Spring 2013 Edition
Physicalism is the thesis that everything is physical, or ascontemporary philosophers sometimes put it, that everything superveneson, or is necessitated by, the physical. The thesis is usuallyintended as a metaphysical thesis, parallel to the thesis attributedto the ancient Greek philosopher Thales, that everything is water, orthe idealism of the 18th Century philosopher Berkeley, that everythingis mental. The general idea is that the nature of the actual world(i.e. the universe and everything in it) conforms to a certaincondition, the condition of being physical. Of course, physicalistsdon't deny that the world might contain many items that at firstglance don't seem physical — items of a biological, orpsychological, or moral, or social nature. But they insistnevertheless that at the end of the day such items are either physicalor supervene on the physical.
Physicalism is sometimes known as ‘materialism’;indeed, on one strand to contemporary usage, the terms‘physicalism’ and ‘materialism’ areinterchangeable. But the two terms have very different histories.The word ‘materialism’ is very old, but the word‘physicalism’ was introduced into philosophy only in the1930s by Otto Neurath (1931) and Rudolf Carnap (1959/1932), both ofwhom were key members of the Vienna Circle, a group of philosophers,scientists and mathematicians active in Vienna prior to World War II.It is not clear that Neurath and Carnap understood physicalism in thesame way, but one thesis often attributed to them (e.g. in Hempel1949) is the linguistic thesis that every statement is synonymous with(i.e. is equivalent in meaning with) some physical statement. Butmaterialism as traditionally construed is not a linguistic thesis atall; rather it is a metaphysical thesis in the sense that it tells usabout the nature of the world as such. At least for the positivists,therefore, there was a clear reason for distinguishing physicalism (alinguistic thesis) from materialism (a metaphysical thesis).Moreover, this reason was compounded by the fact that, according toofficial positivist doctrine, metaphysics is nonsense. Since the1930s, however, the positivist philosophy that under-girded thisdistinction has for the most part been rejected—for example,physicalism is not a linguistic thesis for contemporaryphilosophers—and this is one reason why the words‘materialism’ and ‘physicalism’ are now ofteninterpreted as interchangeable.
Some philosophers suggest that ‘physicalism’ isdistinct from ‘materialism’ for a reason quite unrelatedto the one emphasized by Neurath and Carnap. As the name suggests,materialists historically held that everything was matter —where matter was conceived as “an inert, senseless substance, inwhich extension, figure, and motion do actually subsist”(Berkeley, Principles of Human Knowledge, par. 9). But physics itselfhas shown that not everything is matter in this sense; for example,forces such as gravity are physical but it is not clear that they arematerial in the traditional sense (Lange 1865, Dijksterhuis 1961,Yolton 1983). So it is tempting to use ‘physicalism’ todistance oneself from what seems a historically important but nolonger scientifically relevant thesis of materialism, and related tothis, to emphasize a connection to physics and the physical sciences.However, while physicalism is certainly unusual among metaphysicaldoctrines in being associated with a commitment both to the sciencesand to a particular branch of science, namely physics, it is not clearthat this is a good reason for calling it ‘physicalism’rather than ‘materialism.’ For one thing, manycontemporary physicalists do in fact use the word‘materialism’ to describe their doctrine (e.g. Smart1963). Moreover, while ‘physicalism’ is no doubt relatedto ‘physics’ it is also related to ‘physicalobject’ and this in turn is very closely connected with‘material object’, and via that, with‘matter.’
In this entry, I will adopt the policy of using both terms interchangeably,though I will typically refer to the thesis we will discuss as‘physicalism’. It is important to note, though, thatphysicalism (i.e. materialism) is associated with a number of othermetaphysical and methodological doctrines. We will return to some ofthese when we discussPhysicalism and the Physicalist World-picture.
In approaching the topic of physicalism, one may distinguish what Iwill callthe interpretation question fromthe truthquestion. The interpretation question asks:
The truth question asks:
There is obviously a sense in which the second question herepresupposes an answer to the first — you need to know what a statementmeans before you can ask whether it's true — and I will begin with theinterpretation question. Nevertheless, the issues here turn out to besomewhat technical, and those new to the topic might like to read onlythe first section of my discussion of the interpretation question,which is:Supervenience Physicalism: Introductory, andthen turn directly to the truth question which begins atThe Case Against Physicalism I: Qualia and Consciousness.
The interpretation question itself divides into two sub-questions,which I will callthe completeness question andthecondition question. The completeness question asks:
In other words, the completeness question holds fixed the issue of whatit means for something to satisfy the condition of being physical, andasks instead what it means foreverything to satisfy thatcondition. Notice that a parallel question could be asked of Thales:assuming we know what condition you have to satisfy to be water, whatdoes it mean to say that everything satisfies that condition?
The condition question asks:
In other words, the condition question holds fixed the issue of what itmeans for everything to satisfy some condition or other, and asksinstead whatis the condition, being physical, that everythingsatisfies. Notice again that a parallel question could be asked ofThales: assuming we know what it is for everything to satisfy somecondition or other, what is the condition, being water, that accordingto Thales, everything satisfies? In discussing the interpretationquestion, I will turn first to the completeness question, and thenconsider the condition question.
In attempting to answer the completeness question, it has becomecustomary since Davidson 1970 to look to the notion of supervenience.(The notion of supervenience is historically associated withmeta-ethics, but it has received extensive discussion in the generalmetaphysics and logic literature. For a survey, seesupervenience.)
The idea of supervenience might be introduced via an example due toDavid Lewis of a dot-matrix picture:
A dot-matrix picture has global properties — it issymmetrical, it is cluttered, and whatnot — and yet all there is tothe picture is dots and non-dots at each point of the matrix. Theglobal properties are nothing but patterns in the dots. They supervene:no two pictures could differ in their global properties withoutdiffering, somewhere, in whether there is or there isn't a dot (1986,p. 14).
Lewis's example gives us one way to introduce the basic idea ofphysicalism. The basic idea is that the physical features of the worldare like the dots in the picture, and the psychological or biologicalor social features of the world are like the global properties of thepicture. Just as the global features of the picture are nothing but apattern in the dots, so too the psychological, the biological and thesocial features of the world are nothing but a pattern in the physicalfeatures of the world. To use the language of supervenience, just asthe global features of the picture supervene on the dots, so tooeverything supervenes on the physical, if physicalism is true.
It is desirable to have a more explicit statement of physicalism,and here too Lewis's example gives us direction. Lewis says that, inthe case of the picture, supervenience means that “no two pictures canbe identical in the arrangement of dots but different in their globalproperties”. Similarly, one might say that, in the case of physicalism,no two possible worlds can be identical in their physical propertiesbut differ, somewhere, in their mental, social or biologicalproperties. To put this slightly differently, we might say that ifphysicalism is true at our world, then noother world can bephysically identical to it without being identical to it in allrespects. This suggests the following account of what physicalismis:
(1) Physicalism is true at a possible worldw iffany world which is a physical duplicate ofw is a duplicate ofwsimpliciter.
If physicalism is construed along the lines suggested in (1), then wehave an answer to the completeness question. The completeness questionasks: what does it mean to say thateverything is physical.According to (1), what this means is that if physicalism is true, thereis no possible world which is identical to the actual world in everyphysical respect but which is not identical to it in a biological orsocial or psychological respect. It will be useful to have a name forphysicalism so defined, so let us call itsuperveniencephysicalism.
Supervenience physicalism is relatively simple and clear, but whenconstrued as a formulation of physicalism, it faces five problems: (a)the lone ammonium molecule problem; (b) the modal status problem; (c)the epiphenomenal ectoplasm problem; (d) the blockers problem; and (e)the necessary beings problem. As we will see, while some of these arerelatively easily answered, others are much more challenging.
(Cf. Kim 1993.) Imagine a possible worldW* that is physicallyexactly like our world except in one trivial respect: it has one extraammonium molecule located, say, on Saturn's rings. It is natural tosuppose that atW*, the distribution of mental properties isexactly as it is in the actual world — the presence of an extramolecule does not make that much of a difference. On the other hand,for all (1) says, such a world might beradically different interms of the distribution of mental properties. Since (1) only refersto worlds that are exact minimal physical duplicates of our world, itis silent on worlds that are differenteven in minute details,and hence is silent onW*. It thus leaves open the possibilitythat, inW*, everything has mental properties, or nothing hasor the only things that have mental properties are Saturn's rings! Butthat seems absurd: whileW* is clearly not a physicalduplicate of our world, for it contains an extra molecule, thedistribution of mental properties inW* would neverthelessmatch that of the actual world.
There are a number of different responses to this problemin the literature (cf. Kim 1993). But perhaps the simplest response isthat the problem conflates two issues that are better kept apart: thequestion of what physicalism itself tells us aboutW*, and thequestion of what our general knowledge tells us aboutW*. Itis true that physicalism itself does not tell us anything about thedistribution of mental properties atW*. Nevertheless, we knowindependently what the distribution is — we know independently thatthe presence or absence of molecules on Saturn doesn't affect thingslike who has mental properties here on Earth. But why should one assumethat this last piece of knowledge should be a consequence ofphysicalism? To put the point slightly differently, imagine that wediscover that who has mental properties on Earthis in part afunction of the behavior of molecules on Saturn. That would of coursetell us that we are deeply wrong in our assumptions about how the worldworks. But it would not tell us that we are deeply wrong aboutphysicalism. (For further discussion of this point, see Paull and Sider1992, and Stalnaker 1996.)
Some philosophers (e.g. Davidson 1970) have thought of physicalism as aconceptual or necessary truth, if it is true at all. But most havethought of it as contingent, a truth about our world which might havebeen otherwise. The statement of physicalism encoded in (1) allows away in which this might be so. (1) tells us that physicalism is true ata world just in case the world in question conforms to certainconditions. But it leaves it open whether or not the actual worldconforms to those conditionsas a matter of fact. Perhaps itisnot true of our world that a physical duplicate of it wouldbe a psychological duplicate. If so, physicalism would not be true atour world.
But for some it is puzzling that physicalism is stated using modalnotions (i.e. notions such as possible worlds) and nonetheless iscontingent. To see the problem, notice first that, superveniencephysicalism tells us that the physical truths of the worldentailall the truths; hence
(2) The physical truths entail all thetruths.
Now suppose thatS is a statement which specifies the physical nature of the actual world andS* is a statementwhich specifies the total nature of the world. (It might be thatneitherS norS* are expressible in languages we canunderstand, but let us set this aside.) If supervenience physicalism istrue, it will then be true that:
(3)S entailsS*
On the other hand, (3) is clearly a necessary truth. However, if (3) isa necessary truth, how canphysicalism be contingent? Afterall, (3) seems equivalent to physicalism. But if the two areequivalent, how can one be necessary and the other contingent?
But the response to this problem is straightforward. (3) isnecessary, but it isnot equivalent to physicalism. Rather,(3)follows from physicalism given various contingentassumptions, in particular the assumptions thatS andS*are the statements we say they are — it iscontingent fact, for example thatS* summarizes the totalnature of the world. On other hand, (2)is equivalent tophysicalism but it isnot necessary. (It is important to bearin mind here that not all entailment claims are necessary. Consider‘my aunt's favorite statement entails my uncle's favorite’— that statement is contingent even though it is most naturallythought of as an entailment claim.)
(Cf. Horgan 1983, Lewis 1983.) Imagine a possible worldW thatis exactly like our world in respect of the distribution of physicaland mental properties, but for one difference: it contains some pureexperience which does not interact causally with anything else in theworld —epiphenomenal ectoplasm, to give it a name. Theproblem this possibility presents for (1) is that, if (1) provides thecorrect definition of physicalism, and if physicalism is true at theactual world, then there isno possible world of the kind wejust described, i.e.,W does not exist. The reason is thatW is by assumption a physical duplicate of our world; butthen, if physicalism is true at our world,W should be aduplicate simpliciter of our world. ButW is patently not aduplicate of our world: it contains some epiphenomenal ectoplasm thatour world lacks. On the other hand, it seems quite wrong to say thatW is an impossibility — at any rate, physicalism should notentail that it is impossible.
In order to solve the epiphenomenal ectoplasm problem, we need toadjust (1) so that it does not have the truth of physicalism ruling outW as a possible world. While there are a number of differentproposals about how to do this, one influential proposal is due to Frank Jackson(cf. Jackson 1993. For earlier proposals and further discussion, seeHorgan 1983 and Lewis 1983.) He proposes replacing (1) with:
(4) Physicalism is true at a possible worldw iffany world which is a minimal physical duplicate ofw is aduplicate ofwsimpliciter
By ‘minimal physical duplicate’, Jackson means a possibleworld that is identical in all physical respects to the actual world,but which does not contain anything else; in particular, it does notcontain any epiphenomenal ectoplasm. Unlike (1), (4) does not havephysicalism ruling outW, and so (4) is on the face of it preferable to (1) asa statement of physicalism.
A different proposal is due to David Chalmers (1996). He proposes replacing (1) with:
(5) Physicalism is true at a possible worldw iffany world which is a physical duplicate ofw is apositive duplicate ofw
By ‘positive duplicate’, Chalmers means a possibleworld that instantiates all the positive properties of the actualworld, where in turn a positive property is defined as “one thatif instantiated in a world W, is also instantiated by thecorresponding individual in all worlds that contain W as a properpart” (1996, p. 40). Unlike (1), and like (4), (5) does not havephysicalism ruling outW, and so (5) is on the face of itpreferable to (1) as a statement of physicalism.
(Cf. Hawthorne 2002, Leuenberger 2008) Imagine a possible worldexactly like ours with respect to the distribution of mental andphysical properties, except for this difference: the relation betweenphysical facts and mental facts is weaker thansupervenience—mental facts are entailed by physical facts solong as there are no facts which block that entailment—blockers,as they are called. For example, being in an overall physicalcondition P will necessitate being in pain so long as you do not alsoinstantiate some further property B. If you are in both P and B youare not in pain; but if you are in P and not in B, you will be inpain.
The problem that this possibility raises for superveniencedefinitions of physicalism is as follows. Let us suppose that therelation obtaining at a worldW between the mental and thephysical is one of weak necessity as just defined; that is, supposethat, atW, the mental is necessitated by the physical butonly if certain blockers are absent. Intuitively it would seem thatphysicalism is false atW. On the other hand, if physicalismis defined in the way suggested by Jackson it would be true. Afterall, applied toW, Jackson's definition says that physicalismis true atW just in case any minimal physical duplicateofW is a duplicate simpliciter. But that seems to be trueofW as we have imagined it. Conclusion: if blockers arepossible, physicalism is false atW, and yet it should not befalse on Jackson's definition.
There are a number of possible responses to the blockers problem. Oneis to resist the intuition that physicalism is false atW inthe circumstance described, even if we adopt Jackson's definition ofphysicalism. A different response is to adopt a formulation ofphysicalism that is weaker than supervenience physicalism; this is thestrategy pursued in Leuenberger 2008. A third response is to say thatwhat the blockers problem brings out is that there is a cleardifference between two ways that physicalists have sought to respondfor the epiphenomenal ectoplasm problem; in particular, if one adopts(5) rather than (4) as one's response to the epiphenomenal ectoplasmproblem, this would have the advantage that it does not also face theblockers problem. For if the relation of the mental to the physicalthat obtains atW is one of weak necessity, then not only isphysicalism false but it is also false that any world which isphysical duplicate ofW is a positive duplicate ofW— at some physical duplicate worlds, for example, there will beno psychological properties at all. None of these responses areclearly correct however, and the proper treatment of the blockersproblem (and indeed of the epiphenomenal ectoplasm problem, of whichthe blockers problem is a development) is an open question in theliterature (For simplicity, I will continue with (1), rather than witheither (4) or (5); nothing in what follows will turn on thischoice.)
(Cf. Jackson 1998) Imagine a necessary being — that is, a being whichexists in all possible worlds — which is essentially nonphysical.(Some theists believe that God provides an example of such a being.) Ifsuch a non-physical being exists, it is natural to suppose thatphysicalism is false. But if physicalism is defined according to (1),the existence of such a being is compatible with physicalism. Forconsider: if the actual world is wholly physical, apart from thenecessary non-physical being, any minimal physical duplicate of theactual world is a duplicate simpliciter. Since the non-physical beingexists in all possible worlds, it exists at all worlds which areminimal physical duplicates of the actual world. So we seem to face aproblem: the existence of the non-physical necessary being entails thatphysicalism is false, but the definition of physicalism permits it tobe true in this case.
Like the blockers problem, this problem is not very easily answered. Lyingbehind the problem is a deeper issue about the correct interpretationof necessity and possibility — the modal notions one uses to formulatesupervenience. On one way of interpreting these notions, the existenceof a necessary being of this sort is incoherent. A reason is that itwould violate David Hume's famous dictum that there are no necessaryconnections between distinct existences — the being is distinct fromthe physical world and yet is necessitated by it. On another way ofinterpreting these notions, however, there is nothing incoherent in theidea of such a being. The correct way to think about modal notions,however, is a topic that is well beyond the scope of our discussionhere. The problem seems to be that the supervenience definition ofphysicalism in effect presupposes something like Hume's dictum, in thatit uses failure of necessitation as a test for distinctness. But thismeans that someone who denies the dictum will have to find analternative way of formulating physicalism.
We saw earlier that physicalism is intended as a very general claimabout the nature of the world. Nevertheless, by far the most discussionof physicalism in the literature has been in the philosophy of mind. Thereason for this is that it is in philosophy of mind that we find themost plausible and compelling arguments that physicalism is false.Indeed, as we will see later on, arguments about qualia andconsciousness are usually formulated as arguments for the conclusionthat physicalism is false. Thus, a lot of philosophy of mind is devotedto a discussion of physicalism.
While the issue of physicalism is central to philosophy of mind,however, it is important also to be aware that superveniencephysicalism is neutral on a good many of the questions that are pursuedin philosophy of mind, and pursued elsewhere for that matter. If youread over the philosophy of mind literature, you will often find peopledebating a number of different issues: whether therearemental states at all; whatsort of thing mental states are; towhat extent mental states are environmentally determined. Given themultifariousness of mental states, it is quite likely that the correctposition will be some kind of combination of these positions. But thisis a question of further inquiry that is irrelevant to physicalismitself. So physicalism itself leaves many debates in the philosophy ofmind unanswered.
This point is sometimes expressed by saying thatsuperveniencephysicalism is minimal physicalism (Lewis 1983): it is intended tocapture the minimal or core commitment of physicalism. Physicalists maydiffer from one another in many ways, but all of them must at leasthold supervenience physicalism. (Notice that the idea that (1) capturesthe minimal commitment of physicalism is a distinct idea from that of aminimal physical duplicate which Jackson uses in his attempt to capture minimalphysicalism.)
Two issues here require further comment. First, in some discussionsin philosophy of mind, the term ‘physicalism’ is used torefer to the identity theory, the idea that mental states or propertiesare neurological states or properties (Block 1980). In this use of theterm, one can reject physicalism by rejecting the identity theory — soby that standard a behaviorist or functionalist in philosophy of mindwould not count as a physicalist. Obviously, this is a much morerestricted use of the term than is being employed here.
Second, one might think that supervenience physicalism isinconsistent with eliminativism, the claim that psychological states donot exist, for the following reason. Suppose psychological statessupervene on physical states. Doesn't that mean, contrary toeliminativism, that there mustbe some psychological states?The answer to this question is ‘no.’ For consider: thetelephone on my desk has no psychological states whatsoever.Nevertheless it is still true (though, admittedly, a little odd) to saythat a telephone which is identical to my telephone in all physicalrespects will be identical to it in all psychological respects. In thesense intended, therefore, one thing can be psychologically identical toanother even when neitherhas any psychological states.
To what extent does supervenience physicalism capture minimalphysicalism, the core commitment of all physicalists? In order toanswer this question it is worth comparing and contrastingsupervenience physicalism with two alternative statements ofphysicalism that one finds in the literature: token and typephysicalism.
Token physicalism is the view that every particular thing in theworld is a physical particular. Here is one formulation of thisidea:
Token physicalism:
For every actual particular (object, event or process)x,there is some physical particulary such thatx =y.
Supervenience physicalism neither implies nor is implied by tokenphysicalism. To see that token physicalism does not imply superveniencephysicalism, one need only note that the former is consistent with aversion of dualism, namely property dualism. The mere fact that everyparticular has a physical property does not rule out the possibilitythat some particulars also have non-supervenient mental properties,i.e. mental properties that are only contingently related to thephysical. But supervenience physicalismdoes rule out thispossibility. Since token physicalism does not rule out property dualismbut supervenience physicalism does, the first does not imply thesecond.
To see that supervenience physicalism does not imply tokenphysicalism is more difficult. The crucial point is that tokenphysicalism requires that for every psychological or social particular,there is some physical particular with which it is identical. But thisis by no means obviously true. Consider the United States Court ofAppeals for the Seventh Circuit. This might be thought of as a social orlegal object. But then, according to token physicalism, there must besome physical object for it to be identical with. But there might be nophysical object (in any natural sense of the term) which is identicalto the Court of Appeals for the Seventh Circuit. On the other hand,supervenience physicalism imposes no such requirement, and sosupervenience physicalism does not imply token physicalism (For theclassic presentation of this point, see Haugeland 1983).
The point that supervenience physicalism is logically distinct fromtoken physicalism is an important one. One thing it shows is thattoken physicalism (since it is consistent with property dualism) doesnot capture minimal physicalism, and so the distinction between tokenphysicalism and supervenience physicalism is no objection to thelatter. But the difference between the two theses also raises adifferent question. Given that token physicalism does not capture theminimal commitment of physicalism, why has token physicalism been thesubject of such discussion? One reason is that token physicalismprovides one version of the idea that upper level scientific claimsrequires physical mechanisms. Supervenience physicalism does not onits own entail this. But token physicalism is often seen as a way toensure this requirement. (For the classic presentation of this point,see Fodor 1974; see also Papineau 1996. And for a different viewabout token physicalism see Latham 2001.)
Having considered token physicalism, we can now turn to typephysicalism. Type physicalism is a generalization and extension of theidentity theory, which we considered above. It holds that that everyproperty (or at least every property that is or could be instantiatedin the actual world) is identical with some physical property. Here isa statement of this sort of idea:
Type physicalism:
For every actually instantiated mental propertyF, there issome physical propertyG such thatF=G.
Unlike token physicalism, type physicalism certainlydoesentail supervenience physicalism: if every property instantiated in theactual world is identical with some physical property, then a worldidentical to our world in physical respects will of course be identicalto it in all respects.
Nevertheless the reverse entailment does not hold. Superveniencephysicalism, as we have been understanding it, is a contingent thesisthat is consistent with the possibility (if not the actuality) ofdisembodiment. But type physicalism as defined here is inconsistentwith this possibility. To that extent, supervenience physicalism doesnot entail type physicalism.
Earlier we noted that philosophers such as Davidson have thoughtthat physicalism is a necessary truth. Even on that assumption,however, it is still not completely obvious that superveniencephysicalism entails type physicalism. The reason for this has to dowith questions concerning the logical (or Boolean) closure of the setof physical properties — ifP,Q andR arephysical properties, which of the various logical permutations ofP,Q andR are likewise physical properties?On some assumptions concerning closure and supervenience, superveniencephysicalism (construed as a necessary truth) entails type physicalism;on other assumptions, it doesn't. But the problem is that theassumptions themselves are difficult to interpret and evaluate, and sothe issue remains a difficult one. It is not necessary for our purposesto settle the question concerning closure here. (For further discussionof these issues see Kim 1993, Bacon 1990, Van Cleve 1990, Stalnaker1996.)
Before the development and study of the notion of supervenience,physicalism was often stated as a reductionist thesis. It willtherefore be useful to contrast the supervenience formulation ofphysicalism with various reductionist proposals, and also turn to aquestion that has received a lot of attention in the literature, viz.,whether a physicalist must be a reductionist.
The main problem in assessing whether a physicalist must be areductionist is that there are various non-equivalent versions ofreductionism.
One idea is tied to the notion of conceptual or reductive analysis.When philosophers attempt to provide an analysis of some concept ornotion, they usually try to provide a reductive analysis of the notionin question, i.e. to analyze it in other terms. Applied to thephilosophy of mind, this notion might be thought of entailing the ideathat every mental concept or predicate is analyzed in terms of a physicalconcept or predicate. A formulation of this idea is (6):
(6) Reductionism is true iff for each mental predicateF, there is a physical predicateG such that asentence of the form ‘x isF iffx isG’ is analytically true.
While one occasionally finds in the literature the suggestion thatphysicalists are committed to (6) in fact, no physicalist since beforeSmart (1959) has (unqualifiedly) held anything like (6). Adapting Ryle(1949), Smart supposed that in addition to physical expressions thereis a class of expressions which are topic-neutral, i.e. expressionswhich were neither mental nor physical but when conjoined with anytheory would greatly increase the expressive power of the theory. Smartsuggested that one might analyze mental expressions in topic-neutral(but not physical) terms, which in effect means that a physicalistcould reject (6). It is fair to say that this move is one of thecentral innovations of philosophy of mind, a move to a large extentendorsed and developed later on by functionalists and cognitive scientists.
A different notion of reduction derives from the attempts ofphilosophers of science to explain intertheoretic reduction. Theclassic formulation of this notion was given by Ernest Nagel (1961).Nagel said that one theory was reduced to another if you couldlogically derive the first from the second together with what he calledbridge laws, i.e., laws connecting the predicates of the reduced theory(the theory to be reduced) with the predicates of the reducing theory(the theory to which one is reducing). Here is a formulation of thisidea, where the theories in question are psychology andneuroscience:
(7) Reductionism is true iff for each mental predicateF there is a neurological predicateG such that asentence of the form ‘x isF iffx isG’ expresses a bridge law.
Once again, however, there is no reason at all why physicalists need toaccept that reductionism is true in the sense of (7). Indeed, manyphilosophers have argued that there are very strong empirical reasonsto deny that anything like (7) is going to be the case. The reason isthis. Many different neurological processes (whether in our own speciesor a different one) could underlie the same psychological process —indeed, given science fiction, even non-neurological processes mightunderlie the same psychological process. But if multiple realizability— as this sort of idea is called — is true, then (6) seems to befalse. (Fodor 1974, but for recent alternative views, see Kim 1993).
A third notion of reductionism is more metaphysical in focus thaneither the conceptual or theoretical ideas reviewed so far. Accordingto this notion, reductionism means that the properties expressed by thepredicates of (say) a psychological theory are identical to theproperties expressed by the predicates of (say) a neurological theory— in other words, this version of reductionism is in essence a versionof type physicalism or the identity theory. However, as we have seen,if physicalists are committed only to supervenience physicalism, theyare not committed to type physicalism. Hence a physicalist need not bea reductionist in this metaphysical sense.
A final notion of reductionism that needs to be distinguished fromthe previous three concerns whether mental statements followapriori from non-mental statements. Here is a statement of thissort of idea,
(8) Reductionism is true iff for each mental predicateF there is non-mental predicateG such that asentence of the form “ifx isF thenx isG” isa priori.
What (8) says is that if reductionism is true,a prioriknowledge alone, plus knowledge of the physical truths will allow oneto know the mental truths. This question is in fact a highly vexed onein contemporary philosophy. However, this question is usually debatedin the context of another, viz., the question ofa posteriorianda priori physicalism. It is to that question, therefore,to which we will now turn.
We saw earlier that if physicalism is true then (5) is true, where‘S’ is a sentence that reports the entire physicalnature of the world and ‘S*’ is a sentence thatreports the entire nature of the world:
(5)S entailsS*
Another way to say this is to say that if physicalism is true, then thefollowing conditional is necessarily true:
(9) IfS thenS*
Indeed, this is a general feature of physicalism: if it is true thenthere will always be a necessary truth of the form of (9).
Now, if (9) is necessary the question arises whether it isapriori, i.e. knowable independent of empirical experience, orwhether it isa posteriori, i.e. knowable but notindependently of empirical experience. Traditionally, every statementthat was necessary was assumed to bea priori. However, sinceKripke's classic workNaming and Necessity (1980),philosophers have become used to the idea that there are truths whichare both necessary anda posteriori. Accordingly many recentphilosophers have defended aposteriori physicalism: the claimthat statements such as (8) are necessary anda posteriori(cf. Loar 1997). Moreover, they have used this point to try to disarmmany objections to physicalism, including those concerning qualia andintentionality that we will consider in a moment. Indeed, as we havejust noted, some philosophers have suggested that the necessaryaposteriori provides the proper interpretation of non-reductivephysicalism.
The appeal to the necessarya posteriori is on the surfacean attractive one, but it is also controversial. One problem arisesfrom the fact that Kripke's idea that there are necessary andaposteriori truths can be interpreted in two rather different ways.On the first interpretation — I will call it the derivation view —while there are necessarya posteriori truths, these truthscan be deriveda priori from truths which areaposteriori and contingent. On the second interpretation — I willcall it the non-derivation view — there arenon-derivednecessarya posteriori truths, i.e. necessary truths which arenot derived from any contingent truths (or anya priori truthsfor that matter). The problem is that when one combines the derivationview with the claim that (9) is necessary anda posteriori,one encounters a contradiction. If the derivation view is correct, thenthere is some contingent anda posteriori statementS# that logically entails (9). However, ifS#logically entails (9) then (since ‘IfC, then ifA thenB’ is equivalent to ‘IfC&A, thenB’) we can infer that thefollowing is both necessary anda priori:
(10) IfS &S# thenS*.
One the other hand, if physicalism is true, andS summarizesthe total nature of the world it seems reasonable to suppose thatS#was already implicitly included in S. In otherwords it seems reasonable to suppose that (10) is simply an expansion of(9). But if (10) is just an expansion of (9), then if (10) isapriori, (9) must also bea priori. But that means ourinitial assumption is false: (9) is not a necessaryaposteriori truth after all (Jackson 1998).
How might ana posteriori physicalist respond to thisobjection? The obvious response is to reject the derivation view of thenecessarya posteriori in favor of the non-derivation view.But this is just to say that if one wants to defendaposteriori physicalism, one will have to defend the non-derivationview of the necessarya posteriori. However, the problem hereis that the non-derivation view is very controversial. Indeed, thequestion of which interpretation of Kripke's work is the right one, isone of the most vexed in contemporary analytic philosophy. So it is notsomething that we can hope to solve here. (For discussion, see Byrne1999, Chalmers 1996, 1999, Jackson 1998, Loar 1997, Lewis 1994,Yablo 1999, and the papers in Gendler and Hawthorne 2004)
A further issue needs to be mentioned in connection with thedistinctions between non-reductive and reductive physicalism, and witha posteriori anda priori physicalism. Kim and othershave suggested that non-reductive physicalism is a form ofemergentism, the view that supervenience provides a way tointerpret the relation between the psychological and the physical insuch a way that the psychological isgenuinely novel.(Emergentism was influential in the first forty years of the twentiethcentury, but it is not unfair to say that similar positions aredefended by many contemporary philosophers. For the historicalbackground, see MacLaughlin 1992.)
Now it is difficult to evaluate emergentism because it is unclearwhat genuine novelty is supposed to be. On one interpretation, what theemergentists meant by ‘genuine novelty’ wasnon-predictability in principle, i.e. the idea that no matter how muchphysical information you had about a creature you could not predict onthat basis alone what experiences, if any, they might have. On thisinterpretation, emergentism seems very similar toa posterioriphysicalism (Byrne 1993). On a different interpretation, what theemergentists meant by ‘genuine novelty’ was the idea thatthere was only a contingent connection between psychological states andphysical state, a connection perhaps mediated by contingentpsycho-physical laws. On this interpretation, however, emergentismseems simply to be a denial of physicalism as we have defined ithere.
There is, however, a third interpretation of ‘genuinenovelty’ which requires separate treatment. On thisinterpretation, what the emergentist intends is a position whichweaves together both dualism and physicalism. On the one hand, theemergentist wants to say that mental facts and physical facts aremetaphysically distinct—just as a standard dualist does. On theother hand, emergentist wants to agree with the physicalist thatmental facts are necessitated by, and so supervene on, the physicalfacts. If this sort of combined position is coherent, thenemergentismdoes presents a major problem for superveniencephysicalism. For emergentism seems to be consistent both with dualismand with supervenience physicalism, since according to emergentism,any world physically identical to the actual world will be identicalto it in all respects. However — and here is the key point— we have been assuming all along that supervenience physicalismis inconsistent with dualism. In short, the problem is this: (a)supervenience physicalism is consistent with emergentism; (b)emergentism is consistent with dualism; but (c) superveniencephysicalism isinconsistent with dualism.
How are we to respond to this problem? Some philosophers suppose thatthe problem is so serious for supervenience physicalism that the onlything to do is to give it up in favor of something else. In Kim1998, for example, we find the suggestion that the only form of physicalismworthy of the name is a form of type physicalism. The major burden onthis proposal, however, is that type physicalism was given up for avery good reason, viz., multiple realizability.
A different suggestion has been made by Melnyk (2003), who argues thatphysicalism should be defined not in terms of supervenience but interms of a different relation, realization. For Melnyk, physicalism isrealization physicalism, the thesis that every instantiated propertyis either physical or else is realized by a physical property. Theintent of this proposal is that realization physicalism entailssupervenience physicalism but not vice versa; as such realizationphysicalism may be distinguished from emergentism even whilesupervenience physicalism may not be. (A different notion ofrealization, and of realization physicalism, has been developed byShoemaker 2007; we will set Shoemaker aside here.)
Melnyk's suggestion is a natural one in the face of the emergentismproblem. However, as he himself notes at one point (2003, p. 23),it does face a problem having to do with the definition ofrealization. For Melnyk, a propertyF realizes apropertyG if and only if (a)F is identical to asecond-order property, the property of having some property that has acertain causal or theoretical role; and (b)G is the propertythat plays the causal or theoretical role in question. But what arethe properties involved in spelling out these causal or theoreticalroles? If physicalism is true at all, it must be true of these properties asmuch as any other properties. But then by realization physicalism,these properties themselves will be either physical or realized byphysical properties. If the first option is taken, the realizationphysicalist will stand revealed as holding a version of identityphysicalism (one level up, as it were), and thus will face themultiple realization objection. If the second option is taken, therealization physicalist looks committed to an infinite regress, sincenow we have further properties realized by physical properties and,correlatively, further causal or theoretical roles. To avoid theregress, the realization physicalism might say that these propertiessupervene on physical properties. But now it hard to see thedifference between the realization physicalist and the superveniencephysicalist.
Retreating from supervenience is only one way to respond to thechallenge presented by emergentism. Another way is to attack thebasis of the objection. For the emergentist problem is a problem onlyto the extent that emergentism is a coherent possibility. But itremains a controversial issue whether the emergentist picture can bemade fully coherent (Stalnaker 1996). For example, one sort ofargument against it is that it seems to violate Hume's dictum thatthere are no necessary connections between distinct existences:according to emergentism, the levels of the world are wholly distinctfrom each other, and yet are necessarily connected (Jackson 1993)However, as we saw in our discussion above of the necessary beingsproblem, the proper interpretation of Hume's dictum is itself a matterof controversy, so emergentism remains controversial.
Earlier we distinguished two interpretative questions with respect tophysicalism, the completeness question and the condition question. Sofar we have been concerned with the completeness question. I turn nowto the condition question, the question of what it is for something (anobject, an event, a process, a property) to be physical.
The condition question has received less attention in theliterature than the questions we have been studying so far. But it isjust as important. Without any understanding of what the physical is,we can have no serious understanding of what physicalism is. After all,if we say that, no two possible worlds can be physicalduplicates without being duplicates simpliciter, we don't know whatwe've said unless we understand what it would take to be aphysical duplicate, as opposed (say) to a chemical duplicate or afinancial duplicate. (The point here is a quite general one: if Thalessays that everything is water, or Up-to-Date-Thales says everythingsupervenes on water, we don't understand what he says unless he sayssomething about what water is. The physicalist is in the sameposition.)
So what is the answer to the condition question? If we concentratefor simplicity on the notion of a physical property, we can discern twokinds of answers to this question in the literature. The first ties thenotion of a physical property to a notion of a physical theory, forthis reason we can call it the theory based conception of a physicalproperty:
The theory-based conception:
A property is physical iff it either is the sort of property thatphysical theory tells us about or else is a property whichmetaphysically (or logically) supervenes on the sort of property thatphysical theory tells us about.
According to the theory-based conception, for example, if physicaltheory tells us about the property of having mass, then having mass isa physical property. Similarly, if physical theory tells us about theproperty of being a rock — or, what is perhaps more likely, if theproperty of being a rock supervenes on properties which physical theorytell us about — then it too is a physical property. (The theory-basedconception bears some relation to the notion of physical1 discussed inFeigl 1965; more explicit defense is found in Smart 1978, Lewis 1994,Braddon-Mitchell and Jackson 1996, and Chalmers 1996.)
The second kind of answer ties the notion of a physical property tothe notion of a physical object, for this reason we can call it theobject-based conception of a physical property:
The object-based conception:
A property is physical iff: it either is the sort of property requiredby a complete account of the intrinsic nature of paradigmatic physicalobjects and their constituents or else is a property whichmetaphysically (or logically) supervenes on the sort of propertyrequired by a complete account of the intrinsic nature of paradigmaticphysical objects and their constituents.
According to the object-based conception, for example if rocks, trees,planets and so on are paradigmatic physical objects, then the propertyof being a rock, tree or planet is a physical property. Similarly, ifthe property of having mass is required in a complete account of theintrinsic nature of physical objects and their constituents, thenhaving mass is a physical property. (The best examples of philosopherswho operate with the object-conception of the physical are Meehl andSellars 1956 and Feigl 1965; more recent defense is to be found inJackson 1998.)
It is important to note that both conceptions of the physical remainsilent on the question of whether topic-neutral or functionalproperties should be treated as physical or not. To borrow a phrasefrom Jackson (1998), however, it seems best to treat these propertiesas onlooker properties: given any set of physical properties, one mightadd onlooker properties without compromising the integrity of the set.But onlooker properties should not be treated as being physical bydefinition.
Along with the concepts of space, time, causality, value, meaning,truth and existence, the concept of the physical is one of the centralconcepts of human thought. So it should not be surprising that anyattempt to come to grips with what a physical property is will becontroversial. The theory and object conceptions are no different: eachhas provoked a number of different questions and criticisms. In this section, I willreview some main ones.
One might object that both conceptions are inadequate because they arecircular, i.e., both appeal to the notion of something physical (atheory or an object) to characterize a physical property. But how canyou legitimately explain the notion of one sort of physical thing byappealing to another?
However, the response to this is that circularity is only a problemif the conceptions are interpreted as providing a reductive analysis ofthe notion of the physical. But there is no reason why they should beinterpreted as attempting to provide a reductive analysis. After all,we have many concepts that we understand without knowing how to analyze(cf. Lewis 1970). So there seems no reason to suppose that either thetheory or object conception is providing anything else but a way ofunderstanding the notion of the physical.
The point here is an important one in the context of the conditionquestion. Earlier we said that the condition question was perfectlylegitimate because it is legitimate to ask what the condition of beingphysical is that, according to physicalism, everything has. But thislegitimate question should not be interpreted as the demand for areductive analysis of the notion of the physical. Consider Thalesagain: it is right to ask Thales what he means by ‘water’— and in so doing demand an understanding of the notion of water —but it is wrong to demand of him a conceptual analysis of water.
One might object that any formulation of physicalism which utilizes thetheory-based conception will be either trivial or false. Carl Hempel(cf. Hempel 1970, see also Crane and Mellor 1990) provided a classicformulation of this problem: if physicalism is defined via reference tocontemporary physics, then it is false — after all, who thinks thatcontemporary physics is complete? — but if physicalism is defined viareference to a future or ideal physics, then it is trivial — afterall, who can predict what a future physics contains? Perhaps, forexample, it contains even mental items. The conclusion of the dilemmais that one has no clear concept of a physical property, no conceptthat is clear enough to do the job that philosophers of mind want thephysical to play.
One response to this objection is to take its first horn, and insistthat, at least in certain respects contemporary physics really iscomplete or else that it is rational to believe that it is (cf. Smart1978, Lewis 1994 and Melnyk 1997, 2003). But while there is something rightabout this, there is also something wrong about it. What is right aboutit is that there is a sense in which it is rational to believe thatphysics is complete. After all, isn't it rational to believe that themost current science is true? But even so — and here is what is wrongabout the suggestion — it is still mistaken todefinephysicalism with respect to the physics that happens to be true in thisworld. The reason is that whether a physical theory is true or not is afunction of the contingent facts; but whether a property is physical ornot is not a function of the contingent facts. For example, considermedieval impetus physics. Medieval impetus physics is false (though ofcourse it might not have been) and thus it is irrational to suppose ittrue. Nevertheless, the property of having impetus — the centralproperty that objects have according to impetus physics — is aphysical property, and a counterfactual world completely described byimpetus physics would be a world in which physicalism is true. But itis hard to see how any of this could be right if physicalism weredefined by reference to the physics that we have now or by the physicsthat happens to be true in our world.
A different response to Hempel's dilemma is that what it shows, ifit shows anything, is that a particular proposal about how to define aphysical property — namely, via reference to physics at a particularstage of its development — is mistaken. But from this one can hardlyconclude that we have no clear understanding of the concept at all. Aswe have seen, we have many concepts that we don't know how to analyze.So the mere fact — if indeed it is a fact — that a certain style ofanalysis of the notion of the physical fails does not mean that thereis no notion of the physical at all, still less that we don'tunderstand the notion.
One might object that, while these remarks are perfectly true, theynevertheless don't speak to something that is right about Hempel'sdilemma, namely that for the theory-conception to be complete one needsto say a little more about what physical theory is. Here, however, wecan appeal to the fact that we have a number of paradigms of what aphysical theory is: common sense physical theory, medieval impetusphysics, Cartesian contact mechanics, Newtonian physics, and modernquantum physics. While it seems unlikely that there is anyonefactor that unifies this class of theories, it does not seemunreasonable that there is a cluster of factors — a common oroverlapping set of theoretical constructs, for example, or a sharedmethodology. In short, we might say that the notion of a physicaltheory is a Wittgensteinian family resemblance concept, and this shouldbe enough to answer the question of how to understand physicaltheory.
Hempel's dilemma against the theory-conception is similar to anobjection that one often hears propounded against the object-conception(cf. Jackson 1998). Imagine the possibility of panpsychism, i.e. thepossibility that all the physical objects of our acquaintance areconscious beings just as we are. Would physicalism be true in thatsituation? It seems intuitively not; however, if physicalism is definedvia reference to the object-conception of a physical property then itis hard to see why not. After all, according to that conception,something is a physical property just in case it is required by acomplete account of paradigmatic physical objects. But this makes noreference to thenature of paradigmatic physical objects, andso allows the possibility that physicalism is true in the imaginedsituation.
The first thing to say in response is that the mere possibility ofpanpsychism cannot really be what is at issue in this objection. For nomatter how implausible and outlandish it sounds, panpsychism per se isnot inconsistent with physicalism (cf. Lewis 1983). After all, the factthat there aresome conscious beings is not contrary tophysicalism — why then should the possibility thateverythingis a conscious being be contrary to physicalism? So what is at issue inthe objection is not panpsychism so much as the possibility that theparadigms or exemplars in terms of which one characterizes the notionof the physical might turn out to be radically different from what wenormally assume — for example, they might turn out to be in someessential or ultimate respect mental. If that were so, it certainlydoes seem strange to say that physicalism would or could be true.
Once the problem is put like that, however, it is clear thatthe problem has a rather similar structure to other problems that arisewhen one tries to understand a concept in terms of paradigmatic objectswhich fall under the concept. Suppose one tried to define the conceptred in terms of similarity to paradigmatic red things, such as blood.Pursuing this strategy commits one to the idea that the belief thatblood is red is a piece of common knowledge shared among all those whoare competent with the term. But that seems wrong — someone whothought that blood was green would be mistaken about blood but notabout red. Now this problem is a difficult problem, however — and thisis the crucial point for our purposes — the problem is also a quitegeneral problem, and not particularly tied to the notion of thephysical. So to that extent, the concept of the physical does not seemto be any worse off than the concept of red. (For discussion of thegeneral strategy see Lewis 1997)
One idea that often emerges in the context of the panpsychismproblem, and indeed in the context of Hempel's dilemma, but deservesseparate treatment, is the so-called Via Negativa (see e.g. Monteroand Papineau 2005). The simplest way to introduce the Via Negativa is tointerpret it as a definition of the notion a physical propertysomething like this:F is a physical property if and only ifF is anon-mental property. But there are many reasons to resist such adefinition. Take vitalism. Vitalism isn't true, but it might havebeen true; there is no contradiction in it for example. So imagine aworld in which plants and animals instantiate the key propertyassociated with vitalism, viz., élan vital. It seemsreasonable to say that in that case plants and animals instantiate aproperty that is non-physical, i.e. élan vital is not physical.And yet one should not say on this account that plants and animalsinstantiate a mental property, i.e., élan vital is not mental.In short, élan vital is neither mental nor physical. But theVia Negativa as stated cannot accommodate that fact.
One might try to meet this objection by saying that what proponents ofthe Via Negativa intend is only a partial definition along theselines:F is a physical property only ifF is non-mental. Even soproblems remain. As we have seen élan vital causes a problembecause it is neither mental nor physical. But there might beproperties that are both mental and physical. Consider a version of the identitytheory according to which being in pain just is c-fibers firing. If wesuppose that such a theory is true, is the property of being inpain then mental or physical? Both presumably; but this could not be trueon the Via Negativa construed as a definition of what a physicalproperty is, even a partial definition. For if a property is mentaland physical, then, given the Via Negativa, it will be both mental andnon-mental which (of course) it can't be! Now obviously, there aregood questions about whether an identity theory along these lines isor could be true, but regardless of whether it is true, it should notbe ruled out simply because of a proposal about how to define the wordin which it is stated.
For these reasons, it is unlikely that the Via Negativa will do asdefinition of the notion of a physical property. However, there isdefinitely something right about it. For example, when we think ofproperties that would falsify physicalism we do often think of certainmental properties, e.g., the distinctive properties of ectoplasm orESP. However, this fact—that certain mental properties would,if instantiated, falsify physicalism—can be captured withoutdefining the physical as non-mental. A better way would be to requireany spelling out the notion of the physical, either the object-basedor the theory-based, that it respect that fact that some (perhapsuninstantiated) mental properties will be non-physical.
Finally, an interesting issue concerning the theory and objectconceptions of a physical property concerns the question of whetherthey characterize the same class of properties. There are a number ofdifferent possibilities here, not of all of which we can discuss. Butone that has received some attention in the literature is that physicaltheory only tells us about the dispositional properties of physicalobjects, and so does not tell us about the categorical properties, ifany, that they have — a thesis of this sort has been defended by anumber of philosophers, among them Russell (1927), Armstrong (1968),Blackburn (1992) and Chalmers (1996). However, if this is correct, itwould seem that the physical properties described by the theoryconception are only a sub-class of the physical properties described bythe object conception. For if physical objects do have categoricalproperties, those properties will not count as physical by thestandards of the theory conception. On the other hand, there seems noreason not to count them as physical in some sense or other. If that isright, however, then the possibility emerges that the theory- and theobject-conceptions characterize distinct classes of properties.
Perhaps because of its connection to the physical sciences,physicalism is sometimes construed as an entire package of views, whichcontains the metaphysical thesis I have isolated for discussion as onlyone part. If we want a name for the entire package of views includingthe metaphysical claim we might call it thePhysicalist WorldPicture. I will close our discussion of the interpretationquestion by considering the relation between physicalism (themetaphysical claim) and various other items that at least sometimeshave been thought to be a part of the Physicalist World Picture.
(a) Methodological Naturalism: the ideathat the mode of inquiry typical of the physical sciences will providetheoretical understanding of the world, to the extent that this sort ofunderstanding can be achieved. Physicalism is not methodologicalnaturalism because physicalism is a metaphysical thesis not amethodological thesis.(b) Epistemic Optimism: the idea that the mode ofunderstanding typical of the sciences can be used by us, i.e. by humanbeings, to explain the world in total, to provide a final theory of theworld. Physicalism is not epistemic optimism because, since commitmentto physicalism does not commit you to methodological naturalism, itclearly does not commit you to any optimism about the success of thatmethod in the long run.
(c)Final Theory: the idea that thereis afinal and complete theory of the world, regardless of whether we canformulate it. One might think it obvious that if physicalism is true,there is a final theory of the world. However, because of someunclarity in the notion of a theory, the issues here are not cut anddried. According to some views, something is a theory only if it isfinitely stateable in a language we can understand. If that is so,clearly physicalism does not entail the idea of a final theory. On alooser conception of a theory, however, it is reasonable to say thatphysicalism entails that there is a final theory.
(d)Objectivity: the idea that the final andcomplete theory of world, if it exists, will not involve any essentialreference to particular points of view or experiences. It is reasonableto say that physicalism entails objectivity. However, given thepossibilities of non-reductive ora posteriori physicalismeven here the issues are not settled. On those approaches, it seemspossible to have irreducible points of view or experiences superveningon something physical, which compromises objectivity.
(e)Unity of Science: the idea that all thebranches of sciences developed by us will or should be unified into asingle science, usually (but not always) thought of as physics. Thisthesis is clearly a methodological thesis about how science ought toproceed. As we have seen, however, physicalism is a metaphysical thesisrather than a methodological thesis about how science ought to proceed.Hence it is not equivalent to the unity of science thesis.
(f)Explanatory Reductionism: the idea that allgenuine explanations must be couched in the terms of physics, and thatother explanations, while pragmatically useful, can or should bediscarded as knowledge develops. Physicalism is not explanatoryreductionism because, as we saw in our discussion of non-reductivephysicalism, physicalism is consistent with the idea that specialsciences are quite distinct from physics. One might say that thespecial sciences are concerned with patterns in the physical thatphysicists themselves are not concerned with. For that reason thesubject matter of the special sciences is distinct from the subjectmatter of physics.
(g)Generality of Physics: the idea that everyparticular event or process which falls under a law of the specialsciences (i.e. sciences other than physics) also falls under a law ofphysics. In general, this view presupposes a view about laws andexplanation — for example, it implies or seems to imply that specialsciences have laws. But physicalism does not entail any suchthesis.
(h)Causal Closure of the Physical: the idea thatevery event has a physical cause, assuming it has a cause at all.Strictly speaking, physicalists are not committed to realism aboutcausation, so they are not committed to causal closure. (Of course,many physicalists do think that causal closure is true, as we will seebelow, but their position does not entail causal closure.)
(i)Empiricism: the idea that all knowledge (withthe possible exception of conceptual knowledge) is ultimately foundedon sensory or perceptual experience. Empiricism can be given adescriptive or a normative reading. On its descriptive reading, it ismost likely false. Most of the information that normal humans come todeploy seems to be caused by both experienceand inbornstructure and maturation. On the normative reading, the claim is thatjustification is, at the end of the day, based on experience. But thisepistemological thesis has nothing to do with physicalism.
(j)Nominalism: the idea that there are no abstractobjects, i.e., entities not located in space and time, such as numbers,qualities or propositions. If we assume that abstract objects, if theyexist, exist necessarily, i.e., exist in all possible worlds, thensupervenience physicalism is completely silent on the question ofwhether abstract objects exist. All supervenience says is that if aworld is a minimal physical duplicate of the actual world, it is aduplicate simpliciter. But if abstract objects exist then they clearlyexist in both the actual world and any duplicate of the actual world.What this suggests is that nominalism is a distinct issue fromphysicalism (Schiffer 1987, Stoljar 1996).
(k)Atheism: the idea that there is no God astraditionally conceived. In the 17th and 18th century, physicalism (ormaterialism, as it was then known) was widely viewed as inconsistentwith belief in God (Yolton 1983). Nowadays, this issue is somewhat lessdiscussed. Nevertheless, as we noted previously,if God isthought of as essentially non-physical, then Atheismdoes seemto be a consequence of physicalism, at least on some interpretations ofthe background modal notions.
Having provided an answer to the interpretation question, I now turn tothe truth question: is physicalism (as we have interpreted it so far)true? I will first discuss three reasons for supposing that physicalismis not true. Then I will consider the case for physicalism.
The main argument against physicalism is usually thought to concernthe notion of qualia, the felt qualities of experience. The notion ofqualia raises puzzles of its own, puzzles having to do with itsconnection to other notions such as consciousness, introspection,epistemic access, acquaintance, the first-person perspective and so on.However the idea that we will discuss here is the apparentcontradiction between the existence of qualia and physicalism.
Perhaps the clearest version of this argument is Jackson's knowledgeargument. (There are also a number of other arguments in this area —for a very good recent discussion, see Chalmers 1996). This argumentasks us to imagine Mary, a famous neuroscientist confined to a blackand white room. Mary is forced to learn about the world via black andwhite television and computers. However, despite these hardships Marylearns (and therefore knows) all that physical theory can teach her.Now, if physicalism were true, it is plausible to suppose that Maryknows everything about the world. And yet — and here is Jackson'spoint — it seems she does not know everything. For, upon beingreleased into the world of color, it will become obvious that, insideher room, she did not know what it is like for both herself and othersto see colors — that is, she did not know about the qualiainstantiated by particular experiences of seeing colors. FollowingJackson (1986), we may summarize the argument as follows:
P1. Mary (before her release) knows everything physicalthere is to know about other people.P2. Mary (before her release) does not know everything there is toknow about other people (because she learns something about them onbeing released).
Conclusion. There are truths about other people (and herself) thatescape the physicalist story.
Clearly this conclusion entails that physicalism is false: for if thereare truths which escape the physicalist story how can everythingsupervene on the physical. So a physicalist must either reject apremise or show that the premises don't entail the conclusion.
There are many possible responses to this argument, but here I willbriefly mention only three. The first isthe abilityhypothesis due to Lawrence Nemerow (1988) and developed anddefended by David Lewis (1994). The ability hypothesis follows Ryle(1949) in drawing a sharp distinction between propositional knowledgeor knowledge-that (such as ‘Mary knows that snow is white’)and knowledge-how (such as ‘Mary knows how to ride abike’), and then suggests that all Mary gains is the latter. Onthe other hand, P2 would only be true if Mary gained propositionalknowledge.
A second response appeals to the distinction betweenapriori anda posteriori physicalism. As we saw above, thecrucial claim ofa posteriori physicalism is that (4) — i.e.the claim thatS entailsS* — isaposteriori. Since (4) isa posteriori, you would needcertain experience to know it. But, it is argued, Mary has not had (andcannot have) the relevant experience. Hence she does not know (4). Onthe other hand, the mere fact that Mary has not had (and cannot have)the experience to know (4) does not remove the possibility that (4) istrue. Hencea posteriori physicalism can avoid the knowledgeargument. (It is an interesting question which premise of the knowledgeargument is being attacked by this response. The answer depends onwhether (4) is physical or not: if (4) is physical, then the responseattacks P1. But if (4) is not physical, the response is that theargument is invalid.).
A third response is to distinguish between various conceptions ofthe physical. We saw above that potentially the class of propertiesdefined by the theory-conception of the physical was distinct from theclass of properties defined by the object-conception. But that suggeststhat the first premise of the argument is open to interpretation ineither of two ways. On the other hand, Jackson's thought experimentonly seems to support the premise if it is interpreted in the one way,since Mary learns by learning all that physicaltheory canteach her. But leaves open the possibility that one might appeal to theobject-conception of the physical to define a version of physicalismwhich evades the knowledge argument.
One of the most lively areas of philosophy of mind concerns the issueof which if any of these responses to the knowledge argument will besuccessful. (See the papers in Ludlow, Nagasawa, and Stoljar 2004.)The ability response raises questions about whether know-how isgenuinely non-propositional (cf. Lycan 1996, Loar 1997 and Stanley andWilliamson 2001), and about whether it gets the facts right to beginwith (Braddon Mitchell and Jackson 1996). As againstaposteriori physicalism, it has been argued both that it rests ona mistaken approach to the necessarya posteriori (Chalmers1996, 1999, Jackson 1998), and that the promise of the idea ischimerical anyway (cf. Stoljar 2000). The third response raisesquestions about the distinction between the object and the theoryconception of the physical and associated issues about dispositionaland categorical properties (cf. Chalmers 1996, Lockwood 1992, andStoljar 2000, 2001.)
Philosophers of mind often divide the problems of physicalism into two:first, there are the problems of qualia, typified by the knowledgeargument; second, there are problems of intentionality. Theintentionality of mental states is their aboutness, their capacity torepresent the world as being a certain way. One does not simply think,one thinksof (orabout) Vienna; similarly, one doesnot simply believe, one believesthat snow is white. Just asin the case of qualia, some of the puzzles of intentionality derivefrom facts internal to the notion, and from the relation of this notionto the others such as rationality, inference and language. But othersderive from the fact that it seems difficult to square the fact thatmental states have intentionality with physicalism. There are a numberof ways of developing this criticism but much recent work hasconcentrated on a certain line of argument that Saul Kripke has foundin the work of Wittgenstein (1982).
Kripke's argument is best approached by first considering what isoften called a dispositional theory of linguistic meaning. According tothe dispositional theory, a word means what it does — for example, theword ‘red’ means red — because speakers of the word aredisposed to apply to word to red things. Now, for a number of reasons,this sort of theory has been very popular among physicalists. First,the concept of a disposition at issue here is clearly a concept that iscompatible with physicalism. After all, the mere fact that vases arefragile and sugar cubes are soluble (both are classic examples ofdispositional properties) does not cause a problem for physicalism, sowhy should the idea that human beings have similar dispositionalproperties? Second, it seems possible to develop the dispositionaltheory of linguistic meaning so that it might apply also tointentionality. According to a dispositional theory of intentionality,a mental concept would mean what it does because thinkers are disposedto employ the concept in thought in a certain way. So a dispositionaltheory seems to hold out the best promise of a theory of intentionalitythat is compatible with physicalism.
Kripke's argument is designed to destroy that promise. (In fact,Kripke's argument is designed to destroy considerably more than this:the conclusion of his argument is a paradoxical one to the effect thatthere can be no such a thing as a word's having a meaning. However, wewill concentrate on the aspects of the argument that bear onphysicalism.) In essence his argument is this. Imagine a situation inwhich (a) the dispositional theory is true; (b) the word‘red’ means red for a speaker S; and yet (c) the speakermisapplies the word — for example, S is looking at a white thingthrough rose-tinted spectacles and calls it red. Now, in thatsituation, it would seem that S is disposed to apply ‘red’to things which are (not merely red but)either-red-or-white-but-seen-through-rose-tinted-spectacles. But then,by the theory, the word ‘red’ means (not red but)either-red-or-white-as-seen-through-rose-tinted-spectacles. But thatcontradicts our initial claim (b), that ‘red’ means red. Inother words, the dispositional theory, when combined with a true claimabout the meaning of word, plus a truism about meaning — that peoplecan misapply meaningful words — leads to a contradiction and istherefore false.
How might a physicalist respond to Kripke's argument? As with theknowledge argument, there are many responses but here I will mentiononly two. The first response is to insist that Kripke's argumentneglects the distinction betweena priori andaposteriori physicalism. Kripke often does say that according tothe dispositionalist, one should be able to ‘read off’truths about meaning from truths a physicalist can reject. (For aproposal like this, see Horwich 2000.) However, the problem with thisproposal is, as we have seen, that its background account of thenecessarya posteriori is very controversial. As we saw,aposteriori physicalists are committed to what we called thenon-derived view about necessarya posteriori truths. But thenon-derived view has come under strident attack in recent times.
The second response is to defend the dispositional theory againstKripke's argument. One way to do this is to argue that Kripke'sargument only works against a very simple dispositionalism, and that amore complicated version of such a theory would avoid these problems.(For a proposal along these lines, see Fodor 1992 and the discussion inBraddon-Mitchell and Jackson 1996). A different proposal is to arguethat Kripke's argument underestimates the complexity in the notion of adisposition. The mere fact that in certain circumstances someone wouldapply ‘red’ to white things does not mean that they aredisposed to apply red to white things — after all, the mere fact thatin certain circumstances something would burn does not mean that it isflammable in the ordinary sense. (For a proposal along these lines seeHohwy 1998, and Heil and Martin 1998)
As with the knowledge argument, the issues surrounding Kripke'sargument are very much wide open. But it is important to note that mostphilosophers don't consider the issues of intentionality as seriouslyas the issue of qualia when it comes to physicalism. In differentvocabularies, for example, both Block (1995) and Chalmers (1996)distinguish between the intentional aspects of the mind orconsciousness, and the phenomenal aspects or qualia, and suggest thatit is really the latter that is the central issue. As Chalmers notes(1996; p. 24), echoing Chomsky's famous distinction, the intentionalityissue is aproblem, but the qualia issue is amystery.
The final argument I will consider against physicalism is of a moremethodological nature. It is sometimes suggested, not that physicalismis false, but that the entire ‘project of physicalism’ —the project in philosophy of mind of debating whether physicalism istrue, and trying to establish or disprove its truth by philosophicalargument — is misguided. This sort of argument has been mounted by anumber of writers, but perhaps its most vocal advocate has been NoamChomsky (2000; see also Searle 1992, 1999).
It is easiest to state Chomsky's criticism by beginning with twopoints about methodological naturalism. In general it seems rational toagree with the methodological naturalists that the best hope for atheoretical understanding of the world is by pursuing the methods whichare typical of the sciences. It would then seem rational as a specialcase that our best hope for a theoretical understanding ofconsciousness or experience is by pursuing the methods of the sciences— by pursuing, as we might put it, the naturalistic project withrespect to consciousness. So Chomsky's first point is that it isrational to pursue the naturalistic project with respect toconsciousness.
Chomsky's second point is that the physicalist project in philosophyof mind is on the face of it rather different from the naturalisticproject. In the first place, the physicalist project is, as we havenoted, usually thought of a piece of metaphysics. But there is nothingmetaphysical about the naturalistic project, it simply raises questionsabout what we can hope to explain. In the second place, the physicalistproject is normally thought of as being amenable to philosophicalargument, whereas it is completely unclear where philosophical argumentwould enter the naturalistic project. In short, there doesn't seemanything particularly ‘philosophical’ about thenaturalistic project — it simply applies the methods of science toconsciousness. But the physicalist project is central to analyticphilosophy.
It is precisely at the place where the physicalist project departsfrom the naturalistic project that Chomsky's criticism begins to takeshape. For insofar as it is different from the naturalistic project,there are a number of ways in which the physicalist project isquestionable. First, it is hard to see what the project might be — itis true that throughout the history of philosophy and science oneencounters suggestions that one might find out about the world in waysthat are distinct from the ones used in the sciences, but thesesuggestions have always been rather obscure. Second, it is hard see howthis sort of project could recommend itself to physicaliststhemselves — such a project seems to be a departurefrom methodological naturalism but most physicalists endorsemethodological naturalism as a matter of fact. On the other hand, ifthe physicalist project does not depart from the naturalistic project,then the usual ways of talking and thinking about that project arehighly misleading. For example, it is misleading to speak of it as apiece of metaphysics as opposed to a piece of ordinary science.
In sum, Chomsky's criticism is best understood as a kind of dilemma.The physicalist project is either identical to the naturalistic projector it is not. If it is identical, then the language and concepts thatshape the project are potentially extremely misleading; but if it isnot identical, then there are a number of ways in which it isillegitimate.
How is one to respond to this criticism? In my view, the strongestanswer to Chomsky accepts the first horn of his dilemma and suggeststhat what philosophers of mind are really concerned with is thenaturalistic project. Now, of course, what concerns them is not thedetails of the project — that would not distinguish them from workingscientists. Rather they are concerned with what the potential limits ofthe project are.
This is a theme which has reached its best expression in the work ofThomas Nagel (1980, 1984, 1999) and allied work by Bernard Williams(1984). According to them, any form of scientific inquiry will at leastbe objective, or will result in an objective picture of the world. Onthe other hand, we have a number of arguments — the most prominentbeing the knowledge argument — which plausibly show that there is noplace for experience or qualia in a world that is described in purelyobjective terms. If Nagel and Williams are right that any form ofscientific inquiry will yield a description of the world in objectiveterms, the knowledge argument is nothing less than a negative argumentto the effect that the naturalistic project with respect toconsciousness will not succeed.
If what is at issue is the limits of the naturalist project, why isthe debate so often construed as a metaphysical debate rather than adebate about the limits of inquiry? In answer to this question, we needto sharply divorce the background metaphysical framework within whichthe problems of philosophy of mind find their expression, and theproblems themselves. Physicalism is the background metaphysicalassumption against which the problems of philosophy of mind are posedand discussed. Given that assumption, the question of the limits of thenaturalistic projectjust is the question of whether there canbe experience in a world that is totally physical. Nevertheless, whenproperly understood, the problems that philosophers of mind areinterested in are not with the framework themselves, and to that extentare not metaphysical. Thus, the common phrase ‘metaphysics ofmind’ is misleading.
Having considered one side of the truth question, I will now turn tothe other: what reason is there for believing that physicalism is true?
The first thing to say when considering the truth of physicalism isthat we live in an overwhelmingly physicalist or materialistintellectual culture. The result is that, as things currently stand,the standards of argumentation required to persuade someone of thetruth of physicalism are much lower than the standards required topersuade someone of its negation. (The point here is a perfectlygeneral one: if you already believe or want something to be true, youare likely to accept fairly low standards of argumentation for itstruth.)
However, while it might be difficult to assess dispassionately thearguments for or against physicalism, this is still something we shouldendeavor to do. Here I will review two arguments that are commonlythought to establish the truth of physicalism. What unites thearguments is that each takes something from the physicalistworld-picture which we considered previously and tries to establish themetaphysical claim that everything supervenes on the physical.
The first argument is (what I will call)The Argument fromCausal Closure. The first premise of this argument is the thesisof the Causal Closure of the Physical — that is, the thesis that everyevent which has a cause has a physical cause. The second premise isthat mental events cause physical events — for example we normallythink that events such as wanting to raise your arm (a mental event) causeevents such as the raising of your arm (a physical event). The thirdpremise of the argument is a principle of causation that is oftencalled the exclusion principle (Kim 1993, Yablo 1992). The correctformulation of the exclusion principle is a matter of some controversybut a formulation that is both simple and plausible is thefollowing:
Exclusion Principle
If an evente causes evente*, then there is noevente# such thate# is non-supervenient one ande# causese*.
The conclusion of the argument is the mental events are supervenient onphysical events, or more briefly that physicalism is true. For ofcourse, if the thesis of Causal Closure is true then behavioral eventshave physical causes, and if mental events also cause behavioralevents, then they must supervene on the physical if the exclusionprinciple is true.
The Argument from Causal Closure is perhaps the dominant argumentfor physicalism in the literature today. But it is somewhat unclearwhether it is successful. One response for theanti-physicalist is to reject the second premise and to adopt a versionof what is called epiphenomenalism, the view that mental events arecaused by, and yet do not cause, physical events. The argument againstthis position is usually epistemological: if pains don't cause painbehavior how can it be that your telling me that you are in pain givesme any reason for supposing you are? It might seem thatepiphenomenalists are in trouble here, but as a number of recentphilosophers have argued, the issues here are very far from beingsettled (Chalmers 1996, Hyslop 1999). The crucial point is that thecausal theory of evidence is open to serious counterexamples so it isunclear that it can be used against epiphenomenalism effectively.
A different sort of response is to reject the causal principles onwhich the argument is based. As against the exclusion principle, forexample, it is often pointed out that certain events areoverdetermined. The classic example is the firing squad: both thefiring by soldier A and by soldier B caused the prisoner's death butsince these are distinct firings, the exclusion principle is false.However, while this line of response is suggestive, it is in factrather limited. It is true that the case of the firing squad representsan exception to the exclusion principle — an exception that theprinciple must be emended to accommodate. But is difficult to believethat it represents an exception that can be widespread. A moresearching response is to reject the very idea of causal closure on thegrounds, perhaps, that (as Bertrand Russell (1917) famously argued)causation plays no role in a mature portrayal of the world. Once again,however, the promise of this response is more imagined than real. Whileit is true that many sciences do not explicitly use the notion ofcausation, it is extremely unlikely that they do not imply that variouscausal claims are true.
The second argument for physicalism is (what I will call)TheArgument from Methodological Naturalism. The first premise of thisargument is that it is rational to be guided in one's metaphysicalcommitments by the methods of natural science. Lying behind thispremise are the arguments of Quine and others that metaphysics shouldnot be approached in a way that is distinct from the sciences butshould rather be thought of as continuous with it. The second premiseof the argument is that, as a matter of fact, the metaphysical pictureof the world that one is led to by the methods of natural science isphysicalism. The conclusion is that it is rational to believephysicalism, or, more briefly that physicalism is true.
The Argument from Methodological Naturalism has received somewhatless attention in the literature than the Argument from Causal Closure.But it seems just as persuasive — in fact, rather more so. For howmight one respond? One possibility is to reject its first premise. Butthis is not something that most people are attracted to (or at leastare attracted to explicitly.)
The other possibility is to reject its second premise. However, onceit is appreciated what physicalism is — and, more important, what itis not — it is not terribly clear what this would amount to or whatthe motivation for it would be. In the first place, our earlierdiscussion shows that physicalism is not inconsistent with explanatoryautonomy of the various sciences, so that one should not rejectphysicalism merely because one can't see how to reduce those sciencesto others. In the second place, while it is perfectly true that thereare examples of non-physicalist approaches to the world — vitalism inbiology is perhaps the best example — this is beside the point. Thesecond premise of the Argument from Methodological Naturalism does notdeny that other views are possible, it simply says that physicalism isthe most likely view at the moment. Finally, one might be inclined toappeal to arguments such as the knowledge argument to show thatphysicalism is false, and hence that methodological naturalism couldnot show that physicalism is false. However, this suggestion representsa sort of confusion about the knowledge argument. As we saw above, ifsuccessful the knowledge argument suggests, not simply that physicalismis false but that any approach to the world that is compatible withmethodological naturalism is false. But if that is so, it is mistakento suppose that the knowledge argument gives one any reason to endorseanti-physicalism if that is supposed to be a position compatible withmethodological naturalism.
That completes our discussion of physicalism. It is sketchy incertain places, and this suggests that there is much further work tobe done before we arrive at a final assessment of the doctrine ofphysicalism and the role that it plays in contemporary thought. In the future, however, it would seem to me to be a good idea to think at least about the following:
(a) The prospects for supervenience definitions of physicalism. Whilesupervenience, and more generally modal, definitions of physicalismare attractive, they do raise problems of a difficultnature—such as the blockers problem and the necessary beingsproblem. At the time of writing it is unclear how these difficultieswill be resolved.
(b) The relation between, and the adequacy of, the theory-conceptionand the object-conception of the physical. It seems clear that ourthinking about the physical is anchored in part in the ordinary ideaof a physical object and in part in the idea of physics. But it seemsequally clear that these two ideas can and do come apart. What are theconsequences of this for the notion of physicalism?
(c) The role of physicalism in philosophy of mind. As we have seenfor both Nagel and Williams questions about physicalism are reallyquestions about objectivity in disguise. This raises the more generalissue of whether arguments and theses in philosophy of mind (and otherareas of philosophy) that are formulated in terms of physicalism arenecessarily formulated in that way. Perhaps physicalism here issimply getting in the way of a proper understanding of how theseproblems work?
behaviorism |color |Davidson, Donald |epiphenomenalism |multiple realizability |qualia |supervenience
The author would like to thank Rich Cameron, Robert Pasnau, StewartSaunders, and particularly David Chalmers for their help inconstructing this entry. In addition, the author and editors wouldlike to thank two readers, Joshua R. Stern and Greg Stokley, fordiscovering numerous typographical errors. Their volunteer effortswere entirely unsolicited and very much appreciated.
