0521782333 - Peter McLaughlin - What Functions Explain~ Functional Explanation and Self-Reproducing...

What Functions Explain:Functional Explanation and Self-Reproducing

Systems

CAMBRIDGE UNIVERSITY PRESS

PETER McLAUGHLIN

What Functions Explain

This book offers an examination of functional explanation as it is usedin biology and the social sciences, focusing on the kinds of philosophicalpresuppositions that such explanations carry with them. It tackles suchquestions as: Why are some things explained functionally while othersare not? What do the functional explanations tell us about how theseobjects are conceptualized? What do we commit ourselves to when wegive and take functional explanations in the life sciences and the socialsciences?

McLaughlin gives a critical review of the debate on functional expla-nation in the philosophy of science that has occurred over the last ftyyears. He discusses the history of the philosophical question of tele-ology and provides a comprehensive review of the postwar literatureon functional explanation. The question of whether the appeal tonatural selection sufces for a naturalistic reconstruction of functionascriptions is also explored.

What Functions Explain provides a sophisticated and detailed analy-sis of our concept of natural functions and offers a positive contribu-tion to the ongoing debate on the topic. It will be of interest toprofessionals and students of philosophy, philosophy of science,biology, and sociology.

Peter McLaughlin is Privatdozent in the Department of Philosophy atthe University of Constance.

This page intentionally left blank

cambridge studies in philosophy and biology

General EditorMichael Ruse University of Guelph

Advisory Board

Michael Donoghue Harvard UniversityJean Gayon University of Paris

Jonathan Hodge University of LeedsJane Maienschein Arizona State University

Jess Mostern Instituto de Filosofa (Spanish Research Council)Elliott Sober University of Wisconsin

Alfred I. Tauber: The Immune Self: Theory or Metaphor?

Elliott Sober: From a Biological Point of View

Robert Brandon: Concepts and Methods in Evolutionary Biology

Peter Godfrey-Smith: Complexity and the Function of Mind in Nature

William A. Rottschaefer: The Biology and Psychology of Moral Agency

Sahotra Sarkar: Genetics and Reductionism

Jean Gayon: Darwinisms Struggle for Survival

Jane Maienschein and Michael Ruse (eds.): Biology and theFoundation of Ethics

Jack Wilson: Biological Individuality

Richard Creath and Jane Maienschein (eds.): Biology andEpistemology

Alexander Rosenberg: Darwinism in Philosophy, Social Science and Policy

Peter Beurton, Raphael Falk, and Hans-Jrg Rheinberger (eds.):The Concept of the Gene in Development and Evolution

David Hull: Science and Selection

James G. Lennox: Aristotles Philosophy of Biology

Marc Ereshefsky: The Poverty of the Linnaean Hierarchy

Kim Sterelny: The Evolution of Agency and Other Essays

What Functions Explain

Functional Explanation and Self-Reproducing Systems

PETER McLAUGHLINUniversity of Constance

PUBLISHED BY CAMBRIDGE UNIVERSITY PRESS (VIRTUAL PUBLISHING) FOR AND ON BEHALF OF THE PRESS SYNDICATE OF THE UNIVERSITY OF CAMBRIDGE The Pitt Building, Trumpington Street, Cambridge CB2 IRP 40 West 20th Street, New York, NY 10011-4211, USA 477 Williamstown Road, Port Melbourne, VIC 3207, Australia http://www.cambridge.org Peter McLaughlin 2001 This edition Peter McLaughlin 2003 First published in printed format 2001 A catalogue record for the original printed book is available from the British Library and from the Library of Congress Original ISBN 0 521 78233 3 hardback ISBN 0 511 01247 0 virtual (netLibrary Edition)

For Lea and David

Contents

ix

Acknowledgments page xi

part i functions and intentions 1

1. Introduction 3

2. The Problem of Teleology 16Formal and Final Causes 16Teleology and Modern Science 20Concepts of Teleology in Biology 28Range and Nature of Teleological Statements 33

3. Intentions and the Functions of Artifacts 42Actual and Virtual (Re)assembly 42Benet and Apparent Benet 55

part ii the analysis of functional explanation 63

4. Basic Positions in Philosophy of Science:Hempel and Nagel 65

Function as Cause and Effect 65Limits of the Standard Analyses 73Intrinsic and Relative Purposiveness 75Later Developments 78

5. The Etiological View 82Biological Functions 84Social Functions 91Etiology and Function 93Proper Functions 102Recent Developments and a Recapitulation 114

6. The Dispositional View 118Causal Role Functions 119The Propensity View 125Propensity As a Unication Theory 128Counterexamples 132Heuristics 135

part iii self-reproducing systems 139

7. Artifacts and Organisms 142An Old Analogy 143Nature and Selection 145Design and Natural Selection 150Selection of Parts and Selection of Wholes 153

8. Feedback Mechanisms and Their Beneciaries 162Nonhereditary Feedback 164A Short History of the Notion of a Self-Reproducing

System 173A Naturalistic Fable 179The Limits of Natural Selection 186

9. Having a Good 191Von Wrights Analysis 192The Interests of Plants and Artifacts 194Back to Aristotle 200

10. What Functions Explain 205

Notes 215

Bibliography 237

Index 255

Contents

x

Acknowledgments

This book is the product of a number of years of brooding. The rstdraft of what was to become Parts I and II was written in the fall of1994, when I was a guest of the Max Planck Institute for the Historyof Science. I am grateful to the founding director, Jrgen Renn, and toall the members of the staff for their hospitality and support duringthat period. One of the most important processes of rethinking beganduring a year-long visit to the Philosophy Department at the JohnsHopkins University, where I was able to discuss problems and ideaswith Peter Achinstein and Karen Neander, neither of whom, however,should be blamed for what they could not prevent.

I would like to thank Michael Ruse for nding two referees willingand able to read with understanding a manuscript with which theycould not possibly agree. And I would like to thank those two refereesfor producing trenchant but constructive criticisms, from which I havebeneted greatly.

The penultimate draft of this book was occasioned by a somewhatanachronistic academic ritual called the Habilitation, the Germanequivalent to a tenure hearing without tenure. However, the only realconcession to this academic ritual left in the nal version lies in theoccasionally overpopulous footnotes and in the bibliography, whichcontains all the relevant literature I consulted in writing this book, notjust those works actually cited. I am grateful to Jrgen Mittelstrass andGereon Wolters of the Philosophy Department and to Franz Plank ofthe Linguistics Department, who read the entire draft and provided mewith detailed comments and suggestions for improvement.

For the past ten years, I have had the good fortune to be associatedwith the Center for Philosophy of Science at the University of Con-stance, and I would like to thank the Centers Director, Jrgen Mittel-strass, for his constant support of this and previous projects.

xi

Part I

Functions and Intentions

3We give and take functional explanations not any old functionalexplanation and not of anything and everything. But there are occa-sions when we accept reference to the function of something as a sat-isfactory answer to a genuine why, how, or what question.And we oftendo this without at least knowingly presupposing or implying thatthere is any intentional agency involved. Under the appropriate cir-cumstances certain kinds of things are explained, to the satisfaction ofthose involved, by appealing to their functions. In many areas of thelife sciences, such references may in fact merely be shorthand forhypotheses about the past or present adaptive value of organic orbehavioral traits and about the role of natural selection in their genesis;but this is certainly not always the case. And biology is not the onlydiscipline in which functions are regularly adduced. Functional expla-nation has also been rampant in the social sciences.1

There is a large philosophical literature on functional explanations.2

The statement ascribing a particular function to some entity can beinterpreted as the answer to a number of different questions. We mightask: What does the heart do? What role does it play in the operationsof the body? Which organ pumps the blood? Why do we have a heart?Which organ is it that has the function of pumping the blood? What isthe function of the heart? Why does the blood circulate? To these andmany other questions we might in the appropriate context sensiblyreply: The function of the heart is to pump the blood.3 In some ofthese, or similar cases, we may intend the statement purely descrip-tively; or function may be used as a metaphor or faon de parler.However, in some cases we seem to be offering an explanation, thoughit need not in each of these cases be an explanation of the same kind.It is only the explanatory use of functional ascriptions that will be of

1

Introduction

interest in the following: Some analyses of such ascriptions take us tobe explaining (or attempting to explain) why we have a heart; otherstake us to be explaining what the heart does.

In the philosophical literature there are great and generally self-serving differences in the descriptions and evaluations of the use offunctional explanations. Some consider all (legitimate) appeal to func-tion in scientic explanations to be merely metaphorical, because allgenuine functions presuppose intentionality:Except for those parts ofnature that are conscious, nature knows nothing of functions.4 Othersnot only nd functional vocabulary methodologically unobjectionablebut also take it to be empirically ubiquitous: Furthermore, biologystandardly treats function as a central explanatory concept.5 There isno consensus on what the question is, let alone what the answer oughtto be.

The fault line running through this debate seems to follow the ques-tion of norm and value. Does the attribution of function presuppose avaluation of the end towards which it is a means at least in the sensethat the function bearer is supposed to perform its function? Is valuealways relative to a particular perspective, system, scheme, or lan-guage? Is there intrinsic value? To characterize something as having afunction whether in descriptive or explanatory intent is to view itas a means to an end, as instrumental to or useful for something thatitself is valued or somehow normatively distinguished.

It is, of course, possible that reference to function in an explanationof organic traits or cultural practices is merely metaphorical in otherwords, that it simply evokes some kind of vague analogy to humanintentionality and thus may be somehow psychologically satisfyingwithout being rationally justied. This would be the case if humanintentionality were the only source of purposiveness in the world (andwere itself not explainable in naturalistic terms): All seemingly non-intentional purposiveness would then be due to accident or to our lackof insight into deterministic connections. To ascribe a function tononartifacts would either be merely to talk about them metaphoricallyas if they were products of human intentionality or else to view themliterally as (nonhuman) artifacts and thus to presuppose an intentional(superhuman) creator. This view whether articulated as a metaphorthesis or as implicit creationism must assume that the functions attrib-uted to organs and institutions are functions in just the same sense asthe functions of artifacts. But this is much easier to assert than to arguefor, and it is almost certainly false. In any case, it should be a subject


4

for investigation, not for a priori judgment. In the course of this study,I shall be asking what assumptions about the objects considered makevarious views on function ascriptions satisfying. As it stands, the asser-tion that natural functions are either metaphorical or divine is simplyone particular variant of an antinaturalistic credo and is prima facie noless metaphysical than a commitment to intrinsic value in nature. Thisdoes not make the position wrong; it merely denies it the privilege ofthe default setting. Thus, the question whether human intention by(metaphorical) extension ultimately explains natural or social purpo-siveness or whether a more basic natural purposiveness by specica-tion generates human intentionality should not be prejudged. As onecommentator has put it: it seems at least as plausible that the conceptof intending to is derived by restriction and qualication from a muchbroader concept of direction toward an end.6 Both of these viewsmust be taken seriously. It may in fact turn out that the metaphysicalprice of the second is higher than we would like to pay. However, letus rst nd out what exactly it is.

In one of the stronger accounts of intentionality and action, G. H.von Wright sees all teleological or functional explanation of behaviorto presuppose intentionality. That is, in order to explain some behav-ior teleologically (functionally) we must rst understand it as inten-tional action.7 The explanation that a spider spins its web in order toacquire food is only then (nonmetaphorically) acceptable if we con-ceptualize the spiders spinning behavior as intentional. While vonWrights analysis is extremely plausible in many regards, it is in factoffered only as an analysis of behavior and perhaps (in some extendedsense) of the products of this behavior, not as an analysis of the struc-tures and systems that behave. It is prima facie much less plausible to assert (and von Wright does not) that the organs (parts) of a spidercan only be said to have functions if the spider is viewed as the productof intentional action.8 While it is clear that our talk about functionsinvolves a number of presuppositions that determine the conceptual-ization of the systems whose parts possess functions, it still remains tobe seen what this conceptualization actually involves. We shall in factsee that the conceptualization of systems displaying nonintentionalfunctions is signicantly different from that of systems that are inten-tional artifacts.

The question I shall be dealing with in this book is not so much whata functional explanation is, or what its logical or linguistic form is supposed to be, or whether it is a good type of explanation or not.

Introduction

5

Rather, the question I shall be asking is, what kinds of things can befunctionally explained? Why are some things explained functionallywhile others are not? What is the difference that is supposed to makethe difference? What does the use of function-ascription statements to explain certain kinds of objects tell us about how these objects arereally conceptualized? Thus, I shall not be asking what is the rightmetaphysics for turn-of-the-millenium philosophy and does it counte-nance functional explanation, but rather what are the operative meta-physical presuppositions of an explanatory appeal to functions. Buthere, too, I shall not be interested in all cases and all assumptions. Somemetaphysical assumptions occasion no great or unusual difculty. Forinstance, if we were to nd that a particular kind of explanation onlymakes sense on the metaphysical assumption that there are causal rela-tions among events in the material world, few of us would get excited.Causation presents a metaphysical problem of course, but causation is not the kind of metaphysical presupposition that need move us toreject a theory. A commitment to causation is a metaphysical pricemost of us are willing to pay. Some functional explanations presupposelittle more than causality. I shall deal with some of these in Chapters 4and 6; and then drop them because they are metaphysically unprob-lematic. Other kinds of functional explanation will turn out to be meta-physically more expensive.The strategy of this study will not be to seekout the metaphysically least-problematic use of functional explanationand then recommend it, but rather to pursue those uses of functionalexplanation that are widespread and perhaps metaphysically moreexpensive and then to try to articulate more clearly what metaphysi-cal commitments they demand.

There are many different philosophical analyses of explanation andvarious ongoing controversies about what an explanation is. Theselatter can and need not be settled here. However, I take it that func-tional or teleological explanations are only then genuinely problematicand thus of special signicance, insofar as they are taken to give acausal explanation of why the function bearer is where and what it is.I can see no objection in principle to a noncausal explanatory use offunctions (for instance, as a device for theory unication) and thus noadditional grounds for controversy in such a case due to the appeal to functions. Therefore, I shall generally presuppose that explanationmeans causal explanation. But the interesting question is not so muchwhether functional explanations are reducible to ordinary causal expla-nations under certain conditions. I presume that many may well be in


6

some sense reducible in principle, while they may not be in practice though the usual skepticism as to the preservation of natural kinds in reduction is certainly justied. In such a reduction, we presume thatsome feedback mechanism mediates causality from the effects of afunctional item back to the item itself, and we use a vocabulary that isvague enough that we can arrange the types and tokens appropriately,so that no contradiction, assumption of backwards causality, or otherunpleasant by-product is implied. The really interesting question is, Ithink: What kind of system is S if we can sensibly speak functionallyabout it, even when we believe there are probably appropriate causalmechanisms of some kind? How do we conceptualize a system whoseparts can be function bearers?

Contemporary debate about the analysis and the status of functionalexplanations has reached the stage where it has been characterized bythe dull thud of conicting intuitions.9 Denition attempts, whichonce kept getting longer and more complicated, have now stabilized asquasi-machine-readable reformulations with unexplained notationalconventions are paraded past our intuitions. Counterintuitive counter-examples are suggested: We are asked think about instant organisms,brain tumors that happen to correct hormone imbalances, bullet-stopping pocket bibles, and sewing machines with self-destruct buttonsthat dont work. Some standard types of counterexample have becomeestablished and are traded back and forth between the proponents ofetiological and of dispositional interpretations. And, in fact, each ofthese schools seems to have settled down to live in peace with its coun-terexamples. But this kind of peaceful coexistence with counter-examples is possible only for stipulative denitions of the concept offunction. If we stipulatively dene the term function in biology (say)as the effect of an adapted trait (etiological view) or as the adaptiveeffect of a trait (dispositional view), then intuitive counterexamples tothe usage prescribed by the denition have no force, because theymerely presuppose other conceptions of function. Nonetheless, stipu-lative denitions do, in a sense, relate to everyday usage as at roofsrelate to standing water: However tight they are, they tend to leak. Ifthe prescribed usage of the term goes too much against intuition, it willconstantly tend to be used falsely. We will also begin to ask what workit really does for us.

In the following, I shall argue that the real objection to the variousanalyses on offer, is not that they dont capture some one of ourintuitions about functions, but that, by doing this, they miss the philo-

Introduction

7

sophically interesting point of our adamant proclivity to teleologicalvocabulary when speaking of biological and social systems. I shall notbe collecting intuitively plausible examples and counter-examples butrather looking for the construction principles of such intuitive exam-ples, at the processes they are supposed to be examples of. I shall bearguing not that this or that semiformal translation of a function-ascription statement is better than another, but rather that our use offunctional ascriptions to explain certain kinds of objects can tell ussomething signicant about how we fundamentally conceptualize theseobjects and about the presuppositions we make in doing this. When inthe following I do refer to intuitions or to what we would say orwhat we mean in a particular case, this is not intended as justica-tion of vernacular function ascriptions but merely as description of a practice whose metaphysical presuppositions we want to investi-gate. The bias in deciding what exactly intuition actually says indoubtful cases will always be in favor of the metaphysically moreexpensive alternative. This is a methodological matter of course,because we want to know what we might have to accept if we stick tofunctional vocabulary, not what we might be able to get away with.Thus, my question will not be: Are functional explanations goodexplanations according to some pregiven standard? But rather: Whatkinds of things do we explain functionally? Why do we do this? Andwhat can this tell us about the presuppositions we implicitly makeabout the things we explain in this manner? This is the sense of thetitle: What Functions Explain.

I shall, for the most part, stick to a few standard examples. While itis less entertaining always to use the same boring example of the heartbeating in order to circulate the blood, it nonetheless has the same kindof advantage as mass-use software most of the kinks have beenworked out. We dont have to worry about doubts as to the empiricaltruth or the appropriateness of the example or other factors that mightmuddy the issues. With some other standard examples this is not thecase; for instance to take the most famous example the function ofchlorophyll for photosynthesis in plants is (famously) open to extra-neous questions: whether chlorophyll is really always necessary forphotosynthesis or whether xanthophyll will on occasion do the job; orwhether a chemical substance should be attributed a function at all,instead of the organ that secretes or extracts it. Do things that areinside an organism but are not part of the organism have functions?Do oxygen molecules have the function of nourishing the cells while


8

they are still in the bloodstream, or only on arrival? Do symbiotic par-asites have functions like organs? What if you cant tell the differencebetween the two? However, even if we were to want to answer suchquestions at some point, it would still be appropriate rst to under-stand a simple paradigm case. Perhaps we should also leave open atrst the question of whether particular functional explanations are bestseen as members of a class of explanations or as instantiations of a typeof explanation: Every member of a class is just as much a member asany other, but tokens of a type can instantiate it better or worse. Andto characterize a type we should perhaps best stick to Whewellsdictum: The type must be connected by many afnities with most ofthe others of the group; it must be near the center of the crowd, andnot one of the stragglers.10 Thus, the paradigm generally used to expli-cate functional explanation will be the function of the heart in bloodcirculation, not that of the faulty self-destruct button on your sewingmachine. We can always change our examples later once we haveunderstood the paradigm case.

Functional explanation is considered for various reasons by manyto be illegitimate, and we shall analyze some powerful arguments tothis effect. There are also a number of different attempts to savefunctional explanation by reducing its claims, appealing to differentsenses of explanation and thus separating it from the unsavory teleology that is often associated with it. This is not the tack I shall take. I shall not be looking for a particular use of teleological vocabu-lary that can be reconciled with mechanism or reductionism. There are some such uses, and these will also be considered in Chapter 6 (andto a certain extent in Chapter 4). But as soon as any particular use of such vocabulary turns out to be merely metaphorical, heuristic, orjust descriptive of unproblematical causal connections, it will cease to be of interest and I shall drop it. I am more interested in those uses that are not so reconcilable most particularly, I am interested inthose that can be reconciled with determinism but not with reduction-ism. It will turn out that most genuine functional explanations involvenot so much an illicit appeal to nal causes as an implicit appeal to holistic causality. Furthermore, this holism itself is generally rela-tivized by appealing to various kinds of identity over time, so that the causal relation of a system to the properties of its own parts is interpreted as the relation of a system to the properties of the parts of some successor system. The task of this analysis will be to explicatethe sense and rationale of such implicit assumptions. I shall not be

Introduction

9

justifying or criticizing functional explanation but rather trying toanalyze whats in it.

Recent literature analyzing functional explanations can be dividedinto four major areas:

(1) Biology and its philosophy, where the discussion is concerned pri-marily with questions of adaptation and evolution.

(2) Social science and its methodology, often in connection with thedistinction between latent and manifest functions.

(3) General philosophy of science as shaped by C. G. Hempel andErnest Nagel, where the discussion began as a philosophical reec-tion on the use, abuse, justication, or lack thereof of functionalexplanations in the special sciences. More-recent discussions in thephilosophy of science tend to deal only with biological (and arti-cial) functions and are increasingly often less a second-order reec-tion on problems in scientic explanation than a preliminary to thefourth area of study: naturalistic philosophy of mind.

(4) Naturalistic philosophy of mind, where an explication of biologi-cal function is sought that can help in reducing intentional vocab-ulary to physiological or biological vocabulary. Whereas Hempeland Nagel were very stern with both the biological and the socialsciences in treating functional explanation, contemporary philoso-phy of mind tends to be excessively lenient with biology on thishead. A great deal of the interest in functional explanation is dueto naturalist projects; and I suspect that philosophers of mind aremuch keener on allowing (or encouraging) biologists to use func-tional explanations than are the biologists themselves. Biologistscould in general probably get along fairly well, if they had to,without the term function by substituting either causal role or selec-tive advantage or adaptive value.11 Thus, there is a very real dangerthat the vested interests of philosophy of mind in intentionalitylead it to foist more teleology upon biology than the biologistsneed or want by providing a self-indulgent analysis of functionalexplanation. This is aided and abetted by the tendency of somephilosophers of biology to call any explanation teleological thatadverts to natural selection and to exaggerate the extent of theteleological vocabulary that can actually be legitimated by naturalselection.

My interest is primarily in the third area as a reection on the rstand second, even though, given the state of the literature, the fourth


10

area will play a very signicant role in the following study. I shallanalyze natural or objective functions or, more precisely, the bearers ofsuch functions: The function bearer is a means to the end named bythe function. The heart is a means to the end, blood circulation. TheHopi rain dance is a means to the end, social cohesion. Because func-tional descriptions always involve instrumental, or means-ends, rela-tions, they must thus inevitably display some analogy to descriptionsof human intentional and instrumental actions or to the products ofsuch actions. However, I shall not be concerned primarily with cases of genuine intentionality or goal-directed behavior, where some kindof representation of an effect is part of the causal explanation of thateffect, whether this representation is taken as mental or material orboth. Because much of the literature interested in the fourth area,naturalistic philosophy of mind, is basically doing conceptual analysis,there is a strong focus on providing an analysis of functions that alsoincludes the functions of artifacts and the purposes of actions. Most of the intuitions that this kind of conceptual analysis is supposed tocapture or mobilize are based on discourse about actions and artifacts.I shall not, however, concentrate much on articial functions and theintuitions they support. Even if the ultimate explanation of intentionalpurposefulness should turn out to be naturalistic, there is nothing quiteso intuitively plausible as the (antinaturalistic) distinction betweenbody and mind.And the intuition that (nonreducible) human intentionis the source of genuine purposiveness and is a genuine causal factorin the production of certain material systems (artifacts) is about as anti-naturalistic as you can get. If naturalism should succeed in forcing usto abandon these intuitions, so much the better, but it seems that thenaturalist strategy, too, should demand that natural, nonintentionalfunctions be explained rst without appeal to artifact-based (anti-naturalistic) intuitions, so that the artifactual functions can later bereduced to the natural ones. A too-strong dependence on intuitionsbased on intentionality must cripple the naturalist project from thestart.Thus, whether or not artifactual and natural (intentional and non-intentional) functions are categorically distinct and whether or not theformer can be reduced to the latter, it is natural functions that must beaddressed rst, if we dont want to prejudice the answer.

Thus far, I have merely dened the object of study, the apparent phenomenon of nonintentional purposiveness. Some may take mostexamples of such phenomena to be illusions or confusions based onmetaphor, but the phenomenon of nonintentional purposiveness is

Introduction

11

undeniable. Even in the unlikely event that most apparent examplescan be shown to involve surreptitious appeal to intention, there is still always accidental purposiveness, however seldom this might occur.And furthermore, if natural selection can, for instance, produce appar-ent purposiveness on a regular basis, one might be able to take it as the point of departure for a naturalist reduction. In any case, non-intentional purposiveness is in fact what we rightly or wrongly often mean preanalytically by certain kinds of expressions. This is also the phenomenon that Kant had in mind when he introduced aZweckmigkeit ohne Zweckttigkeit, that is, a purposiveness notdue to purposeful action, into his reections on biology and humanhistory.12

In the following, while I will not always be able to avoid lapsing intotaking sides in various disputes, the primary aim is one of descriptivemetaphysics: to articulate the metaphysical presuppositions or theontological commitments of the explanatory use of function ascrip-tions. Although I shall stick mainly to biological examples, the point isnot to analyze functional explanation as used in biology but functionalexplanation as such. Thus, we want to be able to explicate the explana-tory value and the ontological commitments of talking about the func-tions of wings, beaks, and webbed feet, but also of puberty rites, publicexecutions, and natural language. Both the life sciences and the socialsciences often countenance functional explanation without assumingteleology in either its Aristotelian or its medieval and modern sense.Today, those using functional vocabulary appeal neither to the dunamisor the immanent goal that lies at the essence of a duck and its webbedfeet nor to its representation in the mind of an artisan God. Neither of these views reects the actual ontological commitments of the twentieth-century ornithologist. But what do and dont the commit-ments of this functional vocabulary have in common with the com-mitments of the sociologist?

My own concentration on biological examples can be legitimatedfrom three perspectives: subjective, intersubjective, and objective. Myown scientic background includes ethology but not ethnology, and myspecialized philosophical work has focused more strongly on philoso-phy of biology than on philosophy of the social sciences. Second, thelarger part of the philosophical literature on functional explanationeither deals with it primarily on the example of biology or restrictsitself explicitly to biological explanation. But there is also a third andobjective reason for the focus, which can only be fully legitimated


12

in the course of the presentation: There are two different kinds of(causally intended) functional explanation in biology, only one ofwhich has a correlate in social science. It is often the ambiguity intro-duced by the difference between what may be called evolutionary andphysiological functions that makes many biological examples (thatappeal to intuition) somewhat confusing. As a placeholder for the lateranalysis, let me put it this way: Organisms reproduce themselves bothas tokens (growth, regeneration) and as types (propagation). Social for-mations only reproduce themselves as tokens, as individual entities;they dont as a rule spawn others of their type. Function bearers inbiology contribute either to maintenance or to propagation, to survivalor to generation; function bearers in sociology contribute to mainte-nance (survival). Because the analysis of biological functions coversboth kinds of function and reproduction, it is there that the specic dif-ference should be made clear.

I shall argue that function ascriptions are generally taken to beexplanatory in the causal and unsavory sense only in a particular kindof system, a kind of system encountered primarily in the biological and social sciences. The systems of which some parts are said to havefunctions in an explanatory sense are conceived as self-reproducingsystems, that is, as systems whose identity conditions over time includetheir continual self-reproduction either as the same individual or assuccessive instantiations of the same type. A system that remains thesame only insofar as it re-produces itself by renewing and replacing itsown parts is also temporally prior to (many of) these parts and canthus, without backward causality, be held to be causally responsible forthe existence and/or properties of these parts. Thus, the key concept inexplicating the sense of functional explanation will be the concept ofthe self-reproduction of a system.

One further contingent fact about the current philosophical discus-sion will inuence form and content of the presentation. Philosophicalnaturalism will play a particularly important role in the presentationbecause this is the dominant approach in philosophy of biology morenarrowly conceived and because in much of the other literature a nat-uralistic explanation of intentionality presents the main motive fordealing with the question of functional explanation in the rst place.In fact, some recent literature even has a tendency to be somewhatparochial in this regard so that the philosophical debate is often con-ceived not to be a broad discussion among various positions but istaken to be simply the technical part of an inner-naturalistic project.

Introduction

13

Take the following statement by one very prominent contemporaryphilosopher of science:13

An interesting feature of all extant philosophical accounts of what theconcept of function means is that they are naturalistic. Although the the-ories vary, they all maintain that functional claims are perfectly com-patible with current biological theory.

That something so evidently false as an empirical statement isnt imme-diately recognized as such by its author (or his editor) indicates that atleast some of the participants in the discussion have a rather selectiveperception of what the alternatives are. It is not at all clear that all thefunctional claims we make with regard to biological or social phe-nomena are in fact perfectly compatible with current scientic theoryor with naturalism or scientic materialism as a metaphysical position.For many of us, this is something that we would like to have demon-strated before we are willing to embark on a carefree metaphysicalshopping spree with our teleological credit cards.

Within this framework, one of the most important questions that Ishall attempt to answer is the following: Does natural selection as suchget you all the teleology you need for a naturalistic interpretation offunctional explanation? The answer will turn out to be no. Even thenaturalist needs a bit of Aristotle to reconstruct our functional attri-butions. That is going to be the metaphysical price. I am not going toadvocate simply paying that price, although Aristotle, too, was a natu-ralist. But I will insist that the naturalist who wants to use functionalexplanations without holistic metaphysical commitments cannot reston Darwins laurels. There is more work to be done.

In their explications of the concept of function, many naturalistsattempt to nd a general characterization that covers all actual uses of the term outside of the social eld. Such a characterization, ifachieved, would have only those determinations common to organicand to artifactual functions. A general concept like this would be moreabstract than either the organic or the articial, each of which wouldpossess some particular determinations not found in the other or in thegeneral concept. I dont know precisely what the use of such a generalabstract classicatory concept might actually be, but it would, in anycase, be of no use whatsoever for a naturalistic explanation of inten-tionality. A naturalistic account of horses, zebras, and donkeys doesntlook for an abstract general concept of the genus equus under whichthese species can be subsumed; rather, it looks for a concrete ancestor


14

species out of which they could have developed. Species are notdescended from genera. Any plausible naturalistic account of inten-tionality in terms of biological function must explain how systems withbiological functions can give rise to systems with articial functions(and of course how systems without functions can give rise to systemswith functions). It must explain how natural systems (like organisms orsocieties) whose parts (organs or institutions) have natural functionscan export (or transform) this teleology to external objects so that they,too, possess functions. Taking the biological variant: Somehow, organicfunctions must have given rise to artifactual functions; and there is absolutely no reason to think that only those determinations of biological functions that also apply to artifactual functions (and thuspresumably also to the generic concept) are relevant to producing the organic functions in the rst place. For naturalism, the explanatoryrelation between natural and artifactual functions must be one of thegeneration of the latter from the former not of mutual subsumptionunder a generic term. What a naturalistic reduction needs is not anabstract concept of function that covers all uses, but rather a concreteconcept of biological (or social) function that can help to reconstructthe generation of artifactual functions from natural ones. We shall seethat one basic strategy of recent naturalist argumentation founders on precisely this point. Not only does it fail in its avowed goal of pro-viding such a general overarching concept of function, but the projectis misguided insofar as nothing would be gained for naturalism if it did succeed.

My analysis of functional explanation will be carried out in threesteps. In Part I, I shall attempt to clarify the relation of functions tointentionality and to the traditional problem of teleology. Part II willpresent and critically analyze the current state of discussion on func-tional explanation in the philosophical literature. Part III will developmy answer to the problems that arise out of the analyses of Parts I and II.

Introduction

15

16

formal and final causes

The question of the status of functional explanation is inextricablybound up with the problem of teleology. Teleology is dened some-what infelicitously as we shall see by the Oxford English Dictionaryas the doctrine or study of ends or nal causes. According to count-less accounts of the rise of modern science, it is the rejection of suchcauses that characterizes science in the modern age. Thus, if functionalexplanations are closely associated with nal causality, their scienticstatus would seem to be open to serious doubt.

Teleology, like so many polysyllabic philosophical terms of Greekorigin and so intimately associated with Aristotle, was in fact theproduct of early modern German university philosophy, specically ofthat inimitable conceptual taxonomist and philosophical pedant, Chris-tian Wolff.1 It was introduced to denote a part of physics (or naturalphilosophy) that still lacked a name: namely the study of nal causesas opposed to efcient causes, in particular the study of Gods inten-tions in creating the world and the various things in it. This is preciselythe sort of thing that Descartes and other heroes of the Scientic Revolution had banished from science and its philosophy.

A distinction is commonly made between a Platonic or externalteleology and an Aristotelian or internal teleology.2 In the case ofexternal teleology, the end achieved (or at least striven for) is the enddesired by some intentional agent external to the object created ormodied, and the value or good attained or conferred by achieving thegoal is value for, at least from the perspective of, that agent. The pro-totype agent is the demiurge of Platos Timaeus, who creates the uni-verse, but any artisan and his artifacts can illustrate external teleology.

2

The Problem of Teleology


17

The artisan has a motive for acting and an idea of what the result ofthe action should be. Internal teleology, on the other hand, presupposesneither an external agent nor an external source of valuation for thesystem or process viewed teleologically; and the (internal) agencyinvolved in such teleological processes is generally not intentional atall. Aristotle, of course, acknowledges the existence of external teleol-ogy, but he is also interested in an internal type. The goals or endsinvolved are those of the system itself, not those of its creator; and sim-ilarly, the valuation of a goal as a good is made from the perspectiveof the entity whose good is involved, not from that of some externalagent. The purposiveness is, so to speak, submental and in a sense naturalistic; there need be no ideal anticipation of the result. The prototype is given by organic processes and structures, especiallyembryological development; but cosmic or historical processes alsohave been used often to illustrate internal teleology.3 The reality ofexternal or intentional teleology can scarcely be denied; the reality ofinternal (nonintentional) teleology is what is really at issue.

A further distinction, canonized by the Encyclopedia of Philosophy,is commonly made between goal directed behavior and functional struc-tures or between teleological processes and teleological systems.4 Theterm teleological system may refer either to a system that behaves in agoal-directed manner or to a system that is intelligible only as the resultof a goal-directed or teleological process. I shall be taking the term inthis latter sense. Some people object to speaking of teleology at all in the traditional, supposedly Aristotelian, sense in connection withsystems for example, in connection with the functions of organs, arti-facts, and institutions or anything that isnt a process: Where there isno process change or motion or activity there is no end or telos towhich it can be directed. This point is certainly well taken, but it ratherindicates an ambiguity in the term teleology, which, in spite of itsOxford denition, does not in fact always refer only to nal causes,but sometimes also refers to what were traditionally called formalcauses.

The term teleology and the associated teleological vocabulary havebeen used for all kinds of ideal (intentional) causality (whether or notthe idea involved is a goal or rather a plan of action) as well as forvarious kinds of real ends. To take the traditional example: The nalcause of a house or of the construction of a house is, say, shelter; thehouse is said to be for shelter or for the sake of shelter. That is whyit is there. One aspect, at least, of any satisfying explanation of a house


18

is what it is good for. We could also refer to the idea of or desire forshelter possessed by some agent with artifacts, it is not particularlyimportant to distinguish between the end of the entity and the idealanticipation of this end. But what shall we call the (ideal) blueprint thatthe constructor uses to guide construction? That, too, is certainly partof any reasonable explanation of how the house got to be what it isand where it is. The traditional term for this aspect of the explanationwould be formal cause: The nal cause (shelter) motivates or initiatesthe house building, the formal cause (blueprint) guides the house build-ing, and the nal cause (shelter) once again guides the evaluation ofthe result. More generally, the nal cause gives us the goal for the sakeof which a change is initiated; the formal cause tells us what it is thatis to arise from the change. Sometimes it is hard to distinguish the two.Although the end of a house (and the motive for building it) is shelter,the house may be seen as the end of the process of house building.5

Thus, what we might want to view as the goal of the house-buildingprocess (the house) can also be viewed as the formal cause of thehouse. Aristotle often viewed the formal and nal causes as basicallythe same, especially when dealing with organic development. In embry-ological development, for instance, the form of the mature organism isboth the goal of the process of development and that for the sake ofwhich the development occurs.6 The questions, What? and For thesake of what? have the same answer: the organism. But the twoaspects can be analytically distinguished. The ideal anticipation (blue-print) of the house is its formal cause; the ideal anticipation of thebenet (desire for shelter) of the house is its nal cause. Hence, if wedo allow ourselves to speak teleologically about the structure of exist-ing systems, as opposed to their behavior, it will most often be theformal cause, not the nal cause to which we are appealing. Now, wecould, of course, just ignore traditional usage and talk about twokinds of nal causes: the house and shelter or the idea of the houseand the idea of shelter. However, there is no need for this because thetraditional terms are available and, as we shall see, there are also his-torical examples where the traditional terminology is useful. In anycase, what is important is the distinction itself and the fact that teleo-logical discourse covers both kinds of causes.

Matters are made somewhat more complicated by a slightly obliquedistinction between internal and external teleological relations intro-duced by Immanuel Kant. Kant countenances only Platonic intentional(ideal) teleology as explanatory while accepting something like


19

Aristotelian natural (real) teleology as phenomenally or descriptivelysatisfying in the case of organisms and organisms only. He attempts to provide a mapping of the organic on the intentional. In artifacts, therepresentation of the structure of the whole is part of the formal causeof each of the parts. In organisms, on the other hand, the whole itself(the form) appears to be the efcient cause of (some properties of) theparts, and some way of actually conceptualizing this has to be devisedso that no commitment to holistic causal relations is demanded.7 Thenonintentional teleology of the organism can be simulated heuristi-cally, Kant believes, by interpreting it intentionalistically but suspend-ing judgment on the existence of the artisan presupposed by thepurpose of the object considered. Thus, an organism is viewed as ifit were the product of intentional agency without embracing the actualexistence of the agent. In the Critique of Judgment, Kant distinguishedbetween relative purposiveness (also called external purposive-ness) and internal (intrinsic) purposiveness (also called absolute).8

In Kants sense, relative or external purposiveness refers to the instru-mental relation of one thing for the activity of another: x is purposivefor doing y. If x dependably brings about y, we can say that x is a meansto y; but x is an end only insofar as we view y as an end. This does notmean that x is instrumental to acquiring y only if we desire y; but itdoes mean that the end character of x (as opposed to it means char-acter) is relative to that of y. If y is not an end, then neither is the meansto it x. The relation of relative purposiveness can in principle be iter-ated ad libitum: x can be purposive for y that is purposive for z, etc.But whatever it is that stops such an instrumental regress, it is notviewed as a merely relative purpose. Kants paradigm example is therelative purposiveness of sandy soil for pine trees. This purposivenessdoes not contribute to a causal explanation of how the sandy soil gotto be near the coast; it deals with benecial consequences of the sandysoil for the pine trees, not with causes of the soils origin. Only if forsome reason the existence of pine trees were considered to be a goalof nature would the relative purposiveness of the soil be explanatorilyrelevant for the presence of the soil. However, the very existence ofsuch relative instrumental relations, Kant tells us, points hypotheti-cally to natural purposes. The utility of x indicates that somewheredown the line there is a beneciary y that stops the instrumentalregress.9

Natural or intrinsic purposiveness, on the other hand, is a one-placepredicate. The pine trees, for which the sandy soil is (relatively) pur-


20

posive, may or may not in turn be (relatively) purposive for someoneor something else (like us, for instance). This is an empirical question.But if the trees are the beneciaries of the utility of something, that is,if the regress of means to ends can stop with them, then they are intrin-sically purposive. They are what Kant calls natural purposes. ThisKantian distinction echoes a distinction made by Aristotle in a similarcontext between two senses of the term end:10

That for the sake of which [] is twofold the purpose for which[ upsilonaspertilde] and the beneciary for whom [ ].

The rst sense of end includes Kants relative purpose. The secondsense ts natural purpose fairly well. In any case, two different kindsof instrumental relation are distinguished here: one that can be iter-ated arbitrarily, and a second one that cannot sensibly be iterated butcan stop an instrumental regress. Kants explication and justication ofthe concepts of relative and intrinsic purpose is long and complicated.For our purposes, it sufces to note that he attempts to treat the recip-rocal relative purposiveness of the parts of an organism for each otheras constituting an intrinsic purposiveness of the whole.11 In the follow-ing, I shall use the terms external teleology or external function to referto means-ends relations that can and must be iterated and internalteleology or infernal function to refer to those that stop a functionalregress.

teleology and modern science

Final causes in the stricter sense were banished from science in the seventeenth century by the philosophers of the Scientic Revolutionrather than by the scientists themselves, who in practice often contin-ued to appeal to divine intentions and even to divine interventions.Although such fundamental presuppositions of modern science asmaterialism and actualism were clearly articulated in early modern phi-losophy,12 it was not until the latter nineteenth century that they were(with noteworthy exceptions) generally adhered to in practice. Fur-thermore, the nal causes that were rejected were not the ends or teleof things in Aristotles own philosophy at least as this philosophy hasbeen understood since Hegel but the intentions and purposes of the divine artisan in Christian Aristotelianism. It was not the imma-


21

nent ends of things but the mental representations of these ends in themind of a deity that were banished.13

Aristotles original analysis of four aspects of explanatory discoursehad been codied in Christian Aristotelianism into a system of fourkinds of causes: efcient and material, formal and nal. The rst pairconstituted the material side of causation, the second pair the idealside. As we have noted, however, nal causes in the strict sense makeup only half of the ideal side of causation, and only this half is explic-itly rejected by Descartes (1644) when he asserts:14

And so nally concerning natural things, we shall not undertake any rea-sonings from the end which God or nature set himself in creating thesethings, [and we shall entirely reject from our philosophy the search fornal causes] because we ought not to presume so much of ourselves asto think that we are the condants of his intentions.

Here, Descartes has excluded from science all discussion of Godsintentions (nal causes), but he does not forbid consideration of themental blueprints God might have used (formal causes).

A kind of heuristic teleology long remained in science in the formof an explicitly teleological interpretation of extremal principles suchas Fermats Principle of Least Distance in deriving the law of refrac-tion though such devices were not inconsistent with a thoroughgoingmechanism and were often relegated to the context of discovery. Buteven the intentions of the deity were reimported into science soon afterthe rst-generation prohibition. Where Bacon castigated the search for nal causes as barren, Newton considered it to be one of the pointsof doing physics in the rst place. He believed that drawing infer-ences about God from empirical phenomena does certainly belong to Natural Philosophy.15 Thus, the story about teleology and mod-ern science would seem to be somewhat more complex than usually told.

Early modern scientists and philosophers often spoke of the worldas the machina mundi, and scarcely anyone who was anyone betweenDescartes and Kant neglected to compare the system of the world witha clock.16 But a clock is made by a clockmaker who has some idea ofthe nature of the system he is constructing. Furthermore, the clock-work, with which the world was compared, is a very special kind ofmachine, called a mechanical transmission machine by historians oftechnology. It merely transforms motion or energy from one mechan-ical form to another. It neither generates force or motion like a steam


22

engine, nor does it perform any work like a cotton gin. The fall of aweight is transformed by the clockwork into (among other things) thecircular motion of the hands. In a perfectly constructed machine, suchas Descartes, Leibniz, and others supposed the material universe to be,the initial motion or force applied to matter from without is conservedwithout loss in such transformations, and thus the universe is in a sensea physical perpetual-motion machine.17 But it does not and cannotperform any useful work while it is moving. That is why Descartes canclaim that the nal cause of the world machine is not relevant toscience. God didnt make the world in order to put it to work or tomanufacture anything with it. However, even though a clockwork uni-verse, just like a clock, basically runs idle and doesnt do anything,it does, nonetheless, convey information: A clock shows the time, therelative positions of the planets, etc.18

This is one of the places where external teleology comes back in.The main purpose of the world, we are told by Christian Wolff in thebook that introduces and explicates the concept of teleology,19 is thatwe should recognize Gods perfection. Teleology for Wolff and herehe is completely in line with Newton is that part of physics in whichwe use the empirical study of nature to become the condants of Godsintentions. In the so-called physico-theology of the eighteenth century,empirical science was used to discover the purposes of natural things:organisms and planets, thunderstorms and snow. However, on a fun-damental level this kind of study does not even contradict Descartesinjunction against nal causes, because the intentions of the creator arenot used to explain natural phenomena but are rather themselvesinferred from these phenomena (or simply perceived in them). Thus,while the door is opened wide for pious rhetoric and tenuous theolog-ical speculation, these are simply optional excrescences on a funda-mentally mechanistic science that is, at its cognitive core, independentof such teleological frills. For Wolffs God, the clockwork universe hasmuch the same kind of representational character as the great clocksof the cathedrals and town halls had for the bishops and Brgermeis-ter of the later Middle Ages: They all testify to the grandeur and gloryof their sponsors.

Even those who saw little point in speaking about the nal causesof the world system, planets, or other inanimate objects, often did,however, allow for the nal causes of animate things or more particu-larly of the parts of animate things. While it might be that the sun hasno nal cause (such as to illuminate the earth) and it might even be


23

that plants and animals have no nal causes (such as being useful tohumans), nonetheless the eye and the hand obviously do have nalcauses: the eye is for seeing and the hand for grasping things. The nalcause of such organs, often called their design or intention, need notbe inferred by complex reasonings but could, it was maintained, beexperienced more or less directly:20

For there are some things in nature so curiously contrived, and so exquis-itely tted for certain operations and uses, that it seems little less thanblindness in him, that acknowledges, with the Cartesians, a most wiseAuthor of things, not to conclude, that . . . they were designed for thisuse.

Robert Boyle speaks in this context of design, meaning the purpose for which the divine artisan made an object, not merely the plan heexecuted when making it. But even these teleological reections basedon scientic study do not, as we shall see, change the basically deter-ministic and mechanistic nature of the explanations that scienceoffered.

But putting aside for the moment nal causes, ends, and intentions,there is still the other side to teleology that was not originally excludedfrom science: the formal cause.21 The divine artisan, when making theworld or the things in it, always had two things in mind: a particularpurpose or set of purposes that the created thing was to serve and aplan or blueprint of what the thing had to be like in order to serve thispurpose. This second thing that God had in mind had not been ban-ished from science by the Cartesians; the plan or formal cause wasassimilated to efcient causality. Even Aristotle had been willing toadmit the form of a house existing in the builders mind as (a part of)the efcient cause of the house.22 For modern science, the nal causemight be part of an explanation for the fact that some particular thingexists at all, but it was taken to be nonexplanatory of the nature ofthings. The formal cause, on the other hand, was an integral part ofmany explanations.Thus, most deistic world pictures postulated a prim-itive order or organization of the world system and of biological organ-isms. Some sort of idea of the whole whether of the solar system oran organic body was appealed to in order to explain them.The reasonfor this postulate was not that these systems in their workings appearedto be underdetermined by the (mechanical) laws of nature, that is, bythe mechanical properties and interactions of matters (more) funda-mental particles. It was only the origin of these systems that appeared


24

to be underdetermined to the classical mechanist.23 The functioning ofthe systems presented no special problems, but many were skepticalthat the basic particles of matter would join together of their ownaccord into such complicated systems.

Take, for instance, Robert Boyles (1666) explanation of the originof the planets, animals, and plants:24

I do not at all believe that either these Cartesian laws of motion, or the Epicurean casual concourse of atoms, could bring mere matter into so orderly and well contrived a fabrick as this world; and thereforeI think, that the wise Author of nature did not only put matter into motion, but, when he resolved to make the world, did so regulateand guide the motions of the small parts of the universal matter, as to reduce the greater systems of them into the order they were to con-tinue in; and did more particularly contrive some portions of that matter into seminal rudiments or principles, lodged in convenient recep-tacles (and as it were wombs) and others into the bodies of plants andanimals . . .

The idea of the world system or of the organism in the mind of theartisan deity or opifex optimus, as Copernicus had called him, was partof the efcient cause of the product. It was needed to ensure completedetermination of the origin of the system of the world. Because theconstruction of any system is to be explained by natural laws and initialconditions, any apparent underdetermination of the phenomenon to beexplained must be attributed to some peculiarity of the initial condi-tions in this case, to a kind of original order. Just as Descartes injunc-tion against nal causes is, to a certain extent, a denition of science asan enterprise that studies the nature of what exists but not why (whatfor) it exists, so, too, Boyles appeal to Gods plan presupposes a de-nition of science as an enterprise that studies material systems issuingfrom a certain primitive order.

But this is external teleology:The representation of a system (formalcause) explains how its parts were collected and put together. Itexplains the origin of the system, that is, why certain parts with partic-ular properties were selected and joined together. On the other hand,if the formal cause were interpreted in the sense of internal teleology,it could not be (part of) the efcient cause of the origin of the systembecause it can only exist once the system itself exists. And if the formof the system is to have a causal inuence on the parts, this inuencecan only occur after they have become parts of the system. The notion


25

that the whole can be temporally prior to the parts and thus have acausal impact on them brings up the problem of holism.

Just as it is important for systematic purposes when dealing withteleology to distinguish between nal and formal causes, so, too, is itimportant to keep this distinction in mind when contrasting teleologi-cal and causal explanations. The problems occasioned by teleology inthe biological and social sciences arise not so much from an illegiti-mate appeal to nal causes but rather from an apparent holism involv-ing formal causes. We should make a distinction between causaldeterminism in general and mechanism or reductionism in particular.Determinism of the materialistic variety asserts that material events(processes, entities, etc.) are completely caused and satisfactorilyexplained by an antecedent complex of material states or events. Finalcauses would be explanatorily necessary only if some material event(process, state, entity, etc.) turned out to be underdetermined by its antecedent material conditions (efcient causes).25 Thus, Descartesexclusion of nal causes is simply another way of afrming determin-ism, of asserting that the material world is causally closed.

Determinism as such is, however, completely neutral on the ques-tion of reductionism versus holism. The assertion that the properties ofa containing system are explained by the system-independent proper-ties, and the interactions of its parts is an additional postulate made byspecically mechanistic or reductionistic determinism. Determinism assuch (not necessarily any particular deterministic physics) is quite com-patible with the holistic alternative, that (some of) the properties andinteractions of the parts are determined and thus explained by theproperties of the containing system, that is, that at least some of therelevant explanatory properties of the parts are not independent of the system in which their bearer is to be found.

Now, should in some case the properties or behavior of a (com-pletely determined) system persistently appear to be in some regardunderdetermined by the intrinsic properties and the interactions of itsparts, there are thus two (deterministic) options open to explain thisphenomenon:

(1) Mechanism, on the one hand, appeals to the idea or represen-tation of the whole to explain how the origin of the system can be genuinely underdetermined by the intrinsic properties of itsparts without taking refuge in a nal cause.26 This idea of the whole,strictly speaking, is not a nal cause of the whole but rather a


26

formal one although, as we have seen, it may be viewed as theidea of the end of the process of construction of the whole.

(2) Holism, on the other hand, appeals to the real whole to confersystem-dependent, explanatorily relevant properties on the parts.It is not the idea of the whole that is taken as the formal or nalcause of properties of the parts, but the real system is viewed as anefcient cause of its parts. But the apparent underdeterminationthat occasions the holistic alternative is also not that of the originof the whole but rather that of its functioning.

Thus, the difference between holism and mechanism lies not only inthe interpretation of the causal role of the whole, which mechanismtakes as ideal, holism as real. The two also differ in the kind of under-determination of the system by the intrinsic properties of it parts thatthey are trying to explain: whether it is the production of the systemor its working that occasions the difculties. Mechanism, if it acceptsthe phenomenal underdetermination of the system as genuine, mustpostulate some agent capable of having ideas and of implementingthem whatever other difculties this may occasion. Holism, on theother hand, asserts a real causal inuence of an already existing systemon the explanatorily relevant properties of its parts.This has the further consequence that the two alternatives do not compete as answers tothe same question. While they are indeed two ways of making deter-ministic an apparently underdetermined phenomenon, the kind ofunderdetermination that each alleviates is quite distinct: the underde-termination of the origin or the underdetermination of the functioningof the system in question. One immediate consequence of this differ-ence is that the introduction of an artisan creator by a mechanist couldonly solve the problem of underdetermination of the origin of a system.Should it be the systems functioning, which has the appearance ofunderdetermination, no appeal to the ideas of the artisan who made itis going to remove the mechanistic underdetermination. It is, of course,possible to assert that the artisan creator arbitrarily decides to conferholistic causal powers on a particular system that she has created.However, even if we pretend that the real causality of this whole is ulti-mately due to the creators idea of the causality of the whole, the result-ing causal powers are still taken to be real and must be accommodatedin an explanation.

Let me stress the distinction between causal determinism and mech-anism/reductionism, because it is essential to any understanding of the


27

problem of teleological explanations. A cause, as we conceive of ittoday, must temporally precede its effects. Thus, a nal cause is onlypossible as a faon de parler, as an ideal anticipation of real effects thatitself is or is part of the efcient cause. Or, if the nal cause is itselfconceived to be material, it must be a material representation of theeffect, that is, a component part of the aggregate efcient cause of thateffect. Causality goes forward in time; there is no backward causalityin which a later event determines a prior event. However, it is anentirely different question whether the properties and interactions ofthe parts of a system determine (cause) the properties of the wholesystem or whether some properties of the whole system can insteaddetermine some properties of the parts. The latter would be a kind ofdownward or inward causality. If we assume that the parts are tem-porally prior to the whole, then causality can only ow outward orupward from part to whole. However, if we assume that the whole is temporally prior to the parts, then causation could theoretically godownward or inward from the whole to the parts. Now, although thislatter type of causal action seems extremely counterintuitive, it isnonetheless, as Kant points out in the Critique of Judgment, not inco-herent: It involves no contradiction. This is an important point thatneeds clarication.The extreme plausibility of the mechanistic assump-tion of part-whole determinism has led many partisans of determinismto believe that they were arguing against backward causality, when theywere in fact merely avoiding downward causality, and to think that theywere defending determinism as such, when they were in fact advocat-ing a particular kind of determinism: from part to whole.27 The teleol-ogy that has been virulent in the history of science, especially inbiology, has had very little to do with backward causation. The realproblem has always been holism.

Furthermore, as far as intuitions are concerned, some peculiaritiescan arise: Most social systems have been around longer than the indi-viduals that we take them to be made up of; and any organism oldenough to be reading this page is denitely older than most of the cellsthat make it up. Thus, there would seem to be some sense in which theparts of such systems cannot be said (as mechanism contends) to deter-mine the whole without invoking backward causality. What this senseis will turn out to be connected with the solution to the problem offunctional explanation. We shall have to take up the differences in theidentity conditions over time of parts and wholes that tend to fosterfunctional and teleological perspectives on things.


28

It was the fundamental insight of Kants critique of teleological dis-course in biology that he recognized that the determination of thewhole by its parts is not analytically contained in the concept of causal-ity as he had introduced it. An additional postulate part-whole deter-minism is needed to narrow down the concept to what he calledmechanism. While mechanism, or causal reductionism, may in fact, asKant insisted, be the only game in town with a future, this is nonethe-less an empirical not an analytical question.28

concepts of teleology in biology

In biology and its philosophy, a number of distinctions have been intro-duced in order to grasp more precisely some of the problems relatedto teleology. In particular, a set of distinctions propagated by ErnstMayr has been very inuential. Mayr distinguishes four different non-intentional aspects of traditional notions of teleology and offers termsto keep them distinct: (1) What he calls cosmic teleology is a sort ofgrand historical goal direction of the evolution of the universe or ofthe organic realm. (2) Teleomatic processes seem to be directed to aparticular end by natural laws and boundary conditions. Mayrs stan-dard examples are stones falling down a well and heated pokers coolingdown: The stone comes to rest at the bottom of the well, and the pokerreaches thermodynamic equilibrium at room temperature. (3) Teleo-nomic processes are constrained not only by laws and boundary con-ditions but also by some form of internal material representation ofthe goal state that initiates and guides behavior or development. InMayrs words, a teleonomic process or behavior is one that owes itsgoal-directedness to the operation of a program.29 Examples are thebehavior of lower organisms, where all talk of mental representationis inappropriate, as well as all forms of embryological development. (4)The fourth type of teleology, often associated with teleological or func-tional systems, is according to Mayr only improperly subsumed underthe term teleology, which, he claims, properly applies only to processesnot to systems. In this last case, Mayr proposes to speak instead simplyof adaptations that are in fact due to natural selection and demand nononphysical causation. However, when explaining what he means byrejecting the application of teleology to systems as such, Mayr pulls hispunches considerably, arguing that applying the term teleological tostatic systems involves us in contradictions and illogicalities. No one,


29

however, considers organic systems to be static, so it is unclear whomhe is arguing against.

Because Mayr summarily disposes of (1) and (4), the crucial dis-tinction in his view is between (2) teleomatic and (3) teleonomicprocesses, each representing a kind of prima facie goal-directed processthat implies neither intentionality nor any kind of nonphysical causa-tion. The concept of teleonomy was originally introduced into biolog-ical discussions in the 1960s to characterize end-directed systems andprocesses otherwise often described in teleological vocabulary.30 Thepurpose of this move was clearly to steer a middle course between out-right behaviorism and the methodologically illegitimate, or at leastsuperuous, attribution of mental states to invertebrates and lower ver-tebrates. Thus, a turtle does not merely swim to the shore and lay eggsin the sand. It can be said to swim to the shore in order to lay eggs, ora wasp can be said to hunt bees in order to feed its larvae thoughneither organism is attributed any kind of mental representation orawareness. Mayr later claried some ambiguity in his own positionmarking a difference to the usage of Pittendrigh, Simpson, and othersby explicitly restricting use of the term to processes and behavior.Thus,for Mayr there are no teleonomic systems, only teleonomic activitiesof systems or teleonomic developmental processes that lead to certainsystems.31

On the distinction between teleomatic and teleonomic processes,Mayr somewhat misleadingly presents the primary factor as restingin the question whether the contingent constraints on the action ofnatural laws that give rise to goal directedness lie in the external con-ditions under which the process takes place (teleomatic processes) orwhether they are encoded in some particular part (the program) of thesystem undergoing change (teleonomic processes). In the former case,the appearance of goal direction is due to contingent boundary condi-tions; in the latter case, it is due to the program dependency of thebehavior.

A number of objections have been raised to Mayrs conception ofteleonomy, in particular to his distinction between teleonomic andteleomatic processes. Ernest Nagel formulated three major objec-tions:32 (1) Possession of a program is not the criterion we actually use to distinguish goal-directed processes from other processes; (2)some things for example, reexes are controlled by an inheritedprogram without, therefore, being goal directed; and (3) most impor-tantly, there is no clear distinction between teleomatic and teleonomic


30

processes and thus none between goal-directed and nondirectedprocesses.

On the whole, however, these objections do not seem very grave;they are in the last analysis simply disagreements. (1) Possession of aprogram was not supposed to be a criterion for recognizing or identi-fying goal-directed processes but rather for characterizing theirnature.33 (2) The point about reexes is not only minor but vague. It isunclear in what sense reex action is supposed by Nagel to be teleo-nomic. The construction of the muscles and tendons in the leg is initi-ated and controlled by the genetic program, but the reex seems to bethe necessary result of this structure, and it is initiated not by a programbut by a stimulus. Thus, there is no reason why Mayr should have toconsider them to be teleonomic. (3) Nagels distinction between goal-directed and non-goal-directed does not correspond to Mayrs betweenteleonomic and teleomatic, and Nagels assumption that they do leadshim to see an inconsistency within Mayr instead of between Mayr andNagel. Mayr takes goal directedness descriptively and phenomenally.He explains the phenomenon of goal directedness in the teleomaticcase by boundary conditions that give the actions of natural laws upona system the appearance of direction and in the teleonomic case bya program. The fact that there could be borderline cases Nagel citesa clockwork where it might be legitimate to treat a structure (say, theescapement mechanism) either as a complex boundary condition or asa simple program, does not seem to invalidate the distinction. Nagelsconclusion shows pointedly the misunderstanding:34

I do not know how to escape the conclusion that the manner in whichteleomatic and teleonomic processes are dened does not provide aneffective way of distinguishing between processes in biology that are goaldirected [and] those that are not. In consequence, though the programview notes some important features of goal directed processes, it is notan adequate explication of the concept.

First of all, Mayrs distinction between teleomatic and teleonomicwas not supposed to do what Nagel says it cannot. The point was not to distinguish between two kinds of biological processes, goal-directed and non-goal-directed, but rather between two kinds of apparently goal-directed processes: those that remain causally under-determined without the assumption of an internal (material) repre-sentation of the goal state or the way to it and those that dont. Thedistinction cannot be considered an explication of Nagels concept


31

of goal-directed processes because Mayr does not subscribe to Nagelsconcept. What Nagel can argue is that the same phenomenally goal-directed process could be explained either as teleomatic or as teleo-nomic. Nagel does, however, point to a problem with talk about internalrepresentation insofar as the distinction between teleonomic and teleo-matic seems to turn on a distinction between having a representationof the goal state and being such a representation. If the teleonomicswimming of the turtle to the shore is based on the turtles having arepresentation of the goal state, egg laying, then some specialized part of the turtle is such a representation or representor, but the behav-ior of that part is in turn not teleonomic (under pain of regress) butteleomatic.35

In order to clarify Mayrs intended distinction, let us take a differ-ent example of an apparently goal-directed process: a marble rolling ina round bowl, which tends to come to rest at the bottom and exhibitsgreat plasticity in its efforts to get there. Is this a teleomatic or a teleo-nomic process? Mayrs talk about internal programs and external lawsof nature seems to indicate that this is a case of teleomatic behavior:The teleomatic ball rolling in the bowl is apparently constrained bygravity, friction, and topography. But why should only a set of internalconstraints be called a program, and why should only external con-straints be called boundary conditions? Why shouldnt we interpreta DNA chain as an internal boundary condition and the bowl as anexternal program? The reason is more one of convenience than of prin-ciple. We might just stipulate that a program is a boundary conditionlocated inside the system instead of outside. This is why Mayrs talkabout internal programs and external conditions is misleading.36 Infact, the teleomatic/teleonomic distinction is only really compelling onthe background of Mayrs distinction between proximate and ultimatecausation.

A genuinely goal-directed (teleonomic) process is underdeterminedby what Mayr calls proximate causation unless we assume some mate-rial representation of the end state. How this representation (program)is ultimately to be explained is an entirely different question. Themotion of the marble in the bowl is explained proximately by gravity,friction, and the structure of the bowl and by the fact that somebodytossed the marble into the bowl.Thus, the motion is initiated by humanintention and constrained by an artifact. The process is actually neitherteleomatic nor teleonomic but rather a product of genuine intention-ality: The causal underdetermination of the process is removed by


32

appealing to human intention, which is responsible for making the bowland the marble and putting the marble in the bowl. The winter migra-tion of a warbler, on the other hand, is explained proximately bychanges in temperature and the length of the day, by concomitant phys-iological changes, and by the birds genetic program. The apparentcausal underdetermination of the process is removed by appealing toa program in this case, one ultimately to be explained by evolutionor perhaps by some other process. Mayrs point is that whatever theactual (ultimate) cause of the program and this is a question forempirical inquiry the distinction between the two types of causes isas sensible as the distinction between the physiological or embryolog-ical explanation of a trait and its evolutionary or phylogenetic expla-nation. The distinction among teleomatic, teleonomic, and intentionalis really a distinction among the kinds of assumptions that have to bemade in order to view a process as fully determined. A stone thathappens to fall down a hole in the ground or a piece of metal thathappens to have been exposed to heat and then returns to thermalequilibrium with its surroundings represents a teleomatic process: Goaldirectedness is accidental. A marble intentionally thrown in an inten-tionally molded bowl represents an intentional teleological process:Goal directedness is due to intention. A warbler sent and guided toAfrica by an internal program that must itself be given an evolution-ary explanation represents a teleonomic process: Goal directedness isdue to evolution or, negatively formulated, due neither to accidentnor to intention. Thus, abstracting from talk about internal programs,which may demand either an evolutionary or an intentional explana-tion, we may also conceive the notion of teleonomy such that a processis teleonomic, only if the existence and the nature of the proximatelyunderdetermined boundary conditions that make it appear goaldirected are neither accidental nor intentional. Thus, when Mayr, againsomewhat misleadingly, asserts that the origin of a program is quiteirrelevant for the denition [of teleonomy],37 he means that the ques-tion of which of many possible ultimate causes is actually responsiblefor producing the program is irrelevant. He does not mean that the factthat proximate causes are insufcient to produce the program is irrel-evant. If the program were spontaneously generated, the process itguided would be not teleonomic but teleomatic.

These results tend to vindicate Mayrs distinction between teleo-matic and teleonomic processes, but they dont provide any support forhis rejection of the notion of teleological systems. For these are exactly


33

what we need programs and the distinction between proximate andultimate causality to explain.

range and nature of teleological statements

There are several kinds of states of affairs that can be described in tele-ological or functional terms. In order to locate the particular kinds ofthings that we shall be dealing with, it will be useful to make a surveyof the different kinds of things we talk about in teleological terms. Todo this, I shall start by analyzing the list of teleological descriptionsoffered by Andrew Woodeld in a standard monograph on the subject.This will also provide an opportunity to give a rst illustration of thekind of approach I shall be taking to the examples and counterexam-ples proposed in the literature.

Woodeld gives a seemingly desultory list of examples of teleolog-ical sentences to illustrate the spectrum of possibilities, but this list doeshave some claim to being nonredundantly exhaustive of the kinds ofteleological statements that have actually been studied in the litera-ture. Unfortunately, because Woodeld wants to have a purely gram-matical criterion of what a teleological sentence is, he feels constrainedalways to use the phrase in order to or in order that in every sen-tence which makes some of the examples more than a little stilted.Although he seems basically to be following his linguistic intuitions informed by the discussions in the literature and in fact makes noclaim that the list is either exhaustive or nonredundant, it can be seento be implicitly based on a closed system (which I shall explicate in thefollowing pages) that does in fact generate these and only these kindsof examples. This closed system is what is actually guiding Woodeldsintuitions or at least our assent to their plausibility.

Woodelds examples of teleological sentences are the following:38

(1) The man ran in order to catch the train.(2) The cat opened the door in order to get the cream.(3) The thermostat turned the heater off in order to stop the water

going above a certain temperature.(4) Witchcraft persecutions occur among the Navaho in order to lower

intragroup hostility.(5) The heart beats in order to circulate the blood.(6) Knives have blades in order that they may cut.


34

These examples seem in a number of regards to be somewhat implau-sible, but we shall be able to correct them once we have seen how theyare generated and once we have determined what they are supposedto be examples of.Although Woodeld believes that the average manwould be willing to take all these sentences literally and to accept allof them as candidates for truth,39 it is hard to imagine a reasonableperson (outside certain academic circles) accepting (3), (4), or (5) inanything but a metaphorical sense, and (6) may call for a grain of salt.Furthermore, the peculiarities of the formulations often prevent themfrom illustrating much less illuminating relevant distinctions. However,in some more colloquial, less stilted form, these examples arguably doin fact cover a good deal of the ground needed to include all commonteleological utterances. Woodeld starts work on a provisional classi-cation scheme distinguishing three basic types of teleologicaldescription: purpose, artifactual function, and organic function.40 Buthe gets sidetracked before he is nished. To nish the job for him, weshould try to order these examples and look for a closed system thatcan generate such a list and give us some grounds for believing that itis exhaustive. The kinds of states of affairs actually covered by Wood-elds list are:

(A) Goal-directed behavior of humans (1) and of animals

0521782333 - Peter McLaughlin - What Functions Explain~ Functional Explanation and Self-Reproducing...

Documents

Transcript of 0521782333 - Peter McLaughlin - What Functions Explain~ Functional Explanation and Self-Reproducing...