kim 1992

8/12/2019 Kim 1992

1/27

International Phenomenological Society

Multiple Realization and the Metaphysics of ReductionAuthor(s): Jaegwon KimSource: Philosophy and Phenomenological Research, Vol. 52, No. 1 (Mar., 1992), pp. 1-26Published by: International Phenomenological SocietyStable URL: http://www.jstor.org/stable/2107741

Accessed: 10/11/2008 14:12

Your use of the JSTOR archive indicates your acceptance of JSTOR's Terms and Conditions of Use, available at

http://www.jstor.org/page/info/about/policies/terms.jsp. JSTOR's Terms and Conditions of Use provides, in part, that unless

you have obtained prior permission, you may not download an entire issue of a journal or multiple copies of articles, and you

may use content in the JSTOR archive only for your personal, non-commercial use.

Please contact the publisher regarding any further use of this work. Publisher contact information may be obtained at

http://www.jstor.org/action/showPublisher?publisherCode=ips.

Each copy of any part of a JSTOR transmission must contain the same copyright notice that appears on the screen or printed

page of such transmission.

JSTOR is a not-for-profit organization founded in 1995 to build trusted digital archives for scholarship. We work with the

scholarly community to preserve their work and the materials they rely upon, and to build a common research platform that

promotes the discovery and use of these resources. For more information about JSTOR, please contact [email protected].

International Phenomenological Societyis collaborating with JSTOR to digitize, preserve and extend access to

Philosophy and Phenomenological Research.

http://www.jstor.org
http://www.jstor.org/stable/2107741?origin=JSTOR-pdfhttp://www.jstor.org/page/info/about/policies/terms.jsphttp://www.jstor.org/action/showPublisher?publisherCode=ipshttp://www.jstor.org/action/showPublisher?publisherCode=ipshttp://www.jstor.org/page/info/about/policies/terms.jsphttp://www.jstor.org/stable/2107741?origin=JSTOR-pdf

8/12/2019 Kim 1992

2/27

Philosophy andPhenomenological ResearchVol. LII,No. 1, March 1992

Multiple Realization and theMetaphysicsof ReductionJAEGWONKIMBrown University

I. IntroductionIt is part of today's conventional wisdom in philosophy of mind thatpsychological states are "multiply realizable", and are in fact so realized, ina variety of structures and organisms. We are constantly reminded that anymental state, say pain, is capable of "realization", "instantiation", or"implementation" in widely diverse neural-biological structures inhumans, felines, reptiles, mollusks, and perhaps other organisms furtherremoved from us. Sometimes we are asked to contemplate the possibilitythat extraterrestrial creatures with a biochemistry radically different fromthe earthlings', or even electromechanical devices, can "realize the samepsychology" that characterizes humans. This claim, to be called hereafter"the Multiple Realization Thesis" ('MR",' for short), is widely acceptedby philosophers, especially those who are inclined to favor the functionalistline on mentality. I will not here dispute the truth of MR, although what Iwill say may prompt a reassessment of the considerations that have led toits nearly universal acceptance.

And there is an influential and virtually uncontested view about thephilosophical significance of MR. This is the belief that MR refutespsychophysical reductionism once and for all. In particular, the classicpsychoneural identity theory of Feigl and Smart, the so-called "typephysicalism", is standardly thought to have been definitively dispatched byMR to the heap of obsolete philosophical theories of mind. At any rate, it isthis claim, that MR proves the physical irreducibility of the mental, thatwill be the starting point of my discussion.

Evidently, the current popularity of antireductionist physicalism isowed, for the most part, to the influence of the MR-based antireductionistargument originally developed by Hilary Putnam and elaborated further by

Onoccasion,"MR"willrefer o thephenomenonf multiple ealization atherhan heclaim hat uchaphenomenonxists; here hould enodangerf confusion.

MULTIPLE EALIZATIONND THEMETAPHYSICSFREDUCTION

8/12/2019 Kim 1992

3/27

8/12/2019 Kim 1992

4/27

"If there are infinitely many physical (and perhaps nonphysical)propertieswhich can realize F thenF will not be reducible to abasic physical property.Even if F can only be realized by finitelymanybasic physical properties t mightnot be reducible to a basicphysical property since the disjunction of these propertiesmightnot itself be a basic physical property i.e., occur in a fundamentalphysical law). We will understand 'multiple realizability' as in-volving such irreducibility."

This antireductionist eadingof MR continues to this day; in a recentpa-per, Ned Block writes:7

"Whatever the merits of physiological reductionism, it is notavailable to the cognitive science point of view assumed here.According to cognitive science, the essence of the mental iscomputational,and any computational state is 'multiply realiz-able' by physiological or electronic states that are not identicalwithone another,and so content cannot be identified withany oneof them."

Considerations of these sorts have succeeded in persuadinga large majorityof philosophers of mind8 to reject reductionismand type physicalism. Theupshot of all this has been impressive: MR has not only ushered in "non-reductive physicalism" as the new orthodoxy on the mind-body problem,but in the process has put the very word "reductionism" in disrepute,making reductionisms of all stripes an easy target of disdain and curtdismissals.

I believe a reappraisalof MR is overdue. There is something rightand in-structivein the antireductionistclaim based on MR and the basic argumentin its support, but I believe that we have failed to follow out theimplications of MR far enough, and have as a result failed to appreciateitsfull significance. One specific point that I will argue is this: the popularview that psychology constitutes an autonomous special science, a doctrine7 In "Can the Mind Changethe World?",Meaning and Method: Essays in Honor of Hilary

Putnam,ed. GeorgeBoolos (CambridgeUniversityPress:Cambridge,1990), p. 146.8 They include Richard Boyd, "Materialism Without Reductionism: What Physicalism

Does Not Entair', in Block, Readings in Philosophy of Psychology, vol. 1; Block, in"Introduction:What is Functionalism?" in his anthology just cited, pp. 178-79; JohnPost, The Faces of Existence (Ithaca: Cornell University Press, 1987); Derk Pereboomand Hilary Kornblith, "The Metaphysics of Irreducibility"(forthcoming in Philosoph-ical Studies). Onephilosopherwho is not impressedby the received view of MR is DavidLewis; see his "Review of Putnam" n Block, Readings in Philosophy of Psychology, vol.1.

MULTIPLEEALIZATIONNDTHEMETAPHYSICSFREDUCTION 3

8/12/2019 Kim 1992

5/27

heavily promotedin the wake of the MR-inspiredantireductionistdialectic,may in fact be inconsistentwith the real implicationsof MR. Our discussionwill show that MR, when combined with certain plausible metaphysicaland methodological assumptions, leads to some surprising conclusionsabout the status of the mental and the nature of psychology as a science. Ihope it will become clear that the fate of type physicalism is not among themore interestingconsequences of MR.

II. Multiple RealizationIt was Putnam, in a paper published in 1967,9 who first injected MR intodebates on the mind-body problem. According to him, the classic reductivetheories of mind presupposed the following naive picture of howpsychological kinds (properties,event and state types, etc.) are correlatedwith physical kinds:

For each psychological kind M there is a unique physical(presumably,neurobiological)kind P that is nomologically coex-tensive with it (i.e., as a matterof law, any system instantiatesMat t iff thatsystem instantiatesP at t).

(We may call this "the Correlation Thesis".) So take pain: the CorrelationThesis has it thatpain as an event kind has a neural substrate,perhaps as yetnot fully and precisely identified, that, as a matter of law, always co-occurwith it in all pain-capable organisms and structures. Here there is nomention of species or types of organisms or structures:the neural correlateof pain is invariantacross biological species and structure ypes. In his 1967paper, Putnam pointed out something that, in retrospect, seems all tooobvious:'0

"Considerwhat the brain-statetheorist has to do to make goodhis claims. He has to specify a physical-chemical state such thatany organism (not-justa mammal) is in pain if and only if (a) itpossesses a brain of a suitable physical-chemical structure;and(b) its brain is in that physical-chemical state. This means thatthe physical-chemicalstate in question mustbe a possible state ofa mammalianbrain,a reptilianbrain, a mollusc's brain (octopusesare mollusca, and certainly feel pain), etc. At the same time, it

9 Hilary Putnam,"Psychological Predicates", n W. H. Capitanand D. D. Merrill, eds., Art,Mind, and Religion (Pittsburgh: University of Pittsburgh, 1967); reprinted with a newtitle, "The Nature of Mental States", in Ned Block, ed., Readings in Philosophy ofPsychology, vol. 1 (Cambridge:HarvardUniversity Press, 1980).10 "The Nature of Mental States", p. 228 (in the Block volume).

4 JAEGWONIM

8/12/2019 Kim 1992

6/27

must not be a possible brain of any physically possible creaturethatcannot feel pain."

Putnamwent on to argue that the CorrelationThesis was empirically false.Later writers, however, have stressed the multiple realizability of themental as a conceptual point: it is an a priori, conceptual fact aboutpsychological properties that they are "second-order"physical properties,and that their specification does not include constraints on the manner oftheir physical implementation."' Many proponents of the functionalistaccountof psychological termsandpropertieshold such a view.

Thus, on the new, improvedpicture,therelationshipbetween psychologi-cal andphysical kinds is somethinglike this: there is no single neuralkind Nthat "realizes" pain, across all types of organisms or physical systems;rather,there is a multiplicity of neural-physicalkinds, Nh,N, Nm,... such thatNh realizes pain in humans,N, realizes pain in reptiles, Nm realizes pain inMartians, etc. Perhaps, biological species as standardlyunderstoodare toobroad to yield unique physical-biological realizationbases; the neuralbasisof pain could perhapschange even in a single organismover time. But themain point is clear: any system capable of psychological states (that is, anysystem that"hasa psychology") falls undersome structure ype T such thatsystems with structureT sharethe same physical base for each mental state-kind that they are capable of instantiating (we should regard this asrelativized with respect to time to allow for the possibility that anindividual may fall underdifferent structuretypes at different times). Thusphysical realizationbases for mentalstates must be relativized to species or,better,physical structure-types.We thus have the following thesis:

If anythinghas mentalpropertyM at time t, there is some physi-cal structure ype T and physical propertyP such that it is a sys-tem of type T at t and has P at t, and it holds as a matterof lawthatall systems of type T have M at a time just in case they haveP at the time.

We may call this "the Structure-RestrictedCorrelation Thesis" (or "theRestricted CorrelationThesis" for short).It may have been noticed that neither this nor the correlation thesis

speaks of "realization".'2The talk of "realization"is not metaphysicallyThus, Post says, "Functional and intentional states are defined without regard to theirphysical or other realizations",The Faces of Existence, p. 161. Also comparethe earlierquotationfrom Block.

12 As far as I know, the term "realization"was firstused in something like its present senseby Hilary Putnam in "Minds and Machines", in Sydney Hook, ed., Dimensions of Mind(New York: New York University Press, 1960).


8/12/2019 Kim 1992

7/27

neutral: the idea that mental properties are "realized" or "implemented"by physical propertiescarrieswith it a certainontological picture of mentalpropertiesas derivative and dependent.There is the suggestion thatwhen welook at concrete reality there is nothing over and beyond instantiationsofphysical properties and relations, and that the instantiation on a givenoccasion of an appropriatephysical propertyin the right contextual (oftencausal) setting simply counts as, or constitutes, an instantiationof a mentalproperty on that occasion. An idea like this is evident in the functionalistconceptionof a mentalpropertyas extrinsically characterized n termsof its"causal role", where what fills this role is a physical (or, at any rate,nonmental) property(the latter property will then be said to "realize"themental property in question). The same idea can be seen in the relatedfunctionalist proposal to construe a mental property as a "second-orderproperty"consisting in the having of a physical propertysatisfying certainextrinsic specifications. We will recur to this topic later; however, weshould note that someone who accepts either of the two correlationthesesneed not espouse the "realization"diom. Thatis, it is prima facie a coherentposition to think of mental properties as "first-orderproperties" in theirown right, characterized by their intrinsic natures (e.g., phenomenal feel),which, as it happens, turn out to have nomological correlates in neuralproperties.(In fact, anyone interestedin defendinga serious dualistpositionon the mental should eschew the realization talk altogether and considermental properties as first-order properties on a par with physicalproperties.) The main point of MR that is relevant to the antireductionistargument it has generated is just this: mental properties do not havecomically coextensive physical properties, when the latter areappropriatelyindividuated. It may be thatproperties that are candidatesforreduction must be thought of as being realized, or implemented, bypropertiesin the prospective reduction base;'3that is, if we think of certainproperties as having their own intrinsic characterizationsthat are entirelyindependent of anotherset of properties, there is no hope of reducing theformerto the latter.But this point needs to be argued,and will, in any case,not play a role in what follows.

Assume thatpropertyM is realizedby propertyP. How are M andP re-lated to each other and, in particular,how do they covary with each other?LePoreandLoewersay this:'4

13 On this point see Robert Van Gulick, "Nonreductive Materialism and IntertheoreticConstraints", in Emergence or Reduction?, ed. Ansgar Beckennann, Hans Flohr, andJaegwon Kim (forthcoming from De Gruyter).14 "More on Making Mind Matter",p. 179.

6 JAEGWONIM

8/12/2019 Kim 1992

8/27

"The usual conception is thate's being P realizes e's being F iff eis P and there is a strongconnectionof some sort between P andF. We propose to understand his connection as a necessary con-nection which is explanatory. The existence of an explanatoryconnection between two properties is stronger than the claimthat P -> F is physically necessary since not every physicallynecessaryconnectionis explanatory."

Thus, LePore and Loewer requireonly that the realization base of M besufficientfor M, not both necessaryand sufficient. This presumablyis in re-sponse to MR:if pain is multiplyrealizedin threeways as above, each of Nh,Nr, andNmwill be sufficient for pain, andnone necessaryfor it. This I believeis not a correctresponse, however; the correctresponse is not to weaken thejoint necessity and sufficiency of thephysicalbase, butrather o relativize it,as in the Restricted CorrelationThesis, with respect to species or structuretypes. For suppose we are designing a physical system that will instantiateacertain psychology, and let M1,..., M. be the psychological propertiesrequired by this psychology. The design process must involve thespecification of an n-tuple of physical properties, P.,..., P., all of theminstantiableby the system, such that for each i, Pi constitutes a necessaryand sufficient condition in this system (and others of relevantly similarphysical structure),not merely a sufficient one, for the occurrence of Mi.(Each such n-tuple of physical properties can be called a "physicalrealization"of the psychology in question.15)That is, for each psychologicalstate we must design into the system a nomologically coextensive physicalstate. We must do this if we are to controlboth the occurrenceand non-occurrenceof thepsychologicalstates involved,andcontrolof this kindnecessary if we are to ensure that the physical device will properlyinstantiatethe psychology. (This is especially clear if we thinkof building acomputer; computer analogies loom large in our thoughts about"realization".)

But isn't it possible for multiple realization to occur "locally" as well?That is, we may want to avail ourselves of the flexibility of allowing a psy-chological state, or function, to be instantiatedby alternative mechanismswithin a single system. This means thatPi can be a disjunction of physicalproperties;thus, Mi is instantiated n the system in question at a time if andonly if at least one of the disjuncts of Pi is instantiatedat that time. The up-shot of all this is thatLePoreand Loewer's condition that P -* M holds as amatter of law needs to be upgraded to the condition that, relative to the

Cf. HartryField, "Mental Representation", in Block, Readings in Philosophy of Psy-chology (Cambridge:HarvardUniversity Press, 1981), vol. 2.


8/12/2019 Kim 1992

9/27

species or structure-type in question (and allowing P to be disjunctive),P " M holds as a matter of law."6

For simplicity let us suppose thatpain is realizedin three ways as above,by Nhin humans,N, in reptiles,andN. in Martians.The finitudeassumptionis not essential to any of my arguments: f the list is not finite, we will havean infinite disjunctionrather than a finite one (alternatively,we can talk interms of "sets" of such propertiesinstead of their disjunctions). If the listis "open-ended",that's all right, too; it will not affect the metaphysics ofthe situation. We allowed above the possibility of a realization base of apsychological property itself being disjunctive; to get the discussion going,though, we will assume that these Ns, the three imagined physicalrealizationbases of pain, are not themselves disjunctive-or, at any rate, thattheir status as properties is not in dispute. The proprietyand significance of"disjunctiveproperties" s precisely one of the principal issues we will bedealing with below, and it will make little differencejust at what stage thisissue is faced.

III. Disjunctive Properties and Fodor's ArgumentAn obvious initial response to the MR-based argumentagainst reducibilityis "the disjunctionmove": Why not take the disjunction,Nhv N, v Nm,as thesingle physical substrateof pain?In his 1967 paper,Putnamconsiders such amove but dismisses it out of hand: "Granted, n such a case the brain-statetheorist can save himself by ad hoc assumptions(e.g., defining the disjunc-tion of two states to be a single 'physical-chemical state'), but this does nothave to be taken seriously".17Putnam gives no hint as to why he thinks thedisjunction strategy does not merit serious consideration.

If there is something deeply wrong with disjunctions of the sortinvolved here, that surely isn't obvious; we need to go beyond a sense ofunease with such disjunctions and develop an intelligible rationale forbanning them. Here is where Fodor steps in, for he appears to have anargument for disallowing disjunctions. As I see it, Fodor's argument in"Special Sciences" depends crucially on the following two assumptions:

(1) To reducea special-science theoryTMo physical theoryTp,each"kind" n TMpresumably,representedby a basic predicateof TM)must have a nomologically coextensive "kind" n Tp;

16 What of LePore and Loewer's condition (ii), the requirementthat the realization basis"explain" the realized property?Something like this explanatory relation may well beentailed by the realization relation; however, I do not believe it should be part of thedefinition of "realization"; hat such an explanatory relation holds should be a conse-quence of the realization relation, not constitutive of it.17 "TheNatureof MentalStates", p. 228 (in the Block volume).

8 JAEGWON IM

8/12/2019 Kim 1992

10/27

(2) A disjunction of heterogeneouskinds is not itself a kind.Point (1) is apparentlypromptedby the derivationalmodel of intertheo-

retic reduction due to Ernest Nagel:"8 he reductionof T2 to T. consists in thederivationof laws of T2 from the laws of T1,in conjunction with "bridge"laws or principles connecting T2-termswith T1-terms.Although this charac-terization does not in general requirethateach T2-termbe correlatedwith acoextensive T,-term, the natural thought is that the existence of T1-coextensions for T2-termswould in effect give us definitions of T2-termsnT,-terms,enablingus to rewrite T2-lawsexclusively in the vocabulary of T1;we could then derive these rewrites of T2-lawsfrom the laws of T. (if theycannot be so derived, we can add themas additionalT,-laws-assuming boththeories to be true).

Another thought that again leads us to look for T,-coextensions for T2-terms is this: for genuine reduction, the bridge laws must be construed asproperty identities, not mere property correlations-namely, we must be ina position to identify the propertyexpressedby a given T2-term say, water-solubility) with a property expressed by a term in the reduction base (say,having a certain molecular structure).This of course requires that each T2-term have a nomic (or otherwise suitably modalized) coextension in thevocabulary of the reduction base. To put it another way, ontologicallysignificant reductionrequiresthe reduction of higher-level properties, andthis in turnrequires (unless one takes an eliminativist stance) that they beidentified with complexes of lower-level properties. Identity of propertiesof course requires, at a minimum, an appropriately modalizedcoextensivity.19

So assumeM is a psychologicalkind,and let us agreethatto reduceM, orto reduce the psychological theory containing M, we need a physical coex-tension, P. for M. But why should we suppose that P must be a physical"kind"?But what is a "kind",anyway? Fodor explains this notion in termsof law, saying that a given predicateP is a "kind predicate"of a science justin case the science contains a law with P as its antecedent or consequent.20There are variousproblems with Fodor's characterization,but we don't need

18The Structureof Science (New York:Harcourt,Brace & World, 1961), chap. 11.19 My remarkshere andthe precedingparagraph ssumethat the higher-level theoryrequiresno "correction" n relation to the base theory. With appropriate aveats and qualifications,they should apply to models of reductionthat allow such corrections,or models that onlyrequire he deduction of a suitableanalogue, or "image", n the reductionbase-as long asthe departuresare not so extreme as to warrant alk of replacement or elimination ratherthan reduction. Cf. PatriciaChurchland,Neurophilosophy (Cambridge:The MIT Press,1986), chap.7.20 See "Special Sciences", pp. 132-33 (in Representations).


8/12/2019 Kim 1992

11/27

to take its exact wordingseriously;the main idea is thatkinds, or kind predi-cates, of a science are those thatfigurein the laws of thatscience.

To return to our question, why should "bridge laws" connect kinds tokinds, in this special sense of "kind"?To say that bridge laws are "laws"and that, by definition, only kindpredicatescan occur in laws is not much ofan answer. For that only invites the furtherquestion why "bridge laws"ought to be "laws"-what would be lacking in a reductive derivation ifbridge laws were replaced by "bridge principles"which do not necessarilyconnect kinds to kinds.21But what of the considerationthat these principlesmust represent property identities? Does this force on us the requirementthat each reduced kind must find a coextensive kind in the reductionbase?No; for it isn't obvious why it isn't perfectly proper to reduce kinds byidentifying them with properties expressed by non-kind (disjunctive)predicates n the reductionbase.

There is the following possible argumentfor insisting on kinds: if M isidentified with non-kind Q (orM is reduced via a biconditionalbridge princi-ple "M

8/12/2019 Kim 1992

12/27

IV. Jade, Jadeite, and NephriteLet me begin with an analogy thatwill guide us in our thinkingabout mul-tiply realizable kinds.

Considerjade: we are told thatjade, as it turnsout, is not a mineralkind,contraryto what was once believed; rather, ade is comprisedof two distinctminerals with dissimilar molecular structures, jadeite and nephrite.Consider the following generalization:

(L) Jade s greenWe may have thought,before the discovery of the dual nature of jade, that(L) was a law, a law about jade; and we may have thought,with reason, that(L) had been strongly confirmedby all the millions of jade samples that hadbeen observedto be green (andnone thathadbeen observed not to be green).We now know better: (L) is really a conjunctionof these two laws:

(L,) Jadeite s green(L2) Nephrite is green

But (L) itself might still be a law as well; is that possible? It has thestandardbasic form of a law, and it apparentlyhas the power to supportcounterfactuals: f anythingwerejade-that is, if anythingwere a sample ofjadeite or of nephrite-then, in either case, it would follow, by law, that itwas green. No problem here.

But thereis another standardmark of lawlikeness thatis often cited, andthis is "projectibility", he ability to be confirmed by observation of "posi-tive instances".Any generalized conditional of the form "All Fs are G" canbe confirmedby the exhaustionof the class of Fs-that is, by eliminating allof its potential falsifiers. It is in this sense that we can verify suchgeneralizationsas "All the coins in my pockets are copper" and "Everyonein this room is either first-bornor an only child". Lawlike generalizations,however, are thought to have the following furtherproperty:observation ofpositive instances, Fs thatareGs, can strengthenourcredencein the next F'sbeing G. It is this kind of instance-to-instanceaccretionof confirmation hatis supposed to be the hallmark of lawlikeness; it is what explains thepossibility of confirming a generalization about an indefinitely large classof items on the basis of a finitenumberof favorableobservations. This roughcharacterizationof projectibility should suffice for our purposes.

MULTIPLE EALIZATON N TH METPHYSICSOFREDUCTON 11

8/12/2019 Kim 1992

13/27

Does (L), "Jade is green", pass the projectibility test? Here we seem tohave a problem.23For we can imagine this: on re-examiningthe records ofpast observations, we find, to our dismay, that all the positive instances of(L), that is, all the millions of observed samples of green jade, turn out tohave been samplesof jadeite, and none of nephrite If this should happen, weclearly would not, and should not, continue to think of (L) as wellconfirmed.All we have is evidence stronglyconfirming(L,), and none havinganything to do with (L2). (L) is merely a conjunctionof two laws, one wellconfirmedand the other with its epistemic status wholly up in the air. Butall the millions of greenjadeite samples are positive instances of (L): theysatisfy both the antecedentand the consequentof (L). As we have just seen,however, (L) is not confirmed by them, at least not in the standardway weexpect. And the reason, I suggest, is thatjade is a true disjunctive kind, adisjunctionof two heterogeneousnomic kinds which, however, is not itself anomic kind.24

That disjunctionis implicated in this failure of projectibility can be seenin the following way: inductive projection of generalizations like (L) withdisjunctive antecedents would sanction a cheap, and illegitimate,confirmation procedure. For assume that "All Fs are G" is a law that hasbeen confirmed by the observation of appropriately numerous positiveinstances, things thatare both F and G. But these are also positive instancesof the generalization "All things that are F or H are G", for any H youplease. So, if you in general permit projection of generalizations with adisjunctiveantecedent,this lattergeneralizationis also well confirmed. But"All things that are F or H are G" logically implies "All Hs are G". Anystatement implied by a well confirmed statement must itself be wellconfirmed.25So "All Hs are G" is well confirmed-in fact, it is confirmedby the observationof Fs that are Gs

One might protest:"Look,the very same strategycan be applied to some-thing that is a genuine law. We can think of any nomic kind-say, being anemerald-as a disjunction, being an African emerald or a non-Africanemerald. This would make 'All emeralds are green' a conjunction of two23 The points to follow concerningdisjunctivepredicates were developed abouta decadeago;

however, I have just come across some related and, in some respects similar, points inDavid Owens's interesting paper "DisjunctiveLaws", Analysis 49 (1989): 197-202. Seealso William Seager, "DisjunctiveLaws andSupervenience",Analysis 51 (1991): 93-98.24 This can be taken to define one useful sense of kind heterogeneity: two kinds are het-erogeneous with respectto each other just in case their disjunction s not a kind.25 Note: this doesn't say that for any e, if e is "positive evidence" for h and h logicallyimpliesj, thene is positive evidence forj. Aboutthe latterprinciplethere is some dispute;see Carl G. Hempel, "Studies in the Logic of Confirmation", reprinted in Hempel,Aspects of ScientificExplanation (New York: The Free Press, 1965), especially pp. 30-35;Rudolf Carnap, Logical Foundations of Probability (Chicago: University of ChicagoPress, 1950), pp. 471-76.

12 JAEGWONiM

8/12/2019 Kim 1992

14/27

laws, 'All African emeralds are green' and 'All non-African emeralds aregreen'. But surely this doesn't show there is anything wrong with thelawlikeness of 'All emeralds are green"'. Our reply is obvious: thedisjunction, "being an African emerald or non-African emerald",does notdenote some heterogeneously disjunctive, nonnomic kind; it denotes aperfectly well-behaved nomic kind, that of being an emerald There isnothing wrong with disjunctive predicatesas such; the trouble arises whenthe kinds denoted by the disjoined predicates are heterogeneous, "wildlydisjunctive", so that instances falling under them do not show the kind of"similarity", or unity, that we expect of instances falling under a singlekind.

The phenomenon under discussion, therefore, is related to the simplemaxim sometimes claimed to underlie inductive inference: "similar thingsbehave in similarways","samecause, same effect", andso on. The source ofthe trouble we saw with instantial confirmationof "All jade is green" isthe fact, or belief, that samples of jadeite and sample of nephrite do notexhibit an appropriate"similarity"with respect to each other to warrantinductive projections from the observed samples of jadeite to unobservedsamples of nephrite. But similarity of the requiredsort presumably holdsfor African emeralds and non-African emeralds-at least, that is what webelieve, and that is what makes the "disjunctive kind", being an Africanemeraldor a non-Africanemerald,a single nomic kind. More generally, thephenomenon is relatedto the point often made about disjunctive properties:disjunctive properties, unlike conjunctive properties, do not guaranteesimilarity for instances falling underthem. And similarity, it is said, is thecore of our idea of a property.If that is your idea of a property, you willbelieve that there are no such things as disjunctive properties (or "negativeproperties"). More precisely, though, we should remember that propertiesare not inherently disjunctive or conjunctive any more than classes areinherently unions or intersections, and that any propertycan be expressed bya disjunctive predicate. Properties of course can be conjunctions, ordisjunctions, of other properties. The point about disjunctive properties isbest put as a closure condition on properties: the class of properties is notclosed underdisjunction (presumably,nor under negation). Thus, there maywell be propertiesP and Q such thatP or Q is also a property,but its beingso doesn't follow fromthe mere fact thatP and Q areproperties.26

5 On issues concerning properties, kinds, similarity, and lawlikeness, see W.V. Quine,"NaturalKinds" in Ontological Relativity and OtherEssays (New York:Columbia Uni-versity Press, 1969); David Lewis, "New Work for a Theoryof Universals",AustralasianJournal of Philosophy 61 (1983): 347-77; D. M. Armstrong, Universals (Boulder,Colorado: Westview Press, 1989).


8/12/2019 Kim 1992

15/27

V. Jade and PainLet us now return o pain andits multiplerealizationbases, Nh, N., andNm.Ibelieve the situationhere is instructively parallel to the case of jade in rela-tion to jadeite and nephrite.It seems that we think of jadeite and nephriteasdistinctkinds (andof jade not as a kind) because they are differentchemicalkinds. But why is their being distinct as chemical kinds relevant here? Be-cause many importantpropertiesof minerals,we think,are supervenienton,and explainablein termsof, theirmicrostructure, ndchemical kinds consti-tute a microstructuraltaxonomy that is explanatorily rich and powerful.Microstructure s important,in short, because macrophysical propertiesofsubstances are determined by microstructure. These ideas make up our"metaphysics" of microdetermination or propertiesof minerals and othersubstances, a background of partly empirical and partly metaphysicalassumptionsthatregulateour inductive and explanatorypractices.

The parallel metaphysical underpinnings for pain, and other mentalstates in general, are, first, the belief, expressed by the RestrictedCorrelationThesis, thatpain, or any other mental state, occurs in a systemwhen, and only when, appropriatephysical conditions are present in thesystem, and, second, the corollary belief that significant properties ofmental states, in particular nomic relationships amongst them, are due to,and explainable in terms of, the propertiesand causal-nomic connectionsamong their physical "substrates". will call the conjunction of these twobeliefs "the Physical Realization Thesis".27Whether or not the micro-explanationof the sort indicatedin the second half of the thesis amounts toa "reduction" s a question we will take up later. Apart from this question,though, the Physical Realization Thesis is widely accepted by philosopherswho talk of "physical realization",and this includes most functionalists; itis all but explicit in LePoreand Loewer, for example, and in Fodor.Define a property,N, by disjoining Nh, N,, andNm;that is, N has a dis-junctive definition,Nh v N, v Nm. If we assume, with those who endorse theMR-based antireductionistargument, hat Nh,N, and Nmare a heterogeneouslot, we cannot make the heterogeneity go away merely by introducinga sim-pler expression, "N"; if there is a problem with certain disjunctive

27 This tenn is a little misleading since the two subtheseshave been stated without the term"realization"andmay be acceptable to those who would reject the "realization" diom inconnection with the mental. I use the term since we are chiefly addressing philosophers(mainly functionalists) who construe the psychophysical relation in terms of realization,rather han, say, emergence or brutecorrelation.28 See "Special Sciences", and "Making Mind Matter More", Philosophical Topics 17(1989): 59-79.

14 JAEGWONIM

8/12/2019 Kim 1992

16/27

properties, it is not a linguistic problem about the form of expressions usedto refer to them.

Now, we put the following question to Fodor and like-minded philoso-phers: If pain is nomically equivalent to N, the property claimed to bewildly disjunctiveandobviouslynonnomic,whyisn'tpain itself equallyheterogeneous nd nonnomic s a kind?Whyisn'tpain'srelationshipo itsrealization bases, Nh,Nr, andNm analogous to jade's relationship to jadeiteand nephrite?If jade turns out to be nonnomicon account of its dual "reali-zations" in distinct microstructures,why doesn't the same fate befall pain?After all, the group of actual and nomologically possible realizations ofpain, as they are describedby the MR enthusiastswith such imagination,isfar moremotley than the two chemicalkindscomprisingjade.

I believe we should insist on answersto these questionsfrom those func-tionalists who view mental properties as "second-order"properties, i.e.,properties that consist in having a property with a certain functionalspecifications9Thus, pain is said to be a second-orderproperty n thatit is theproperty f having omepropertywitha certain pecificationn termsof itstypical causes and effects and its relation to other mental properties; callthis "specificationH". The point of MR. on this view, is that there is morethan one property that meets specificationH-in fact, an open-ended set ofsuch properties, t will be said. But painitself, it is argued,is a moreabstractbut well-behaved propertyat a higher level, namely the propertyof havingone of these propertiesmeeting specificationH. It shouldbe clear why a po-sition like this is vulnerable to the questions that have been raised. For thepropertyof having propertyP is exactly identical with P, and the propertyof havingoneof theproperties,P1.P2,..., Pn,is exactly identicalwith the dis-junctive property,PI v P2v.. .v Pn.On the assumptionthatNh,Nr, andNmareall the properties satisfying specification H, the property of having apropertywith H, namely pain, is none other than the property of having ei-therNhor Nr or Nm30-namely, the disjunctiveproperty,Nhv Nr v Nm Wecannot hide the disjunctive characterof pain behind the second-orderexpres-sion, "thepropertyof having a property with specification H". Thus, on theconstrualof mentalproperties as second-orderproperties, mentalpropertieswill in general turn out to be disjunctions of their physical realizationbases. It is difficult to see how one could have it both ways-that is, tocastigate Nh v Ni v Nm as unacceptably disjunctive while insisting on theintegrityof pain as a scientific kind.

Moreover, when we thinkabout makingprojections over pain, very muchthe same worry shouldarise about theirproprietyas did for jade. Considera29 See, e.g., Block, "Can he MindChangetheWodd?",p. 155.30 We might keep in mind the close relationship between disjunction and the existentialquantifierstandardlynoted in logic textbooks.


8/12/2019 Kim 1992

17/27

possible law: "Sharp pains administeredat randomintervals cause anxietyreactions". Suppose this generalizationhas been well confirmed for humans.Should we expect on that basis that it will hold also for Martianswhosepsychology is implemented (we assume) by a vastly different physicalmechanism?Not if we accept the Physical Realization Thesis, fundamentalto functionalism, that psychological regularities hold, to the extent thatthey do, in virtue of the causal-nomological regularities at the physicalimplementation level. The reason the law is true for humans is due to theway the human brain is "wired";the Martianshave a brainwith a differentwiring plan, and we certainly should not expect the regularityto hold forthem just because it does for humans.3" Pains cause anxiety reactions"mayturn out to possess no more unity as a scientific law than does "Jade isgreen".Suppose that in spite of all this Fodor insists on defending pain as anomic kind. It isn't clear thatthat would be a viable strategy.For he wouldthen owe us an explanationof why the "wildly disjunctive"N, which afterall is equivalent to pain, is not a nomic kind. If a predicate is nomicallyequivalent to a well-behaved predicate, why isn't that enough to show thatit, too, is well behaved, and expresses a well-behaved property?To say, asFodor does,32that "it is a law that..." is "intensional"and does not permitsubstitution of equivalent expressions ("equivalent" n various appropriatesenses) is merely to locate a potentialproblem,not to resolve it.

Thus, the nomicity of pain may lead to the nomicity of N; but this isn'tvery interesting. For given the Physical Realization Thesis, and the priorityof the physical implicit in it, our earlier line of argument, leading from thenonnomicityof N to the nonnomicity of pain, is more compelling. We must,I think, take seriously the reasoning leading to the conclusion that pain, andother mental states, might turn out to be nonnomic. If this turns out to bethe case, it puts in serious jeopardy Fodor's contention that its physicalirreducibility renders psychology an autonomous special science. If painfails to be nomic, it is not the sort of property n terms of which laws can beformulated; and "pain" is not a predicate that can enter into a scientifictheory that seeks to formulate causal laws and causal explanations. And thesame goes for all multiply realizable psychological kinds-which,according to MR, means all psychological kinds. There are no scientifictheories of jade, and we don't need any; if you insist on having one, you can31 It may be a complicated affair to formulate this argument within certain functionalistschemes; if, for example, mental properties are functionally defined by Ramseyfying atotal psychological theory, it will turn out that humans and Martians cannot share any

psychological state unless the same total psychology (including the putative law inquestion) is true (or held to be true) for both.32 "SpecialSciences",p. 140 (in Representations).

16 JAEGWONKiM

8/12/2019 Kim 1992

18/27

help yourself with the conjunction of the theoryof jadeite and the theoryofnephrite. In the same way, there will be theories about human pains(instances of Nh), reptilianpains (instancesof Nr), and so on; but there willbe no unified, integratedtheory encompassing all pains in all pain-capableorganisms, only a conjunction of pain theories for appropriatelyindividuated biological species and physical structure-types. Scientificpsychology, like the theoryof jade, gives way to a conjunctionof structure-specific theories.If this is right, the correctconclusion to be drawnfrom theMR-inspired antireductionistargumentis not the claim that psychology isan irreducible and autononomousscience, but something thatcontradictsit,namely that it cannot be a science with a unified subjectmatter. This is thepicture that is beginning to emerge from MR when combined with thePhysical RealizationThesis.Thesereflections havebeen promptedby the analogywith the case of jade;it is a strong and instructiveanalogy, I think, and suggests the possibility ofa general argument. In the following section I will develop a directargument,with explicit premises and assumptions.

VI. Causal Powers and Mental KindsOne crucialpremisewe need for a directargument s a constrainton conceptformation,or kind individuation,in science that has been aroundfor manyyears; it has lately been resurrectedby Fodor in connectionwith content ex-ternalism.33A precise statementof the constraintmay be difficult and con-troversial,but its main idea can be put as follows:

[Principleof CausalIndividuationof Kinds] Kinds in science areindividuated on the bass of causal powers; that is, objects andevents fall under a kind, or share in a property, insofar as theyhave similarcausal powers.

I believe this is a plausible principle,and it is, in any case, widely accepted.We can see that this-principleenables us to give a specific interpretationto the claim thatNh,Nr,andNmare heterogeneous as kinds: the claim mustmean thatthey are heterogeneousas causal powers-that is, they are diverseas causal powers and enter into diverse causal laws. This must mean, giventhe Physical RealizationThesis, thatpain itself can show no more unity as acausalpowerthanthe disjunction,Nhv N, v Nm.Thisbecomes especially clear

3 See, e.g., Carl G. Hempel, Fundamentals of Concept Formation in Empirical Science(Chicago: University of Chicago Press, 1952); W.Q. Quine, "NaturalKinds".Fodor givesit an explicit statement in Psychosemantics (Cambridge: MI Press, 1988), chap. 2. Aprinciple like this is often invoked in the currentexternalismfintemalism debate aboutcontent; most principalparticipants n this debate seem to accept it.


8/12/2019 Kim 1992

19/27

if we set forth the following principle, which arguably is implied by thePhysicalRealization Thesis (butwe need not makean issue of this here):

[The Causal Inheritance Principle] If mental property M isrealized in a system at t in virtue of physical realizationbase P.the causal powers of this instance of M are identical with thecausalpowersof P.34

It is important to bear in mind that this principle only concerns the causalpowers of individual instances of M; it does not identify the causal powersof mental propertyM in general with the causal powers of some physicalproperty P; such identification is precluded by the multiple physicalrealizability of M.

Why should we accept this principle? Let us just note that to deny itwould be to accept emergent causal powers: causal powers that magicallyemerge at a higher-level and of which there is no accounting in terms oflower-level propertiesand their causal powers and nomic connections. Thisleads to the notorious problemof "downwardcausation" and the attendantviolation of the causal closure of the physical domain.33I believe that aserious physicalist would find these consequences intolerable.

It is clear that the Causal InheritancePrinciple, in conjunction with thePhysical Realization Thesis, has the consequence that mental kinds cannotsatisfy the Causal Individuation Principle, and this effectively rules outmental kinds as scientifickinds. The reasoningis simple: instances of M thatarerealizedby the samephysical base mustbe grouped underone kind, sinceex hypothesi the physical base is a causal kind; and instances of M withdifferentrealizationbases mustbe grouped under distinctkinds, since, againex hypothesi, these realization bases are distinct as causal kinds. Given thatmental kinds are realized by diverse physical causal kinds, therefore, itfollows thatmental kinds are not causal kinds, and hence are disqualifiedasproperscientific kinds. Each mental kind is sunderedinto as many kinds asthere are physical realization bases for it, and the psychology as a sciencewith disciplinary unity turnsout to be an impossible project.

34 A principle like this is sometimes put in terms of "supervenience"and "superveniencebase" ratherthan "realization"and "realizationbase". See my "Epiphenomenaland Su-pervenientCausation",MidwestStudies in Philosophy 9 (1984): 257-70. Fodorappears oacceptjust such a principle of supervenientcausationfor mentalproperties n chap. 2 of hisPsychosemantics. In "The Metaphysics of Irreducibility" Pereboom and Kornblithappear o rejectit.

35 For more details see my "'Downward Causation' in Emergentism and NonreductivePhysicalism", orthcoming n Emergence or Reduction?,ed. Beckermann,Flohr,andKim,and "The Nonreductivist's Troubles with Mental Causation", forthcoming in MentalCausation, ed. John Heil and Alfred Mele (OxfordUniversity Press).

18 JAEGWONIM

8/12/2019 Kim 1992

20/27

What is the relationshipbetween this argumentand the argumentadum-bratedin our reflections based on the jade analogy? At first blush, the twoargumentsmight seem unrelated: he earlierargumentdepended chiefly onepistemological considerations,considerationson inductiveprojectibilityofcertainpredicates, whereasthe crucialpremiseof the second argument s theCausal Kind Individuation Principle, a broadly metaphysical andmethodological principle about science. I think, though, that the twoargumentsare closely related, and the key to seeing the relationshipis this:causal powers involve laws, and laws are regularities that are projectible.Thus, if pain (orjade) is not a kind over which inductiveprojections can bemade, it cannotenterinto laws, andthereforecannotqualifyas a causalkind;and this disqualifies it as a scientific kind. If this is right, the jade-inspiredreflections provide a possible rationale for the Causal IndividuationPrinciple. Fleshing out this rough chain of reasoning in precise terms,however, goes beyond what I can attempt n this paper.

VII. The Status of Psychology: Local ReductionsOur conclusion at this point, therefore,is this: If MR is true,psychologicalkinds are not scientific kinds. What does this imply about the statusof psy-chology as a science? Do our considerations show that psychology is apseudo-sciencelike astrologyandalchemy?Of course not. The crucialdiffer-ence, from the metaphysical point of view, is thatpsychology has physicalrealizations, but alchemy does not. To have a physical realization is to bephysically groundedand explainablein terms of the processes at an underly-ing level. In fact, if each of the psychological kinds posited in a psychologi-cal theory has a physical realization for a fixed species, the theory can be"locally reduced" to the physical theory of that species, in the followingsense. Let S be the species involved; for each law Lmof psychological theoryT, S + Lm(the proposition that Lmholds for members of S) is the "S-re-stricted"version of Lm;and S -4 Tm s the S-restrictedversion of Tm,the setof all S-restricted laws of Tm.We can then say that Tmis "locally reduced"for species S to an underlying heory,Tp, ust in case S -* Tms reduced to Tp.And the latterobtains just in case each S-restricted law of Tm,S < Lm,36 isderivable from the laws of the reducing theory Tp, taken together withbridge laws. What bridge laws suffice to guarantee the derivation?Obviously, an arrayof S-restrictedbridge laws of the form, S -+ (Mi*-4 Pi),for each mentalkindMi.Justas unrestrictedpsychophysicalbridge laws canunderwrite a "global" or "uniform"reduction of psychology, species- orstructure-restrictedbridge laws sanction its "local" reduction.36 Or an appropriately orrectedversion thereof (this qualificationapplies to the bridge lawsas well).


8/12/2019 Kim 1992

21/27

If the same psychological theory is true of humans, reptiles, andMartians, the psychological kinds posited by that theory must haverealizationsin human, reptilian, and Martianphysiologies. This implies thatthe theory is locally reducible in three ways, for humans, reptiles, andMartians.If the dependenceof the mentalon the physical meansanything,itmust mean that the regularities posited by this common psychology musthave divergent physical explanationsfor the threespecies. The very idea ofphysical realization involves the possibility of physically explainingpsychological properties and regularities, and the supposition of multiplesuch realizations, namely MR, involves a commitment to the possibility ofmultiple explanatory reductions of psychology.37The importantmoral ofMR we need to keep in mind is this: if psychological properties are multiplyrealized, so is psychology itself. If physical realizations of psychologicalproperties are a "wildly heterogeneous" and "unsystematic" lot,psychological theory itself must be realized by an equally heterogeneousand unsystematiclot of physical theories.I am inclined to-think that multiple local reductions, rather than globalreductions, are the rule, even in areasin which we standardly upposereduc-tions are possible. I will now deal with a possible objection to the idea oflocal reduction, at least as it is applied to psychology. The objection goeslike this: given what we know about the differences among members of asingle species, even species are too wide to yield determinaterealizationbases for psychological states, and given what we know about thephenomenaof maturationand development, braininjuries,and the like, thephysical bases of mentality may change even for a single individual. Thisthrows into serious doubt, continues the objection, the availability ofspecies-restrictedbridge laws needed forlocal reductions.

The point of this objection may well be correct as a matter of empiricalfact. Two pointscan be madein reply,however.First, neurophysiologicalre-searchgoes on because thereis a shared,andprobably well grounded,beliefamong the workers that there are not huge individual differences within aspecies in the way psychological kinds arerealized. Conspecifics must showimportant physical-physiological similarities, and there probably is goodreason for thinking that they share physical realization bases to a sufficientdegree to make search for species-wide neuralsubstrates for mental statesfeasible and rewarding. Researchers in this area evidently aim for

37 In "Special Sciences" and "Making Mind MatterMore" Fodorappearsto accept the localreducibilityof psychology and other special sciences. But he uses the terminology of localexplanation, rather than reduction, of psychological regularitiesin terms of underlyingmicrostructure. thinkthis is because his preoccupationwith Nagelian uniform reductionpreventshim from seeing that this is a form of inter-theoreticreduction f anything is.

20 JAEGWONIM

8/12/2019 Kim 1992

22/27

neurobiological explanationsof psychological capacities and processes thatare generalizableover all or most ("normal")members of a given species.

Second, even if therearehuge individualdifferencesamong conspecifics asto how their psychology is realized, that does not touch the metaphysicalpoint: as long as you believe in the Physical Realization Thesis, you mustbelieve that every organismor system with mentalityfalls undera physicalstructure-type such that its mental states are realized by determinatephysical states of organisms with that structure. It may be that thesestructuresare so finely individuatedand so few actual individuals fall underthem that research into the neural bases of mental states in these structuresis no longer worthwhile, theoretically or practically. What we need torecognize here is that the scientificpossibility of, say, humanpsychology is acontingent fact (assuming it is a fact); it depends on the fortunate fact thatindividual humans do not show huge physiological-biological differencesthat are psychologically relevant. But if they did, that would not change themetaphysics of the situation one bit; it would remain true that thepsychology of each of us was determined by, and locally reducible to, hisneurobiology.

Realistically, there are going to be psychological differences among indi-vidual humans: it is a commonsense platitude that no two persons areexactly alike-either physically or psychologically. And individualdifferences may be manifested not only in particularpsychological facts butin psychological regularities. If we believe in the Physical RealizationThesis, we must believe thatour psychological differencesare rooted in, andexplainable by, ourphysical differences, just as we expect ourpsychologicalsimilarities to be so explainable. Humans probably are less alike amongthemselves than, say, tokens of a-Chevrolet model.38 And psychologicallaws for humans, at a certain level of specificity, must be expected to bestatistical in character,not deterministic-or, if you prefer, "ceteris paribuslaws" rather than "strict laws". But this is nothing peculiar to psychology;these remarkssurely apply to human physiology and anatomy as much ashuman psychology. In any case, none of this affects the metaphysical pointbeing argued here concerning microdetermination and microreductiveexplanation.

38 Compare J. J. C. Smart's instructive analogy between biological organisms and super-heterodyne radios, in Philosophy and Scientific Realism (London: Routledge & KeganPaul, 1963), pp. 56-57. Smart's conception of the relationbetween physics and the specialsciences, such as biology and psychology, is similar in some respects to the position I amdefendinghere

MULT[PLE EALIZATIONNDTBEMETAPHYSICSFREDUCION 21

8/12/2019 Kim 1992

23/27

VIII. Metaphysical ImplicationsBut does local reduction have any interestingphilosophicalsignificance, es-pecially in regard to the status of mental properties? If a psychologicalproperty has been multiply locally reduced, does that mean that theproperty itself has been reduced?Ned Block has raised just such a point,arguing that species-restricted reductionism (or species-restricted typephysicalism) "sidesteps the main metaphysicalquestion: 'What is commonto the pains of dogs andpeople (andall otherspecies) in virtue of which theyarepains?"'.39

Pereboomand Kornblithelaborate on Block's point as follows:"...even if there is a single type of physical state that normallyrealizes pain in each type of organism, or in each structure ype,this does not show that pain, as a type of mental state, is re-ducible to physical states. Reduction,in the presentdebate, mustbe understoodas reductionof types, since the primary object ofreductive strategies is explanations and theories, andexplanations and theories quantifyover types....The suggestionthat there are species-specific reductions of pain results in theclaim thatpains in differentspecies have nothingin common. Butthis is just a form of eliminativism."40

Thereareseveral related but separable ssues raised here. But first we shouldask:Must all pains have "somethingin common"in virtue of which they arepains?

Accordingto the phenomenologicalconceptionof pain, all pains do havesomething in common: they all hurt. But as I take it, those who hold thisview of pain would rejectany reductionistprogram, ndependentlyof the is-sues presentlyon hand.Even if therewere a species-invariantuniformbridgelaw correlatingpains with a single physical substrate across all species andstructures, they would claim that the correlation holds as a brute,unexplainable matterof fact, and that pain as a qualitative event, a "rawfeel", would remain irreducibly distinct from its neural substrate. Manyemergentistsapparentlyheld a view of this kind.

39 "Introduction:What is Functionalism?"in Readings in Philosophy of Psychology, pp.178-79.40 In their "The Metaphysics of Irreducibility".See also Ronald Endicott, "The Species-Specific Strategy", orthcoming.In personalcorrespondenceEarlConee andJoe Mendolahave raised similar points. There is a useful discussion of various metaphysical issuesrelating to MR in CynthiaMacdonald,Mind-BodyIdentity Theories (London and NewYork:Routledge, 1989).

22 JAEGWONiM

8/12/2019 Kim 1992

24/27

I presumethatBlock, and PereboomandKornblith,arespeakingnot froma phenomenological viewpoint of this kind but from a broadly functionalistone. But from a functionalistperspective, it is by no means clear how weshould understand he question"Whatdo all pains have in commonin virtueof which they are all pains?"Why should all pains have "somethingin com-mon"?As I understand t, at the core of the functionalistprogram s the at-tempt to explain the meanings of mental terms relationally, in termsof in-puts, outputs, and connections with other mental states. And on the view,discussed briefly earlier,that mentalpropertiesare second-orderproperties,pain is the property of having a property with a certain functionalspecificationH (in termsof inputs,outputs,etc.). This yields a shortanswerto Block's question: what all pains have in common is the pattern ofconnections as specified by H. The local reductionist is entitled to thatanswer as much as the functionalist s. Comparetwo pains, an instanceof Nhand one of Nm: what they have in common is that each is an instance of aproperty that realizes pain-that is, they exhibit the same patternof input-output-other nternalstate connections,namely the patternspecifiedby H.But some will say: "ButH is only an extrinsic characterization;what dothese instances of pain have in common thatis intrinsic to them?"The localreductionist must grant that on his view there is nothing intrinsic that allpainshave in commonin virtue of which they arepains (assumingthatNh,Nr,andNm"have nothing intrinsic in common"). But that is also precisely theconsequence of the functionalist view. That, one might say, is the wholepoint of functionalism: the functionalist, especially one who believes inMR, would not, and should not, look for something common to all painsover and above H (the heartof functionalism, one might say, is the beliefthatmental states have no "intrinsicessence").But there is a furtherquestion raised by Block et al.: What happens toproperties that have been locally reduced? Are they still with us, distinctand separate from the underlyingphysical-biological properties?Granted:humanpain is reduced to Nh, Martianpain to Nm, and so forth,but what ofpain itself It remains unreduced. Are we still stuck with the dualism ofmentaland physical properties?

I will sketch two possible ways of meeting this challenge. First, recallmy earlierremarksabout the functionalist conception of mental propertiesas second-orderroperties: ain is thepropertyof havinga propertywithspecificationH, and, given thatNh,Nr, andNmare the propertiesmeeting H,pain turnsto be the disjunctiveproperty,Nhv N, v Nm. If you hold the sec-ond-orderpropertyview of mentalproperties,pain has been reduced to, andsurvives as, this disjunctive physical kind. Quite apartfrom considerationsof local reduction,the very conceptionof pain you hold commits you to theconclusion thatpain is a disjunctivekind, and if you accept any form of re-

MULTIPLEEALIZATIONNDTn METAPHYSICSFREDUCTION 23

8/12/2019 Kim 1992

25/27

spectable physicalism (in particular, he Physical RealizationThesis), it is adisjunctive physical kind. And even if you don't accept the view of mentalproperties as second-orderproperties,as long as you are comfortable withdisjunctive kinds andproperties,you can, in the aftermathof local reduction,identify pain with the disjunction of its realizationbases. On this approach,then, you have another, more direct, answer to Block's question: what allpains have in common is that they all fall under the disjunctive kind,NhvN vNm.If you are averse to disjunctivekinds, there is anothermore radical,andinsome ways more satisfying, approach.The startingpoint of this approach sthe frankacknowledgement that MR leads to the conclusion thatpain as apropertyor kind must go. Local reduction after all is reduction,and to be re-duced is to be eliminated as an independententity.You might say: global re-duction is different in that it is also conservative-if pain is globallyreduced to physical propertyP. pain survives as P. But it is also true thatunder local reduction,pain survives as Nhin humans,as Nr in reptiles,and soon. It must be admitted, however, that pain as a kind does not survivemultiple local reduction. But is this so bad?

Let us returnto jade once again. Is jade a kind? We know it is not a min-eral kind;but is it any kind of a kind? That of course depends on what wemean by "kind". There are certain shared criteria, largely based onobservable macropropertiesof mineral samples (e.g., hardness, color, etc.),that determine whether something is a sample of jade, or whether thepredicate "is jade" is correctly applicable to it. What all samples of jadehave in common is just these observable macrophysicalproperties hat definethe applicabilityof the predicate "is jade". In this sense, speakers of Englishwho have "jade" n their repertoireassociate the same concept with "jade";and we can recognize the existence of theconcept of jade and at the same timeacknowledge that the concept does not pick out, or answer to, a propertyorkind in the naturalworld.

I think we can say something similar about pain and "pain": there aresharedcriteria for the applicationof the predicate "pain"or "is in pain",andthese criteria may well be for the most part functionalist ones. Thesecriteria generate for us a concept of pain, a concept whose clarity anddeterminacy depend, we may assume, on certain characteristics (such asexplicitness, coherence, and completeness) of the criteria governing theapplication of "pain".But the concept of pain, on this construal, need notpick out an objective kind any more thanthe concept of jade does.

All this presupposes a distinction between concepts and properties (orkinds). Do we have such a distinction?I believe we do. Roughly, conceptsarein the same ball park as predicates, meanings (perhaps, something likeFregean Sinnen), ideas, and the like; Putnam has suggested that concepts be

24 JAEGWONKIM

8/12/2019 Kim 1992

26/27

identified with "synonymy classes of predicates",41 and that comes closeenough to what I have in mind. Propertiesand relations, on the other hand,are "out there in the world"; they are features and characteristicsof thingsand events in the world. They include fundamentalphysical magnitudes andquantities, like mass, energy, size, and shape, and are part of the causalstructure of the world. The property of being water is arguably identicalwith the property of being H20, but evidently the concept of water isdistinct from the concept of H20 (Socrates had the former but not thelatter). Most of us would agree that ethical predicates are meaningful, andthatwe have the concepts of "good", "right",etc.; however, it is a debatableissue, and has lately been much debated, whether there are such propertiesasgoodness and rightness.42f you find that most of these remarksmake sense,you understand the concept-property distinction that I have in mind.Admittedly, this is all a little vague and programmatic,and we clearly needa better articulatedtheory of propertiesand concepts; but the distinction isthere, supported by an impressively systematic set of intuitions andphilosophical requirements.43

But is this second approacha form of mental eliminativism?In a sense itis: as I said, on this approach no propertiesin the world answer to general,species-unrestricted mental concepts. But remember: there still are pains,and we sometimes arein pain, just as therestill aresamplesof jade. We mustalso keep in mind that the present approach s not, in its ontological implica-tions, a form of the standardmental eliminativism currentlyon the scene.44Without elaborating on what the differences are, let us just note a fewimportant points. First, the present view does not take away species-restrictedmentalproperties,e.g., human pain, Martianpain, canine pain, andthe rest, although it takes away "pain as such". Second, while the standardeliminativism consigns mentality to the same ontological limbo to whichphlogiston, witches, and magnetic effluvia, have been dispatched, he positionI havebeen sketchingviews it on a par with jade, tables, and adding machines.To see jade as a nonkind is not to question the existence of jade, or the legiti-macy and utility of the concept of jade. Tables do not constitute a scientifickind; there are no laws about tables as such, and being a table is not a causal-explanatory kind. But that must be sharply distinguished from the falseclaim that thereare no tables. The same goes for pains. These points suggest41 In "The Natureof Mental States".42 I of course have in mind thecontroversy concerningmoral realism; see essays in Geoffrey

Sayre-McCord,ed., Essays on Moral Realism (Ithaca:Cornell University Press, 1988).43 On concepts and properties, see, e.g., Hilary Putnam, "On Properties",Mathematics,Matter and Method (Cambridge: Cambridge University Press, 1975); Mark Wilson,"PredicateMeets Property",Philosophical Review 91 (1982): 549-90, especially, sectionIII.44 Such as the versions favoredby W.V. Quine, StephenStich, and PaulChurchland.

MULTIPLEEALIZATIONNDT1E METAPHYSICSFREDUCTION 25

8/12/2019 Kim 1992

27/27

the following difference in regardto the status of psychology: the presentview allows, and in fact encourages, "species-specificpsychologies", but thestandard eliminativism would do away with all things psychological-species-specific psychologies as well as global psychology.45

To summarize, then, the two metaphysical schemes I have sketched offerthese choices: either we allow disjunctivekinds and construepain and othermental properties as such kinds, or else we must acknowledge that our gen-eral mental terms and concepts do not pick out propertiesand kinds in theworld (we may call this "mental property irrealism").I should add that Iam not interestedin promoting either disjunctive kinds or mental irrealism,a troublingset of choices to most of us. Rather, my main interesthas been tofollow out the consequences of MR and try to come to terms with themwithin a reasonablemetaphysicalscheme.I have alreadycommentedon the status of psychology as a science underMR. As I argued,MR seriously compromisesthe disciplinary unity and au-tonomy of psychology as a science. But that does not have to be taken as anegative message. In particular, the claim does not imply that a scientificstudy of psychological phenomena is not possible or useful; on the contrary,MR says that psychological processes have a foundation in the biologicaland physical processes and regularities, and it opens the possibility ofenlightening explanationsof psychological processes at a more basic level.It is only that at a deeper level, psychology becomes sundered by beingmultiply locally reduced. However, species-specific psychologies, e.g.,human psychology, Martian psychology, etc., can all flourish as scientifictheories.Psychology remainsscientific, though perhapsnot a science. If youinsist on having a global psychology valid for all species and structures,youcan help yourself with that, too; but you must think of it as a conjunction ofspecies-restricted psychologies and be careful, above all, with yourinductions.46

4s The approach to the mind-body problem being adumbrated here is elaborated in my"Functionalism as Mental Irrealism"(in preparation).46 This paper is descended from an unpublishedpaper, "The Disunity of Psychology as aWorking Hypothesis?", which was circulated in the early 1980s. I am indebted to thefollowing persons, among others, for helpful comments: Fred Feldman, HilaryKornblith, Barry Loewer, Brian McLaughlin,Joe Mendola, Marcelo Sabates, and JamesVan cleve.

kim 1992

Documents