JENNIFER HORNSBY

ON F U N C T I O N A L I S M , AND ON J A C K S O N , P A R G E T T E R AND P R I O R ON F U N C T I O N A L I S M

(Received 15 February, 1983)

0.1. I want to take issue with 'Functionalism and Type-Type Identity TheOries', in which Jackson, Pargetter and Prior argued that proponents of functionalism ought to accept mental-physical type-type identities) I shall not be concerned with the portions of their paper where they attempt to show that the functionalist is better placed to reply to objections if he adopts a sort of type-type theory. I am happy enough if those portions are correct. For I claim that the functionalist ought not to adopt such a theory, and (like at least one of Jackson, Pargetter and Prior) I am not a functionalist. In Section 2 I introduce an analogy as a means of displaying what I take to be the errors of Jackson, Pargetter and Prior's line. The analogy is with cricket. Replying to a group of Australians, with the Ashes recently lost by nay countrymen, I try to turn that game to my advantage. But I hope that the analogy is of independent usefulness - enabling us to think in a fairly concrete way about matters that are usually presented at a very high level of abstraction. In Section 4 I use the analogy to demonstrate that functionalists are committed to more claims than they usually represent themselves as believing. These claims are strange, stranger perhaps than has been generally recognized; but they are not type-type identity claims. 0.2. Since it is part of my case that functionalists, if they ought to assert any mental/physical identities, ought at most to assert identities of things that are particulars, I need a word for such particulars. The word 'states' serves at least as well as any other; I shall use it only of non-repeatable things, of things which do not have instances (of things which have come to be called "tokens"). Jackson, Pargetter and Prior themselves, used 'state' of such things as states (in my sense) are instances of, which things I shall always call

Philosophical Studies 46 (1984) 75-95. 0031-8116/84/0461-0075502.10 9 1984 by D. Reidel Publishing Company

76

JENNIFER

HORNSBY

types of state. I shall adjust quotations from them so that they accord with this terminological policy of mine. The distinction between states and types they are of is a distinction between particulars and universals. We need also to recognize a distinction within the realm of universals, a distinction between universals that we see states (of people) as instantiating and universals that we see people (themselves) as instantiating. I shall reserve the word 'mental property' for the things we attribute to people. Thus mental properties are (some of the) properties possessed by persons. I f a person has some mental property, there is a state of that person - a state which is of some certain mental type. If a person, say, believes that p at t, then she has the property of believing that p at t; in virtue of that, some state of hers at t is a belief that p, which is to say that a state of the type belief that p exists in this person at t. 2 I shall follow Jackson, Pargetter and Prior in speaking sometimes of types of state as more or less specific or generic. The type which a state is of in virtue of its being a's belief that p at t is a more specific type than the type which a state is of in virtue of being a belief that p, for example. If one considered organisms of other kinds than persons, one might see varying degrees of specificity of mental properties by distinguishing, for example, pain for Jumbo, pain for elephants, pain for organisms. But here I shall confine attention to persons as the subjects of mental properties, and I shall assume that the functionalist works with a theory which he takes to be true of them.

1

1.1. Jackson, Pargetter and Prior hold that the functionalist can and should use the following two premises to obtain mental-physical identities, by the transitivity of identity (cp. p. 210). (i) (ii)

Pain for O at t = the type which occupies the pain-functional role at t for O The type which occupies the pain-functional role at t for O = type B of brain state

We may call the argument from (i) and (ii) the alleged functionalist argument. Jackson, Pargetter and Prior suppose that functionalists think that premises like (i) flow from claims in the philosphy of mind, and that premises like (ii) could, in principle, be discovered empirically - by investigation of O at t.

ON FUNCTIONALISM

77

Inasmuch as the denotata of the phrases employed in (i) and (ii) are types of state, the identities that would be established by the argument are identities between types of state. So if the functionalists accept the premises, they ought to believe in type-type identities. But ! believe that the functionalists cannot use this argument. 1.2. My initial doubt concerns whether, as Jackson, Pargetter and Prior claim, (i) records the functionalists's thesis about meaning. The functionalist's idea is that we can say w h a t p a i n is - what the concept of pain is the concept of - using our knowledge of what 'pain' and other mental terms mean. Their thesis about the meaning of 'pain' is to the effect that (roughly) pain is that type of state instances of which would be brought about by such and such sorts of input in the presence of states of such and such mental types, and instances of which would interact with states of such and such mental types thus and so ..., etc.. They believe that such an account of pain is a component of a complete psychological theory, a theory dealing with all the types of mental state there are. A complete psychological theory not only deals with all types of mental state, it applies to all persons at all times. Since it quantifies universally over persons and times, it tells us such things as what it is for a person (any person) to be in pain at a time (any time). But then if the theory can tell us what it is for me to be in pain now (say), it can do so only through its perfectly general account of what p a i n and all the other types of mental states are. It seems, then, that although the theory will indeed tell us (by universal instantiation) about the role of pain for me now, it will not distinguish between that role and the role of pain for you now, or between that role and the role of pain for me at some time different from now. If (i) did record the functionalist's thesis about meaning, then we should have to understand 'pain for O at t' in (i) as standing for the generic state of pain. (Notice that we can understand it so: a philosopher can make a point by saying that pain for me now is the same as pain for you tomorrow - that pain is pain.) Jackson, Pargetter and Prior intended mental/physical identities to concem more specific types than pain. And in (ii) at least, 'pain for O at t' must stand for something specific, because the identity recorded in (ii) is supposed to be established by investigating simply the person O at the time t. But if a thesis about meaning could tell one about nothing more specific than pain, then how is premise 0) supposed to be established?

78

JENNIFER HORNSBY

What I shall attempt to show (in Section 3) is that nothing that could be couched in terms of roles could serve to pinpoint the specific types of state in question in the alleged functionalist argument.

2.1. Consider now the team on the field in a game of cricket. It is to be the analogue of a subject of mental properties. The individual members are to be the analogues of the mental states of a subject. Each member will be in some position - bowler, first slip, mid-off etc. - , and these positions are to be the analogues of types of mental state. The team, then, has cricketing properties, just as a person has mental properties; and if a team satisfies some cricketing property (say, having a second slip), then there is a member of it of a certain type (second slip), just as if a person has a mental property (say, believing that p), then there is a state of hers of a certain type (belief that p). Individual members of a team will be of types in addition to any that we could know about from knowing only the cricketing properties of the team, just as, in the functionalist's view, individual mental states are of types, viz. physical types, in addition to any we could know about from knowing only the mental properties of a person. Here I shall consider cricket players as being not only of cricketing types but also of such types as being six foot tall (though I acknowledge that it is far from obvious which of the many types that cricket players instantiate should best be taken as the analogues of physical types of states in the mental case). It goes without saying that there are differences between cricket and the mind. (One of these will be important for the argument in due course [Section 4].) But what makes the analogy useful is that it seems that roles in cricket, like roles of mental things, need, in a certain sense, to be simultaneously defined. 2.2. It seems that we can state the roles of individual players in cricket games by reference to the positions that they occupy. To do this, we must begin with a completely general characterization of how teams of players occupying the various positions would behave in various possible cricketing circumstances. We may suppose that anything we mention in our general account of cricket follows from our theoretical knowledge of the nature of the game,

ON F U N C T I O N A L I S M

79

rather as the functionalist thinks that our general theory of psychology is got from our competence with psychological language. So we need a theory which applies to any fielding side at any time,just as the functionalist begins from a theory of the mental which applies to any person at any time. It will be a theory which speaks of such interdependencies between fielders' endeavors and movements of the ball as can be stated by reference to the positions of fielders on the pitch. 3 We should appreciate the fact that at any time in any actual game not all the fielding positions would actually have players occupying them, just as for any person at any time, not all types of mental state will be instantiated by actual states of the person (I am not in pain now, for example). And our general characterization of fielding teams should make no particular assumptions about the past history of any particular game or the whims of any particular captain, just as the functionalist's theory of mental things makes no assumptions about the vicissitudes of the mental life of any individual. It says, e.g. how fielders occupying certain positions wouM behave if they were members of teams in which certain other positions were occupied. 2.3. With the theory imagined to have been constructed, we can focus on a particular player in a particular cricket game at a particular time. Call him Botham, and imagine that he is playing second slip for England at noon on Wednesday. Botham has a role in the game in virtue of being of the type second slip, which role we ought to be able to read off from our general theory. Part of the specification of this will be such things as 'to endeavour to catch a bali which is disposed to land at place P if it be the case that there is no-one on the field of type N ...'. Suppose though that we speak to Botham himself, and ask him what his role is. In giving his answer, he is unlikely to go in for any such abstract considerations as the general theory involves. He may say such things as 'My role is to endeavour to catch the ball if it is disposed to land in such and such region'. If we were to say to him 'But isn't your role such that if there were no third slip, then you would endeavour to catch the ball if it were disposed to land in some larger region', then he might reply 'Well, it's true that I'm at second slip, and true that that might be something that the player at second slip had to do. But I thought you only wanted to know about the role I actually have.' Indeed Botham might say 'I can't tell you what my role would

80

JENNIFER HORNSBY

be if there were not a third slip; if there were not a third slip, then perhaps I'd be bowling, perhaps I wouldn't be a member of the team ...; certainly I wouldn't occupy what I think of as my present role'. What this suggests is that we need to distinguish between two roles that Botham occupies. First, there is what we might call Botham's counterfactual role, which it seems no cricket player would naturally work with: its specification requires the use of counterfactuals whose antecedents Botham would know to obtain or not to obtain ('If there were not a person of such and such other type, then ...'). Second, there is the role that Botham has in mind when he tells us what his role is, which we might call Botham's actual role. Its specification will require some use of counterfactuals (what Botham wouM do if the ball did such and such); but its specification will pay heed to what is actually true of Botham's team (of what other types the persons in his team are of). 2.4. We can if we want call each of the two roles that Botham occupies 'the role of second slip for the English team at t'; but then we shall have to distinguish between two ways of understanding this description. The distinction evidently corresponds to two ways of reading 'second slip for the English team at t'. According to the first reading, this phrase denotes second slip simply; and indeed the counterfactual role which Botham occupied was a role that anyone at any time at second slip would have occupied. According to the second reading, the phrase denotes a much more specific type than second slip - a type that someone instantiates at some time only if he is playing for a team which is composed in exactly the way that the English team was composed when we consulted Botham - ; and indeed the actual role that Botham thought of himself as occupying was a role that one could not know about unless one knew (as Botham did) the particular arrangement of his team at the relevant time. We emphasize the specificity of the latter role by calling it Botham's actual role. But if we are to speak of Botham's actual role, we must bear it in mind that the only thing about Botham in virtue of which he occupies this role is that he is second slip in a team which has the very same cricketing properties as the English team actually had at noon on Wednesday. This, then, is a role that someone other than Botham might have occupied for England at the time in question (if, say, at 11 a.m. the English fielders had been swopped around without altering the overall arrangement of the team). And it is a role

ON F U N C T I O N A L I S M

81

that someone from a different team at a different time might have occupied (if, say, England had decided to give Australia some of their own medicine, and had arranged their field to be exactly like Australia's had been at noon on Tuesday).

3.1. By looking at Botham, and by appreciating that qua second slip he occupies an actual role as well as a counterfactual role, we have come to see that there is a relatively specific way of understanding 'the role of second slip for England at t'. Returning to the philosophy of mind, we see that there must similarly be a specific way of understanding 'pain for O at t' - so that the type of state it denotes has a role specification that pays heed to how O actually is, mentally speaking, at t. What might now be suggested is that this relatively specific way of understanding 'pain for O at t' is all that Jackson, Pargetter and Prior need. The problem for their alleged functionalist argument (exposed in Section 1.2) was that a theory about the meanings of mental terms could confer on 'pain for O at t' no interpretation more specific than that of 'pain'. But if we Understand 'the role of pain for O at t' by analogy with Botham's actual role, then it will denote a role that a state of pain has in anyperson with exactly the same mental properties as O. And just as someone could play second slip without occupying Botham's actual role, so, on this new understanding of 'pain for O at t', it has such a sense that it can denote a different type of state from 'pain for N at t' or 'pain for O at u '. In the present section, I shall try to establish that this suggestion is in fact of no help to Jackson, Pargetter and Prior. Tile alleged functionalist argument could not be made to work even if its first premise did avail itself of an analogue of the specific role that emerged from the cricket case (Section 3.3). Moreover the functionalist can have no use for any role still more specific than that (Section 3.5). (I believe in fact that there is a real problem about finding, in the mental case, a usable analogue of Botham's specific, actual role. But I leave this problem for Section 4.) 3.2. We need first to notice a difficulty about matching up our talk of Botham's roles with Jackson, Pargetter and Prior's talk of pain for O at t's role. Both of the roles which Botham occupied, between which we distin-

82

JENNIFER

HORNSBY

guished, were roles whose occupants were, and had to be, particulars. But evidently we are required to understand types as occupants of roles in the alleged functionalist argument (cp. premises (i) and (ii) above). As the word 'role' is commonly used, particulars are the things that occupy roles. Thus in the case of cricket, it was certainly natural to speak of roles as occupied by individual cricket players like Botham, and not by any abstract things. It is true that we speak about the role of a type of property; but we do so because roles are the sorts of things that particulars occupy only by virtue of being of some certain type. One role of Botham's was the role of second slip. This is a role had by that type, the type second slip. But what occupies the role (and "has" it in another sense) is Botham, the man. 4 In the psychological case that concerns us now, it seems reasonable as well as natural to suppose that what primarily occupy roles are states, in the sense of particulars. Functionalists often speak of causal roles. And that makes it seem that roles have as occupants things that stand in causal relations. But states and events are the things that stand in causal relations, s It is true that we speak of types of state and event as causes - as when we say (e.g.) that pain gives rise to belief that one is in pain. But, if we accept that there are states (in the sense of particulars), and we think of causal relations as obtaining between particulars in the world, then, if we say that one type of state is causally related to another type of state, what we mean is that states of the one type are related causally to states of the other type. (If pain causes belief that one is in pain, then, given that there are particular states of persons' being in pain on occasion, one of the things that [typically at least] is causally related to a state of someone at some time's believing she is in pain is a state which is her being in pain.) Well, we could understand the talk of roles occupied by types if we could define what it is for a type of state to have a role in terms of what it is for states of that type to have a (related, other) role. 6 To avoid all danger of letting the typical ambiguity in 'role' tempt us to confuse states with types of state, we should make a typographical distinction: roles are things whose occupants are particulars (cricket players or states); ROLES are things whose occupants are types (positions in cricket, or types of mental state). Then the idea would have to be that the ROLE occupied by a type T of state is that ROLE which a type occupies by virtue of its instances occupying the role that things of type T occupy (by virtue of being of type T). 7

ON F U N C T I O N A L I S M

83

3.3. Consider the role that any player of the type second slip occupies what we called Botham's counterfactual role. We can try to specify a ROLE whose occupant is the type second slip. It seems it must be a ROLE whose occupant has instances that occupy Botham's counterfactual role. This, then, is obviously not a ROLE whose occupant could be identified with any firstorder type. We know that things of quite different types - being six foot tall, being six foot six inches tall, say - can occupy Botham's counterfactual role; for we know that all that is required of one if one is to occupy that role is that one play second slip. The analogue of this point is made, in effect, by Jackson, Pargetter and Prior when they say that no theorist should identify being in pain with having any sort of brain state, because being in pain is a second order property of organisms (pp. 212-213). Well, similarly, being a pain is a second-order property of states; and there need be no one type of brain state such that states of precisely that type have the second-order property required of a state if it is to be a pain. But we should turn our attention to Botham's more specific, actual role, to the role that Botham occupied not by virtue of being second slip simply, but by virtue of being second slip in a team having a certain composition at a certain time. The suggestion was that this notion of role might serve Jackson Pargetter and Prior's argument. So we have to ask: can this (by analogy) enable us to understand 'the ROLE occupied by pain for O that t' in such a way that its occupant can be identified with a physical type? The answer is No. There is the same problem for Botham's actual role as we have just seen there is for his counterfactual role. The ROLE occupied by the specific type that Botharn instantiated playing for England at noon on Wednesday is a ROLE whose occupant is a type whose instances occupy Botham's actual role. But once again, and for exactly the same sort of reason, this is not a ROLE whose occupant could be identified with any first-order type. We saw (at the end of Section 2.4) that Botham's actual role was a role whose second-order specification allowed that it might have been occupied by a different member of the English team, or by a member of the Australian team at some other time. And the various players who might have occupied that role would be of different types (of different heights). There is, then no such thing as "the" first-order occupant of the ROLE occupied by the type second slip in a team composed exactly as the English team was at t. Analo-

84

JENNIFER

HORNSBY

gously we must suppose that there is no such thing as "the" occupant of pain for O at t's ROLE, even where pain for O at t is construed as pain in a person mentally exactly as 0 was at t.

3.4. The versatility of cricketing sides has been exploited in reaching this conclusion. And it may be said 'Of course players of different heights can be at second slip, because what second slip has to do does not depend upon his being of any particular height; but what a state has to be like to be of a mental type places enormous constraints upon the physical types it could be of. ~

This is right, of course. But it would be wrong to think that there is no analogue in the mental case of the versatility of cricket teams. There is such an analogue in the plasticity of the nervous system. Functionalists sometimes remind us of such plasticity; and (more generally) they tell us that they believe in the variable realization of mental properties: they think that it is a contingent ("variable") matter how a person's mental properties are realized at a time.8 Functionalists not only allow that O and N might be in pain and O's pain be of one physical type and N's of another: they also allow that even if some state of the physical type B (say) is actually O's pain at t, it is possible that O should have been in pain at t without participating in any state at t that was of type B. What is more, functionalists allow that two persons could be exactly mentally alike and yet differ (significantly) physically. (They sometimes compare this with two machines whose software states are of exactly the same types, but whose hardware states are of altogether different types.) They think, that is, that O and N could be mentally indiscernible at t and O's pain be of one physical type and N's of another; and that even if some state of the physical type B is actually O's pain at t, then it is possible that O should have been mentally exactly as she was at t without then participating in any state of type B. Thus, just as we all think that Botham's actual role could be occupied by players of many different heights, so functionalists think that even the specific type pain for someone exactly mentally like 0 was at t is a type whose instances can fail to be of any common (significant) physical types. The analogy goes through. The point about variable realization helps to makes it clear why a functionalist cannot use the alleged functionalist argument. If O's pain at t occupies a role which can be occupied by states of pain in which N partici-

ON F U N C T I O N A L I S M

85

pates and by states of physical types which O did not actually participate in, then we know that investigation of how 0 actually is at t could not on its own serve to show which physical type pain for 0 at t is (identical with). If one were to identify pain for 0 at t with a physical type, it seems one would require not only knowledge of persons who at any time are in fact mentally indiscernible from O, but also knowledge of the different ways that O might physically have been at t, and knowledge of all persons, who, though never actually mentally indiscernible from 0 at t, might at some time have been mentally indiscernible from 0 at t. Knowledge of 0 at t could never be enough for a type-type identity, as Jackson, Pargetter and Pollock assumed. 9 3.5. It is an implicit assumption of the alleged functionalist argument that it is possible to confer upon 'the role of pain for O at t' such a determinate sense that the expression has a unique physical reference. And we have seen that a functionalist will wish to deny that assumption if he understands 'the role of pain for O at t' in either of the two ways suggested by our analogy, But perhaps Jackson, Pargetter and Prior think that there is some novel reading for this phrase which can be used in the premises of their alleged functionalist argument: perhaps they think that 'pain for O at t' can be understood in a way which connects up with neither of the two ways suggested by our analogy. But how might that be? Section 1.2 showed why a thesis purely about meaning could take a functionalist at most to a specification of the role of instances of a generic type of state like pain. We have seen how a thesis about the meaning of mental terms might combine with knowledge of mental properties of persons at times to take one to the specification of the role of instances of much more specific types of mental state. Thus: take O who is actually in pain at t; then (we have assumed) there is a role which O's pain actually occupied - which O's pain occupied taking account not only of the meaning of 'pain' but also of all the actual facts about what mental types were instantiated in O at t. It seems impossible to conceive of any yet more specific role for instances of pain for O at t. For if we have brought to bear all the knowledge we have of the nature of psychological states, and all the knowledge we have of O's psychological properties at t, then what more could we possibly bring to bear upon the question what it is for something to be O's pain at t? Let it not be said that investigation of O's physical properties can give us a still more specific understanding of 'the role of pain for O at t'. The func-

86

JENNIFER HORNSBY

tionalist's idea was that we should use our knowledge of which mental things qua mental things occupy which roles in order to be able to come to tell which physical things those mental things are. And physical investigation of 0 at t will not increase our knowledge of the mental types that O's states are of. I conclude that there is no way, along functionalist lines at least, to establish a version of (i) which can be used in the alleged functionalist argument.

4

4.1. We have seen that neither of the two notions of a role for an individual at a time between which we distinguished in Section 2 will serve the alleged functionalist argument. Let us now ask which notion should be used in a genuine functionalist argument, with a view to determining what the functionalist's commitments are. We may assume (pace Jackson, Pargetter and Prior) that a genuine functionalist argument will not lead to identity conclusions other than the state identities which are all that many functionalists actually assert. Apparently, then, such an argument will use premises parallel to those of the alleged argument, but will speak of states where the alleged argument spoke of types of state. Thus: (1) (2)

The state which is O's pain at t = that state which occupies the role which O's pain at t occupies This brain state (of type B, say) = that state which occupies the role which O's pain at t occupies.

Someone who asserted (1) and (2) would be taking it for granted that O was actually in pain at t. And of course someone who wanted to arrive at a state identity claim would have to take that for granted, because if O were not in pain at t, there would be no question of discovering which physical type O's pain at t was of. This makes it seem that the role mentioned in (1) and (2) might be understood by analogy with Botham's actual role. For interpreting 'the role that O's pain at t occupies' by analogy with Botham's actual role, we say in (1) that O's pain occupies a role occupied by any state that is pain in someone with mental properties exactly as O's are at t; and if O is not in pain at t, there is no such role as that. (Compare: if there were no mid-off in

ON F U N C T I O N A L I S M

87

the English team at noon on Wednesday, then there would be no actual role for players at mid-off in a team composed as that team was then.) Certainly counterfactual roles cannot be meant in (1) and (2). For counterfactual roles are specified in a general theory which takes no account of how O is at t; and types of state in which O does not participate at t - types which have no instances to be identified with a physical state - have counterfactual roles for O at t. But using the notion of an actual role, a pair of premises on the pattern of (1) and (2) would be forthcoming for each of the types of mental state actually instantiated in O at t; and in the case of O at t those are all the types whose instances have to be identified with physical states. We should make a mistake, however, if we supposed that we could simply import the notion of an actual role to arrive at identities. In order to avoid that mistake, we need now to notice one of the imperfections of the analogy between cricket and the mind. 4.2. In cricket, when players do such things as their roles demand of them, that does not have the effect of changing the cricketing properties of their teams. Of course a captain can decide that his team's cricketing properties are to be changed, and, in making decisions of this sort, he will take account of how his players have been performing. But it is never the effect simply of the bali's being bowled (say) that the team come to have someone playing second slip, and never the effect simply of the player at second slip's taking a catch (say) that the team ceases to have someone playing mid-off (cp. Note 6). With psychological happenings, it is otherwise. Sometimes, for instance, it is the effect of something's being present in O's environment that O revises her beliefs - she ceases to have some beliefs she previously had, and she comes to have some beliefs that she previously did not have. Thus whereas cricket teams do not have altered cricketing properties as a result of the role-suited goings on that their members participate in, persons do have altered mental properties from moment to moment, the alterations being effected by rolesuited interactions of the mental states they participate in. The point may be illustrated with a very simple example. Suppose that a theory about persons tells us that if someone were in a mental state of type M1, but were not in a state of type M2, then if she were subjected to input of type 17, she would as a result come to be in a mental state of

88

JENNIFER

HORNSBY

type M2. Suppose also that, knowifig exactly which types of mental states are instantiated in a at t, we know that she participates in a state of type M1, and that she does not participate in any state of type M2. We want to give an account of the actual role of a's state of type M1. Evidently we cannot specify the role of this state without saying how a would be affected by input of type I7; and in saying this, we have to speak of states of the type M2. Our account of the role of M1 as it exists in a at t then makes mention of a type of state of which no instance exists in a at t. But this is not the end of the story. For not only must the account o f a ' s M1 state's role make mention of a role which a's M2 state would have if it existed, also it must make mention of the role of instances of any type of state in an instance of which O might participate. Imagine that a has indeed been subjected to input of type I7 at t. Then, inasmuch as our functional theory had told us anything about the physical character of a's M1 state, it would have told us something about the physical character of the M2 state that a has come to be in. But of course our general theory of the mind has things to say about how mental states of yet other types are related to states of the M2 type. So inasmuch as our theory had told us about the physical character of a state of the M2 type, it would have to have told us something about the physical character of states of yet other types .... The ramified character of the interdependencies between mental states, on which the functionalists insist, ensures that one can give no specification of the actual role of mental states of any one type without speaking of the roles of states of every other type. 4.3. (1) and (2) concern O and t andpain ; and statements on the pattern of (1) and (2), concerning O and t, can be constructed for states of all the mental types which are instantiated in O at t. But we have seen that the functionalist will have to speak also of states of all the mental types which are not instantiated in O at t : he must say which physical types those states would be of if, perhaps contrary to fact, O did come to participate in them; and no actual state-state identities will be forthcoming unless he can say this. Consider then a type of mental state which is not instantiated in me now - belief that Australia is the same size as Wales. Unless the functionalist is in a position to say something about this type, he will have nothing to say about the states in which I actually participate. Well, presumably the functionalist says something like this:

ON FUNCTIONALISM

89

If you did come to believe that Australia was the same size as Wales, then that would come about as a result of a series of inputs of some determinate kind or other. The philosopher of mind knows how to specify, from any starting point, the set of series of inputs which would have brought it about that someone came to believe that Australia is the same size as Wales; and he can therefore say which sets of series of inputs would bring it about that someone mentally as you are now came to believe this. The scientist with knowledge of the physical states of your brain can say how your states would be affected by any series of inputs. Putting the philosophy of mind and scientific knowledge together, one could know the various physical types that a state of believing that Australia is the same size as Wales would be of in you, according as that state had been brought into existence in one way or another. It seems then that a functionalist must be prepared to claim in principle to k n o w , in the case o f any type o f mental state M and any person and any time, which type o f physical state the person would come to participate in an instance o f if she came to participate in M as a result o f such and such possible inputs, which physical state she would come to participate in an instance o f if she participated in M as a result o f such and such other inputs ... and so on. The functionalist's claim to arrive at mental/physical state identities from causal role claims is in fact a claim to arrive at identities simultaneously with indefinitely m a n y counterfactual claims that go indefinitely far into indefinitely various possible futures. 10 The point may emerge plainly b y contrast with the case o f cricket. When we move from a specification o f Botham's counterfactual role to a specification o f his actual role, we merely feed into our general, counterfactually given theory statements about the positions occupied in the team o f which Botham is a m e m b e r at the relevant time. But when we move from the specification o f the role o f instances o f pain (say) to a specification o f the role o f an instance o f pain in O, who has such and such mental properties at t l , we cannot feed in statements about the types o f mental states actually instantiated in O at t l , and then' forget about all other types o f mental state. Rather we have to let our knowledge o f what types o f mental state are actually instantiated in O at t~ feed back into the counterfactually given theory so as to say o f O what states she would be in, if at t2 ..., or if at t2 .... and if at t3 ..., or i f at t3 ..., .... Whether O's pain now is to be identified with a state o f some physical type does not depend simply upon h o w O is mentally and physically now, in the way that whether the occupant o f second slip for England now is a man o f a certain height depends upon how the English team is arranged now and what heights its members are. Whether O's pain at t is to be identified with a state o f some physical t y p e depends upon how O would be, both mentally and

90

JENNIFER

HORNSBY

physically, if she were confronted with such and such inputs, upon how she would be both mentally and physically if she were confronted with such and such other inputs .... and so on, for every one of the possible series of inputs which might confront a person. 4.4. It seems to me that only someone with an antecedent commitment to a physicalist position which embraces far more than state identities would find it plausible that, even in principle, anyone could have the extraordinary quantity of knowledge of counterfactuals which the functionalist would require state/state identity claims to be based upon. But it is not my business here to assess functionalism. I am only anxious that the extent of the functionalists' commitments should be clear. It is not, as Jackson, Pargetter and Prior claimed, that the functionalist ought to believe type identities. But the functionalist ought to believe, in addition to state identities, endless counterfactual claims. For each of the types of mental stae in an instance of which O does not participate at t, there must be accepted a huge variety of counterfactuals giving all of the many possible ways that O would be physically if she came to participate in an instance of a state of that type. Moreover no-one could be in a position to assert any actual identities, unless he had this extraordinarily extensive counterfactual knowledge correlating mental types that states would be of with indefinitely many physical types that states would be of.

APPENDIX

In considering what I have called the alleged functionalist argument, I have considered an argument which Jackson Pargetter and Pollock claim that a functionalist can use. I append some consideration of the passage in which the authors present the positive defence of their claim that functionalists ought to accept type/type identities - in which they purport to prove that a functionalist m u s t use an argument which establishes the conclusion of the alleged functionalist argument. T h e functionally inspired identity theorist can spell o u t ... a set o f s t a t e m e n t s for a given organism at a given time o f which the following is an illustrative m e m b e r . When there exist in an organism states o f types M2, M3, M14 a n d M46, a n d t h a t organism is subjected to events o f i n p u t types 14 and 16, the probability is 0.9 that there will come to be states in it o f types M30 and M31, and

ON FUNCTIONALISM

91

that there will continue to be in it states of types M2 and M14, and that there will be events of output types 04 and 070. A sufficiently large set of statements like this combined with knowledge of the internal workings of the organism will enable us to solve what is in effect a large set of simultaneous equations for the values of the Ms .... [I] n order to solve these equations we will need perforce to interpret the Ms as types of state .... [F] or instance M14 is that type which, when it has instances which combine with instances of, say M6, M24 and M30, has instances which play such and such causal role; while when it has instances which combine with instances of M73 and M1 has instances which play so and so a role. But this makes no sense if M14 is a state of some mental type, instead of the type of state itself .... [A] state makes its appearance (if any) in the world exactly once. There is no question of a state which is of a certain type playing one role at the time it is combined with a set of states of certain other mental types and another when combined with states of different set of types, just because it is a state, a particular. (pp. 220-221.) I have amended this passage to conform with m y own terminological practice (cp. Section 0.2). The amendments have ensured that Jackson, Pargetter and Prior's claim that the Ms here are types of state is unexceptionable. What is in dispute is not that claim, but the claim that the functionalist thinks that one can 'solve what is in effect a large set of simultaenous equations' so as to find physical values for the Ms. I am not sure what is meant to be the force of saying that we have 'a set of statements for a given organism at a given time'. We are provided with an illustrative member of such a set. But examination of this suggests that, if it states a truth about the types of states it speaks of, then it states a truth which holds good for any organism at any time. (Cp. Section 1.1 and 4.2 above.) But perhaps we can make sense of a theory which concerns a given organism at a given time, by assuming that a functionalist who hoped to derive any actual mental/physical identities would take account of the mental properties of that organism at that time. He might start with a list of types of state which were instantiated in the organism then. For any state on that list, he can say (counterfactually) that its instance would interact thus and so if... ; for any type of state not on that list, he can say (doubly counterfactually, as it were) that i f there were an instance of it, then it would interact thus and so

i f . . . . I suggest that Jackson, Pargetter and Prior intended us to consider a set of statements which was for an organism at a time in the sense of a set which spoke counterfactuaUy about types of state that were instanfiated in the organism then, and spoke doubly counterfactually about such types of state as were n o t instantiated in the organism then.

92

JENNIFER HORNSBY

Unless we accept this suggestion, I think that we cannot make sense of the rest of what the authors say. They point out that there is no question of a state playing one role at one time and another role at another time. But they evidently think that there is a question of something - a type of state presumably - having one role at one time and another role at another time. If they had it in mind that the set of statements for a given organism at a given time incorporates the facts about the mental properties of that organism at that time, then we could understand the idea that what is said about roles in the set is different from what is said about roles in another set, concerning with the organism at some different time. (Here we are making available to them the notion of an actual role: the actual role of O's pain at tl may be different from the actual role of O's pain at t2. See Section 2; but see also Note 10.) But reading them in this way, we shall now feel free to contradict what we must take them to have meant in saying that there is no question of a state of a certain mental type playing one role at one time (in the presence of states of certain types) and another role at another time (in the presence of states of certain different other types). Of course there is no question of a state's actually interacting in two ways at a single time. But still we have to allow that a state might by virtue of being of some mental type occupy one role at one time and another at another time. For consider the state in a at t which is of the M46 type; and suppose that there continues to be a state of the M46 type in a until at least t', even though some changes in a's mental properties take place between t and t'. Then in the only sense of 'role for a type of state at a time' available to the authors, the role of states of the M46 type may change between t and t'. 11 Now when Jackson, Pargetter and Prior argued against the possibility that we have just seen that they must acknowledge - the possibility of instances of some type of mental state having different role properties at different times - they were, I think, preparing us for the conclusion that state/state identitites could not be derived directly, but could only be derived via typetype identities. But the functionally inspired identity theorist will not be prepared for that conclusion. He will say that, for an organism at a time, the scientist could, in principle, find brain states whose relations to one another and interactions make good the set of statements stating functional roles. (If Section 4 above is correct the relations and interactions here will include possible relations to and possible interactions with states of types that are not actually found in the organism at the time.) The brain states discovered by

ON F U N C T I O N A L I S M

93

t h e s c i e n t i s t will (he alleges) h a v e t h e e x t r a o r d i n a r i l y c o m p l e x s e c o n d - o r d e r p r o p e r t i e s w h i c h t h e i r b e i n g states o f m e n t a l t y p e s r e q u i r e o f t h e m , a n d t h e y will b e seen t o h a v e t h e s e p r o p e r t i e s (he will say) b y b e i n g seen t o b e o f c e r t a i n p h y s i c a l t y p e s . B u t t h e r e is n o r e a s o n f o r t h i n k i n g t h a t t h e b r a i n t y p e s w h i c h t h e s e s t a t e s p r o v e t o b e o f are identical w i t h m e n t a l t y p e s . I n t h e t e x t I a t t e m p t e d t o s h o w t h a t i f w e were to m o v e t o t y p e - t y p e i d e n t i t y claims ( w i t h o u t a n y r e a s o n , it n o w seems), t h e n w e s h o u l d b e m o v i n g against t h e grain o f f u n c t i o n a l i s t t h o u g h t .

NOTES 1 Philosophical Studies 42 (September 1982), pp. 2 0 9 - 2 2 5 . Discussions with Michael Smith were a great help to me in arriving at the view presented here. 2 Since a functionalist presumably thinks that the satisfaction of some mental properties requires the existence of an event, rather than a state, one should strictly read 'state or event' for 'state' in the preceding two sentences, and in much of what follows. I do not say that the ontological assumptions imported with this talk of states are obviously correct, but only that they are assumptions that functionalists have to make. (And I am aware that assumptions about the existence of states are more controversial than assumptions about the existence of events; see Note 5.) Two more terminological points are worth noting. (a) I speak of people participating in states. O participates in a state iff that state is O's where the relevant sense of the apostrophe's' in 'O's' is just the sense that we l-rod used in such a description of a state as 'O's belief that p '. (It might be asked why I do not speak more briefly of people being in states. The reason is that the use of 'in' next door to 'a state' forces on 'state' its construal as true of universals - a construal inconsistent with my terminological policy, that is.) (b) I shall treat realization as a relation between mental properties and states (particular states, albeit states which will have to be seen to be of physical types for realization theses to hold good). 3 The theory of cricket I intend is supposed to be free of all considerations about cricketing tactics and strategies. 4 It is a potential source of confusion that 'second slip' (say, or 'the belief that p') can be used to stand either for a type or for a thing of that type. s I am careful not to say that states and events are the relata of the predicate 'cause'. For I think that some people take the view (and not without good reason) that only events are literally causes. But by taking this view, these people are not forced to reject the ontological assumption made at the outset, that there are states which stand to one another and to events in genuine relations. For there could be relations of ontological dependence between states, deriving from the operation of causality, even ff no state was ever itself a relatum of 'cause'. 6 Another way to understand such talk is as involving a simple and natural ellipsis: by 'the role occupied by type T', one means the role of instances o f type T. But there can be no such ellipsis in the alleged functionalist argument: Jackson, Pargetter and Prior's talk of roles occupied by types has to be taken at face value, and literally, because the argument is supposed to establish a mental/physical identity by showing a mental type and a physical type alike to occupy a certain functional role. -

94

JENNIFER HORNSBY

I shall try to avoid all possible confusion and not go in for the elliptical way of speakhag myself. I pay the price, which is longwhadedness, below. 7 It is a good question what sort of quantification should be read into 'instances' here: 'all and only instances', 'most instances', 'typical instances'? Luckily my argument will not depend upon our returning one answer rather than another to this question. I notice that Jackson Pargetter and Prior introduce probabilistic notions into statements of functional roles (see the quotation from them in the Appendix below). The introduction of probability must be intended to make due allowance for the fact that what we say about roles for instances of types of states need not ha fact be true of 100% of instances of those types. (Consider: it is the role of British military bandsmen to bear stretchers in times of war, even if the occasional bandsman is incapable of acting as a stretcher bearer.) If we do make the role statements themselves probabilistic, then we may suppose that we are free to say that all and only instances of a type T occupy the T-role. (If it sounds wrong to say that the role a man has qua bandsman confers on him a 98% chance of being a stretcher bearer, then so much the worse for using probability to convey an idea of what is typical.) 8 I assume that when functionalists speak of variable realization, they intend us to think of possibilities which include possibilities that are thought not to be actual (such as the possibility that there should be things morphologically very different from human beings, but of which functionalist psychological theories were nonetheless true). And I assume that when functionalists speak of the plasticity of the nervous system, they intend us to think of certain possibilities that we suppose to be actual (such as the possibility that physical states of different neural types [differently located in the brain, say] should subserve some common task). Nervous system plasticity is then a special case of variable realization. And the thesis labelled 'variable realization' might as well be called 'variable realizability': it tells us of possibilities. I take it that functionalists do not mean to rule certain possibilities out where they do not explicitly rule these possibilities in. Thus I believe that even the functionalists who do not speak o f variable realization or o f nervous system plasticity need not object to the idea that (very roughly) such connexions between mental and physical things as might be empirically discovered are contingent connexions. 9 It may be responded to this that sufficient empirical investigation of people like O would make it possible to identify pain for O at t with a physical type of state, even if the physical type would need to be specified as a disjunction. But this response can be made only by someone who thinks that interpersonal projectible psychophysical theories are obtainable : without such theories, no amount of knowledge of how people actually are physically could give one knowledge how people might have been physically at some time if they had at that time been mentally indiscernible from O at t. 10 I have assumed throughout this section that in arriving at state-state identities the functionalist will employ knowledge of how O is mentally at t. What I am suggesting here is that a functionalist with knowledge of a person's actual mental states at a time can say not only what actual roles those states have, but also what actual role states of other mental types would have if states of those types came to exist in that person via such and such changes. I have made the assumption that identities are arrived at via knowledge of how a person is mentally at a time because I wanted to find a sense for Jackson Pargetter and Prior's claim that we can have a functional theory for a given organism at a given time (see the Appendix). But there may be functionalists who think that knowledge of how people are mentally is not required for state identities. Such a functionalist will have no use for the notion of an actual role. For in place of premises on the pattern of (1) and (2) (which can be stated only for states of types actually instantiated ha O at t), he can use premises which say what counterfactual role states of a type would occupy in O at t (and he will have such premises for states of every possible mental type). On his view,

ON F U N C T I O N A L I S M

95

investigation of the brain might itseff determine what mental properties a person has. Loar's thesis in Mind andMeaning (Cambridge University Press, 1981) is that there are mental/physical type-type correlations relativized to persons and times. This thesis could be used to characterize the position sketched in the text; but I think that Loar's own position is the one suggested in the previous paragraph of this note. It is clear that either of the two positions would require actual identities to be based upon extraordinarily extensive counterfactual knowledge - as described in the text. 11 In consequence of this possibility, one has to allow that it might be that at t a's M46type state is of some physical type (B, say) and that at t' a's M46-type state is of some different physical type (C, say). If one assumes that criteria for the individuation of states can be given in exclusively physical terms, then one thereby allows that a might participate in a state of the M46 type throughout a period, even though there is no one state with which it could be identified throughout that period. I suspect that Jackson, Pargetter and Prior's reminder that 'states make their appearance in the world only once' was intended to make us think we could not allow the idea of such temporal relativity for state identity claims. I do not say that the idea is unproblematic. But a functionalist presumably thinks that it is no more problematic than the idea that a cricket team might never cease to have a player at second slip in a certain period even though at different times within this period different people are at second slip. (At any time within the period, there is a person [state of the brain of some type] that is the same as the person at second slip [state of the M46 type]; but it is not the case that there is a person [state of some determinate physical type] such that at any time within the period she [it] is at second slip [is of the M46 type] .)

Corpus Christi College University of Oxford, Oxford OX1 4JF, United Kingdom