Neander Karen Functions as Selected Effects the Conceptual Analyst's Defense

8/13/2019 Neander Karen Functions as Selected Effects the Conceptual Analyst's Defense

1/18

http://www.jstor.org/stable/187457.

Your use of the JSTOR archive indicates your acceptance of the Terms & Conditions of Use, available at.http://www.jstor.org/page/info/about/policies/terms.jsp

.JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide range of

content in a trusted digital archive. We use information technology and tools to increase productivity and facilitate new forms

of scholarship. For more information about JSTOR, please contact [email protected].

.

The University of Chicago Pressand Philosophy of Science Associationare collaborating with JSTOR to

digitize, preserve and extend access to Philosophy of Science.

http://www.jstor.org

This content downloaded from 171.67.129.231 on Thu, 16 Jan 2014 17:29:46 PMAll use subject to JSTOR Terms and Conditions
http://www.jstor.org/action/showPublisher?publisherCode=ucpresshttp://www.jstor.org/action/showPublisher?publisherCode=psahttp://www.jstor.org/stable/187457?origin=JSTOR-pdfhttp://www.jstor.org/page/info/about/policies/terms.jsphttp://www.jstor.org/page/info/about/policies/terms.jsphttp://www.jstor.org/page/info/about/policies/terms.jsphttp://www.jstor.org/stable/187457?origin=JSTOR-pdfhttp://www.jstor.org/action/showPublisher?publisherCode=psahttp://www.jstor.org/action/showPublisher?publisherCode=ucpress


2/18

FUNCTIONS AS SELECTED EFFECTS:THE CONCEPTUAL ANALYST'S DEFENSE*KAREN NEANDERtDepartment of PhilosophyAustralian National University

In this paper I defend an etiological theory of biological functions (accordingto which the proper function of a trait is the effect for which it was selected bynaturalselection) against three objections which have been influential. I argue,contrary to Millikan, that it is wrong to base our defense of the theory on arejection of conceptual analysis, for conceptual analysis does have an importantrole in philosophy of science. I also argue that biology requires a normativenotion of a proper function , and that a normative notion is not ahistorical.

1. Introduction. This paper defends the etiological theory of properfunctions, according to which, roughly speaking, biological proper func-tions are effects for which traits were selected by naturalselection.' Ac-cording to this theory, for instance, hearts have the proper function ofpumping blood, because pumping blood is what hearts did that causedthem to be favoredby naturalselection. This theoryis now familiarenough(indeed it is fast becoming the consensus) but while it is mostly acceptedthat something of the sort happens to capture the actual reference class,the theory has been rejected as conceptual analysis. It is only becauseconceptual analysis itself has plummeted in the popularity stakes that thepopularity of the etiological theory has, inversely, been allowed to rise.It is this to which I object.The supposedly problematic feature of the etiological theory as con-ceptual analysis is that (trueto its name) it makes a trait's function dependon its history, more specifically (and supposedly worse) on its evolu-tionary history. There are three standardobjections to this which I willmention now and discuss later. (i) It seems blatantly inaccurate histori-

*Received January 1989; revised January 1991.tI am grateful to many people for their helpful comments on earlier drafts of this paper,especially Paul Griffiths, William Lycan, Ruth Millikan, Robert Pargetter, Huw Price,Elliott Sober, and Kim Sterelny.'I defended this theory in my unpublished but widely circulated paper, Teleology inBiology , first read at the A.A.P. Conference, Christchurch, New Zealand, 1980, andlater in my Ph.D. dissertation, La Trobe, 1983. Some of the material in this paper appearsin these works. See also Millikan (1984, 1989) who independently developed a very similartheory. The most notable forerunner of the etiological theory was provided by Wright(1973, 1976). His work greatly influenced me.Philosophy of Science, 58 (1991) pp. 168-184.Copyright C 1991 by the Philosophy of Science Association.

168

http://www.jstor.org/page/info/about/policies/terms.jsp


3/18

THE CONCEPTUAL ANALYST'S DEFENSE 169cally. When Harvey announced the function of the heart in 1616 he knewnothing of Darwin's theory, and so he clearly didn't mean, or have inmind, that the circulation of blood was the effect for which hearts wereselected by naturalselection (Wright 1976, 97; Boorse 1976, 74; Nagel1977, 284). (ii) Defining the notion of a proper function in terms ofnatural selection begs empirical and theological questions, and is ana-lytically arrogant because it would seem . . . to suggest that it is im-possible by the very nature of the concepts-logically impossible-thatorganismic structures and processes get their functions by the consciousintervention(design) of a Divine Creator (Wright 1976, 96-97. See alsoBoorse 1976, 74; and Bigelow and Pargetter1987, 188.) (iii) It is stronglycounterintuitivethat a creaturewhich lacked a history would thereby lackfunctions. Suppose we discovered that the whole lion species freakishlycoalesced into existence one day, without evolution or design of any kind,by an unparalleledsaltation (Boorse 1976, 74). Or consider . . . thepossible world identical to this one in all matters of laws and particularmattersof fact, except that it came into existence by chance (or withoutcause) five minutes ago (Bigelow and Pargetter1987, 188). Surely, wecan correctlyascribe biological functions to any such complex, intricatelyintegrated organisms, despite their lack of history and their accidentalgenesis. Or so the argumentgoes.Ruth Millikan has defended an etiological theory of proper functionsfrom these objections (1989). However, she does so by turningher backon conceptual analysis. In her view, to use her own rhetoric, conceptualanalysis is . . . a confused program, a philosophical chimera, a squaringof the circle, the misconceived child of a mistaken view of the natureoflanguage and thought (1989, 290).2 Millikan argues that the standardobjections to the etiological theory are thus underminedby the fact thatthey presuppose that our task is conceptual analysis.

While Millikan's defense throws into stark relief our need to be clearabout our analytic aims, I believe it is wrong to base our defense of theetiological theory on a rejection of conceptual analysis. In my view, thesins of conceptual analysis do not measure large enough for it to be dis-counted as a witness in this case. Moreover, if we are clear about ouranalytic aims, we can confronthead-on these standardobjections and suc-cessfully defend the etiological theory as conceptual analysis. Towardthis, I have found it useful to follow Millikan in making a distinctionbetween conceptual analysis and something she calls theoretical defi-nition . The first section of this paper begins with a discussion of thismeta-analyticdistinction, and explains why I believe conceptual analysis

2Millikan defends these strong words to some extent in Millikan (1984, see especiallychap. 9). I address her criticisms of conceptual analysis in section 2 of this paper.



4/18

170 KAREN NEANDERcannot be disregarded.Following this, I elaborateon the etiological theorybeing defended and then consider each of the standardobjections in turn.2. Conceptual Analysis or Theoretical Definition, or Both? The dif-ference between conceptual analysis and theoretical definition is hard tostate in uncontroversial terms, but the basic gist of the difference, as Iwill be using this meta-analytic distinction in this paper, is as follows.Conceptual analysis is an attempt to describe certain features of therelationship between utterances of the term under analysis, and the be-liefs, ideas, and perceptions of those who do the uttering. It involvestrying to describe the criteria of application that the members of the lin-guistic community generally have (implicitly or explicitly) in mind whenthey use the term. In contrast, a theoretical definition is an attempt toexplain some aspect of the thing referred to, or some aspect of the re-lationship between utterances of the term and the actual world. Whichparticularaspect depends on one's semantictheoryand on what one thinksis most central for semantic theory. Millikan argues (this is a crude ren-dition) that a theoretical definition should describe what a term refers tothat explains the use of the term and why the term has survived and con-tinued to be used (1984). She also suggests that theoretical definitionsshould describe the underlying phenomenon that explains the surfaceanalogies by which we have recognized that things are things of a kind(1989, 293). Thus wateris a liquid with the molecular structure HOHand gold is the element with the atomic number 79 are theoretical def-initions. (Liquid HOH is the underlying phenomenon to which we use-fully refer, and it also explains water's opacity, its thirst quenching pow-ers, and so on.)Some would add two furtherfeatures, which I have purposely omitted,to my characterizationof conceptual analysis. They would add that con-ceptual analysis is thought to constitute a search for meaning, and that itis a search for necessary and sufficient conditions for the application ofthe termunderanalysis. Add these, and I too will probablyabandonship.Since Millikan includes these features in her characterization of concep-tual analysis (1989, 290) some of our disagreement may be merely ver-bal. However, her defense of the etiological theory requires us to rejectconceptual analysis as I have more weakly characterizedt, and my weakercharacterization describes an analytic exercise that is more readily de-fended.

Dispute over whether conceptual analysis or theoretical definition bestcaptures meaning (properly so called) may be largely responsible for therejectiorn f conceptual analysis. So let me concede a point or two. It maybe that meanings are not inside people's heads, not wholly, and perhapsnot even partly. And since conceptual analysis is an attemptto understand



5/18

THE CONCEPTUAL ANALYST'S DEFENSE 171what goes on inside speakers' heads-in that it is a search for the criteriaof application that people generally have in mind-it may be that, if weare searching for meaning, conceptual analysis is a confused approach.But we might be interested in something other than, or in addition to,meaning. So let us define conceptual analysis as a search for the criteriaof applicationthatpeople generally have in mind when they use the termunder analysis, and discuss the merits of such a search, leaving the issueof meaning aside. Perhaps attempting to understand how biologists im-plicitly understandhe notionof a proper unction can be fruitful, whetheror not this gives us the meaning of the term.Conceptual analysis has also often been associated with a search fornecessary and sufficient conditions, and the idea that such conditions arerequiredhas long been disreputable. But such a requirementis no moreinherent to the aims of conceptual analysis than it is to the aims of the-oretical definition. The criteria of application that the relevant linguisticcommunity generally has in mind might be better expressed in terms ofa family resemblance, similarity to prototypes, or Minskian frames.3 In-deed, necessary and sufficient conditions are probably more likely to befound for theoretical definitions (consider those given for water andgold ).

Admittedly, the criteriaof applicationthatpeople actually use are oftenvague, shifting, highly context-sensitive, highly variable between indi-viduals, and often involve perceptual data of a kind that is inaccessible,at least to philosophical methods. This is certainly daunting. But it doesnot follow that we cannot discover anythinguseful about these criteria ofapplication. Conceptual analysts can describe those criteriaof applicationthatpeople standardly apply, most of the time, in the most standardcon-texts. Also, when the relevant linguistic community consists of special-ists, and the term under analysis is one of their specialist terms, and isalso abstract(nonperceptual)and embedded in well-articulatedtheory, theseverity of each of these factors will be greatly reduced. Since this is thecase for contemporary biologists and their notion of a properfunction ,we have more reason to expect success in this case than in most.4 Theproblems with conceptual analysis are many, but they amount to a needto concede that it will be rare, if ever, that a definitive, exhaustive anduniversal analysis can be given. This is not to say that some insight into

3Minskianframes and prototypes are popular but controversial models among ArtificialIntelligence researchers and cognitive scientists, respectively. For seminal papers on these,see Minsky (1974) and Rosch (1973, 1975).4Another reason why it might be thought that conceptual analysis is impossible is thatour concepts are endlessly interwoven. But this point, interesting as it is, is out of placehere. A given conceptual analysis does not have to stand on its own. It connects with thiswhole web of ideas by analyzing one notion in terms of other, implicitly understood, butunanalyzed (or analyzed elsewhere) notions.



6/18

172 KAREN NEANDERrelevant criteria of application cannot be achieved in many cases. Sincephilosophersseem to have been providing nsights along these lines, strongerargumentis needed to show that they cannot have been doing so.Of course it would be futile tryingto do conceptual analysis if we didn'thave any use for it; but we do. We need it to clarify thought and com-munication, which are the time-honored reasons for engaging in the ac-tivity. While this point epitomizes the unoriginal, we may need to remindourselves that theoretical definition cannot entirely replace conceptualanalysis in this respect. This point can be made most vividly with respectto nonreferring terms, but it also applies more generally.Given that conceptual analysis is an attempt to describe what peoplethinkthey arereferning o, we can providea conceptualanalysisfor witch ,entelechy and phlogiston because we can explain what people thoughtwitches, entelechies and phlogiston were. A theoreticaldefinition, in con-trast, is (roughly) an attemptto describe the things referredto, so wherereference fails there is no theoretical definition. If we wanted to persuadea Medieval theologian that witches don't exist, no theoretical definitioncould refine our understandingof his or her understandingof witch .Good communication and debate with this benighted individual demandsthat we know more than a theoretical definition can tell us (which is justthatwitches don't exist). To successfully engage in debate we would needto know what precisely a witch was supposed to be. Admittedly, we arenot likely to meet many theologians of this kind, but similar sorts ofdebates occur often enough in philosophy. For instance, to persuade usthat there is no pain, Daniel Dennett once argued that our criteria of ap-plication are inconsistent (Dennett 1978). He engaged (as he must) inconceptual analysis to do this. Similar styles of argument can also befound regarding libertarianfree-will, to give one other example.Nor is theoretical definition independent of conceptual analysis, sincethe latter is required to delimit the scope of the former. Furthermore,when the relevant linguistic community is very well informed about whatthey speak of, we should expect conceptual analysis and theoretical def-inition to closely correspond. For example, while it was once false asconceptual analysis that water meant liquidwith the molecular structureHOH 1,he criteriaof applicationhave changed and kept abreast with ourknowledge. Now a conceptual analysis of water would have to includethat water is HOH. I could be deceived into thinking that some otherclear and thirst-quenching liquid was water. But if I learned that it wasnot HOH, I would then deny that it was water. So being HOH is nowmy criterion of water. Closer to the task at hand, if we were to offer aconceptual analysis of water as used by modern Western chemists wewould certainly have to mention that the molecular structureHOH wasa defining characteristicfor them. Given that modern biologists are in



7/18

THE CONCEPTUAL ANALYST'S DEFENSE 173possession of a fairly full and correct set of beliefs aboutbiological properfunctions (more so than we are anyway) it is highly likely thatconceptualanalysis and theoretical definition will closely correspondin this case too.The moral for Millikan's defense of the etiological theory is this: if theetiological theory is implausible as conceptual analysis, then it is therebyimplausible as theoretical definition. We can gain no immunity from thestandardobjections to the etiological theory by insisting that we are onlyinterested in theoretical definition. When the relevant linguistic com-munity thoroughly understands the nature of what it speaks of, there isa strong prima facie case in favor of its having integratedthis understand-ing into the basis of its linguistic competence. Our understanding of theworld, and the conceptual framework we employ in talking about it, arevirtually the same thing.I see conceptual analysis and theoretical definition as complementaryand interdependent, describing different but related aspects of language.In keeping with the views expressed in this section, the claims made inthe next, where I explain the etiological theory in a little more detail, areintended to hold true from the standpoint of both theoretical definitionand conceptual analysis. In later sections, the differences between thesetwo modes of analysis are again the focus of attention, for different crit-icisms andjustifications are appropriatedepending on our analytic aims.3. Proper Functions as Selected Effects. In biology, as well as in othercontexts, the most explicit use of the notion of a properfunction is insentences that express function attributions,such as It is normal for pen-guins to be myopic on land since the properfunction of their optic lensesis sharp underwatervision , or A function of the hypothalamus is themonitoringof blood oxygen and sugar levels . Such function attributionsare also made regardingartifacts(the function of an armchair s to providecomfortable seating, the function of the carburetor s to mix fuel with air,and so on). Especially in biology, the notion of a proper function hasother less obvious but essential roles that derive from these basic functionattributions (more on this in the last section). However, these explicitfunction attributions are a naturalplace to start when we seek to under-stand the notion. The problem is to discover what exactly is attributedwhen we attribute a proper function (or more loosely a function) in sen-tences of the form The/a proper function of X is Z .My claim is that the properfunction of a trait is to do whatever it wasselected for. In so far as our organismic structures and processes are theresult of selection they are the result of naturalselection,5 so I claim that

5Elliott Sober (1984, 147-155) argues that natural selection does not explain the traitsof individuals, as opposed to the distribution of traits in a population. His argument is



8/18

174 KAREN NEANDERthe biological proper function of such an item is to do whatever items ofthat type did that caused them to be favored by natural selection. Morecarefully expressed, and assuming that the unit of selection is a genotype,the idea is as follows (where the relevant items are evolved biologicalparts and processes):

It is the/a proper function of an item (X) of an organism (0) to dothat which items of X's type did to contributeto the inclusive fitnessof O's ancestors, and which caused the genotype, of which X is thephenotypic expression, to be selected by naturalselection.For example, it is the function of your opposable thumb to assist in grasp-ing objects, because it is this which opposable thumbs contributed to theinclusive fitness of your ancestors, and which caused the underlying ge-notype, of which opposable thumbs are the phenotypic expression, toincrease proportionally in the gene pool. In brief, grasping objects waswhat the trait was selected for, and that is why it is the function of yourthumb to help you to grasp objects.There are certain constraints on the proper functions of evolved traitsthatdo not apply to most artifacts, and which I have built into the offereddefinition. Differences between biological and artifact functions arise be-cause of the different nature of the selection processes involved in thetwo cases. Biological properfunctions belong primarilyto types and onlysecondarily to tokens because natural selection does not operate on in-dividuals or their biological parts and processes. A particular piece ofgenetic material, or a particular nstance of a trait (your thumb, Reagan'snose) cannot be selected by naturalselection which operates over wholepopulations. Also, biological functions of evolved traits must be selectedon the basis of actual past performances (by, or in, ancestralorganisms)of the effect that becomes the function. In contrast, because artifacts aregenerally the result of a very different sort of selection process-the in-tentional design of an agent with forethought-they can have idiosyn-

based upon an illuminating distinction between what he calls selectional explanationsand developmental explanations . According to Sober, developmental explanations ex-plain how an individual comes to possess a trait, whereas selectional explanations can onlyexplain why a population has a certain distributionof individuals with a given trait. I haveelsewhere argued (Neander 1988) that it is a mistake to suppose that the two forms ofexplanation are mutually exclusive. I argue that a full developmental explanation of anindividual's trait will often contain a selectional explanation. Very roughly, the idea is thatyou and I have opposable thumbs because we come from a long line of ancestors all ofwhom had opposable thumbs, and no countervailing mutation intervened to prevent youand me from inheriting the trait. By Sober's own lights, a selectional explanation is neededto explain why all of our recent ancestors had opposable thumbs, and this selectional ex-planation is thus part of the full developmental explanation of why we have opposablethumbs.



9/18

THE CONCEPTUAL ANALYST'S DEFENSE 175cratic functions, like the details of James Bond's cars. And intentionalselection can be made without any performanceof the function-effect everbeing realized (either in the past, present or future)because the mere hintof a hope of a desired effect is enough to inspire intentional selection byan agent with forethought.My main concern in this paper is to address the three standardobjec-tions to the etiological theory which I mentioned in the introduction.Hopefully, this brief elaborationof the theory will suffice for this limitedpurpose, along with just one additional point. It is debatable whether anetiological theory of biological functions must explicitly mention naturalselection. Since this feature of the theory is found particularlyobjection-able by some people, this is worth noting. It might be thought that thetheory could be expressed more neutrally, in terms of selection simpli-citer, for instance. The idea might be that a proper function of a traitwas, simply, whatever effect that trait was selected for, leaving openwhat kind of selection process was causally operative. This might havethe double advantage of providing a univocal account for biological andartifactualfunctions, as well as being neutral with respect to evolutionand Creationism.6While I haven't the space to explore this possibility atlength, my view is that this is indeed the vaguer, unifying, everydaynotion of a proper unction from which the biological notion is derived.Nonetheless, the peculiarities of natural selection impose certain con-straintsupon a more detailed and precise analysis of the biological notion(as mentioned above), constraints which do not apply to the everydaynotion employed for artifacts.Moreover, the standardobjections to overt-ly preferringnatural selection in a conceptual analysis of biological func-tions are misguided, or so I will now attemptto show (in sections 4 and5).4. The Historical Accuracy of an Evolving Notion. The first of theobjections to be considered is the weakest of the three. It charges theetiological theory with historical inaccuracy, since the theory claims thatthe notion of a properfunction in biology should be analyzed in termsof natural selection. Presumably, logically derivative notions must bechronologically antecedent notions. So, the etiological theory is chargedwith the implication that natural selection was discovered before the no-tion of a (biological) proper unction was used, andthis is plainly false.Harvey, for one, proclaimedthe properfunction of the hearttwo hundredyears in advance of Darwin's On the Origin of the Species.First note that Harvey's ignorance of natural selection is not directly

6Wright attempted to provide an analysis that was both univocal and neutral in theserespects (1976, 73-98) but this led him into difficulties (Boorse 1976, 72).



10/18

176 KAREN NEANDERrelevant to a theoretical definition of properfunction (Millikan 1989,290). Water would be HOH whetheror not anyone knew it, and similarly,proper functions might happen to be effects for which traits were selectedby naturalselection, whether or not Harvey or anyone else was aware ofthe fact. But now what is the reply of the conceptual analyst who is, afterall, concerned with the criteria of application that people have in mind?Although conceptual analysis is often equated with something calledordinary anguage philosophy it need not be an analysis of ordinary asopposed to specialist language. We can attemptto understandhow mem-bers of a scientific community, like contemporary biologists, use a termby searching for their criteriaof application. This would still be concep-tual analysis, not theoretical definition, as I have defined them, and thisseems to be the task before us here. It is a relevant philosophical com-monplace, therefore, that scientific terms are shaped by backgroundthe-ories. So how biologists understandthe notion of a proper functionmight shift significantly with dramaticchanges in these backgroundthe-ories. In other words, it is unproblematicif Harvey's notion of a properfunction , before the DarwinianRevolution, was different from the closelyrelated notion used by biologists today, after the Darwinian Revolution.Scientific notions are not static. Harvey obviously did not have naturalselection in mind when he proclaimed the function of the heart, but thatdoes not show that modern biologists do not have it in mind. Just so,Newton did not have the theory of General Relativity in mind when hetalked of time, but that does not show that modem physicists do not haveit in mind when they speak of time, or space-time.Of course I do not claim that modern biologists have naturalselectionconsciously or explicitly in mind when they use the notion of a properfunction . If that were required, conceptual analysis could be done bydeed poll. That a grammaticalrule might come as a surpriseto us is notproof that we do not employ it, andby the same token, no explicit knowl-edge is requiredhere either. What matters is only thatbiologists implicitlyunderstand properfunction to refer to the effects for which traits wereselected by natural selection.Larry Wright was the first (to my knowledge) to propose and be per-suaded by this objection. Yet as he himself suggested, Harvey will havesupposed that biological parts and processes were the result of some sortof selection process (such as design by God). Add that scientific conceptsare not set in stone-that radical shifts in backgroundtheory can modifya scientific community's understanding of its terms-and this first ob-jection is more than adequately answered.5. Analytic Aims and Pretensions. The second objection is more sub-tle. Analyzing the biological notion of a properfunction in terms of



11/18

THE CONCEPTUAL ANALYST'S DEFENSE 177natural selection is said to be analytically arrogant because it begs em-pirical and theological questions. More precisely, the etiological theoryseems to entail the following disjunct: natural selection theory is true orthere are no properfunctions. Given that there obviously are proper func-tions, this entails that the theory of natural selection is true as a matterof logic rather than contingent fact.Again, begin by noting that this objection has no direct bite against theetiological theory offered as theoretical definition. Briefly, theoreticaldefinitions are legitimately in the business of answering empirical ques-tions (Millikan 1989, 290). The scientist begs no empirical questions intelling us that water is HOH, or that gold is the element with atomicnumber 79. Nor does the philosopher who suggests that pain is, as amatterof fact, C-fibers firing. These may be false empirical claims, butthey are not empirical questions begged. Conceptual analysis, however,is in the business of conceptual, not empirical truths; so it seems to bein the firing line of this objection, at least until we take a more carefullook.The etiological theory, offered as conceptual analysis, entails a slightly,yet significantly, different disjunct, which is as follows: natural selectiontheory is true or there are no proper functions in the modern biologists'sense of properfunctions . By comparison, if I were to offer a con-ceptual analysis of time as understoodby Newtonian physicists, I sup-pose my analysis would refer to some axioms of Newtonian physics, suchas that time is absolute. Now we can ask whether this would entail thateither time is absolute or there is no time. And it is obvious that it doesnot. Rather it entails that either time is absolute or there is no time in theNewtonian physicist's sense of time . Since conceptual analysis is aimedat discovering the criteria of application that people have in mind, andsince these can and often do change over time, conceptual truths willalways be relative to a linguistic community at a given time.The persuasive force of this second objection stems from our beliefthat there would be proper functions even in the unlikely event that thetheory of natural selection was falsified. However, this belief statementis ambiguous between opaque and transparent(de dicto and de re) read-ings of the proposition involved. The proposition involved is compatiblewith my claims if what we believe is that ourreference to properfunctionswill not radically fail, as did our reference to phlogiston, entelechies andwitches. But this is just to say that, even if the theory of natural selectionis falsified, and the etiological theory as theoretical definition is therebyshown to be untrue, some other theoretical definition of properfunctionswill be true. It would not follow that the etiological theory as conceptualanalysis would also be falsified, simply because the truth of that does notdepend upon the truthof evolutionary theory. The relevant truthfor con-



12/18

178 KAREN NEANDERceptual analysis is that biologists today believe that a theory of naturalselection is true, and they do. By comparison again, Newtonian physicswas indeed falsified, and so the Newtoniantheoreticaldefmitionof timewas falsified. But some other theoretical definition of time , or at leastof space-time is presumably true. (Time still exists.) Further, the fal-siflcation of Newtonian physics leaves untouchedthe claim that Newtonianphysicists understood time in the light of Newtonian theory. All thatmatters for the conceptual analysis is that the physicists of the day be-lieved that time was absolute, and they did.6. The Status of Hypothetical Examples. Both objections just dis-cussed are directed against the explicit mention of naturalselection in theanalysis. The third and last objection has a broader aim. It is aimed atshowing that function is an ahistorical notion. The essence of the ob-jection is the claim that we could ascribe proper functions to creaturesthat were just like actual creatures, except for the fact that they have nohistory and are not the result of any kind of selection process, eithernatural or supernatural. Suppose, for example, that we discovered thatall lions came into being as the result of one unparalleled saltation .Let us again begin by considering how the etiological theory fares whenoffered as a theoretical definition. The theoreticaldefiner might reply thattheoretical definitions are not obliged to caterfor the merely hypothetical,since they are concerned with mattersof fact (Millikan 1989, 292-293).This may be too quick, however. If a theoretical definition is supposedonly to describe the actual reference class, then hypothetical cases arebeside the point. But if it is supposed to describe the reference class ina way that reveals how the notion has been useful and has survived, thensomething more is required. A theoreticaldefinition should then elucidatethe theoretical role that the notion has enjoyed, and hypothetical casescan be instrumental n this respect. It is arguablethat the mere possibilityof Martian pain, for example, was a legitimate problem for the earlymind-brain identity theorists-if we take them to have been offering atheoretical definition of this kind-when they asserted that pain is (as acontingent matter of fact) C-fibers firing (or some other type of neuro-physical state). Whether or not Martians (or any other aliens) actuallyexist, the mere possibility of Martianpain highlighted the fact that moraltheory requiresa notion of pain that transcendsthe biological peculiaritiesof creatures on this planet. It is arguable that a theoretical definition ofpain expressed in abstract functional terms, which explicitly allows forthe possibility of radical variation in physical instantiation, better eluci-dates the theoretical role of pain . (This was not the case with water ,since the theoretical role of water requires chemical specificity. Were



13/18

THE CONCEPTUAL ANALYST'S DEFENSE 179Twin Earthwaterto exist, it would not reallybe waterif it was not HOH.)7Both theoretical definer and conceptual analyst are in the same boatfor this objection. Conceptual analysis notoriously involves a parading ofexamples past our intuitions. If we are trying to discover the criteria ofapplication upon which we implicitly rely, this is not a bad place to start.However, since these criteriaare not-by hypothesis-already explicitlyknown to us, we might make mistakes about how we would consistentlygo on in the same way , especially when the examples become wildlydivorced from actual facts and natural laws. Intuitions should not there-fore be considered the final arbiters for the conceptual analysts either.They too must be constrained by the theoretical role of the notion beinganalyzed. After all, if we successfully use a notion to perform a giventheoretical role, we must implicitly understandthe notion in a way thatallows it to performthatrole. As I will arguein the next section, waywardintuitions about instant lions might have to be revised in the light of thefact that biology has and needs a notion of a properfunction that isnormative.But first let's tamper with the intuition pump to see how robust theseintuitions really are.8 Suppose there are no lions. Then suppose that halfa dozen lions pop into existence, we know not how. Having stared atthem in stupefied amazementfor some time, we eventually begin to won-der about their wing-like protuberanceson each flank. We ask ourselveswhether these limbs have the proper function of flight. Do they? Whenwe discover that the lions cannot actually fly because their wings arenot strong enough, we are tempted to suppose that this settles the matter,until we remember that organismic structures are often incapable of per-forming their properfunction because they are deformed, diseased, atro-phied from lack of use, or because the creatureis displaced from its nat-ural habitat (the lions could perhaps fly in a lower gravitational field).

7Moreover,he objectionneed not be based on purelyhypothetical ases. A variationof the sameobjectionappeals o piggyback raits. Forexample,it is believed that somellamas,due to a certainproperty f theirhemoglobinmolecules, possessedthe capacityto survive n higheraltitudes hantheirconspecifics,beforetherewas anyneed to moveto highergroundand thereforebeforethis capacityenhanceditness. Some of us maybetempted o say thatthesehemoglobinmoleculeshad the proper unctionof enabling hellamas to survive n high altitudesbeforethe migrationup the mountainsactually ookplace. (I am grateful o an anonymous efereeof thisjournal or this pointandthis ex-ample.)8Ourntuitions egarding iggyback raitsareeven less robust.Indeed,somepiggybacktraitsaretaken n theliteratureo be paradigmases of nonfunctions:heheart's humpingactingas a diagnosticaid, andthe bridgeof the nose keeping up ourspectacles, or in-stance.(Norshould t be saidthat hesecasescan be ignoredbecause heyinvolveartifactsandaretherefore artificial . onsider heEnglish itthatevolveda beakspeciallyadaptedfor sipping rom milkbottles.)



14/18

180 KAREN NEANDEROn the other hand, often enough there are complex structuresthat haveno functions, for instance, the vestigial wings of emus and the humanappendix. The puzzle is where among these various categories are we toplace the lions' wings . I contend that we could not reliably place themin any category until we knew or could infer the lions' history. And ifwe were to somehow discover that the lions had no history, and were theresult of an accidental and freak collision of atoms, they would definitelynot belong in any of our familiar functional categories. They are not thendysfunctional either because of disease, deformity, lack of use, or be-cause they are exiled from their naturalenvironment. All of these requirea past. Nor did they once have a function thatthey have now lost. Withouthistory the usual biological/functional norms do not apply.9This point requires more argument, and luckily we need not allow thedebate to further decay into the dull thud of conflicting intuitions (touse Bigelow and Pargetter'snice expression; 1987, 196). In the next sec-tion, I argue that ahistorical theories of functions do not provide a notionof proper unction which is appropriatelynormative, and so they cannotaccommodate the theoretical role required of the notion in biology.7. Biological Norms and Their Role in Biology. The notion of a properfunction is the notion of what a part is supposed to do. This fact iscrucial to one of the most important theoretical roles of the notion inbiology, which is that most biological categories are only definable infunctional terms (Beckner 1959, 112-118). For instance, heart cannotbe defined except by reference to the function of hearts because no de-scription purely in terms of morphological criteriacould demarcate heartsfrom non-hearts. Biologists need a category that ranges over differentspecies, and hearts are morphologically diverse: fish have a single pumpwith only one auricle, but amphibians and most other reptiles have thesingle heartwith two auricles, and while many reptiles have the ventriclepartly partitioned, only crocodiles, birds and mammals have the two sep-arateventricles. Highly significant, moreover, is that for the purposes ofclassifying hearts, what matters is not whether the organ in question man-ages to pump blood, but whether that is what it is supposed to do. Theheart that cannot perform its proper function (because it is atrophied,clogged, congenitally malformed, or sliced in two) is still a heart.Some theorieswhich imply thatinstant ions (and piggyback traits)wouldhave properfunctions do not capturethe distinction between what an itemdoes and what it is supposed to do, and so they do not describe a notionof a properfunction that is capable of generating these biological cat-

91 make this and several of the following points in my dissertation (Neander 1983, 73-140). See also Millikan (1989, 294ff.).



15/18

THE CONCEPTUAL ANALYST'S DEFENSE 181egories which embrace both interspecies and pathological diversity. Abrief discussion of the three most interesting of these alternative theories(mentioning just one problem with each regarding biological norms) willfurther illustrate my point that the biological norms are determined bythe history of traits.(i) According to one importanttheory of functions, they are causal con-tributions to complexly achieved overall activities of the containing or-ganism or system (Cummins 1975). According to this theory, functionsare relative to our interests, and both the boundaries of the containingsystem, and which of its overall activities we focus upon, will vary withour current concerns. Now instant lions could be analyzed as complexcausal systems, in effect organized toward the achievement of certainactivities-like hunting, consuming deer, birthing cubs, and so on. Soinstant lions would have proper functions if this theory accurately de-scribed what proper functions were. But note that this kind of causalanalysis can be done on any complex causal system (plate movementsthat culminate in earthquakes, intragalactic motion, for examples). So,this theory bestows proper functions on instant lions only at the expenseof bestowing them upon a vast range of systems to which we do notnormally attribute them.

We could usefully refer to Cummins's notionof a function as a causalrole function . This notion has importantuses. Causal role functions haveplayed an important part in functionalist theories of mind, for instance(although some of us have arguedthat we should putthe function backinto functionalism by employing the more biological notion of a properfunction ).'0 However, causal role functions are not equivalent to properfunctions. While this account offered by Cummins allows us to chooseamong causal roles according to our interests, causal role functions arenonetheless purely descriptive (not normative) in the sense that the causalrole function of X is some subset of X's actual causal roles. Yet in biologythe properfunction of a trait is not necessarily anything it actually does.By way of illustration of the difficulties with this account, consider thatone thing that many organisms do, unfortunately, is die of cancer. Dyingof cancer is complexly achieved; it involves chromosome replication andcell reproductionin the growth of tumors, and as the tumors spread, theyreplace healthy tissue and functional integrity progressively deteriorates.If a tumor presses on an arteryto the brain, this is an actual causal rolethe tumor has in this pathological process, and one in which we are veryinterested. However, this causal role is not the tumors' proper function;tumors simply don't have properfunctions. Moreover, in this process ofdeterioration, proper functioning will be disrupted, and so the proper

10Seeespecially Sober (1985).



16/18

182 KAREN NEANDERfunctions of many diseased organs will cease to be among their actualeffects. The brain may fail to coordinatebodily movements, for instance,in which case this will be a proper function of the brain that is not anactual causal contribution by the brain to any complexly achieved overallactivity of the system.(ii) Most writers, not so long ago, wanted to specify certain goals towhich proper biological functions causally contribute. William Wimsatt(1972, 62-65) suggested thatthe goal is provided by background theoriesand ChristopherBoorse (1975) suggested that they are given by the con-text of enquiry in combination with these backgroundtheories. The bi-ologically given goal has been variously identified as the survival andreproductionof the individual organism within physiology (ibid., 57) andas the fitness, or long range survival of the evolutionary unit, which isbasicallya temporallydefined conceptof a breedingpopulation Wimsatt1972, 6, f.n.11). Perhaps the goal should be specified as the inclusivefitness of the individual organism. In any case, the general idea was thata proper function was a causal contributionto the specified goal.So far described, such theories suffer the same problem as that en-countered in (i). The problem is that if a biological part dysfunctionsbadly enough, it is no longer capable of making any causal contributionto fitness (or whatever), so it lacks a proper function if having a properfunction means having a causal contribution to fitness. This is unaccept-able because items that are dysfunctional are dysfunctional precisely be-cause of their incapacity to perform their proper function.An appeal to statistical normality is therefore introduced so thatproperfunctions are defined as contributions to fitness that are species-typical ,or standard ,or statisticallynormal within a class of organisms (seeWimsatt 1972, 50; and Boorse 1976, 557).1 Although a particularkidneymight fail to filter blood, kidneys typically manage to performtheirproperfunction, and so they standardlycontribute to the fitness of organisms ofthe type. On this statistical account, therefore, a dysfunctional kidneydoes have a proper function that it fails to perform because it fails tocontribute what kidneys standardlycontribute toward fitness. However,biological norms cannot be reduced to statistical norms, as we can seeby noting that dysfunction can become widespread within a populationthrough epidemics or major environmental disasters. A statistical defi-nition of biological norms implies that when a trait standardly fails toperform ts function, its function ceases to be its function;so that if enoughof us are stricken with disease (roughly, are dysfunctional) we cease tobe diseased, which is nonsense. 12

1 Thus hese theories are also forced to acknowledge that biological proper functionsapply primarily to types and only secondarily to tokens, but the fact does not fall neatlyout of the logic of the theory; it is introduced to ward off a plethora of counterexamples.12ChristopherBoorse and William Lycan have both suggested, in correspondence, that



17/18

THE CONCEPTUAL ANALYST'S DEFENSE 183(iii) Much the same problem arises for the closely related propensitytheory that Bigelow and Pargetter (1987) have proposed, in which abiological characterhas a function if it has a disposition that is apt forselection, or that enhances survival systematically, in a creature's nat-ural habitat .While this theory distinguishes properfunctions from actualcausal roles (the creatures may not actually be in their natural habitat, forinstance), it links them instead to actualdispositions and categorical basesfor these dispositions. But the problem is that dysfunctional traits do notactually have the disposition to perform their proper function. The im-paired kidney does not have a disposition that is apt for selection, or adisposition to enhance survival systematically, so according to the pro-pensity theory it does not have a proper function. But, once again, it isprecisely because it has a function that it is incapable of performing thatit counts as dysfunctional.13To conclude this section and the paper, for a trait to have a properfunction is not for it presently to have any actual causal role, statisticallytypical contribution to fitness, or disposition. Instead, a traithas a properfunction if there is something that it is supposed to do. According to myetiological theory, a trait is supposed to do whatever it was selected forby naturalselection. So understood,we can easily accommodatethe above

deviations from biological norms. It is not the function of tumorsto presson arteries to the brain because tumors have not been selected for thateffect. It is the function of kidneys (both normal aindabnormal)to filterwastes from blood because that is what kidneys did in ancestralorganismsthat caused them to be favored by natural selection (and this fact remainstrue even if renal failure becomes universal). Functional norms seem tobe determined by a history of selection, and this would explain why thebiological categories that are based on functional norms are stable, rel-ative to currentvagaries in demographic, environmental, or pathologicalfeatures. Given that biologists use the notion of a properfunction to gen-erate these stable biological categories, biologists must implicitly under-stand the notion of a properfunction as an historical notion.

REFERENCESBeckner, M. (1959), The Biological Way of Thought. New York: Columbia UniversityPress.

biological norms might be statistically relative to a larger time-slice of a population (suchas the past millennium or two). I have argued elsewhere that the suggestion will not work(Neander 1983, 85), but I need not pursue the idea here since I am here concerned withtheories that might support the claim that properfunction is an ahistorical notion.'3In another paper (Neander forthcoming), I further argue that the propensity theoryparadoxically fails to capture the forward-looking nature of proper functions, and so failsto endow proper functions with explanatory power, contrary to the positive claims whichBigelow and Pargettermake on behalf of their theory (1978, 181-182, 189-191).



18/18

184 KAREN NEANDERBigelow, J. and Pargetter, R. (1987), Functions , The Journal of Philosophy 84: 181-196.Boorse, C. (1976), Wright on Functions , The Philosophical Review 85: 70-86.Cummins, R. (1975), Functional Analysis , The Journal of Philosophy 72: 741-765.[Excerpt reprintedin N. Block (ed.), (1980), Readings in Philosophy of Psychology.Cambridge, MA: HarvardUniversity Press, pp. 185-190.]Dennett, D. (1986), Brainstorms: Philosophical Essays on Mind and Psychology. FourthPrinting. Cambridge, MA: MIT Press.Millikan, R. G. (1984), Language, Thought and Other Biological Categories: New Foun-dations for Realism. Cambridge, MA: MIT Press.(1989), In Defense of Proper Functions , Philosophy of Science 56: 288-302.Minsky, M. (1974), A Framework for Representing Knowledge . Cambridge, MA: MITAl Lab Memo 306. [Excerpts reprintedin P. Winston (ed.), (1975), The Psychologyof Computer Vision. New York: McGraw-Hill, pp. 211-277; other excerpts reprintedin J. Haugeland (ed.), (1981), Mind Design. Cambridge, MA: MIT Press, pp. 95-

128.]Nagel, E. (1977), Teleology Revisited , The Journal of Philosophy 84: 261-301.Neander, K. (1983), Abnormal Psychobiology . Ph.D. Dissertation, La Trobe Univer-sity.. (1988), Discussion: What Does Natural Selection Explain? Correction to Sober ,Philosophy of Science 55: 422-426.. (forthcoming), The Teleological Notion of 'Function ', Australasian Journal ofPhilosophy.Rosch, E. (1973), Natural Categories , Cognitive Psychology 4: 328-350.. (1975), Cognitive Representations of Semantic Categories , Journal of Experi-mental Psychology 104: 192-233.Sober, E. (1984), The Nature of Selection: Evolutionary Theory in Philosophical Focus.Cambridge, MA: MIT Press.. (1985), Panglossian Functionalism and the Philosophy of Mind , Synthese 64:165-193.Wimsatt, W. (1972), Teleology and the Logical Structure of Function Statements , Stud-ies in the History and Philosophy of Science 3: 1-80.Wright, L. (1973), Functions , The Philosophical Review 82: 139-168.. (1976), Teleological Explanations: An Etiological Analysis of Goals and Func-tions. Berkeley and Los Angeles: University of California Press.

Documents

Neander Karen Functions as Selected Effects the Conceptual Analyst's Defense