Karola C., 189 Strasburger, Eduard, 112–113 Strohman, Richard, 5 Structural genes, 4 Structuralism, 99–132, 201–204, 209–

213; compared with other perspectives, 226–228; form in, 130; foundations of, 124–126; genome in, 129; key concepts in, 126–129; shared assumptions with functionalism, 221; subjects in, 221–222; vs. functionalism, 102–103. See also Functionalism Structures: and history, 129–130; organisms as, 126–129. See also Form Sulston, John, 7, 15, 17 Swenson, Rod, 159 Szathmáry, Eörs, 215–216, 227, 325 Tatum, E. L., 249–250 Taylor, Charles, 7 Temporality, and digital codes, 167– 168 Theory of the Gene, The (Morgan), 77 Timoféeff-Ressovs.ky, N. W., 78–79, 81–82 Tissues: adhesive structures in, 55– 57; epithelioid, 54–58; as excitable media, 49; inherent properties of, 62–63; mesenchymal, 54–58 Transcription factors, 24. See also rna Transformation, in structuralism, 126–129 Turvey, M. T., 159 Type, concept of, 105–106, 109 Unpredictability, 341–342 Van der Weele, Cor, 225–226, 284 Varela, F. J., 341 Vasiev, B., 339–340 Vogt, Oskar, 81 Von Baer, 104, 105 Von Bertalanffy, Ludwig, 136 Von Linné, Carl, 136 Von Neumann, John, 155 Waddington, Conrad, 4, 82–83, 129, 142, 226 Watson, J. D., 78–79, 250

377

Index Webster, Gerry, 8 Weijer, C. J., 339, 340 Weinberg, Robert, 136–137 Weingarten, M., 146–147 Weismann, August, 170, 206, 325; and the evolutionary paradigm, 111–117; on form, 113; on germ plasm, 112–113, 115–116, 121; on inheritance, 100–101, 111–112 What Is Life? (Schrödinger), 78 White, John, 17 Wholeness: notion of, 136–137; in structuralism, 126–129; in think-

ing about organisms, 103–104, 106, 341–342 Wiener, Norbert, 155 Wilden, Anthony, 168 Wilkins, A. S., 142 Williams, G. C., 212, 241 Wolf, Ulrich, 8 Wolpert, Lewis, 4, 116, 121–122, 175, 298 Woltereck, Richard, 80 Wood, W. B., 15 Ycas, Martynas, 250 Zwilling, E., 119

Library of Congress Cataloging-in-Publication Data Genes in development : re-reading the molecular paradigm / edited by Eva M. Neumann-Held and Christoph Rehmann-Sutter. p. ; cm. — (Science and cultural theory) Includes bibliographical references and index. isbn 0-8223-3656-1 (cloth : alk. paper) isbn 0-8223-3667-7 (pbk. : alk. paper) 1. Developmental genetics. I. Neumann-Held, Eva M. II. RehmannSutter, Christoph, 1959– . III. Series. [dnlm: 1. Developmental Biology—methods. 2. Growth and Development—genetics. 3. dna—genetics. 4. Evolution. 5. Genetic Determinism. 6. Genetic Processes. qu 450 g327 2005] qh453.g47 2005 571.8'5—dc22 2005021633

3 78