Introduction: What Is Functionalism? Ned Block

T IS DOUBTFUL whether doctrines components. The article by Robert Cum­ I known as "functionalism" in fields as mins (chapter 12) describes functionalism disparate as anthropology, literary criti­ in this sense. (See also Fodor, 1965, 1968a, cism, psychology, and philosophy of psy­ 1968b; Dennett, 1975.) chology have anything in common but the name. Even in philosophy of psychology, Computation-representation func­ the term is used in a number of distinct tionalism. In this sense of the term, senses. The functionalisms of philosophy "functionalism" applies to an important of psychology are, however, a closely knit special case of functional explanation as group; indeed, they appear to have a defined above, namely, to psychological common origin in the works of Aristotle explanation seen as akin to providing a (see Hartman, 1977, especially chap. 4). computer program for the mind. What­ Three functionalisms have been enor­ ever mystery our mental life may initially mously influential in philosophy of mind seem to have is dissolved by functional and psychology: analysis of mental processes to the point where they are seen to be composed of Functional analysis. In this sense of computations as mechanical as the primi­ the term, functionalism is a type of expla­ tive operations of a digital computer­ nation and, derivatively, a research strat­ processes so stupid that appealing to them egy, the research strategy of looking for in psychological explanations involves no explanations of that type. A functional hint of question-begging. The key notions explanation is one that relies on a decom­ of functionalism in this sense are represen­ position of a system into its component tation and computation. Psychological parts; it explains the working of the sys­ states are seen as systematically represent­ tem in terms of the capacities of the parts ing the world via a language of thought, and the way the parts are integrated with and psychological processes are seen as one another. For example, we can explain computations involving these representa­ how a factory can produce refrigerators tions. Functionalism in this sense of the by appealing to the capacities of the vari­ term is not explored here but is discussed ous assembly lines, their workers and in volume 2, part one, "Mental Represen­ machines, and the organization of these tation." 172 Ned Block

Metaphysical functionalism. The that what binds certain types of things last functionalism, the one that this part is together is a physical property. mainly about, is a theory of the nature of Metaphysical functionalists charac­ the mind, rather than a theory of psycho~ terize mental states in terms of their causal logical explanation. Metaphysical func­ roles, particularly, in terms of their causal tionalists are concerned not with how relations to sensory stimulations, behav­ mental states account for behavior, but ioral outputs, and other mental states. rather with what they are. The function­ Thus, for example, a metaphysical func­ alist answer to "What are mental states?" tionalist theory of pain might characterize is simply that mental states are functional pain in part in terms of its tendency to be states. Thus theses of metaphysical func­ caused by tissue damage, by its tendency tionalism are sometimes described as func­ to cause the desire to be rid of it, and by tional state identity theses. The main con­ its tendency to produce action designed to cern of metaphysical functionalism is the separate the damaged part of the body same as that of behaviorism (see part one, from what is thought to cause the damage. "Behaviorism") and physicalism (see part What I have said about metaphysi­ two, "Reductionism and Physicalism"). cal functionalism so far is rather vague, All three doctrines address themselves to but, as will become clear, disagreements such questions as "What is pain?" -or at among metaphysical functionalists pre­ least to "What is there in common to all clude easy characterization of the doc­ pains in virtue of which they are pains?" trine. Before going on to describe meta­ It is important to note that meta­ physical functionalism in more detail, I physical functionalism is concerned (in shall briefly sketch some of the connec­ the first instance) with mental state types, tions among the functionalist doctrines not tokens-with pain, for instance, and just enumerated. One connection is that not with particular pains. (For further ex­ functionalism in all the senses described planation of this distinction see Boyd, in­ has something to do with the notion of a troduction to part two; Davidson, chap­ Turing machine (described in the next ter 5; Fodor, chapter 6.) Most functional­ section). Metaphysical functionalism of­ ists are willing to allow that each particu­ ten identifies mental states with Turing lar pain is a physical state or event, and machine "table states" (also described in indeed that for each type of pain-feeling the next section). Computation-represen.. organism, there is (perhaps) a single type tation functionalism sees psychological of physical state that realizes pain in that explanation as something like providing a type of organism. Where functionalists computer program for the mind. Its aim is differ with physicalists, however, is with to give a functional analysis of mental ca­ respect to the question of what is common pacities broken down into their compo­ to all pains in virtue of which they are nent mechanical proces~es. If these me­ pains. The functionalist says the some­ chanical processes are algorithmic, as is thing in common is functional, while the sometimes assumed (without much justi­ physicalist says it is physical (and the be-· fication, in my view) then they will be haviorist says it is behavioral).l Thus, in Turing-computable as well {as the Church­ one respect, the disagreement between Turing thesis assures US).2 Functional functionalists and physicalists (and be­ analysis, however, is concerned with the haviorists) is metaphysical without being notion of a Turing machine mainly in that ontological. Functionalists can be physi­ providing something like a computer pro­ calists in allowing that all the entities gram for the mind is a special case of func­ (things, states, events, and so on) that tional analysis. exist are physical entities, denying only Another similarity among the func- Part Three Introduction 173 tionalisms mentioned is their relation to fied by the machine table is described by physical characterizations. The causal the machine table and is a realization of structures with which metaphysical func­ the abstract automaton specified by the tionalism identifies mental states are real­ machine table. (This definition actually izable by a vast variety of physical sys­ characterizes a finite automaton, which is tems. Similarly, the information process­ just one kind of Turing machine.) ing mechanisms postulated by a particular One very simple version of machine computation-representation functionalist functionalism states that each system that theory could be realized hydraulically, has mental states is described by at least electrically, or even mechanically. Finally, one Turing machine table of a certain functional analysis would normally char­ specifiable sort; it also states that each acterize a manufacturing process abstract­ type of mental state of the system is iden­ ly enough to allow a wide variety of types tical to one of the machine table states of machines (wood or metal, steam-driven specified in the machine table (see Put­ or electrical), workers (human or robot or nam, chapter 17; Block and Fodor, chap­ animal), and physical s~tups (a given ter 20). Consider, for example, the Turing number of assembly lines or half as many machine described in the following "Coke dual-purpose assembly lines). A third machine" machine table (compare Nelson, similarity is that each type of functional­ 1975): ism described legitimates at least one no­ tion of functional equivalence. For exam­ 5, ple, for functional analysis, one sense of functional equivalence would be: has ca­ nickel Emit no output Emit a Coke pacities that contribute in similar ways to input Go to 5, Go to 5, the capacities of a whole. ----- In what follows, I shall try to give the dime Emit a Coke Emit a Coke and reader a clearer picture of metaphysical input Stay in o5J a nickel functionalism. ("Functionalism" will be Go to 51 used to mean metaphysical functionalism in what follows.) One can get a crude picture of the simple Machine Versions of Functionalism version of machine functionalism de­ Some versions of functionalism are scribed above by considering the claim couched in terms of the notion of a Turing that 5, = dime-desire, and 52 = nickel­ machine, while others are not. A Turing desire. Of course, no functionalist would machine is specified by two functions: one claim that a Coke machine desires any­ from inputs and states to outputs, and one thing. Rather, the simple version of ma­ from inputs and states to states. A Turing chine functionalism described above machine has a finite number of states, in­ makes an analogous claim with respect to puts, and outputs, and the two functions a much more complex machine table. specify a set of conditionals, one for each Machine versions of functionalism combination of state and input. The con­ are useful for many purposes, but they do ditionals are of this form: if the machine is not provide the most general character­ in state 5 and receives input I, it will then ization of functionalism. One can achieve emit output a and go into next state 5'. more generality by characterizing func­ This set of conditionals is often expressed tionalism as the view that what makes a in the form of a machine table (see below). pain a pain (and, generally, what makes Any system that has a set of inputs, out­ any mental state the mental state it is) is puts, and states related in the way speci- its having a certain causal role. 3 But this Block

at the price of turn , causes brow wrinkling. The Ramsey precise formulation sentence of T is 4 knc'iril,r(},aW-"~ as follows. Let T be a Ex1Ex,(Xl is caused by pin pricks and causes n"'VLltLn'~,,'--- theory (of either common x, and emission of loud noises & X2 causes sense or scientific psychology) that tells us brow wrinkling) . (among other things) the relations among pain, other mental states, sensory inputs, Relative to T, pain is the property ex­ and behavioral outputs. Reformulate T so pressed by the predicate obtained by add­ that it is a single conjunctive sentence with ing a conjunct as follows: all mental state terms as singular terms; Ex,Ex2[ (Xl is caused by pin pricks and for example, 'is angry' becomes 'has an­ causes x, and emission of loud noises ger'. Let T so reformulated be written as & x, causes brow wrinkling) & y has XlI. T(Sl .. . sn) That is, pain is the property that one has when one has a state that is caused by pin where Sl .•• Sn are terms that designate mental states. Replace each mental state pricks, and causes emission of loud noises, term with a variable and prefix existential and also causes something else, that, in quantifiers to form the Ramsey sentence turn, causes brow wrinkling. of the theory We can make this somewhat less cumbersome by letting an expression of Ex! ... XnT(Xl . .. xn). the form '%xFx' be a singular term mean­ [In this anthology, the ordinary "E" is ing the same as an expression of the form used instead of the backward "E" as the 'the property of being an x such that x is existential quantifier.J Now, if Xi is the F, that is, 'being F. So %x(x is bigger variable that replaced 'pain', we can de­ than a mouse & x is smaller than an ele­ fine 'pain' as follows: phant) = being bigger than a mouse and smaller than an elephant. Using this nota­ y has pain if and only if tion, we can say EXl ... xn[ T(XI ... xn) & y has xiJ. pain = % yEx,Ex,[ (x, is caused by pin That is, one has pain just in case he has a pricks and causes x, and emission state that has certain relations to other of loud noises & x, causes brow states that have certain relations to one wrinkling) & y has x,j , another (and to inputs and outputs; I have rather than saying that pain is the prop­ omitted reference to inputs and outputs erty expressed by the predicate for the sake of simplicity). It will be con­ venient to think of pain as the property ExIEx,[(x, is caused by pin pricks and expressed by the predicate 'x has pain', causes x, and emission of loud noises that is, to think of pain as the property & x, causes brow wrinkling) & y has x,] . ascribed to someone in saying that he has It may be useful to consider a non­ pain. 5 Then, relative to theory T, pain can mental example. It is sometimes supposed be identified with the property expressed that automotive terms like 'valve-lifter' or by the predicate 'carburetor' are functional terms. Any­ thing that lifts valves in an engine with a Ex! ... xn[ T(x! ... xn) & y has xJ certain organizational structure is a valve­ For example, take T to be the ridicu­ lifter. (,Camshaft', on the other hand, is a lously simple theory that pain is caused "structural" term, at least relative to by pin pricks and causes worry and the 'valve-lifter'; a camshaft is one kind of emission of loud noises, and worry, in device for lifting valves.) Part Three Introduction 175

Consider the "theory" that says: 'The cone if one is proffered, other things being carburetor mixes gasoline and air and equal. But, as functionalist critics have sends the mixture to the ignition chamber, emphasized, the phrase "other things be­ which, in turn ... " Let us consider 'gaso­ ing equal" is behavioristically illicit, be­ line' and 'air' to be input terms, and let Xl cause it can only be filled in with refer­ replace 'carburetor', and X2 replace 'igni­ ences to other mental states (see Putnam, tion chamber'. Then the property of being chapter 2; the point dates back at least to a carburetor would be Chisholm, 1957, chap. 11; and Geach, 1957, p. 8). One who desires an ice-cream % Y EXI ... xn[ (The x, mixes gasoline cone will be disposed to reach for it only and air and sends the mixture to the x" if he knows it is an ice-cream cone (and which, in turn ... ) & Y is an x,] . not, in general, if he believes it to be a That is, being,a carburetor = being what tube of axle-grease) , and only if he does mixes gasoline and air and sends the mix­ not think that taking an ice-cream cone ture to something else, which, in turn ... would conflict with other desires of more This identification, and the identifi­ importance to him (such as the desire to cation of pain with the property one has lose weight, avoid obligations, or avoid when one is in a state that is caused by pin cholesterol). The final nail in the behav­ pricks and causes loud noises and also iorist coffin was provided by the well­ causes something else that causes brow known "perfect actor" family of counter­ wrinkling, would look less silly if the the­ examples. As Putnam argued in convinc­ ories of pain (and carburetion) were more ing detail (see chapter 2), it is possible to complex. But the essential idea of func­ imagine a community of perfect actors tionalism, as well as its major weakness, who, by virtue of lawlike regularities, can be seen clearly in the example, albeit have exactly the behavioral dispositions rather starkly. Pain is identified with an envisioned by the behaviorists to be asso­ abstract causal property tied to the real ciated with absence of pain, even though world only via its relations, direct and in­ they do in fact have pain. This shows that direct, to inputs and outputs. The weak­ no behavioral disposition is a necessary ness is that it seems so clearly conceivable condition of pain, and an exactly analo .. that something could have that causal gous example of perfect pain-pretenders property, yet not be a pain. This point is shows that no behavioral disposition is a discussed in detail in "Troubles with Func­ sufficient condition of pain, either. tionalism" (Block, chapter 22; see Shoe­ Functionalism in all its forms differs maker, chapter 21, and Lycan, forthcom­ from behaviorism in two major respects. ing, for critiques of such arguments). First, while behaviorists defined mental states in terms of stimuli and responses, Functionalism and Behaviorism they did not think mental states were Many functionalists (such as David themselves causes of the responses and Lewis, D. M. Armstrong, and J. J. c. effects of the stimuli. Behaviorists took Smart) consider themselves descendants mental states to be "pure dispositions." of behaviorists, who attempted to define a Gilbert Ryle, for example, emphasized mental state in terms of what behaviors that "to possess a dispositional property is would tend to be emitted in the presence not to be in a particular state, or to under­ of specified stimuli. E.g., the desire for an go a particular change" (1949, p. 43). Brit­ ice-cream cone might be identified with a tleness, according to Ryle, is not a cause set of dispositions, including the disposi­ of breaking, but merely the fact of break­ tion to reach out and grasp an ice-cream ing easily. Similarly, to attribute pain to Ned Block s6ineone is not to attribute a cause or ef­ that would make functionalism a version fect of anything, but simply to say what of behaviorism. Even functionalists have he would do in certain circumstances. Be­ offered definitions of 'behaviorism' that haviorists are fictionalists about the men­ would make functionalists behaviorists. tal, hence they cannot allow that mental For example, if we defined 'behaviorism' states have causal powers. Functionalists, as the doctrine that mental states (such as by contrast, claim it to be an advantage of pain) can be characterized in nonmental their account that it "allows experiences terms, versions of functionalism along the to be something real, and so to be the ef­ lines of the Ramsey sentence version fects of their occasions, and the causes of sketched above (held by Lewis, Arm­ their manifestations (Lewis, 1966, p. 166). strong, Smart, and Sydney Shoemaker) Armstrong says that "r when I think] it is would qualify as versions of behaviorism not simply "that I would speak or act if (since all of the original mental state terms some conditions that are unfulfilled were are replaced by variables in the Ramsey to be fulfilled. Something is currently go­ sentence). Many other definitions of 'be­ ing on. Rylean behaviorism denies this, haviorism' count functionalism as a type and so it is unsatisfactory" (chapter 13). of behaviorism. But it would be ludicrous­ The second difference between func­ ly literal-minded to take such definitions tionalism and behaviorism is that func­ very seriously. Clear and general formu­ tionalists emphasize not just the connec­ lations of functionalism were not avail­ tions between pain and its stimuli and re­ able until recently, so standard definitions sponses, but also its connections to other of behaviorism could hardly be expected mental states. Notice, for example, that to draw the boundaries between behav­ any full characterization of 5, in the ma­ iorism and functionalism with perfect ac­ chine table above would have to refer to curacy. Furthermore, given an explicit 52 in one way or another, since it is one of definition of behaviorism, logical inge­ the defining characteristics of 5, that any­ nuity can often disguise a functionalist thing in 5, goes into 52 when it receives a account so as to fit the definition (see nickel input. Another example, recall that Bealer, 1978; Thomas, 1978, for accom­ the Ramsey sentence formulation identi­ plishments of this rather dubious variety). fies pain with Definitions of behaviorism that count functionalism as behaviorist are misguid­ % yEXl ... xn[ T(XI ... xn) & y has Xi} ed precisely because they blur the distinc­ where the variable Xi replaced 'pain', and tions between functionalism and behav­ the rest of x, . . . Xn replaced the other iorism just sketched. A characterization mental state terms in T. So the function­ of pain can hardly. be counted as behav­ alist expression that designates pain in­ iorist if it allows that a system could be­ cludes a specification of the relations be­ have (and be disposed to behave) exactly tween pain and all the other mental states as if it were in pain in all possible circum­ related to it, and to inputs and outputs as stances, yet not be in pain.6 well. (The role of inputs and outputs would have been better indicated had I Is Functionalism Reductionist? written T as Functionalists sometimes formulate their claim by saying that mental states can only be characterized in terms of other explicitly including terms for inputs and mental states. For instance, a person de­ outputs.) sires such and such if he would do so and Behaviorism is a vague doctrine, and so if he believed doing so and so will get one that is sometimes defined in a way him such and such, and if he believed do- Part Three Introduction 177 ing so and so would not conflict with oth­ itself to the existence of mental states by er desires. This much functionalism brings the use of quantification over mental in no reductionism, but functionalists states, or some equivalent device.? have rarely stopped there. Most regard mental terms as eliminable all at once. The Varieties of Functionalism Armstrong says, for example, "The logi­ Thus far, I have characterized func­ cal dependence of purpose on perception tionalism without adverting to any of the and belief, and of perception and belief confusing disagreements among function­ upon purpose is not circularity in defini­ alists. I believe that my characterization is tion. What it shows is that the corre­ correct, but its application to the writings sponding concepts must be introduced of some functionalists is not immediately together or not at all" (1977, p. 88). Shoe­ apparent. Indeed, the functionalist litera­ maker says, "On one construal of it, func­ ture (or, rather, what is generally, and I tionalism in the philosophy of mind is think correctly, regarded as the function­ the doctrine that mental or psychologi­ alist literature) exhibits some bizarre dis­ cal terms are in principle eliminable in a agreements, the most surprising of which certain way" (chapter 21). Lewis is more has to do with the relation between func­ explicit, using a formulation much like tionalism and physicalism. Some philoso­ the Ramsey sentence formulation given phers (Armstrong, 1968, 1977, chapter above, which designates mental states by 13; Lewis, 1966, chapters 15, 18; Smart, expressions that do not contain any men­ 1971) take functionalism as showing that tal terminology (see chapter 15 for de·· physicalism is probably true, while others tails). (Fodor, 1965; Putnam, 1966; Block and The same sort of point applies to Fodor, chapter 20) take functionalism as machine functionalism. Putnam says, showing that physicalism is probably "The 5i, to repeat, are specified only im­ false. This is the most noticeable differ­ plicitly by the description" (chapter 17). ence among functionalist writings. I shall In the Coke machine automaton described argue that the Lewis-Armstrong-Smart above, the only antecedently understood camp is mistaken in holding that function­ terms (other than 'emit', 'go to', and so alism supports an interesting version of on) are the input and output terms, 'nick­ physicalism, and furthermore, that the el', 'dime', and 'Coke'. The state terms '5/ functionalist insight that they share with and '5/ in the Coke machine automaton­ the Putnam-Fodor-Harman camp does as in every Turing machine-are given have the consequence that physicalism is their content entirely in terms of input probably false. I shall begin with a brief and output terms (+ logical terms). historical sketch. Thus functionalism could be said to While functionalism dates back to reduce mentality to input-output struc­ Aristotle, in its current form it has two tures (note that 51 and 52 can have any main contemporary sources. (A third natures at all, so long as these natures source, Sellars's and, later, Harman's connect them to one another and to the views on meaning as conceptual role, has acceptance of nickels and dimes and dis­ also been influential.) bursement of nickels and Cokes as de­ scribed in the machine table). But func­ Source I tionalism gives us reduction without elim­ Putnam (1960) compared the mental ination. Functionalism is not fictionalist states of a person with the machine table about mentality, for each of the function­ states of a Turing machine. He then re­ alist ways of characterizing mental states jected any identification of mental states in terms of inputs and outputs commits with machine table states, but in a series Ned Block 178 .··t·'j the years he moved closer 1977) have in my view failed to meet this o f. ar IC es over -.,., h 'dentification, a pattern cul- challenge, t nsuc an 1 -. t' l'n "psychological Predicates" If we could formulate a machine table mma mg (1967, reprinted as chapter 17, this vol- for a human, it would be absurd to iden­ ume), In this article, Putnam came close tify any of the machine table states with a to advocating a view--which he defended type of brain state, since presumably all in his philosophy of mind lectures in the manner of brainless machines could be late 1960s-that mental states can be iden­ described by that table as well. So if pain tified with machine table states, or rather is a machine table state, it is not a brain disjunctions of machine table states. (See state. It should be mentioned, however, Thomas, 1978, for a defence of roughly that it is possible to specify a sense in this view; see Block and Fodor, chapter which a functional state F can be said to 20, and Putnam, chapter 7, for a critique be physical. For example, F might be said of such views.) to be physical if every system that in fact Fodor (1965, 1968a) developed a sim­ has F is a physical object, or, alternative­ ilar view (though it was not couched in ly, if every realization of F (that is, every terms of Turing machines) in the context state that plays the causal role specified of a functional-analysis view of psycho­ by F) is a physical state, Of course, the logical explanation (see Cummins, chap­ doctrines of "physicalism" engendered by ter 12). Putnam's and Fodor's positions such stipulations should not be confused were characterized in part by their oppo­ with the version of physicalism that func­ sition to physicalism, the view that each tionalists have argued against (see note 8). type of mental state is a physical state.B Jaegwon Kim objects that "the less Their argument is at its clearest with re­ the physical basis of the nervous system gard to the simple version of Turing ma­ of some organisms resembles ours, the less chine functionalism described above, the temptation there will be for ascribing to view that pain, for instance, is a machine them sensations or other phenomenal table state. What physical state could be events" (chapter 19). But his examples de­ common to all and only realizations of Sl pend crucially on considering creatures of the Coke machine automaton described whose functional organization is much above? The Coke machine could be made more primitive than ours. He also points of an enormous variety of materials, and out that "the mere fact that the physical it could operate via an enormous variety bases of two nervous systems are different of mechanisms; it could even be a "scat­ in material composition or physical orga­ tered object," with parts all over the nization with respect to a certain scheme world, communicating by radio, If some­ of classification does not entail that they one suggests a putative physical state cannot be in the same physical state with common to all and only realizations of S1, respect to a different scheme." Yet the it is a simple matter to dream up a nomo­ functionalist does not (or, better, should logically possible machine that satisfies not) claim that functionalism entails the the machine table but does not have the falsity of physicalism, but only that the designated physical state. Of course, it is burden of proof is on the physicalist. Kim one thing to say this and another thing to (chapter 19) and Lewis (chapter 18; see prove it, but the claim has such over­ also Causey, 1977, p,149) propose species­ whelming prima facie plausibility that the specific identities: pain is one brain state burden of proof is on the critic to come up in dogs and another in people. As should with reason for thinking otherwise. Pub­ be clear from this introduction, however, lished critiques (Kalke, 1969; Gendron, this move sidesteps the main metaphysical 1971; Kim, 1972; Nelson, 1976; Causey, question: "What is common to the pains Part Three Introduction 179 of dogs and people (and all other pains) in identity theory. As far as an argument for virtue of which they are pains?" the identity theory went, he relied on con­ siderations of simplicity. It was absurd, Source II he thought, to suppose that there should The second major strand in cur­ be a perfect correlation between mental rent functionalism descends from Smart's states and brain states and yet that the early article on mind-body identity (1959). states could be nonidentical. (See Kim, Smart worried about the following objec­ 1966; Brandt and Kim, 1967, for an argu­ tion to mind-body identity: So what if ment against Smart; but see also Block, pain is a physical state? It can still have a 1971, 1979; and Causey, 1972, 1977, for variety of phenomenal properties, such as arguments against Kim and BrandL) But sharpness, and these phenomenal proper­ Lewis and Smart's Australian allies (nota­ ties may be irreducibly mental. Then bly D. M. Armstrong) went beyond Smart and other identity theorists would Smart, arguing that something like topic­ be stuck with a "double aspect" theory: neutral analyses could be used to argue pain is a physical state, but it has both for mind-brain identity. In its most per­ physical and irreducibly mental proper­ suasive version (Lewis's), the argument ties. He attempted to dispel this worry by for physicalism is that pain can be seen analyzing mental concepts in a way that (by conceptual analysis) to be the occu­ did not carry with it any commitment to pant of causal role R; a certain neural the mental or physical status of the con­ state will be found to be the occupant of cepts.9 These "topic-neutral analyses," as causal role R; thus it follows that pain = he called them, specified mental states in that neural state. Functionalism comes in terms of the stimuli that caused them (and by way of showing that the meaning of the behavior that they caused, although 'pain' is the same as a certain definite de­ Smart was less explicit about this). His scription that spells out causal role R. analysis of first-person sensation avowals Lewis and Armstrong argue from were of the form "There is something go­ functionalism to the truth of physicalism ing on in me which is like what goes on because they have a "functional specifica­ when ... ," where the dots are filled in by tion" version of functionalism. Pain is a descriptions of typical stimulus situations. functionally specified state, perhaps a In these analyses, Smart broke decisively functionally specified brain state, accord­ with behaviorism in insisting that mental ing to them. Putnam and Fodor argue states were real things with causal effi­ from functionalism to the falsity of physi­ cacy; Armstrong, Lewis, and others later calism because they say there are func­ improved his analyses, making explicit tional states (or functional properties), the behavioral effects clauses, and includ­ and that mental states (or properties) are ing mental causes and effects. Lewis's for­ identical to these functional states. No mulation, especially, is now very widely functional state is likely to be a physical accepted among Smart's and Armstrong's state. adherents (Smart, 1971, also accepts it). The difference between a functional In a recent review in the Australasian state identity claim and a functional speci~ Journal of Philosophy, Alan Reeves de­ fication claim can be made clearer as fol­ clares, "I think that there is some consen­ lows. Recall that the functional state iden­ sus among Australian materialists that tity claim can be put thus: Lewis has provided an exact statement of pain = % yE.x, ... Exn[ T(XI ... xr() their viewpoint" (1978). & yhasx,]; Smart used his topic-neutral analyses only to defeat an a priori objection to the where x, is the variable that replaced Ned Block 180 'pain'. A functional specification view the negation of physicalism owes simply could be stated as follows: 1o to ambiguities in 'functionalism' and 'physicalism'. In particular, it may seem pain = the x ,E.x, ... EXn T(x, ... xn) . that the functional specification view jus-· In terms of the example mentioned ear­ tifies token physicalism (the doctrine that lier, the functional state identity theorist every particular pain is a physical state would identify pain with the property one token), while the functional state identity has when one is in a state that is caused by view justifies the negation of type physi­ pin pricks and causes loud noises and also calism (the doctrine that pain is a type of something else that causes brow wrin­ physical state). kling. The functional specifier would de­ This response oversimplifies matters fine pain a~ the thing that is caused by pin greatly, however. First, it is textually mis­ pricks and causes loud noises and also taken, since those functional specifiers something else that causes brow wrin­ who see the distinction between type and kling. token materialism clearly have type ma­ According to the functional specifier, terialism in mind. For example, Lewis the thing that has causal role R (for exam­ says, "A dozen years or so ago, D. M. ple, the thing that is caused by pin pricks Armstrong and I (independently) pro­ and causes something else and so forth) posed a materialist theory of mind that might be a state of one physical type in joins claims of type-type psychophysical one case and a state of another physical identity with a behaviorist or functionalist type in another case. The functional state way of characterizing mental states such identity theorist is free to accept this claim as pain" (chapter 16; emphasis added). as well, but what he insists on is that pain More important, the functional specifica­ is not identical to a physical state. What tion doctrine commits its proponents to a pains have in common in virtue of which functional state identity claim. Since the they are pains is causal role R, not any latter doctrine counts against type physi­ physical property. calism, so does the former. It is easy to see In terms of the carburetor example, that the functional specification view functional state identity theorists say that commits its proponents to a functional being a carburetor = being what mixes state identity claim. According to func­ gas and air and sends the mixture to some­ tional specifiers, it is a conceptual truth thing else, which, in turn ... Functional that pain is the state with causal role R. specifiers say that the carburetor is the But then what it is to be a pain is to have thing that mixes gas and air and sends the causal role R. Thus the functional speci­ mixture to something else, which, in turn fiers are committed to the view that what ... What the difference comes to is that pains have in common by virtue of which the functional specifier says that the car­ they are pains is their causal role, rather buretor is a type of physical object, though than their physical nature. (Again, Lewis perhaps one type of physical object in a is fairly clear about this: "Our view is that Mercedes and another type of physical the concept of pain ... is the concept of a object in a Ford. The functional state iden­ state that occupies a certain causal role./I) tity theorist can agree with this, but he I suspect that what has gone wrong in insists that what it is to be a carburetor is the case of many functional specifiers is to have a certain functional role, not a simply failure to appreciate the distinction certain physical structure. between type and token for mental states. At this poin t, it may seem to the read.. If pain in Martians is one physical state, er that the odd disagreement about wheth­ pain in humans another, and so on for er functionalism justifies physicalism or pain in every pain-feeling organism, then Part Three Introduction 181 each particular pain is a token of some they are pains, just as saying that the win­ physical type. This is token physicalism. ning number is 37 does not suggest that 37 Perhaps functional specifiers ought to be is what all winning numbers have in com­ construed as arguing for token physical­ mon. Many of Lewis's opponents disagree ism (even though Lewis and others explic­ about the rigidity of 'pain', but the dispute itly say they are arguing for type physical­ is irrelevant to our purposes, since Lewis ism). I shall give three arguments against does take 'having pain' to be rigid, and so such a construal. First, as functional state he does accept (he tells me) a functional identity theorists have often pointed out, property identity view: having pain = a nonphysical state could conceivably having a state with such and such a typi­ have a causal role typical of a mental cal causal role. I think that most function­ state. In functional specification terms, al state identity theorists would be as will­ there might be a creature in which pain is ing to rest on the thesis that having pain is a functionally specified soul state. So a functional property as on the thesis that functionalism opens up the possibility pain is a functional state. that even if our pains are physical, other In conclusion, while there is consid­ pains might not be. In the light of this erable disagreement among the philoso­ point, it seems that the support that func­ phers whom I have classified as meta­ tionalism gives even to token physicalism physical functionalists, there is a single is equivocal. Second, the major argu­ insight about the nature of the mind to ments for token physicalism involve no which they are all committed. functionalism at all (see Davidson, chap­ ter 5, and Fodor, chapter 6). Third, token Notes physicalism is a much weaker doctrine 1. Discussions of functional state iden­ than physicalists have typically wanted. tity theses have sometimes concentrated on In sum, functional specifiers say that one or another weaker thesis in order to avoid functionalism supports physicalism, but issues about identity conditions on entities they are committed to a functionalist an­ such as states or properties (see, for example, swer, not a physicalist answer, to the Block and Fodor, chapter 20). Consider the question of what all pains have in com­ following theses: mon in virtue of which they are pains. (1) Pain = functional state S. And if what all pains have in common in (2) Something is a pain just in case it is a virtue of which they are pains is a func­ (token of) S. tional property, it is very unlikely that (3) The conditions under which x and y pain is coextensive with any physical are both pains are the same as the con­ state. If, on the contrary, functional speci­ ditions under which x and yare both fiers have token physicalism in mind, tokens of S. functionalism provides at best equivocal (1) is a full-blooded functional state identity support for the doctrine; better support is thesis that entails (2) and (3). Theses of the available elsewhere; and the doctrine is a form of (2) and (3) can be used to state what it rather weak form of physicalism to boot. is that all pains have in common in virtue of Lewis's views deserve separate treat­ which they are pains. 2. Dennett (1975) and Rey (1979) make ment. He insists that pain is a brain state this appeal to the Church-Turing thesis. But if only because he takes 'pain' to be a non­ the mechanical processes involved analog rigid designator meaning 'the state with rather than digital computation, then the pro­ such and such causal role'Y Thus, in cesses could fail to be algorithmic in the sense Lewis's view, to say that pain is a brain required by the Church-Turing thesis. The ex­ state should not be seen as saying what all periments discussed in volume 2, part two, pains have in common in virtue of which "Imagery" suggest that mental images are (at Ned Block

"p'itrtiaIly} analog representations, and mental properties are second-order physical the computations that operate on images properties (in various senses of that phrase). are (at Jeast partially) analog operations. 9. As Kim has pointed out (1972), Smart 3. Strictly speaking, even the causal role did not need these analyses to avoid "double formulation is insufficiently general, as can be aspect" theories. Rather, a device Smart intro­ seen by noting that Turing machine function­ duces elsewhere in the same paper will serve alism is not a special case of causal role func­ the purpose. Smart raises the objection that if tionalism. Strictly speaking, none of the states afterimages are brain states, then since an after­ of a Turing machine need cause any of the image can be orange, the identity theorist other states. All that is required for a physical would have to conclude that a brain state can system to satisfy a machine table is that the be orange. He replies by saying that the iden­ counterfactuals specified by the table are true tity theorist need only identify the experience of it. This can be accomplished by some causal of having an orange afterimage with a brain agent outside the machine. Of course, one can state; this state is not orange, and so no orange always choose to speak of a different system, brain states need exist. Images, says Smart, are one that includes the causal agent as part of the not really mental entities; it is experiences of machine, but that is irrelevant to my point. images that are the real mental entities. In a 4. Formulations of roughly this sort similar manner, Kim notes, the identity theo­ were first advanced by Lewis, 1966, 1970, rist can "bring" the phenomenal properties into 1972; Martin, 1966. (See also Harman, 1973; the mental states themselves; for example, the Grice, 1975; Field, 1978; Block, chapter 22.) identity theorist can concern himself with 5. See Field, 1978, for an alternative states such as John's having a sharp pain; this convention. state is not sharp, and so the identity theorist is 6. Characterizations of mental states not committed to sharp brain states. This tech·· along the lines of the Ramsey sentence formu­ nique does the trick, although of course it lation presented above wear their incompati­ commits its perpetrators to the unfortunate bility with behaviorism on their sleeves in that doctrine that pains do not exist, or at least that they involve explicit quantification over men­ they are not mental entities; rather, it is the tal states. Both Thomas and Bealer provide havings of sharp pains and the like that are the ways of transforming functionalist definitions real mental entities. or identifications so as to disguise such trans­ 10. The functional specification view I parent incompatibility. give here is a much simplified version of Lew­ 7. The machine table states of a finite is's formulation (see chapter 15). automaton can be defined explicitly in terms of 11. A rigid designator is a singular term inputs and outputs by a Ramsey sentence that names the same thing in each possible method, or by the method described in Thom­ world. 'The color of the sky' is nonrigid, since as (1978). Both of these methods involve one it names blue in worlds where the sky is blue, or another sort of commitment to the existence and red in worlds where the sky is red. 'Blue' is of the machine table states. rigid, since it names blue in all possible worlds, 8. 'Physical state' could be spelled out even in worlds where the sky is red. for these purposes as the state of something's having a first-order property that is expressible References by a predicate of a true physical theory. Of Armstrong, D. M. 1968. A Materialist Theory course, this analysis requires some means of of Mind. London: Routledge & Kegan Paul. characterizing physical theory. A first-order --1970. The Nature of Mind. In C. V. property is one whose definition does not re­ Borst, ed., The Mind/Brain Identity The­ quire quantification over properties. A second­ ory. London: Macmillan. Reprinted as order property is one whose definition requires chapter 13, this volume. quantification over first-order properties (but -- 1977. "The Causal Theory of the not other properties). The physicalist doctrine Mind." In Neue Heft fur Philosophie, no. that functionalists argue against is the doctrine 11, pp. 82-95. Vendenhoek and Ruprecht. that mental properties are first-order physical Bealer, G. 1978. "An Inconsistency in Func­ properties. Functionalists need not deny that tionalism." Synthese 38:333-372. Part Three Introduction 183

Block, N. 1971. "Physicalism and Theoretical Harman, G. 1973. Thought. Princeton: Prince­ Identity." Ph.D. dissertation, Harvard Uni­ ton University Press. versity. Hartman, E. 1977. Substance, Body and Soul, ---- 1978. "Troubles with Functionalism." Princeton: Princeton University Press. In C. W. Savage, ed., Minnesota Studies in Kalke, W. 1969. "What Is Wrong with Fodor Philosophy of Science. Vol. 9. Minneapolis: and Putnam's Functionalism?" Nous 3: 83- University of Minnesota Press. Reprinted as 93. chapter 22, this volume. Kim, J. 1966. "On the Psycho-physical Identity --- 1979. "Reductionism." In Encyclopedia Theory." American Philosophical Quarterly of Bioethics, New York: Macmillan. 3, no. 3:227-235. Block, N., and J. A. Fodor. 1972. "What Psy­ --- 1972. "Phenomenal Properties, Psy­ chological States Are Not." Philosophical chophysical Law, and the Identity Theory." Review 81, no. 2:159-182. Reprinted as Monist 56, no. 2:177-192. chapter 20, this volume. Lewis, D. 1966. "An Argument for the Identity Brandt, R., and J. Kim. 1967. 'The Logic of Theory." Reprinted in D. Rosenthal. ed., the Identity Theory." Journal of Philosophy Materialism and the Mind-Body Problem, 64, no. 17:515-537. Englewood Cliffs, N.J.: Prentice-Hall, 1971. Causey, R. 1972. "Attribute Identities in Mi­ --- 1969. "Review of Art, Mind and Reli­ cro-reductions." Journal of Philosophy 69, gion." Journal of Philosophy 66, no. 1:23- no. 14:407-422. 35. Reprinted in part as chapter 18, this vol­ ---1977. Unity of Scienre. Dordrecht: Rei­ ume. del. --- 1970. "How to Define Theoretical Chisholm, R. M. 1957. Perceiving. Ithaca: Terms." Journal of Philosophy 67, no. 13: Cornell University Press. 427-444. Cummins, R. 1975. "Functional Analysis." ---1972. "Psychophysical and Theoretical Journal of Philosophy 72, no. 20:741-764. Identification." Australasian Journal of Phi­ Reprinted in part as chapter 12, this volume. losophy 50, no. 3:249-258. Reprinted as Dennett, D. 1975. "Why the Law of Effect chapter 15, this volume. Won't Go Away." Journal for the Theory of Lycan, W. In press. "A New Lilliputian Argu­ Social Behavior 5:169-187. ment against Machine Functionalism." Phil­ Field, H. 1978. "Mental Representation." Er­ osophical Studies. kenntniss 13:9-61. Martin, R. M. 1966. "On Theoretical Con­ Fodor, J. A. 1965. "Explanations in Psychol­ stants and Ramsey Constants." Philosophy ogy." In M. Black, ed., Philosophy in Amer­ of Science 31:1-13. ica. London: Routledge & Kegan Paul. Nagel, T. 1970. "Armstrong on the Mind." --- 1968a. "The Appeal to Tacit Knowl-­ Philosophical Review 79:394-403. Reprinted edge in Psychological Explanation." Journal as chapter 14, this volume. of Philosophy 65:627-640. Nelson, R. J. 1975. "Behaviorism, Finite Au­ ---1968b. Psychological Explanation. tomata and Stimulus Response Theory." New York: Random House. Theory and Decision 6:249-267. Geach, P. 1957. Mental Acts. London: Rout­ ---1976. "Mechanism, Functionalism and ledge & Kegan Paul. the Identity Theory." Journal of Philosophy Gendron, B. 1971. "On the Relation of Neuro-­ 73, no. 13:365-386. logical and Psychological Theories: A Cri­ Putnam, H. 1960. "Minds and Machines." In tique of the Hardware Thesis." In R. C. S. Hook, ed., Dimensions of Mind. New Buck and R. S. Cohen, eds., Boston Studies York: New York University Press. in the Philosophy of Science. Vol. 8. Dor­ --- 1963. "Brains and -Behavior." Reprint­ drecht: Reidel. ed in Mind, Language, and Reality: Philo­ Grice, H. P. 1975. "Method in Philosophical sophical Papers. Vol. 2. London: Cambridge Psychology (from the Banal to the Bizarre)." University Press, 1975 . Proceedings and Addresses of the American ._- 1966. "The Mental Life of Some Ma­ Philosophical Association. Newark, Del.: chines." Reprinted in Mind, Language and American Philosophical Association. Reality; Philosophical Papers. Vol. 2. Lon- Ned Block

don: Cambridge University Press, 1975. Rey, G. 1979. "Functionalism and the Emo­ .~. 1967. "The Nature of Mental States" tions." In A. Rorty, ed., Explaining Emo­ (originally published as "Psychological Pred­ tions. Berkeley and Los Angeles: University icates"). In W. H. Capitan and D. D.Mer­ of California Press. rill, eds., Art, Mind, and Religion. Pitts­ Ryle, G. 1949. The Concept of Mind. London: burgh: University of Pittsburgh Press. Re­ Hutchinson. printed as chapter 17, this volume. Sellars, W. 1968. Science and Metaphysics. --- 1970. "On Properties." In Mathemat­ London: Routledge & Kegan Paul. chap. 6. ics, Matter and Method: Philosophical Pa­ Shoemaker, S. 1975. "Functionalism and Qua­ pers. Vol. 1. London: Cambridge University lia." Philosophical Studies 27:271-315. Re­ Press. printed as chapter 21, this volume. --- 1975. "Philosophy and Our Mental Smart, J. J. C. 1959. "Sensations and Brain Life." In Mind, Language and Reality: Philo­ Processes." Philosophical Review 68:141- sophical Papers. Vol. 2. London: Cambridge 156. University Press. Reprinted as chapter 7, --- 1971. "Reports of Immediate Experi­ this volume. ence." Synthese 22:346-359. Reeves, A. 1978. "Review of W. Matson, Sen­ Thomas, S. 1978. The Formal Mechanics of tience." Australasian Journal of Philosophy Mind. Ithaca: Cornell University Press. 56, no. 2 (August):189-192.