HOWTO KEEP THE 'PHYSICAL' IN PHYSI(:ALISM* Physicalism is roughly the thesis (1) that every entity is either itself

a physical entity or is exhaustively composed, ultimately, of physical entities, and (2) that every propertyis either itself a physicalproperty or is realized, ultimately,by physical properties.lNever mind whether alizationare the best notions for exhaustive composition and rea cating the idea that everything physicalist to employ in expliis physical better, for thanthe notions of token instance, identity and supervenience;2and mindalso how these notions never are to be precisely understood. The problemI wish to discuss is what physicalists can and meanby 'physical' in their should formulations of physicalism, and this problem arises on any reasonable view as to the answers to these questions. A physicalist'sdefinition of 'physical'can perfectlywell be lative the problem is not one stipuof plausible it must at least meet the conceptual analysis but to be following two conditions: when plugged into an otherwise satisfactory formulation of physicalism, it must yield a thesis (1) that is not obviously false and (2) possesses that content determinable by us. (A physicalismwhose content was not determinable by us would presumablybe for us to support empirically, impossible and might, for all we know, not even exclude from existence the sort of paradigmaticallynonphysical items for example, souls, entelechies, ghosts which have physicalists traditionallyrefused to countenance.) There is a however, dilemma, apparentlyowed to Carl G. thought to show that no physicalist Hempel,3 that is sometimes definition of 'physical' can
* am I grateful for helpful comments on earlier drafts to William Lycan,Peter Markie, Geoffrey Hellman, 'See,for example, my and Paul Weirich. aBeing a Physicalist: How and (More Importantly) Why," Philosophical Studies, LXXIV (1994): 221-41. 2See,for example, John F. Post, The Faces of 1987). Existence (Ithaca: Cornell, 3aComments on Goodman's Ways of Worldmaking," Synthese, XLV (1980): 193-99, especially pp. 194-95. See also his aReduction: Ontological and Linguistic Facets," in Sidney Morgenbesser, Patrick Suppes, eds., and Morton White, Philosophy, Science, and Method: Essays In Honor of Ernest St. Martin's, Nagel (New 1969), especially pp. 180-83. Compare J. J. C. Smart, aTheYork: tent ofPhysicalism," ConPhilosophical Quarterly, XXVIII (1978): 339-41; and Chomsky, Language and Mind (New York: Harcourt Brace Jovanovich, Noam p. 98. 1972), 0022-362X/97/941 2/622-37

622

THEJOURNALOF PHILOSOPHY

C) 1997 TheJoumal of Philosophy, Inc.

HOW TO KEEP THE

'PHYSICAL' IN PHYSICALISM

meet these two conditions simultaneously.4 It is nicely expressed by GeoffreyHellman:3

...currentphysicsis surely incomplete inaccurate(in its laws). This poses a (even in its ontology) as well as dilemma:either physicalist ples are basedon cuITent principhysics,in which case there is eveIyreasonto think they are false;or else they are not, in which case it is, at cult to interpretthem, since they are based on a "physics" best, difflthatdoes not exist yet we lack any general criterionof "physical object, property,or lawX framedindependentlyof existing physicaltheory (ibzd., p. 609).

623

Soif aphysical" entities and inthe laws and theories of propertiesare those mentioned as such current physics,then physicalismis very probably false; but if they are those mentioned as such in the laws and theories of completed physics,then, since we have no idea what completed physicswill look like, the resulting formulation of physicalism will lack content determinableby us. I shall argue here the first horn of this Hempelian that dilemma is blunt, and that therefore it remainsopen to a physicalist to understandby aphysical" ties and propertiesthose entimentioned as such in the lawsand of current theories physics.6 Here is what is supposed to be wrong with defining 'physical' terms of currentphysics.Past in theories in the standpointof currentphysics,have physics,whenjudged from usuallyturned out to be both false and incomplete; it is therefore very likely (though not, course, of absolutelycertain) that current complete.7 But if so, and if physicalismphysicsis both false and informulatedin terms of curSee, for example,T. Craneand D. H. Mellor,'@There ism," is No Questionof PhysicalMind, xc (1990): 185-206, especiallyp. 188;and Crane,aWhy on Supervenience," Indeed?Papineau Analysis, Ll (1991): 32-37, especiallyp. 34. See also Bas F>assen, aScience,Materialism, C. van and False ed., Consciousness," in Jonathan L. Kvanvig, WarrantIn Contemp(rrary Epistemology: Essays in Honor of edge Pluntinga's Theotyof (Lanham, MD:Rowmanand Littlefield, 1996), especiallypp. 163-70and Knowl5 "Determination and LogicalTruth," 173-74. thiSJoURNAL, 607-16, LXXXII, 11 (November p. 609. 1985): 6 For a criticalsurveyof other responses to the very dilemma, plus a positiveproposal different from mine, see JeffreyPoland, tions Physicalism: The Philosophical (New York: Oxford, 1994), ch. 3. Founda7 is It not entirely clear what an of Hempel's dilemma could mean by 'incomplete' here;advocate should or even physics is incomplete if and only but I assume the intuitive idea is that current if it fails to mention which some entity or property (a) exists and which (b) would, if discovered, physical unhesitatinglybe classifiedas such a thing as a new particle with mass, charge, and spin. A with suchan account, of course, is difficulty that it the contraryto the conclusion of dilemma, that a viable conception of presupposes, 'physical' is now vocate of the dilemma can no doubt availableto us, but an adget around this problem.
4

rent physicsassumesthe truth and completenessof current then it is veiy likely (though physics, not, of course, absolutely certain) that physicalism is false, too8 which requiresone to cease to be a calist. physiMyreplyto this argumentis to challenge its final step, that is, the inferencethat a physicalist should abandon physicalism just because physicalism is verylikelyfalse. The argumentassumesthat a physicalistis someone who must assigna high, or even veryhigh, tothe thesis of physicalism; probability whereforeit is unreasonablefor a calist to define 'physical'in terms physiof current physicsif doing pliesthat physicalism is so imimprobable. But I deny this claiming assumption, that a physicalist need notassigna high probability to physicalism, and can therefore comfortably live with the result that physicalismhas a very low probability. There developing such a reply to the argument, are different ways of their positive account of what attitude however,waysthat varyin to physicalisma physicalist should take.For instance,one could say that a physicalist is someone who merelytreatsthe thesis of physicalism as some sort of regulaiive ideal for science, a role it could playwhile indeed, Hellman seems to say something being literallyfalse;and, just like this in his own response to Hempel's dilemma (op. cit., p. 610). Alternatively,one could say that a physicalistis someone who holds that physicalism, while literallyfalse, is nevertheless closer to the truth, a better aw proximation to the truth, than its rivals. But both these suggestions have drawbacks. The firstrequiresus to abandon the intuition that a physicalist is someone who takes some sort of attitude truth toward the of physicalism. The second can only be as good as the of account verisimilitude or approximationto the truth on which it relies, and these notions are notoriously hard to contrast, By the account I shall give here explicate saiisfactorily. respects the intuition that a physicalist is someone who takes some sort of attitude toward truth of physicalism;and it has the no dependence on the concept of verlslml ltUC .e.
. . . .

624

THEJOURNALOF PHILOSOPHY

Of course, the past historyof physicaltheorizing is not the only vant to current physics' evidence rele(that is, nonhistorical, likelihood of being true, since there is also independent observational) evidence for the true claim totake into account; and when this other evidence is that current physicsis far from taken into account, it is obvious that current physicswill Michael Levin, uOn Theory-Changeand emerge on balance as very unlikely (see XLVI Meaning-Change," (1979):407-24, especially pp. Philosophy of Science, 42S21). lows. One reason for doing so is that there But I waive Levin's point in what folseems to be no independent for the that physics is claim evidence complete which could outweigh the for the that it is claim historicalevidence incomplete; so sponse to the first horn of Hempel's Levin'spoint could not be the whole of a redilemma.
8

HOWTO KEEP THE

Mydevelopmentof 625 a differentbasis for the claim that a need not regard physicalism physicalist as more givingan account, not of the thesisof probable than not will involve be a physicalist. but of what it is to Here is my argument physicalism, in outline. (P1)
To be a physicalist is to tude is-toward the take the same attitude-whatever that attihypothesis of have broadly scientific realist and physicalismwhich those who wardwhatthey antirelativist intuitions take to(P2) The attitude regardas the best of currentscientific that those who have hypotheses. tirelativist intuitions take toward broadlyscientificrealistand anwhat they regard as currentscientific the best of is identicalwith an fined later) that I hypotheses attitude (to be deshall call the SR (C1) Therefore, to be a physicalistis attitude. to take the SR physicalism. attitude toward (P3) But to take the SR attitude toward a hypothesis regardingit as likelyto be true (let alone very does not require (C2)Therefore,to be likelyto be true). a does not ism as likelyto be physicalist true (let alone very requireregardingphysicallikelyto be true). (P1) is very

'PHYSICAL' IN PHYSIGALISM

plausible,once physicalism is viewedas less than a scientific hypothesis, albeit one with no more and no Surely, it can be so viewed. unusual features. It does not, of any law,nor does it admittedly, assert the holding belong to any one claim science. But it does concerning the make a like the claim that genes constitution/realization of the world are which, made of DNA (though larger scale), is entirely obviously on a contingent and which is surely rather than, say, scientific commonsensical or end ofmy final section, religious.As I shall show at the physicalismcan be planatory in just the waywe exhibited as being exexpect scientific rate, if physicalismis hypothesesto be. At any not a scientific one in respect hypothesis, then it every relevantto current resembles to be a triviality, purposes. (P3) will turn once the nature of out the will SR attitudeis made shortly be done. (P2), clear, as however, which attitudes urges the identification not obviously of identical, is far from much trivialand will require elucidationand defense.
II The elucidation and defense of both an (P2) and (P3) must explanation of whatI mean by begin with the 'SR sis, attitude'towarda however. Here is a stipulative hypothedefinition:

(SR) To take the SRattitude towarda hypothesisis pothesis (1) to regardthe hyas true or false in world is, and (2) to assign virtueof the waythe mind-independent the hypothesisa that of its relevant higher probability rivals. than

626

OF PHILOSOP THEJOURNAL

Here is another stipButwhat are the relevantrivalsto a hypothesis? definition: ulative

(RR)HypothesisH1 is a relevantrivalto H2 if and only if (a) H1 is sensiblyintended to achievea significantnumber of H2's theoretical goals; (b) the hypotheses,H1 and H2, fail to supetweneon one another;and (c) H1 has actuallybeen formulated.

Clauses (a) through (c) require some unpacking and motivation. Take clause (a) first. The theoretical goals of a hypothesis will inexplanationof certain phenomclude such things as the satisfactory ena and the solution of certain problems. The reference to a number"of such shared goals is supposed to do (rough) asignificant justice to the fact that advocates of rival hypotheses almost never completely agree what theoreticalgoals their respectivehypotheses can reasonablybe expected to achieve;insisting that all theoretical goals be sharedfor two hypothesesto count as rivalswould therefore leave almost no pairs of hypothesesas rivals.The reference to goals intended for a hypothesisto achieve is supposed that are "sensibly" to ensure that the hypothesisthat, say, the moon is made of green just because cheese does not qualifyas a relevantrivalto Darwinism some lunatic thinks it can account for the origin of species. Let us now turn to clause (b). It is included so that hypothesesat different levels of explanation (for example, folk psychology and scientific psychology),whose theoreticalgoals arguablyoverlap considerably, are not mistakenlyclassified as rivals;presumablyfolk psychology does superveneupon scientific psychology.It also servesto exclude hypotheses that are merely notationalvariantsof one another from counting as relevantrivals.Finally,clause (c). I have no full account of what it is for a hypothesisto be formulated,but two points are crucial. First,to count as formulated,a hypothesisneed not have been formulated in any great detail, but we must have been told something about its basic principles.So, for instance, I would count creationismas havingbeen formulated,on the grounds thatwe are told about the basic mechanismit hypothesizesto account for sornething life, even though creationistsare notoriouslystingywith suggestions as to the details of God's plans for flora and fauna. Secondly, we must distinguishformulatinga hypothesisfrom referringto it. The expression 'the set of laws accepted by people at Hatvardwho call in 2097' very probablyrefers to a hypothesis, themselves"physicists" but it does not formulateorle;for it tells us nothing about the basic principles of the hypothesis (if any) referred to, which hypothesis thereforefailsto count as a relevantrivalto any currenthypothesis.

HOWTO KEEPTHE

627 One especially important negationof a hypothesis, consequence of (RR) is that the sheer unsupplementedby any other notcount as a claims,does relevant rival to the mentednegation of a hypothesis,since the unsupplehypothesisfails to meet not sensibly be condition (a): it canintended to achieve the hypothesis. theoretical goals Simplydenying the existence of electrons, for of the goes no waytoward instance, accountingfor the troduced phenomena electronsare into explain. Of course, the negaiion of a hypothesis tainly be partof-one can cerand I suspect that thisconjunctof a relevantrivalto the is hypothesis; typically so sis when it appearsthat a has a pure negation as a rival.So, for hypothetrons' example, the denial of existence conjoined with elecappropriate generalizations can perfectlywell be phenomenological a relevantrivalto ory. Similarly,while (RR) electron theimplies that relevant atheism unadorned is rivalto theism,it can not a allowthat contemporary science is a relevantrival atheismplus the findingsof to theism. According to (RR), the relevantrivalsto a certain hypothesiswill be (1) predecessorsin the historyof the branchof the hypothesisbelongs, (2) science to which potheses in the branch of certain current, actuallyformulatedhyand (3) certain current, science to which the hypothesis belongs, actuallyformulated crackpot hypotheses hypotheses-call which in some to the sociological sense do not them branch of science to belong which the hypothesis they may once have done belongs (though so). So, for tude instance, to take the SR towardthe hypothesis attiof evolution by tion is (a) to regard gradualistnaturalselecthe hypothesisas way true or the mind-independentworldis, and (b) false in virtue of the be true to regardit as than (at least) Lamarckianism, the punctuated likelierto model, and creationism. equilibrium In thelight of the stipulative tolerably clear at leastwhat (P2)definitions (SR) and (RR), it is now and (P3) are true? Take claiming.But are they (P3) first. Since, according to clause (2) of the SR attitude toward a (SR), taking more requires only regarding likely to be true thanhypothesis it as its relevantrivals,9 rivals toa hypothesisdo and since the not include the relevant sheer pothesis, it is possible to take the SR attitude negation of that hytowarda hypothesis
9 Notice that although regardinga hypothesisas vals may not require more likely than its being able to estimate are, it is relevantrihow likely the quite consistent with being that able to do so; nothing hypotheses in question gthe testing of theories yields here assumesor gIs only a the Best Good Enough?"in David comparativewarrant"-see Peter implies York: Lipton, Papineau, Oxford, ed., ThePhilosophyof 1996), p. 93. Science(New

'PHYSICAL' IN PHYSICALISM

628

THEJOURNALOF PHILOSOP

without regardingit as likely, still less very likely, to be true: a hypothesis might be unlikely,and yet still more likely than its relevant rivals.l What about (P2)? Is it true, as (P2) asserts,that the SR attitude, as stipulatively defined, can be identifiedwith the attitudethat those who have broadlyscientific realist and antirelativist intuitions take toward what they regard as the best of current scientific hypotheses?I shallnow arguethat it can. Whatis the deepest intuitivecommitmentof those who would call themselvesscientificrealistsand antirelativists? It consists,I suggest, in the respectfulwayin which they regard (certain) currentscientific hypotheses. By and large, they regard these current scientific hypotheses as:
(1) true or false in virtueof the waythe mind-independent worldis (2) oyectively superior, in some truthwonnectedsense, to earlier hypotheses in the field, so thatscience has, in this sense, progressed (3) objecovelysuperior,in the same sense, to currentrivalscientifichypotheses (4) objectivelysuperior, in the same sense again, to current rival hypotheses advocatedby people outside the scientificestablishment (5) such that whether the regardfor a hypothesisembodied in (1)-(4) is appropriateis generallyindependent of whether or not the hypothesispostulatesentiiies and propertiesthat cannot be obselved

Someone with these attitudes to the best of current scientific hypotheses will therefore find repugnant each of the following three ideas: (a) the (allegedly) nebKantian idea that the postulatesof a scientific hypothesis are somehow conjured into existence by the widespreadacceptance of the hypothesis;(b) the epistemologically egalitarian idea that all hypotheses, past or present, scientific or fringe, are more or less on a cognitive parwith one another (even if their political influence is unequal); and (c) the empiricistidea that the distinction between observableand unobservablemarks a distinction of greatepistemologicalsignificance. The argument for (P2) that is, for identifying the stipulativelydefined SR attitudewith the attitudethat those who havebroadlyscientific realist and antirelativist intuitions take toward what they regardas the best of currentscientifichypotheses has two premises. The first premiseis that the attitude that those who have broadlyscientific realist and antirelativist intuitions take towardwhat they re10 If the relevantrivalsof a hypothesisdid include the sheer negation of the hypothesis, then taking the SR attitude towarda hypothesis regardingit as likelier than all its relevant rivals would entail regarding it as likelier than its negation, and hence regardingit as likelierthan not, and hence as likely.

IN PHYSiICALISM HOWTO KEEP THE 'PHYSICAL'

629

gard as the best of currentscientifichypothesesis adequatelycharacterizedby (1)-(5). (One who denies this premise has only to specify something that has been left out or wronglyincluded.) The second premise is that the SR attitude, as stipulativelydefined, should be identified with the attitude characterizedby (1)-(5). Five observations together provide evidence for this identification.First,clause (1) of (SR) saysexactlywhat (1) above says.Second, to take the SR attitude towarda hypothesisis surely one way of taking a truth-connected attitude (as mentioned in (2)-(4)) toward it. It is not, of course, to have an allKr-nothingbelief that the hypothesis is true; nor is it even to believe that the hypothesis is closer to the truth than enjoys greater verisimilitude than other false hypotheses. But it is, in part, to assign to the hypothesisa higher probabilityof beirlg truethan is assignedto its relevantrivals.Third, to take the SR attitude towarda hypothesis,and in particularto assign it a higher probability than any of its relevantrivals,is surelyone wayof regarding it as objectivelysuperior (as mentioned in (2)-(4) ) to certain rivals, at least on the assumptionthat assignmentsof probabilityare answerableto objectiveconstraints,such as would be supplied by a reliabilisttheory of confirmation(or by a Bayesiantheory, appropriately supplemented by some account of objectiveconstraintson assigning prior probabilities).Fourth, the relevantrivalsthat the SR attitudeconcerns includejust the sort of rivalsthat (2)-(4) concern. Finaily,nothing in the SR attitude rules it out that one could perfectlywell take the SR attitudetowarda hypothesispostulatingunoW servables. (P2)'s identificationof the SR attitudewith the attitude that those taketoward sympathies with broadlyscientificrealistand antirelativist the best of current hypotheses evidently rests upon the idea that nothing compels a scientific realist/antirelativistto assign a high probability to the theoriesshe picksout as best.Yetit is hard,I admit, requiresthe assignmentof a to dislodge the intuition that something sugBut what?Not the need to rejectthe egalitarian high probability. gestion that all theories,past or present,scientificor fringe, are really on a cognitivepar with one another,since that suggestioncan be rejected merelyby adopting an attitudethat, like the SR attitude,orlly to favoredhypotheses.Here assignsa higher, not a high, probability must believe the realist/antirelativist is a bettersuggestion:"Ascientific theory just is to theoriesshe picks out as best, and since to believe a swiftly the high probabilityrequirement assign it a high probability, follows." Indeed it does, but neither premise is veryplausible.To bemust gin with,it is not at all clearwhya scientificrealist/antirelativist

630

THEJOURNALOF PHILOSOPHY

believe the theories she picks out as best; certainly,belief is not required to explain practicalreliance on the theories;for a sufficient basis for action can be the assignmentof only a low probability,as when I carrya sparetire, even though I certainly do not believe that I will have a flat. Whereasit might plausiblybe claimed that scientific realists/antirelativists must in some sense accept the theories they favor, argument is needed to show that this attitude of acceptance amountsto anythingmore than the SR attitude.More seriously,the identificationof belief with the assignmentof a high probability runs into the problem of the LottetyParadox,in which, given the identification,an apparently quite rationalperson mustbe regardedas holding contradictotybeliefs:a person assignsa high probability to each propositionsaying,of one lotteryticket, that it will lose, and also to the propositionthat some ticketwill not lose; but if beliefjust is the assignmentof high probability, then the person must believe of each ticket that it will lose, and also that some ticketwill not lose which commitsher to a contradiction.ll Nor is it true that the assignmentof a high probabilityis even a logically necessarycondition of belief: surely,it is logically possible for someone who believes ten (probabilistically independent) propositions,to each of which he assignsa probabilityof 0.9, to believe also the conjunction of those propositions, even though the probabilityof the conjunction is low, being the productof the probabilities of each conjunct.l2 Here is a second suggestion for supportinga high-probability requirement:"Evenif believing a theoxydoes not logicallyrequire aF signing it a high probability, rationally believingit surelydoes. So let us assume some analysisof belief which makesbelief quite independent of probabilitrassignments.l3 Then, since a scientificrealist/antirelativist must believe the theories she picks out as best, and since rationally believinga theoryrequiresassigningit a high probability, a scientific realist/antirelativist must on pain of irrationality assign a high probabilityto her favoredtheories."But this second argument for a high-probability requirementis also inconclusive.It retainsthe nonobviousassumptionof its predecessorthat a scientificrealist/antirelativistmust believe the theories she picks out as best. But the premise that rational belief requires assigning a high probability seems clearlywrong;for surelyit is alwaysrationalto believe the im" For elaboration and defense, see Mark Kaplan, DecisionTheory as Philosophy (New York:Cambridge,1996), pp. 9S98; see also PatrickMaher, Betting on Theories (NewYork:Cambridge,1993), pp. 134-35. 12 For the argumentas applied to acceptance, see Maher,pp. 137-39and 152-55. 13 For such analysesof a belief-like notion of acceptance, see, for example, Kaplan, ch. 4, and Maher,ch. 6.

HOWTO KEEP THE 'PHYSICAL' IN PHYSICALISM

631

mediate logical consequences of what one rationallybelieves. But then the person who rationallybelieves ten (probabilistically independent) propositions,to each of which he assignsa probabilityof 0.9, is rationalin believingtheir conjunction,even though, as already noted, this conjunctionhas a low probability.l4 Let me also note that there are theoriesof rationalacceptancewhich leave open the possibility of accepting hypotheses that are improbable,and they could presumably be modifiedto cover rationalbelief, non-probabilisticallyconstrued.l5 Neither MarkKaplan'snor PatrickMaher'stheoty is at all reliabilistin spirit, but a reliabilist theory which also yields the same resultis easy to imaginein vagueoutline: if one's cognitivegoal is not just truths,but truths that, say, provide a basis for prediction and explanation,then a belief-formingmethod could be rationalto adopt, and its productscount as rational,even though unlikelyto be true,just so long as they made up for their improbability, as it were, by constitutinga superiorbasis for prediction and explanation.Unformulatedtheories, or the sheer negations of formulatedtheories, might be likelierto be true,but cannot providesuch a basis.l6 In the absence of any furtherreasonfor insisiingon a high-probabilit requirement,(P2)'sideniificaiionof the SRattitudewith the atiitude that those who have broadlyscientific realistand aniirelativist intuiiionstake toward whatthey regardas the best of currentscientific hypothesescan stand.l7 It maybe possibleto go on the offensivehere,
Again, see Maher,pp. 137-39and 152-55. See Kaplan,ch. 4, and Maher,ch. 6.3. This is a good point at which to respond to the complaint that my answerto the Hempelian dilemma offers no solace to the physicalistwho believesphysicalism(rather thanjust taking the SR attitude toward it). In one way,of course, the complaint is misconceived:the thrust of the entire paper is that a physicalistdoes not have to believephysicalism.A more substantial reaction is to note that, even if it is insisted that a physicalistis one who believes physicalism,Hempel's dilemmacan be avoided so long as belief in physicalismcan be rational despite physicalism's improbability, which would be true if some general thesis held that belief in an improbablehypothesiscan still be rational. 16 Perhapsscientific realist/antirelativist intuitions include the idea that the best of current scientific hypothesesare things we know. Perhaps;but it is far from clear that one knows that p only if it is likelier to be true than false that p. These are accounts of knowledge (for example, reliabilistones) that do not imply that this is so, and a staunch defender of scientific knowledge might favorsuch an account indeed, might be expected to do so. 17 A differentkind of objection to (P2) claims that, if a hypothesis were very,very bad but still better than its relevantrivals,one could take the SR attitude towardit without also favoringit in the manner distinctiveof a scientificrealist/antirelativist. But, naturally,I doubt that this could happen, and I know of no example. I conjecture that any putativeexamplewould either tacitlyassume that favoringa hypothesis in the distinctive scientific realist/antirelativist manner requires believing it (which I have denied), or would be a case in which many relevant rivals no less likelycould easilybe formulated(even though ordinarilyno one would bother to).
4 15

632

THEJOURNALOF PHILOSOPHY

however. Resisting a high-probability requirement offersscientificreaL ists/antirelativists an attractive replyto the so-calledpessimistic induction which claimsthat, since most past theorieshave turnedout to be false,currenttheoriesare probably false,too.l8 Moreover, it is arguable that insistingon a high-probability requirement would be unreasonable,so that the SR attitudetoward is all we can decently hope for. hypotheses(or somethinglike it) Suppose a scientific realist/antirelativist insiststhat we should regardour best current theories as more likely to be true thanfalse.Thatimplies regarding them as more likely tobe true than the disjunctionof their rivals,and hence more likely thaneach disjuncttaken individually. But some of those disjunctsare unborn hypotheses,that is, hypothesesthat have not yet been formulatedand perhaps never will be.'9How could our current evidence make it reasonableto regarda current hypothesisas likelierthan an unformulated hypothesis? So we maybe going beyond any attitudeit could possiblybe reasonableto takeif we claim that a currenthypothesis is likelierto be true than not. My response to Hempel's dilemmais now complete. be formulatedin terms of currentphysics.Then, Let physicalism given that a physicalist is simplysomeone who takes the SR attitude towardphysicalism, the mere fact that the history of physical theorizing makes physicalism unlikelyto be true providesno reason by itself to abandon being a physicalist; one can remain a physicalist, just so long as physicalism, though unlikely,is still more likely than its relevant rivals. This is not simplyspecialpleading on behalf of physicalism; for the SR attitude,as I have argued,is none other than the attitudescientific realists/antirelativists take towardtheir favoredhypothesesin particular branchesof science. Of course, the histoxyof physical theorizing still constitutes evidence against physicalism,in the sense of lowering the likelihood that it is true, and presumably it does not in the same wayconstitute evidence against physicalism'srelevant rivals. So it still threatens physicalism. But any evidence againstphysicalism must obviouslybe weighed againstevidence for it, and the balanceof probabilities may yet leave physicalism likelier than its relevantrivals.Naturally, physicalists insist that there is strong evidence for capable in principleof giving the probability physicalism,evidence of physicalismthe necAs noted earlier,I discovered,in Maher,p. 137. l9 See LawrenceSklar,aDo Unborn Hypotheses Have Rights?" Pacific Philosophical QXarterly, LXII (1979): 17-29.
18

III

HOWTO KEEPTHE 'PHYSICAL' IN PHYSICALISM

633

essaryboost. Such counteractingevidence, moreover,does not have to raisephysicalism's probability to 0.5 or higher;given my proposed account of whatit is to be a physicalistX it need only lift it above that of its closest relevantrival,an easier requirementto fulfill. In fact, and contraryto the presumpiionof the last paragraph, the histoxy of physical theorizingmaybe as damagingto physicalism's relevantrivalsas it is to physicalism. Relevant rivals to physicalism are actually formulatedhypothesesthat are sensiblyintended to addressthe problemthatit is the centraltheoreiical goal of physicalism to address: givingan accountof the relaiionsamong the ontologiesthat the many sciences(including folk psychology and folk physics)respectively postulate,in light of such crosascientific regularities as have been discovered empirically.20 An exampleof one sort of crudecrossdsientific regularity mightbe thatnothingis everin a mentalstateunlessit is in some simultaneousbrainstate;but the sciences (for example,the neurosciences) presentmore refinedregularities. To identifyphysicalism's relevantrivals,it is helpfil to viewphysicalism as the conjunctive thesisthat
(1) There is some science, S, distinct from the totality of all the sciences, such that everyentity (property)is either itself mentioned as such in the lawsand theories of S or is ultimatelyconstituted (realized) by entities (properties)meniioned as such in the lawsand theories of S.

and that:
(2) S is currentphysics.21
See my op. czt., pp. 222-24. Understanding physicalism in thisway,as the conjunctionof (1) and (2), has the possiblyunsettlingconsequence that,say,ThomasHobbeswas not a physicalist, since he had no notion of currentphysics.But, surely,Hobbesianmaterialism nevertheless hasmuch in commonwith physicalism. The mostimportantcommonality is the belief thatsome science distinctfrom the bare conjunctionof the manysciencesis basicin some metaphysical sense. But there is also the factthatHobbes pickedas the basicscience somethingthat turned out to be, sociologically and historically speaking,the ancestor of current physics.Similarly,if tomorrownew evidence forces a revision of physics,it will be possible to formulateanother relevantrivalto physicalism in categoxy(A), by lettingtomorrow's(superior)physicstakethe place of currentphysicsin (2); and if the evidencefor takingtomorrow's physicsto be basic is as good as the evidence for takingcurrentphysicsto be basic,which is highly likely,the resultingdoctrine will be superior.But it will not, strictlyspeaking,be the doctrineI have called physicalism, though it will, however,havemuch in common with it. Specifically, it will implement the generic idea that there is a basicscience distinctfrom the bare conjunction of the manysciences,and it will do so by appealto a science thatis an immediate descendent of current physics.So if we are looking for something to serve as athe spirit of physicalism" which transcendsparticularformulations,then commitment to (1) might be the heart of a good candidate,though the spiritof physicalism wouldthen be cognitiveand not attitudinal (see vanFraassen, pp. 16970).
20 21

634

THEJOURNALOF PHILOSOPHY

Relevantrivalsto physicalismtherefore fall into two categories:(A) those which endorse (1) without endorsing (2), by agreeing that there is a basic science to which all the other sciences stand in some specialrelation,while proposingthat this basic science is something other than physics;and (B) those which deny (1), maintainingthat no science is basic, since all sciences are on an ontological par, linked to one another merelyby variousfundamentallaws.22 Do relevant rivalsof either sort gain any advantageover physicalism from evidence constitutedsolelyby the historyof physicaltheorizing? Take rivalsin category (A). Suppose one such rival asserts that there is a basic science, but that it is, say,biology. Such a view is evidentiallyquite untouched by the historyof physicaltheorizing. On the other hand, the trackrecord of biological theorizingis arguably no better than that of physicaltheorizing,which evens things out, and obviouslythere is other evidence decisive againsttakingbiology to be the basic science. But the best-known relevantrivalin category (A) is traditionaldualism,which I interpretas the view that, to put it verycrudely,physicalism is true of everythingexcept the mind: there is a basicscience, but it is the conjllnction of physicsand folk psychology (presumably linked to one anotherby fundamentalpsychophysical laws). The impact of the history of physical theorizing on traditionaldualism, relativeto that on physicalismX is trickierto assess. To the extent that the historyof physical theorizing makes it likely that current physicsis false, there is exactly the same evidence against traditionaldualismas there is against physicalism,given my interpretation of traditionaldualismas the view that the conjunction of physicsand folk psychologyconstitutesthe basic science. It might appear,though, that historicalevidence that currentphysicsis incompletewouldleave traditionaldualismunharmed,since traditionaldualism, unlike physicalism, is not committed to regarding current physicsas complete. But, in fact, what the historyof physicaltheorizing makeslikely is that currentphysicshas left out something like a new kind of particle,with mass and charge, and traditionaldualism zscommittedto regardingcurrent physicsas complete in respect of that sort of thing. What about relevant rivals in category (B)? The egalitarian and pluralistview that there is no basic science, the view, in effect, that cross-scientificregularitiesshould be treated as fundamental laws, appears initially to be committed neither to the truth nor to the completenessof currentphysics; so on evidence constitutedsolelyby
22See Crane and Mellor.

HOWTO KEEP THE 'PHYSICAL' IN PHYSICALISM

635

the histoxyof physicaltheorizing,physicalism appearsless likelythan antiphysicalist pluralism.Actually,though, on pain of simplysaying nothing at all about interscientific relations,antiphysicalist pluralism is committedto lawstatementsconnecting events as characterized by current physicsto events as characterized by each of the special sciences; and if current physicsis probablyfalse, then so, surely, are those law statements. Surprisingly,then, antiphysicalistpluralism mayderive no advantageover physicalism from evidence constituted solelyby the histoxyof physicaltheorizing. Finally,let me give the briefestoutline no more than a hint-of the evidencefor physicalism that I mentioned a few paragraphs back. The evidence for physicalism, as I understandit, consistsin empirically discoveredregularitiesof varioussorts, the best explanationof which is physicalism. One importantsort of regularityis that which links physicalconditions with simultaneousnonphysical (for example, mental, biological,geological) conditions;so, for instance,there are physicalconditions of a person's brain and environmentwhich are regularlyaccompaniedby a certain simultaneousmental condition.23 Now, physicalismcan explain a regularityof this kind by hypothesizing that the mental condition is identical with a functional condition, and that this functional condition is physicallyrealized, and physically realized,in particular, by the physicalcondition of the person's brain and environmentempiricallyfound to be sufficient for it. On this hypothesis,we would expect to observethatwhenever the physicalcondition obtains, so, too, does the mental condition: the physicalcondition sufficesfor the playingof a certainfunctional role, and that something playsthat role suffices for the obtaining of the mental condition, so that the physicalcondition sufficesfor the mental condition. (Evidently, the physicalcondiiion must be capable of playing the relevantrole, and whether it is must be determined empirically,or else the explanation must be rejected.) But if physicalism is not just an explanation but also the best explanation of such a regularity, then its powerto explain the regularitr is some evidence for it.24
23 There are other empiricallydiscovered regularities that can arguablybe explained by physicalism but not othetwise;for instance, I believe the argumentfrom overdeterminationfor physicalism(see, for example, my op. cit., pp. 22S31) points us toward some. Yet other regularitiespotentiallyproviding evidence for physicalism are the lawsof the special sciences;see my "TwoCheers for Reductionism;Or, The Dim Prospects For Non-Reductive Materialism,"Philosophy Of Science,LXII (1995): 37S88. 24 For elaboration, see my "Being a Physicalist:How and (More Importantly) Why," pp. 231-35.

636

THEJOURNALOF PHILOSOPHY

Furtherlight can be shed on how regularitiesof this kind provide evidence for physicalism by consideringthe egalitarianand pluralist rival to physicalismwhich maintainsthat there is no basic science. Since this rival assertsthe holding of, and hence (trivially)entails, the veryregularitiesthat I claim provideevidence for physicalism, it might be objected that these regularitiesthereforefail to provideevidence that favorsphysicalismover its rival;the evidence is neutral between physicalismand its rival. But there are obvious replies to this objection.The firstis that the rivalis still inferior to physicalism in respect of explanatory power. Although the egalitarianrivalindeed entails the regularitiesthat physicalismcan explain, it entails them quite trivially, by explicitlyassertingthat they hold; so it does not explain them at all, but instead treatsthem as brute,fundamentallaws, on a parwith the fundamentallawsof physics.A second replyis that, both because the egalitarianrival treats these regularitiesas brute laws to be accepted in additionto the fundamental laws of physics, and because it postulatesso many propertiesneither identical with nor even realized by physical properties, it is less economical than physicalism. Certainly, antiphysicalists are awareof these replies, but they tend to disparagethem. So here is a reason to take them seriously. On my account of the evidence for physicalism, the empirical case for physicalism is stronglyanalogousto thatfor any physicaltheoxyaiming to give an account of whatunderliessome (relatively)oW seIvableregularity(for example, a physicaltheoxyaiming to explain whywaterboils at 100 C). Now, a physicaltheorywill always have an empirically equivalentrival,namely,the view that the obseIvableregularitiesit aims to explain should be treatedas brute lawsthat simply have no explanation.But manyof us suppose that the theoxyis nevertheless to be preferred, on the grounds of some superempirical virtueit enjoys,like explanatory poweror economy;and whereasthe philosophical understanding and justification of appeals to such virtuesis obviouslycontested, that we make such appeals, and take ourselvesto be reasonablein doing so, seems hard to deny.25 But if these appeals are legitimate when used to break the tie between a physicaltheory and its empiricallyequivalentphenomenological rival, then why not also in the case of physicalism and its pluralistand
25 I suspect that hostilityto takingthe superempirical virtuesof a theoryas counting in favor of its truth stems from the unstated assumptionthat methodological principlesmust be a prioriin character;for it must be conceded that it is no neceF saxytruth that the world exhibits the sort of economy and connectedness of structure that would make the explanatoxypower and economy of a theory count in favor of its truth. But from the perspectiveof an a posteriori,reliabilistepistemology, surelythe matterlooks quite different.

HOW TO KEEPTHE 'PHYSICAL' IN PHYSICALISM

637

egalitarianrival?For the case is simplyanother instance of a theory and its empirically equivalentrival.26 At first sight, Hempel's dilemma seems to guarantee a swift and decisivevictory to the enemies of physicalism.But closer examination revealsthat it is merely one offensivein a campaignwhose outcome is at best undecided.
ANDREW MELNYK

University of Missouri/Columbia

26See Smart, zSensations and Brain Processes," PhilosophicalRe/view, LXVIII ( 1959): 141-56, especially pp. 155-56.