Mental Representation Author(s): Hartry H. Field Reviewed work(s): Source: Erkenntnis (1975-), Vol. 13, No.

1, Philosophy of Language (Jul., 1978), pp. 9-61 Published by: Springer Stable URL: http://www.jstor.org/stable/20010621 . Accessed: 24/06/2012 07:49
Your use of the JSTOR archive indicates your acceptance of the Terms & Conditions of Use, available at . http://www.jstor.org/page/info/about/policies/terms.jsp JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide range of content in a trusted digital archive. We use information technology and tools to increase productivity and facilitate new forms of scholarship. For more information about JSTOR, please contact support@jstor.org.

Springer is collaborating with JSTOR to digitize, preserve and extend access to Erkenntnis (1975-).

http://www.jstor.org

HARTRY

H. FIELD

MENTAL

REPRESENTATION*

Any

requires not only that there be no interesting version of materialism objects, but also that there be no irreducibly mental irreducibly mental the idea that although people and certain higher animals do properties: not contain any immaterial substance, nonetheless they have certain in physical terms, is that are completely unexplainable properties an idea that very few people who regard themselves as materialists would of postulating find satisfying.1 The unsatisfactoriness irreducibly mental of in the philosophy is the source of the two main problems properties

mental

mind.

The first and more widely discussed problem is the problem raised for instance the property of feeling pain: a by experiential properties, a believable account of such properties accord? materialist needs to provide ing to which those properties are is the lem, raised by Brentano,2 believing, desiring, properties more precisely, they properties: entities not irreducibly mental. The second prob? problem of intentionality. Many mental and so forth - appear to be relational appear to relate people to non-linguistic

who takes believing and called propositions. So any materialist at face value - any materialist who admits that belief and desire desiring are relations between people and propositions any such materialist must show that the relations in question are not irreducibly mental. Brentano felt that this could not be done; and since he saw no alternative to viewing he concluded that materialism belief and desire as relations to propositions, must be false. problem to be unsolvable) of giving account of believing, desiring, and so forth. adequate I take it as unquestionable (given what we know about of Brentano's The first half of this paper is an examination that is, of the problem (which Brentano believed a materialistically Unlike

Brentano, for the the world) that materialism is true; I also take it as unquestioned two of this paper that people do believe and desire. These purposes that Brentano's problem assumptions together amount to the assumption can be solved; what I shall be interested in is the question of how it can be solved, that is, of what the assumption that it can be solved shows about the nature of belief and desire. I will be particularly concerned to show that

Erkenntnis 13 (1978) 9-61. All Rights Reserved Copyright ? 1978 by D. Reidel Publishing Company, Dordrecht, Holland

10 one cannot

HARTRY solve Brentano's

H.

FIELD

the kind of problem merely by adopting of belief and desire that Robert Stalnaker and David theory In the second half of the paper I shall elaborate Lewis have advocated.3 on the position suggested in the first half, and discuss some implications functionalist for the philosophy of language. of my conclusions about intentionality that people do in fact I should say at the outset that my presupposition that an instrumental believe and desire is intended to be a presupposition is inappropriate. I am presupposing and desire-talk in other words, that belief and desire ascriptions can be literally true, rather than being merely useful devices that we adopt for various purposes but istic view of belief-talk for which are highly in materialism It is clear that belief talk of truth and falsity is inappropriate. even together with the view that belief-talk and desire-talk useful instruments is not sufficient to entail that Brentano's

so the adoption of an instrumentalistic view of is solvable;4 problem is a possible way to save materialism and yet and desire-talk belief-talk instru? to solve Brentano's To my mind, however, avoid having problem. it seems to me that if a psychological is unappealing: mentalism theory that postulates
that works

beliefs
better,

and desires works well,

then barring strong

and there is no available

reasons to the contrary,

theory
we

specific

should regard that theory as literally true. That it should be so regarded is a working hypothesis under which I am operating in this paper.
I. THE BASIC ARGUMENT

to reply to Brentano's argument would be to develop a theory of etc. according to which the objects of belief and desire were belief, desire, If we but something more not propositions, say sentences. 'accessible', construed belief and desire as relations between people and sentences, - at least if we also demanded that the sentences a believer or d?sirer then - it to were sentences in a language he was familiar with were related One way to give an account of these relations in a way. I will have more to say about this method impeccable materialistically of dealing with Brentano's problem later on. For now I want to ignore it. Suppose then that we accept the idea that belief and desire are relations would presumably be possible that we and suppose, for definiteness, people and propositions; to be sets of possible in taking propositions Lewis and Stalnaker worlds. The question arises, how can one give a materialistically acceptable so construed ? account of a relation between a person and a proposition between follow

MENTAL

REPRESENTATION

11

The question is neither silly nor impossible to answer. One might think silly: "If you accept the ontology of propositions, plus a minimal amount of set theory, then a relation between people and propositions is it were are people and whose merely a set of ordered pairs whose first members are propositions. second members Or better - since the same person can be in the belief relation at one time but not at another - the belief relation a function that assigns to each time a set of ordered pairs of A function assigning sets of ordered pairs is not and propositions. people a mental entity at all, let alone an irreducibly mental one, so what's the to say, this misses the point. If this way of dismissing problem" ?Needless is merely Brentano's question were other prima facie difficulty work, so that materialism would be a totally vacuous doctrine. For instance, account of pain, one could say, if someone asked how to give a materialist a set of people, viz. those people that "What's the problem? Pain is simply we say 'feel pain' or 'are in pain'. Or better - since the same person can feel pain at one time but not at others - pain is a function that assigns to each time the set of people who are in pain at that time. Such a function is a purely set-theoretic entity; it is not mental, let alone irreducibly mental, I take it to be obvious that hence it carries no problem for materialism". to the theory of pain is totally unsatisfying. approach reason why the set-theoretic And think of any to the approach theory of belief and desire should be thought to be any better. Part of the explanation of why the set-theoretic account of pain is is that it treats pain as a set rather than a property. What is unsatisfying I can't of that property of people that 'pain' stands for - we need to which is materialistically show that property is acceptable not irreducibly mental. What the set-theoretic account does is to ignore the question of the property altogether, and to focus instead on the set of - since the same (Or things that have that property. thing can have the at one time but not another - it focusses on the function-in property extension mapping each time into the set of things that have the property at the time.) Obviously there is no substantive question of this set (or this function) but that just shows that the being irreducibly mental; interesting question concerns not the set but the property. This diagnosis extends to the case of belief: the interesting question about belief is a question about the relation that 'belief stands for, where by 'relation' needed is an account this set-theoretic legitimate, we for materialism could equally well dismiss any without doing any substantive

12 I mean needs But not

HARTRY

H.

FIELD

to show

'set of ordered pairs but '2-place property'. What a materialist is that that property is not irreducibly mental. is there any hope of giving a materialistically acceptable account of relation?

In particular, is there any hope of doing this when we to be a relation between people and propositions, and construe in terms of possible worlds? It might be thought that the propositions answer is 'obviously no', and that this fact shows the untenability of the belief take belief treating belief as a relation between a person and a set of possible worlds. I believe however that such a dismissal of the possible-worlds view of to give a belief would be too quick: for I believe that it is possible materialistically adequate account of the belief relation, even if that rela? tion is construed as a relation to sets of possible worlds. I will now sketch a possible such account. We will see later that the main elements of this account can be preserved even if we abandon the assumption that belief is a relation between people and sets of possible worlds. that the belief relation is a com? The account involves the assumption

posite of two other relations: first, a relation between a person and a the sentence that the person understands; second, a relation between sentence and a set of possible worlds. This will be refined shortly, but the claim is this : unrefined, (1) X that p if and only that X believes* S and S means believes is a relation between if there that p. is a sentence S such

Here own

believes*

language: I believe* the English sentence 'Snow is green', and no one ignorant of English believes* either of these sentences.5 The effect of adopting (1) is to divide the problem of adequate account of the belief relation into two giving a materialistically subproblems: subproblem (a): the problem of explaining what it is for a person to believe* a sentence (of his or her own language). subproblem (b): the problem of explaining what it is for a sentence to mean that p. (The problem of giving a materialistically adequate account of

a person and a sentence in his sentence 'Snow is white' but not the

the

MENTAL desire relation would

REPRESENTATION in the analogous

13

by introducing The rough idea of how to give an account of (a) should be clear enough: I believe* a sentence of my language if and only if I am disposed to employ that sentence in a certain way in reasoning, deliberating, and so on. This is very vague of course, but providing that the vagueness can be eliminated and providing that a physical basis can be found for the dispositions in? then believing* will not be a relation that poses any problems for voked, on in this section I will state a more precise version of the materialist, iater a dispositional account of believing*, one which will make it pretty clear what a physical basis for the disposition invoked would have to be like. But I hope that even the vague remarks above are enough to predispose the reader to think that believing* is not a relation that should be a particular worry to amaterialist (even amaterialist impressed by Brentano's problem). On the other hand, anyone impressed with Brentano's problem is likely to be impressed with subproblem (b), for unlike (a), (b) invokes a semantic relation (of meaning that). Before looking into the question of whether it is reasonable to expect that a materialistically account of (b) could be given, let me acceptable note two auxiliary advantages of adhering to (1), and then go on to con? sider a necessary refinement in what I have said. The first auxiliary advan? the problem of explaining what it is to believe that p tage of approaching via (1) is that with this approach we need not assume from the outset the that the objects of belief are propositions. That assumption is, ifwe later come up with a way to explain what it is to mean that/? which does not make meaning that a relation between sentences and propositions, also adopted (1) we would have an account of believing that to which it was not a relation between people and propositions according either. In fact, one can adhere to (1) without assuming that the occurrences of '/?' in it are occurrences of a quantifiable variable; they can be taken to be just instances of a schematic letter, so that talk of the objects of belief (as such talk has heretofore been construed) makes no literal sense. Such a possibility will be discussed in Section III; for now however I will con? tinue to adopt Lewis' and Stalnaker's assumption that '/?' is a quantifiable variable and stands for entities that are to be explained in terms of possible
worlds.

a notion

be split into two subproblems of desiring*.)

way,

controversial

then if we

second auxiliary

advantage

of (1), for possible world

theorists,

is that

14 it would

HARTRY

H.

FIELD

of belief which are give them a plausible way to make distinctions intuitively there but which cannot be made on the possible worlds account as it stands. Consider who has never formulates believe a more any set theorist (who believes the axiom of choice) of the Banach-Tarski if someone now theorem; the theorem for him (without telling him that it is a theorem), heard

he will almost

surely not believe it to be true. (In fact, he is almost sure to it false6; to believe it true in the circumstances described would be likely sign of irrationality than to believe it false.) There is a big

difference, then, between believing set theory and believing set theory plus the Banach-Tarski theorem; yet set theory and set-theory-plus-Banach are logically equivalent, hence must be represented by the same set Tarski of possible worlds. Now, David Lewis has pointed out that a possible worlds of the axioms of set theorist need not say that the conjunction means the same as the conjunction of those axioms plus the Banach theory7 Tarski possible theorem: worlds.8 A he can recognize set of possible than sets of finer grades of meaning can be called a coarse-grained is a certain complicated kind of proposition worlds worlds, but which
What

a,fine-grained proposition; function defined out of possible

to speak, the structure of a

has coded
a sentence

into it, so
means, on

specific

sentence.

Lewis'

;but what a person believes is coarse-grained, account, is fine-grained so believing are set theory and believing set-theory-plus-Banach-Tarski the same thing. precisely Now formally grained

speaking, it would be possible to simply take these fine? and use them in a way that Lewis does not, as the propositions to do so unless we think of of belief. But it is quite implausible objects in the head by sentences or something like beliefs as being represented have the structure of specific sen? sentences, for fine-grained propositions Lewis and Stalnaker accept tences coded right into them. Consequently, the idea that to believe set theory and to believe set-theory-plus-Banach Tarski are the same things. Stalnaker feelings to the contrary as follows : has tried to explain away our strong

true one and the true propositions, are only two mathematically the necessarily false. But that the first is true and the second false one, and we all know necessarily is expressed which of the two propositions the functions that determine by a given to give rise to are just the kind that are sufficiently statement mathematical complex it Hence which is expressed doubt about reasonable by a statement. proposition There seems tions mine). reasonable about to take the relations the objects of belief statements between and doubt and what in mathematics they say ([25], to be proposi? p. 88; italics

MENTAL REPRESENTATION But this will not do. Let conditional whose antecedent 'the Banach-Tarski conditional'

15 stand for the

(including choice) and whose Consider a person who doubts or disbelieves the Banach-Tarski result, but who knows the semantic rules that relate sentences in the language of set theory to propositions. really doubt or disbelieve conditional, is the proposition (i) the to Stalnaker, such a person doesn't According the proposition expressed by the Banach-Tarski since that is a logical truth; what he does doubt or disbelieve expressed semantic by rules for the language of set theory to the necessary truth. relate

is the conjunction of the axioms of set theory is the Banach-Tarski Theorem. consequent

the Banach-Tarski But

conditional

since the person knows what the semantic rules for the language of set theory are, the only way he can doubt or disbelieve the proposition ex? would be to doubt or disbelieve the proposition pressed by (i) expressed by (ii) the semantic rules-relate truth, the Banach-Tarski conditional

to the necessary where

in the blank goes a statement of the semantic rules for the language of set theory. Unfortunately ex? for Stalnaker, however, the proposition a necessary truth, and hence on Stalnaker's coarse? pressed by (ii) is itself

grained possible worlds view there is no way that anyone could doubt or disbelieve it. The ascent from mathematical to metalinguistic propositions has gained nothing. propositions I take it, then, that it is clearly desirable more finely than sets of possible worlds to be able to divide propositions : and Lewis' analysis shows us how to do this, if we can accept the idea that believing a proposition involves an attitude toward sentences. To believe set theory is very roughly to the conjunction is very roughly of the axioms; to believe set-theory-plus-Banach to believe* the conjunction of those axioms plus theorem. The sentences believed* have different fine? analysis; hence to the two believers. (1) assigns different fine?

believe* Tarski

the Banach-Tarski grained meanings,

on Lewis's

grained propositions It might be thought however that this sentential approach induces too a distinction it makes the belief that either among beliefs; fine-grained Russell was hairless or snow is white differ from the belief that if Russell was not hairless then snow is white; how then can we confidently pass from the claim that someone has the first belief to the claim that he has

16 the second? These

HARTRY

H.

FIELD

are two possible lines of answer. The first, which I believe to be rather ad hoc and artificial, is to introduce a slight coarsen? classes under by taking equivalence propositions, ing of the fine-grained a suitable equivalence relation. The second and more satisfactory is to grant that the beliefs are indeed distinct, but to explain why someone who has one belief will nearly always have the other. Such an explanation is to give: it involves the idea that belief* is dispositional. I easy enough will on the dispositional nature of belief* (thus returning to subproblem (1)); in the course of doing so, it will become clear why someone who has one of the two beliefs about Russell will nearly always now elaborate the other. The

have

account crudest approach to the problem of giving a materialistic : this approach would involve of belief * would be wholly non-dispositional that a person could believe* a sentence only if the sentence the assumption were explicitly stored in the person's head. (Similarly for desire*. If we to this non-dispositional suppose that there approach we must and storage-as-desire. are at least two kinds of storage, storage-as-belief or stored-as The precise details of how sentences are stored-as-beliefs adhere desires is of course not a matter that can be settled without detailed investigation.) A defect of this approach is that people neurophysiological if so, then if we adopt (1) believe infinitely many propositions; apparently it follows assume that the relation of meaning that is not one-many), (and no way that sentences. But there is that people believe* infinitely many can be explicitly stored in a finite head, so the is false. be explicitly stored to be believed* to that some kind of dispositional This argument approach suggests can in fact be argued for without the belief* is required. (This conclusion that we have infinitely many beliefs.) The simplest kind of assumption infinitely many sentences idea that a sentence must dispositional stored within account (suggested by Dennett9) it certain explicitly represented core beliefs', and what we say a person believes are just the obvious con? of his consequences [Not: the obvious sequences of his core beliefs. that even very unobvious for then we could show recursively beliefs', inferences of beliefs had also to be beliefs (since unobvious consequences can be broken down into a large number of obvious ones).] If the core approach than the notion is correct, then it may be the notion of core belief rather of belief that would enter into any detailed psychological is that an organism has calls beliefs, which Dennett

belief

MENTAL REPRESENTATION

17

theory. But even if this is so, it is the notion of belief rather than the notion of core belief which is more useful to us in everyday life, for it is much The easier reason a person believes than what he core-believes. the proposition A is an obvious enough con? of one's core beliefs, one is almost sure to add A to one's belief to know what is clear: when

sequence core whenever

then, any question as to the truth of A arises; behaviorally, this case from the case where A was present it will be hard to distinguish in one's belief core all along. But there are many cases where it is perfectly

is not part of one's belief core. For instance, suppose I tell you that no one dug a tunnel from here to China through the center of the earth in 1953. I'm sure that by telling you this I'm not telling you something you didn't already believe, but I'm equally sure that itwas not i.e., not one of your explicitly part of your belief core represented before I told it to you. beliefs clear that a belief idea of core belief gives us a very natural proposal for solving sub problem (a). The idea is simply to say that one believes* a sentence if and of sentences that are consequence only if that sentence is an obvious This explicitly opposed someone would to beliefs as of storage appropriate (in that manner to say desires).10 This kind of account would make clear why who believes that either Russell was hairless or snow is white stored

certainly also believe that if Russell was not hairless then snow is white: almost any stock of core beliefs from which the former was an obvious consequence would also be a stock of core beliefs from which the latter was an obvious consequence.11 (This solves the problem from which we began.) The account also makes clear why there is a certain almost in ascriptions of belief that are far removed vagueness or indeterminacy beliefs show a deep in? from the core (e.g., of why, when a philosopher's there is no clear point to asking whether he believes or dis? consistency, a proposition about which he has expressed no opinion but which and refutable by equally unobvious from is both provable reasoning comes in himself to). The indeterminacy things that he has committed because of the use of the notion of obviousness in defining belief from believes core belief.12 explains I think attractive. It is a striking advantage these facts. then that the core-belief I do not want of the core-belief proposal that it

However, sible to develop a more

to subproblem approach (a) is quite to insist on it: conceivably it is pos? of dispositional subtle kind approach to belief*

18

HARTRY

H.

FIELD

perhaps an approach which would require fewer sentences to be explicitly stored. (It seems quite unlikely, however, that one could do without stored no stand on whether an alternative sentences altogether.) I take to the core belief approach to make it plausible
solved.

would

be preferable, for I think I have that in some way or other subproblem

said enough (a) could be

- the to subproblem (b) problem of giving a material? account of the relation of meaning that (which we are istically adequate to relate sentences to sets of possible worlds) - I temporarily assuming Before turning the long-promised refinement that is needed in (1). The should mention and need of the refinement arises from the fact that dogs, chimpanzees, children presumably have beliefs but have no language. One could young simply insist that beliefs and desires are not to be attributed to these or any other languageless creatures. To do so however does not seem very plausible. After all, we do often explain the behavior of languageless of course by postulating beliefs and desires; and while in some such cases can be are clearly anthropomorphic and the behavior the explanations in simpler ways, there are many cases involving psychologically explained complicated organisms where these are the only sorts of explanation avail? are false, able. It seems rather rash to conclude that all such explanations creatures and therefore we must modify (1). Fortunately only a minor modification is needed, if we accept the widely held view13 that though a dog has no and language, he does have some kind of system of internal representation, some system of internal representation that that it is only because he has Let us then he can represent to himself, and believe, any proposition. that the entities which are the requirement (1) by abandoning modify let us require instead only that they be believed* be literally sentences, - where either sentences or sentence-analogs by a sentence analog I simply some psychological mean (or more entity which represents propositions neutrally, have). (1') which has the kind of meaning then, by or content which sentences (1) should be replaced, X believes

that p if and only if there is a sentence or sentence analog S such that X believes* S and S means that p.

This

with adult humans virtue in connection shift has an independent that they might have that do have a language: it allows for the possibility beliefs that go somewhat beyond the expressive capacity of their languages.

MENTAL

REPRESENTATION

19

are hostile to postulating systems of internal repre? Many philosophers I will try to remove some of that sentation in languageless organisms; some of the later on in the paper. But perhaps I can undercut hostility in advance by noting that on my view, the system of internal hostility in which a chimpanzee believes and desires would presum? representation - its have a much simpler structure than human languages have ably would be much than English simpler syntactically 'sentence-analogs' sentences are, and by calling them 'sentence-analogs' I do not mean to be denying the existence of these structural disanalogies. How simple could an animal's of internal representation' be and still count 'system as a system of representation? This seems to me an uninteresting question of terminology. If you like, you could even allow that an animal's 'system of internal representation' consisted of five sentences, each of them syn? point things unless one attributes a reasonably complex system of beliefs and desires to it; and (according to system of beliefs and desires will require a (1')) a reasonably complex So there is no point in system of representation. reasonably complex at all unless one attributes attributing a system of internal representation or desires complex one. But there say "something must be at least this internal representation"; where such of how widely one wants to extend pendent of such a reasonably is no need to draw a precise line and to count as a system of complicated a line would go is merely a matter the term 'belief. My claim is inde? However, tactically unstructured. in saying that an animal believes I don't think that there is much

issues: it is that organisms which are terminological the notions of belief and desire to be clearly sufficiently complicated for in which the sentence applicable have systems of internal representation structure. analogs have significant grammatical Couldn't we an alternative

avoid introducing systems of internal representation, by modification of (1)? The alternative revision would be to that an organism believe* a sentence only if he give up the requirement it: we could then say that dogs believe* sentences of human understands languages like English this approach: (A) The and Serbo-Croatian. There are three problems with

is not very useful as a step toward solving approach Brentano's problem: on this approach, believes* would become a semantic relation - it would be a relation between organisms

20 and sentences

HARTRY

H.

FIELD

of English which holds in (or Serbo-Croatian) of the meanings of those sentences. Consequently, the to (1) would divide the original belief alternative proposed virtue into two semantic organisms and propositions means that is also semantic) ;and since semantic (since relations are just the sorts of relations that Brentano thought were inexplicable, this would appear to be a doubling of our relation between relations (B) original problem. The second problem the suggestion is that the whole to aid us in explain? point of the notions of belief and desire is ing behavior. It seems rather bizarre to explain a chimpanzee's in a human language behavior via his relation to sentences with

facts about English sentences have that he doesn't understand; to do with why the chimpanzee behaves as nothing whatever he does. So it is important to leave English sentences out of the account The third problem with cause of these difficulties, tion of (1). of belief and desire for chimpanzees. in Section III. Be? the suggestion is mentioned I think that (1') is the only reasonable modifica?

Let us turn, finally, to subproblem (b), which because of the shift from a materialistically acceptable account (1) to (1') is the problem of giving in a system of internal of what it is for a sentence or a sentence-analog to mean that p. Recall that we are temporarily construing representation to sets of possible means that as relating sentences (and sentence-analogs) to slightly more structured entities built up out of functions worlds (or let us ignore this more but for simplicity defined on possible worlds; of principle for a account - it raises no new difficulties complicated set of possible worlds does a sentence of belief.) What materialist theory It means the set of possible worlds at which (or sentence analog) mean? the sentence account is true. If then we between of the relation can give sentences a materialistically acceptable and worlds of being true at, acceptable account of (b).

to give a argued that it is possible account of truth (i.e., of truth at the actual adequate materialistically The idea of how this is done is simplest for the very simple lan? world). there it involves giving a Tarski guages that logicians usually discuss;

then we will have a materialistically in another paper14 I have Now,

MENTAL

REPRESENTATION

21 is for a a name to have

it Such a truth-definition type truth-definition. explains what sentence to be true (at the actual world) in terms of what it is for to denote an object (at the actual world), what it is for a predicate a certain extension (at the actual world), and so forth. In other such a truth-definition reduces

to be true (at the actual world) of the sentence to stand (at the actual world) for certain entities or sets of entities. I argued that we have every reason to think that the relations of standing for (i.e., the relations of denoting, having as an extension, etc.) are explainable in a materialistically acceptable way, and so the truth definition a materialistically account of acceptable gives us derivatively truth (at the actual world). The above remarks are for sentences of the simple languages to which Tarski's methods apply directly; but following

words, the problem of what it is for a sentence to what it is for the one-word components

Davidson1&

and many other philosophers, I believe that such methods so that for sentences in natural are generalizable to natural languages, languages too we can say that truth can be explained in a materialistically if we acceptable way. And the same should be true for sentence-analogs, suppose that they are sufficiently like the sentences in natural languages. I have claimed that we can give a materialistically adequate account of truth at the actual world. But what we need, if we are to explain truth, i.e.,

what

a Stalnaker-type is a materialistically proposition, account of truth at w for an arbitrary possible world w. And this, adequate it may seem, is much harder to come by: for if we relativize Tarskian semantics to an arbitrary possible world w, we find that it explains what it is to mean it is for a sentence to be true at w in terms of what it isfor a name to

refer to an entity at w and what it isfor a predicate to have a set of entities as its extension at w. That is, we need to invoke relations of words (in the actual world) standing for entities or sets of entities in the possible world w; and it is difficult to see how to explain such trans-world relations acceptable way. is a very misleading way of looking at the matter: for in actual fact one can define truth at w purely in terms of relations between words (in the actual world) and entities in the actual world. For instance, for a sentence of form ^(b)1 where P is a predicate and b a, name, the This however definition will read if there is an object x that b in a materialistically

rPCi)1 is true at w if and only

22 denotes

HARTRY

H.

FIELD

(in the actual world) and a property Z that P stands for (in the actual world), and w is a world in which x exists and has Z.

is hairless' is true at any possible world w 'Bertrand Russell For example, in which Bertrand Russell (i.e., the person denoted in the actual world by exists and is hairless (i.e., has the property the name 'Bertrand Russell') that the word 'hairless' stands for in the actual world). So truth at w is definable in terms of the relation of denotation for names of standing for predicates, where it is now that a predicate stands not for a set (its extension) but for a property exists in the actual world.16 the relation together with to be understood that

So the problem of giving a materialistically adequate account of truth at w reduces to (i) the problem of giving a materialistically adequate for names, and (ii) the problem of account of actual-world-denotation adequate account of actual-world-standing-for giving a materialistically for predicates, where as I've said predicates are now to be construed as that arose rather than sets. But (i) is a problem standing for properties even in the account of truth in the actual world ; I'm assuming (since I've article) that we have good reason to argued it in the aforementioned is solvable. With believe that that problem (ii) the situation is a bit dif? is a in any account of truth-in-the-actual-world account of the relation between predicates and adequate materialistically is let us then assume that that problem their actual world extensions; that is formally stronger; it requires solvable. But (ii) requires something ferent. What we need a materialistically adequate account of the relation between predicates and since these properties determine but are not deter? and properties; of the predicates, what we need is formally mined by the extensions semantics would have to give us. A little stronger than what actual-world to is bound semantics that actual-world reflection will show however as well. For what could a materialistically give the formally stronger thing like? Such an adequate account of the relation 'is the extension of be account would have to tell us the kinds of facts about the way we use things is the extension of the is that it would be that? One possibility word 'hairless'. How could it do state that some kind of causal it would a causal theory of reference: connection between hairless individuals and our use of the word 'hairless' words in virtue of which the set of hairless

MENTAL

REPRESENTATION

23

is responsible for our word having the set of hairless individuals as its extension. But surely we have not come into causal contact with all members of the extension of 'hairless' - after all, there are hairless things on remote planets, there are hairless things that exist only in the distant could causal contact with certain hairless future, and so forth. How individuals help in solving the problem of what makes our word 'hairless' have as its extension a set which includes such future hairless creatures (and which excludes the future hairy ones) ? The only possible answer, I of the term 'hairless' is determined think, is that the extension by a is directly associated with the predicate is not its extension, property. What but a certain property. Perhaps the property of hairlessness is associated with the term 'hairless' by causal connection with a number of hairless individuals, as a causal theory of reference would have it; or perhaps there theory of reference which determines how the property of hairlessness is associated with the word. But however that association is made, the association of an extension with the word 'hairless' is a derivative one: the extension of 'hairless' is simply the set of objects (in the actual world) which have that property which in the actual world is associated with the word 'hairless'. If it were impossible to give a material? istically adequate account of how properties were associated with predi? to give a materialistically cates, then it would be impossible adequate of how impossible truth. So is some non-causal

sets are associated with predicates, and so it would be to give a materialistically account of actual-world adequate the problem of giving a materialistically adequate account of is no more difficult than the task of giving a truth-at-a-possible-world account materialistically adequate of truth-in-the-actual-world; funda? these apparently distinct problems are one and the same. mentally, I now have all the pieces : let me put them in place. The task with which I have been occupied for most of Section I has been to explain ont possible to giving a materialistically approach adequate account of belief; and I have been assuming for the time being that belief is to be construed as a relation between people and sets of possible worlds. The 'other? character of the assumed objects of belief might appear to rule worldly' out the possibility of a materialistic account of the belief relation, but, I have argued, that appearance is illusory. For we can say that a person believes a set/? of possible worlds (or sentence-analog) S, and p if and only if he believes* is the set of possible worlds some sentence at which S is

account

24 true. And possible

HARTRY

H.

FIELD

genuine us to explain what makes actual world for whatever is doubtless

of possible worlds) truth at a defining (given the ontology world requires no more resources than are required for defining truth, i.e., truth at the actual world: both of these tasks require the names in S denote whatever items in the in S stand the predicates they do denote, and what makes in the actual world they do stand for,17 and that properties a difficult task; but I know of no reason to believe that the

task is hopeless, and in fact (as I've argued in the aforementioned article) I think there is good reason to think it can be solved. The account of believing that has just been elaborated has only been put forward as a possible account. It is in fact an account that I will want to revise in certain respects later on e.g., by dropping all reference to and possible worlds. But I will not revise the two central propositions features of the account given so far: the idea that belief involves a relation or to something very much like sentences, and the idea that semantics (including a theory these sentences must be given a Tarski-like the primitives). To a large extent, my adherence to these of reference for to of what a solution two ideas is based on an inability to conceive to sentences problem that did not rely on these ideas would be like. I do account of to have proved that there could be no alternative but did not have the two features that solved Brentano's belief problem I have never seen even a clear sketch of what above. However, mentioned such an alternative account would be like. The fact that, the kind of ac? Brentano's not claim solve Brentano's problem, and to the problem are known, seems to me to provide a very strong reason for tentatively accepting the kind of account sketched.18 alternative At this point the reader may feel that there is a well-known count that I have sketched above would that no alternative solutions account of belief that will either solve Brentano's problem, or show it to to my hypo? account (according That alternative be a pseudo-problem. thetical reader) is the account that Lewis and Stalnaker actually advocate: In the next section I will examine the alternative. it is the functionalist question problem.
II. FUNCTIONALISM

of whether

functionalism

does

indeed solve or dissolve

Brentano's

The Lewis-Stalnaker is a functional

to belief involves the idea that believing approach state. Putting the view very roughly, a state of an organism

MENTAL is a state of believing

REPRESENTATION

25

that p if that state plays an appropriate role in the to inputs and that is, if it is causally connected organisms's psychology: states in the right sort of way. Now it outputs and to other psychological theory of belief allows for certainly seems at first as if such a functional the possibility of beliefs independent of language or of inner representa? is in a state tions: that is, why couldn't it be the case that an organism which to inputs and outputs and to other states in is causally connected the right sort of way for that state to satisfy the functional conditions for being a belief that p, even though that state does not involve a relation to any sentence or to any sentence-like sentation? Stalnaker has put the point
It is conceivable and

item in any system of inner repre? as follows:

beliefs

or not it is true) that there are rational creatures who have (whether but who do not use language, and who have no internal repre? which attitudes have a linguistic form. I think this is true of many sentations ones even some rather cases. animals but there might be clearer stupid Imagine on some other creatures that we discovered who did not perhaps living planet desires, of their was predictable, but whose behavior for the most communicate, part, on the hypoth? in highly esis that they engaged theoretical further deliberation. sophisticated Imagine our hypotheses: that we had this indirect evidence that the beliefs that supporting our cases, hypothesis in terms were for attributed of their correlated their to sensory these creatures and could that inputs; be causally in many explained, to them by the the desires attributed

hypothesis

quirements lating the environments

survival.

or and by satisfying as predicted, I think we would to attribute behave be tempted to these creatures not and desire but beliefs and desires We would themselves. just belief not, analogues, have to hypothesize in a mental that they thought any reason however, language, or in any language at all ([25], p. 82).

re? for the most the physical part, with that we test the hypothesis Finally, imagine by manipu? of these creatures, them misleading 'evidence' say by feeding some of their alleged If they continued desires. to frustrating appropriately,

Is Stalnaker In order functional makes

right about this ? to evaluate Stalnaker's

theory of belief much it. According to the crude description above, a state of an organ? ism is a state of believing that p if the state is causally connected to inputs and outputs and to other psychological states in the right sort of way; but until we know what the other psychological states are that it must

claims, we need to make the idea of a clearer than my crude description above

be causally connected to, and what is 'the right sort of way' for it to be we are in no position to them, connected to determine whether such a functional theory of belief requires a system of inner representation. (E.g., why couldn't one of the other states to which a state of belief that

26

HARTRY H.

FIELD

p must be causally connected be a state of having an inner of the proposition /??) This fact must be borne in mind about the creatures on another planet. Stalnaker's story clear that in the situation Stalnaker describes, we could not certain But

representation in considering It is perfectly be absolutely

that the creatures thought in any internal system of representation. certain that it is also perfectly clear that we could not be absolutely desires. We couldn't be certain of this because (a) they had beliefs and let alone all of their we have not observed all of their actual behavior, and (b) because even if we had, a claim about beliefs possible behavior, it is a claim about how is not simply a claim about behavior, is produced. Of course, we do not need to have detailed that behavior the organ? of how that behavior is produced to know whether knowledge - the whole is to provide of functionalism has beliefs and desires ism point inner states, so that of an organism's a fairly abstract representation and desires certain kinds of information about how to the question some knowledge do need irrelevant to know
organism's

of whether about
is a

the organism's it has beliefs that behavior

inner

states will

be

and desires.

But we

that it has beliefs

inner states

and desires:
claim about

is produced in order a functionalist's claim about an

that organism's inner states, and

actual and possible to any claims about the organism's It comes out very I hope, common knowledge. behavior. (That fact is, of the functionalist claim, which I clearly in Lewis' precise explication when we conclude of the creatures on will give shortly.) Consequently, is not reducible planet that they have beliefs and desires, we do so because is the best explanation we can find of their behavior. But now, if we that can know that the creatures have beliefs and desires because belief-and the other desire couldn't explanation theory is the best available we also know that the creatures have of their behavior, why a system of inner repre? the first inference has better

sentation, by the same means? Whether that can be settled than the second is not something inductive warrant to : it depends on whether it is possible by a casual appeal to intuitions come up with an adequate functional account of belief and desire accord? So ing to which belief and desire do not require internal representation. from Stalnaker the last sentence of the quotation (or, if you like, the of the last two sentences) simply begs the question at issue. conjunction is correct, but we can show this only by a detailed its conclusion Perhaps a functional of what examination theory of belief would be like.

MENTAL A full discussion sion would belief would built involve have

REPRESENTATION

27

of this topic would be two-fold. Part of such a discus? of whether the question any adequate theory of to have about internal representation assumptions and I strongly inclined to think it would, in Section I and the remarks about

explicitly think my

into it. I am about

remarks

core beliefs

propositions ultimate proof

in Section

logical theory. a more undertake approach alternative to

III provide some support for this claim; but the the claims lies in the detailed development of of psycho? In this section I will not deal with these matters, but will modest task: I want to show that the functionalist does an not provide desiring some system of representation.

of believing and to the idea that belief requires theories

In saying this I mean to be saying much more than that functionalism is compatible with there being some system of inner representation (in? sentences of natural that volving language and/or sentence-analogs): point is certainly not one that either Lewis or Stalnaker would deny. Both of them would and found certainly agree that if we opened up a person's head a blackboard inside on which various English sentences were

written, and if we also found that the occurrence of those sentences on that blackboard entered in the right way into that person's behavior, then that would be strong reason to suppose that (for that person anyway) belief involved inner representation. This would of course not conflict with for what we would be discovering is simply that the functionalism, state which (in that person) is causally connected in the right sort of ways to inputs and outputs and to other states happens to be a state which involves

inner representations. It is uncontroversial, then, that functional? ism is compatible with the inner representation hypothesis. But I want to argue for a much stronger claim, one which I suspect Lewis and Stalnaker would deny (though I infer this more from the general tone of their dis? than from anything I suspect Lewis they explicitly say). What and Stalnaker hold is that without opening up people's heads, we can have little or no reason to think that believing anything involves a system of inner representation; the hypothesis that belief does involve a system then, is simply unfounded neurophysiological representation, in Section 11 have given a,prima facie argument that it speculation. Now, was not just unfounded the inner rep? neurophysiological speculation: resentation I claimed, was the only known possibility for hypothesis, and this in itself provides good reason for solving Brentano's problem, of inner cussions

28

HARTRY

H. FIELD

tentatively accepting it. But in arguing that, I ignored the possibility of the functionalist approach. What we must investigate now, then, is whether the functionalist approach alters the conclusion. My claim is that it does not, and that is what I mean when I say that functionalism provides no to the view that belief requires some system of representation. alternative To put my claim in a sentence, then, it is that functionalism does not provide a solution the problem. That The whole abstract point to Brentano's is the claim problem, nor does it in any way dissolve to be argued in this section. is to provide a fairly a representation that structure of particular that if two organisms

in psychology of psychological representation properties, is not tied too closely to the details of the physical in fact, a functional organisms; theory guarantees are, in a suitable

of functionalism

then they have pre? sense, psychologically isomorphic, the same psychological however different they may be cisely properties, in those aspects of physical structure that are not relevant to establishing A functionalist the psychological does not say that the isomorphism.

proper? physical structure of an organism is irrelevant to its psychological are materialists, that is, they believe that all ties : nearly all functionalists for properties require a physical realization.12 This means psychological that for any organism X and any time t, X can have a 1-place property 0 at t only if X has at / some physical property psychological R(i/j) which realizes *p(in a sense soon to be made precise). But different instance physical different organisms properties. realizations properties can realize in the (or organisms of very different the same psychological in properties same organisms so at different times); structure can have the same psychological

It is because the functionalist allows the possibility of different - not because he of a realization up the requirement gives - that he achieves the goal of abstracting from the physical altogether structure of particular organisms.20 It will realization be in what follows important of a psychological property itself is. The following account to have a clear idea of what a is, and of what a psychological is David Lewis's,21 except for a in footnotes) which make no difference to apply at for instance

property few minor

(mentioned divergences to the argument at hand. is Suppose that <f> a psychological which

any time t to all organisms

theory that is intended are of type 2 at t. (<? might

MENTAL

REPRESENTATION

29

be a theory intended to apply to all adult humans, or to all organisms let capable of feeling pain, or to all rational beings.)218. For simplicity, us suppose that <f> finitely axiomatized; then we can represent it as a is single formula which I abbreviate as A{%, t), where x is a variable ranging over organisms and t a variable ranging over times. [Then the theory is true of all the organisms in its intended range if and only if the following claim holds: For any t and any x> if x is of type S at t then A(%, t).

that is properly speaking true or false, It is this last claim, rather than <j>, so you might prefer to use the term 'theory' for the last claim rather than omit of psychological for </>.]In giving crude formulations theories we often but they must be understood the variables (and initial quantifiers), as implicit: we say 'pain has such and such a causal role' when what we

really mean is 'For any t and any x of type 2 at t, pain has such and such a causal role in x at t'. If we do not write the theory in this way, we cannot properly define the notion of a realization.22 Suppose Tl9..., I will that Tn; assume then we in </>are primitives specifically psychological23 can write as A(Tl9..., For simplicity, Tn; x, t). <f> are all predicates.24 that Tx,...,Tn Let us say that an in Pn> of properties realizes <f> organism X at time t if and the

?-tuple <Pl5..., and Yn; x, t) is true of <Pl5.. .,Pn, X,t}; only if the formula A(Yl9..., that such an ?-tuple uniquely realizes <f> X at tif it and no other ?-tuple in realizes </> X at t. in

Now if </>is a psychological theory with n primitive psychological predicates we can use it to define n functional properties. Suppose for in? stance that Tj is a predicate that stands for a 1-place property of organisms, sense to say that the the kind of property (like pain) of which it makes organism has it at one time but this sort 'states',25 but I prefer If Tj is a predicate of this sort, with <f> the property ^ defined is (2) has 0 at t =df P such that X P is the jth X at t not at another. to reserve (Lewis calls properties of this term for a different use.) property associated

then the jth functional as follows: is some

there

1-place physical26

property

(i)

component

of a unique27

realization

of

in </>

30

HARTRY

H.

FIELD

(ii)

X has P at t

we If iffis the yth functional property associated with <f>, can then say that a realization ofif? in X at t is simply they'th component of a unique realiza? tion of <f> X at t. From this and (2) we derive in (2') (i) X has 0 at f if and only property P such that P realizes 0 in X at t, and if there is some 1-place physical

(ii)

X has P at t.
sense to the general remarks about pain claims is that

This machinery enables us to give a precise of five paragraphs back. What functionalism the property

of pain28 is a functional property associated with some theory 'physical' replaced by by (2) (or by the analog of (2) with the word <f)p - see note 'non-functional' in this way we can 26). By taking functionalism to in the general make precise sense of various vague notions appealed and we can also isomorphism), (e.g., the notion of psychological is true, then for an verify the really important claim that if materialism to have the psychological of pain it must have some property organism remarks physical Now assume property in the organism. property that realizes that psychological to the case of believing, where we let us apply this machinery as before that belief is a relation between an organism and a

relation associated with Belief, then, will be a functional proposition. some theory <f>b which the term 'believes' occurs, say as the yth psycho? in logical term. We cannot of course apply the schema (2) to define such a that functional property, since (2) made sense only in defining properties to 1-place predicates, correspond it to entities of another kind: (3) (i) (ii) No Xbears R ait X bears R to p at t of (2) to the case of 2-place thing to note about this functional is that properties but it is clear enough how to generalize

= df there is some 2-place physical ifftop at t R such that is the yth component of a unique realization

property in of </> X

other generalization is possible. Now the important

it existentially

MENTAL

REPRESENTATION

31

over physical If relations between people and propositions. quantifies sort between a person there is no physical relation of an appropriate to (3) the person cannot stand in the then according and a proposition, The functional functional relation ^ to the proposition. relation ^ is not itself a physical relation; but if i/j is to relate an organism to a proposi? tion at a time, there must be some physical relation Rty) which realizes and which related the organism to the proposition. Thus even if we take t/j relation, we have to solve Brentano's problem: we have to show that there could be physical relations (non-functional) That is what I meant when I said earlier between people and propositions. belief to be a functional does not either solve or dissolve Brentano's problem. The it says of relevance to that problem is something that probably only thing no one ever doubted anyway: that the physical relation that relates me now that functionalism to the propositions I believe can differ from the one that relates dogs to the propositions they believe; and even from the one that relates other people to the propositions they believe, and from the one that related me to the propositions I believed then. In other words: twenty years ago it is indeed legitimate to solve Brentano's problem in different ways for different organisms; but this does not remove the need for solving the prob? lem within an organism. I don't mean to downgrade the importance of the in the same problem of the way in different organisms : that fact is crucial to the plausibility of believing put forth in Section I, for no one could plausibly model claim that the relation of believing* is physically the same across different All I am saying is that to admit the allowability and importance species. of such variation across organisms does not in any way diminish the force of the tentative It remains relation argument for the model true that if no other model of believing put forth in Section I. can be given of what a physical observation that we don't need to solve Brentano's

would be like, then we should people and propositions tentatively accept the model there proposed. The argument I have just given is I think an extremely obvious one: one would expect it to be obvious to anyone who thought about func? tionalism for a moment. Yet I have found in conversation that the con? clusion of the argument is one that nearly all functionalists all of them are convinced that functionalism somehow oppose: nearly undermines the

between

argument of Section I. One reason, I suspect, is a tendency to slip from to an extreme behaviorism functionalism to which nothing according

32

HARTRY

H.

FIELD

Another

its psychological inside an organism is relevant to determining properties. reason is perhaps a Leibnizian tendency to regard relations as less real than 1-place properties - not out of conscious doctrine but merely because of the fact that the word 'relation', unlike the word 'property', doubles as a word for a certain kind of set. But there is a third source of error that I now want to consider. It is equally crude, but I think that there are deep and subtle confusions that lead people to implicitly make it. The third source of error is that even among people who explicitly the view that belief is a relation between people and proposi? advocate

tions, there seems to be a tendency to sometimes fall into the 'orthographic accident' view: the view that an adequate theory of belief could treat 'X 'X believes that either Russell was that Russell was hairless', believes hairless without and do is white', etc., as primitive 1-place predicates, 6X believes that p' entirely. (The fact that the 2-place predicate the term 'believes that' occurs in both 1-place predicates would then be, from a theoretical view, of no significance, a mere orthographic accident; likewise be an 'Russell was hairless' would that fact that both contain or snow

It is not easy to take such a view seriously. But accident.) orthographic the account would let us suppose it were true: what would follow? Well, and relation between the need for a physical obviate people clearly relation between since it didn't talk of a psychological propositions: people people course and propositions, and propositions does not refute it is clear would relation between that no physical in a realization. But this of be needed

that if you do construe of the belief relation.

is in this section, which the point I was making belief relationally, you need a physical realization

I think that it is an easy one to In spite of the crudity of this mistake, make implicitly. In fact, in the opening paragraph of this section, when I obviated the need of a system the view that functionalism tried to motivate talking in a way that strongly suggested is a state I said "a state of an organism of believing that p if that state plays the appropriate role in the organism's Now, for this to make any sense, the letter '/?'here must be psychology". a specific sentence - say 'Either Russell was as abbreviating understood role" of the hairless or snow is white'. Now, what is "the appropriate of representation, the orthographic I found myself accident view: that either Russell state of believing no sense can be made do not say that was hairless or snow is white? I of such talk: if we give a functional

MENTAL account of the relation

REPRESENTATION

33

of believing which holds between organisms and an account will certainly have implications about the such propositions, state of believing that Russell was hairless or the particular proposition snow is white. I think however state of believing can give a direct that talk of "the appropriate role" of the this particular proposition strongly suggests that we functional definition of this particular state. And that that the kind of procedure used for 'pain' can be applied was hairless or snow is white': in other

strongly suggests to 'believes that either Russell

words, it suggests that believing that either Russell was hairless or snow is white can be regarded as a functional 1-place property defined by schema That however is the orthographic accident view, for it presupposes (2). that the theory <f> from which believing that either Russell was hairless or is to be functionally defined contains a primitive term that this property. The moral is that if you want to avoid the represents accident view, you should not regard 'X believes that p0' orthographic for specific sentences p0 as functionally definable in the way that 'X is in pain' is: you should regard them as defined non-functionally from a relational predicate 'X believes the p' which is functionally defined by (3). And that means that we must invoke physical relations between organisms and sets of possible worlds. snow is white

I believe that there is a deeper source of the tendency to slip into the accident view; it has to do with functional theories that orthographic invoke psychological 'state' is not used in Lewis' sense states, where note 25) but in the sense of 'type of inner occurrence'. There is nothing (cf. wrong with such functional theories, but they must be treated with care, as I will now explain. The reader may however wish to skip the explana? tion and move directly to Section III (p. 36). Let us first consider the case of pain. Linguistic usage suggests the view: following (4) (i) (ii) Now X feels pain at t if and only o such that X feels o at t, and o is a pain. that we give a theory of <f>'p the sortal property being a pain if there is an internal occurrence

suppose

34 which was used to functionally

HARTRY

H.

FIELD

in clause (ii) of (4). It is easy to see that the proper way define being a pain from such a theory is this:

(5) (i) (ii)

o is a pain if and only if there of inner occurrences such that P is the jth component at t, and o has the property P.

is some physical realization

property

of a unique

in of <f>p X

(Here

t is the time of occurrence Putting

of o and X

is the organism

in which

occurred.)

this together with

(4), we get

(6) (i) (ii) (iii)

X feels pain at t if and only if there is an internal occurrence and a physical property P such that X feels o at t o has the property P, and of a unique P is the jth component
at t.

realization

in of <f>'p X

Now

of the word 'state' for physical properties if P is a state-type let us use it for state-types; (That is, that have P.) If we also then the state-tokens are the inner occurrences ' introduce the expression 'is inP at t for 'feels at t some internal occurrence of type P (i.e., with the property P)', then we can rewrite (6) as introduce if there is a physical

suppose we occurrences. inner

(7) (i) (ii)

X feels pain at t if and only that X P

att.

state P

such

is in P at t, and is the yth component

of a unique

realization

in of </>'p X

The

of this kind in X at problem of finding a realization of pain-theory /; physical t, then, is the problem of finding a physical state of pain in Zat state in the sense of state-type, that is, in the sense of property of inner

occurrences.

MENTAL Now analogue (4') (i) the same of (4) is X believes o such that o occurs that pattif thing done

REPRESENTATION here for pain can be done for belief.

35 The

and only if there is an inner occurrence

in X at t, and

(ii)

o is a belief thatp

to (5) above, defining the relation is a belief that, analogously Functionally and combining the result with (4'), we get (in analogy to (6) above) (6') X believes that p at tif and only if there is an internal occur? rence o and a physical relation R between internal occurrences and propositions such that in X at t, o bears R to p, and R is the yth component
at t.

(i) (ii) (iii)

o occurs

of a unique

realization

of <f>'b X in

So far so good: the result makes clear that in using a psychological theory of belief of this sort, we need to find physical realizations of a relation between internal occurrences and propositions. Now suppose we want to 'state' for a 1-place property of inner occurrences. to doing so: for each sentencep0, There is no objection there is a 1-place property of bearing R to p0, and we can call this a state (meaning of course a state-type). And we can go on to define what it is for an organism to be introduce the term in such a state: X at t some But now if we inner occurrence is in the state go at t if and only if there occurs in X of the type g0 (i.e., with the property g0). try to reformulate (6') in terms of state-types, we find that

we get the following: (7') (i) (ii) X believes go X that p0 at t if and only if there is a physical state

such that is in go at t, and the relational component of g0 is the jth component in unique realization of </>'b X at t of a

(go is the property of bearing R to the specific sentence p0; by the rela? tional component of g0 I mean the relation R.) This is the correct way to introduce talk of state-types into a functional theory of belief.

36 But,

HARTRY

H.

FIELD

I suspect, many people do not bother to think the matter through. as (7), they immediately for pain formulated jump Seeing functionalism to the conclusion that belief-theory in terms of states should formulated read as (7*) (i) (ii) X believes go X that p0 at t if and only if there is a physical state

such that is in go at t, and go is the yth component

att.

of a unique

realization

in of <f>b X

Formulated

in this way, <f>'b to be construed as a theory in which a term has for being a state of believing p0 (for the specific sentence p0) appears as a accident view. primitive. And that of course is the orthographic The moral of all this is that the view that functionalism either solves the confusion results from confusion: problem of belief with an orthographic-accident theory theory of a relation at all. I am sure that any belief according to which belief is not theories functionalist would emphatically reject the idea that psychological Brentano's of a relational of belief of the orthographic-accident format; but if one thinks on the as having functional roles not derivative of believing one is implicitly adhering to functional role of the belief relation, then an orthographic accident account. should have states III. DISPENSING WITH PROPOSITIONS or dissolves

At

the beginning of Section suggested that a functional

III quoted an argument of Stalnaker's which theory of belief would obviate the need for a

I remarked that in order to fully evaluate system of internal representation. it was necessary to get clear on what a functional account the argument, sense: I of belief would be like. I have done that only in a very minimal have looked only at those features of a functional account of belief which follow more from the nature of functionalism involves a relation between in getting be like. For we have than this is involved that belief together with the (supposed) fact But much people and propositions. a functional account

clear about what

of belief would

of belief, the relation of believing a certain theory </>b. Part of what's

seen that on any functional account is a functional relation associated with involved in getting clear on what a

MENTAL REPRESENTATION functional

37

account of belief would be like, then, is getting clear on what would be like. I am strongly inclined to think that any adequate (f>b <f>b and that this would have to explicitly postulate a system of representation, inde? provides a reason to believe the internal representation hypothesis considerations about Brentano's problem. I will not how? pendently of any ever try to argue this claim here. Instead I will merely note a converse: that since (as I've argued) we appear to need a system of representation in order to solve Brentano's problem, we shouldn't have too much reluctance such a system into our psychological about explicitly incorporating theory if doing so seems helpful. </>b a system of representation One advantage of explicitly incorporating into one's psychological theory is that it enables us to obtain most of the of regarding belief and desire as relations between people and the attendant without that for some liabilities. Suppose propositions, reason or other we do not want to quantify over propositions; in that advantages case, then apparently (8) There we will be unable to say things about like him, and none of

are many things she believes them are at all complimentary

or (9) No one can perceive an object things about it; without coming to believe

various

for apparently the 'things' quantified I know of no reason why However wouldn't be equally well rather than of belief: e.g. to believe* various things about

over in (8) and (9) are propositions. our purposes in uttering (8) and (9) over objects of belief* served by quantifying 'No one can perceive an object without coming

it' ; here the 'things' quantified over are not propositions, but sentences in an internal system of representation. Of course, we could accept this reformulated version of (9) in our psycho? the inner representation theory only if we incorporated hypothesis into the theory. explicitly I think that by now the asterisks have become tiresome: so let us intro? logical duce the terminological convention that the word 'believes' is to be used that I have heretofore used 'believes*'. On this way of talking, of belief are sentences or sentence-analogs, and these sentences or sentence-analogs have content or meaning. Contrary to the suggestion in the way the objects

38

HARTRY

H.

FIELD

in the first paragraph of Section I, this way of talking does not really remove Brentano's that problem rearises as the problem of problem: account of having content. Unless such an account giving a materialistic of content can be given, much of what we say about belief (e.g. that cer? tain beliefs are about Caesar, that certain beliefs are true, and so forth) no sense at all. I have however suggested, at the end of Section I, account seems of giving a materialistic of content the problem one way to manage it is to give a Tarski-like account of manageable: a theory of reference. truth, supplemented by makes that 'believes*' does not accord very My use of the word 'believes' to mean in English 'believes' is well with the use of the term 'believes' in English: with 'believes that', and we can say even of synonymous pretty much that they believe that snow is white. who know no English organisms us then introduce a new technical term, 'believes that', which will Let serve the purposes that 'believe' and 'believe that' serve in English: let an English us say that a person 'believes that p' (where '/?' abbreviates some sentence in his system of internal sentence S) if that person believes representation whose translation into English is S. In effect then 'believes organisms and English sentences. of translation employed in the definition of 'believes that' is a loose and sloppy one. For even those of us who are thesis are bound to recognize that hostile to Quine's radical indeterminacy translation between languages of very different structure or expressive between It must be realized that the notion power is highly indeterminate. (Part of the reason is that a sentence in as any sentence in not have exactly the same meaning language Lx need language L2, so that a translator of Lx into L2 has to settle for approximate of meaning; and which that' is to be used for a relation

one picks depends on approximation Since the notion of believing that complicated pragmatic considerations.) is is to be applied to organisms whose system of internal representation and different from ours doubtless e.g., chimpanzees, Martians, quite is right) humans whose spoken and written language differs (if Whorff - then it would be absurd not to recognize that the significantly from ours involved in the definition of 'believes that' is highly indeter? translation sameness minate. For this reason the notion of 'believing that' defined above is itself a highly indeterminate it is a notion that we notion, and consequently theorizing.29 Instead, we should use ought to avoid in our psychological

MENTAL the notion semantic avoiding theories I have

REPRESENTATION

39

together perhaps with various believed. (Another reason for use of 'believing that' (as defined above) in our psychological in Section I - cf. (B) on p. 20.) has already been mentioned introduced conventions about the use of 'believes' and 'believes of believing (i.e., believing*), notions applied to the sentences in to of of

to sentences to which the first term relates organisms that' according and the second relates organisms their own system of representation in English. These terminological conventions to rule out the possibility that we need propositions belief and desire: it may be that we need propositions meaning however for the sentences is that are believed sentences are not in the analysis in our account

intended

and desired. My

own view

(a)

that talk of propositions is best avoided (except as a dispensable manner of speaking) unless it can be shown to serve purposes that cannot condition be served otherwise; ismet us to semantic theses which (indepen? scruples) may well be false. and it is doubtful that this

(b)

that such talk commits dently of ontological

take (a) first. That one should not posit entities needlessly is, I : to do so would be to indulge in "unfounded it, uncontroversial speculation". And that there is no particular need to intro? ontological - at duce propositions construed least, no need to introduce propositions - was in terms of possible worlds implicit in my discussion at the end of Let's take I argued that if one wanted to use the idea of possible as a set of possible worlds, if one construed a proposition worlds, then the correlation of a particular proposition with a sentence posed no Section I. There and problem as long as we had an adequate truth-theoretic semantics for the Another way of putting this point is that talk of propositions language. adds nothing of semantic that is semantically of interest; everything interest is already there in the truth-theoretic semantics. (I'm speaking of course of a truth-theoretic semantics that assigns properties rather than related sets to predicates.) In other words, instead of saying that a person is to the set of possible worlds in which Russell was hairless (or to some fine-grained out of possible worlds), why constructed proposition

40 not

HARTRY

H.

FIELD of a name stands for

say instead that he is related to a sentence that consists a predicate stands for Russell with that copulated hairlessness ?29a that

One might I try to respond to this question by holding (implausibly, But this response, while that names are non-rigid think) designators. to the question just asked, does not undercut the point I am adequate trying to make (as has in effect been observed already, in note 16). For is non-rigid, then to believe that Russell was hairless is to some sentence that contains a definite description, some sentence an individuating of in which iF' is of the form 'G^xFx)' description of this Russell and 'G' stands for hairlessness. A truth-theoretic analysis if 'Russell' believe sentence contains all the semantic so why information that is contained in a they possible world analysis; seem to be excess ontological Under bring in possible worlds? Again, serving no semantic role.

baggage (b) there are two points to be made. The first is that if one talks must be explicable and also assumes that propositions of propositions that then one will be driven to postulate in terms of possible worlds, are non-rigid terms like 'temperature' and theoretical at best. The second point is controversial and this postulate designators; talk is that whether or not one adheres to the possible worlds approach, names and it commits one to a general relation of synonymy, of propositions I will relation. that there is any such general is not at all obvious these points, the second one, in Section VI. Their especially develop

proper

upshot

involves not only unfounded is that the postulate of propositions semantic but highly controversial speculation speculation ontological as well. To me both and the semantic claims that the ontological

underlie

seem far more dubious than the of propositions the postulate since this of a system of internal representation, especially postulate last postulate appears to be a necessary one for the solution of Brentano's problem. Let us say then that belief tions, but toward meaningful and desire sentences toward proposi? in a system of internal representa? for a system account of meaning semantics; I will as well. suggest in But first representa? are not attitudes

tion. Presumably part of any adequate is a truth-theoretic internal representation of Section I would
tion.

VI

that

there

is another

like to try to clarify

aspect of meaning the idea of a system of internal

MENTAL

REPRESENTATION HYPOTHESIS

41

IV. REMARKS ON THE INNER REPRESENTATION I have

sentation;

spoken of belief and desire as involving a system of internal repre? that have a but I have allowed that in the case of organisms

genuine language, the system of internal representation might either be the language or include the language as a part.30 This combination of claims may seem puzzling: if the only representation is in natural language, If there is internal representation, what why call it internal representation? sense does itmake to say that the representation is in a natural language? The answer to these questions lies in the distinction between types and I have talked of an organism as believing (i.e., believing*) sentence But I said in Section I that (neglecting the complication about core types. if and only if he employs that beliefs) a person believes a sentence-type in an appropriate way in reasoning, deliberating, etc. ; and sentence-type tokens. the only way to employ a sentence-type is to employ some of its tokens. it is clear, that in order to believe (or core believe) a sentence I Now, can't be required to employ spoken or written tokens of it: no one writes down all of his beliefs sentences of those natural if I believe (or all of his core beliefs). Consequently, of my language, what I employ has to include internal tokens sentences. That explains why, even if all representation is in

some of languages, we have to speak of internal representation: are certainly internal. the tokens This brings us to the second question: does it make sense to speak of as representation internal representation in natural language! It makes sense if, and only if, it makes sense to speak of internal tokens as being

of the same type as spoken or written tokens. One might argue that any claim of type-identity between internal tokens and spoken or written tokens is highly implausible: after all, it seems pretty absurd to suppose that there is much the one hand course would and of a physical resemblance between internal tokens on tokens on the other. But this of spoken or written

be a very bad argument: after all, a spoken token can be said to be of the same type as a written token, and yet spoken and written tokens bear little physical resemblance to each other. My own view is that the issue of whether we speak of internal tokens as is partly a verbal issue, but that that underlie it, of which in which) the linguistic

to spoken or written tokens type-identical there are interesting empirical questions most important is the extent to which

(and the manner

42

HARTRY

H.

FIELD

involves conceptual To put the point very development development. if (as seems to me quite likely) learning a first language involves vaguely: an initial representational extending system to include an isomorphic copy of the language being learned, then I think it is quite natural to view as establishing a criterion of type-identity between isomorphism internal tokens and spoken or written tokens. However, the issues here are pretty complicated. Harman for an interesting discussion (See [10] the which more or less supports the pro-type-identity position, and Fodor [7] in support of the other side.) My (Ch. 2) for an interesting discussion than with Fodor on this matter, sympathies are much more with Harman

but I will

try to remain neutral on the question in the rest of the paper.31 But there are other questions that need to be considered. First, I have talked of 'internal tokens'. What kinds of entities are these? Presumably

they are inner occurrences of some kind, but what kind ? Second, although I have begged off the question of what it is for an internal token to be of the same type as a spoken or written token, there is also the question of what it is for two internal tokens to be of the same type as each other; and since a psychological theory will clearly need to use the notion of between different tokens, this question cannot very well be type-identity internal Third, there is the question of syntactically characterizing ignored. tokens - what is it for an internal token to belong to a given syntactic category (e.g., what is it for it to be a sentence token) ?Again, we will need to appeal to syntactic characterizations of inner tokens

in developing the is a fairly pressing one. theory, so this question psychological There are two possible strategies in dealing with such questions. The first strategy, which I think is the wrong one, is to try to answer the above the psychological theory. Such a strategy questions prior to developing it is hard to see how to carry it out without doing a deal of neurophysiological e.g., we would apparently speculation: great have to specify two neurophysiological properties Px and P2 and a neuro? relation R, and say that an inner occurrence is an expression physiological is a bad one because if in addition it has if and only if it has P1 and is a sentence-token two inner occurrences with property Px are of the same type P2, and that if one bears R to the other. The task of specifying Pl9 P2 and R is certainly and I not a task we are equipped for in our current state of knowledge, token don't ment see how we could ever become of psychological theory. equipped for it prior to the develop?

MENTAL

REPRESENTATION

43

The second strategy, which is the one I advocate, is to develop the syntax as part of the psychological of the system of representations theory: we can then use the psychological answer to the theory to give a functionalist raised above. That this is the right way to proceed seems com? questions pletely obvious : it is simply an instance of the general rule that psycho? logical theories ought to be construed functionally. Let me be a bit more explicit about what is involved in this functionalist If one were to write out in detail a theory (f>that postulates a approach. as 'jc is an expression such predicates system of inner representation, and 'x and y are expression-tokens of the token', 'x is a sentence-token', or be explicitly definable same type' would either appear as primitives (within set theory or higher order logic) from other such syntactic primi? tives. Such syntactic primitives are to be included among the psychological is primitives when we functionalize' (/>: that is, a realization of <f> to be an of properties and relations corresponding to the syn? ?-tuple consisting tactic predicates psychological the answers of the theory If we predicates. as well as to the more 'functionalize' straightforwardly the theory in this way then can be read right off of the

that 'x and y are tokens theory of the same type' is a primitive of </>. we want to know what it is for two If in an organism X at a time t to be tokens of the same internal occurrences type, the answer is simple : c and d are tokens of the same type if and only if there is a physical relation R which is the appropriate of a component unique tokens in realization of <f> X at t, and c bears R to d.33,3? (If 'x and y are of the same type' is a defined term rather than a primitive, the answer is slightly more complicated; but again, it can be read right off of the theory.) The upshot is that there is no need in developing psychological theory to specify what R is; we can leave that to future neurophysiology. (Moreover, we can allow that there are different physical type-identity to organisms.) The fact that we leave the question future neurophysiology does not imply any unclarity in our theory: in some sense, the theory implicitly specifies what it isfor two inner tokens in a system of representation to be tokens of the same type. relations in different is one final point to be made about theories that postulate a of internal representation. I have said that the syntax of a system system of internal representation should be explicitly stated in a psychological theory of belief and desire. Should the semantics of the system of internal There

to the questions raised above For instance, suppose for simplicity <f>.32

44

HARTRY H. FIELD

also be stated as part of the psychological representation theory? That on what we want psychological for. If the task of psychol? theory depends ogy is to state (i) (ii) the laws by which an organism's beliefs and desires evolve as he is subjected to sensory stimulations, and the laws by which those beliefs and desires affect his bodily
movements,

then I think

that it is clear we do not need

to use

the semantics

of the

in stating the psychological laws: the sentences system of representation as in the system of internal representation might as well be meaningless is concerned.35 This is not the only way to view a far as the psychology - a broader conception of a belief-desire psychology theory psychological will be suggested in the next section, and in it semantic notions would play a genuine role. But it isworth stressing the narrow kind of psychology For we have seen that the syntax and type-identity at least momentarily. should be regarded for a system of internal representation conditions as functionally characterized by a psychological theory in which they that is, the appear; and we can take that theory to be narrow psychology, that does not employ any semantic characterizations kind of psychology This is important, for it of the sentences in a system of representation. means of type-identity for the system of that the syntax and conditions representation considerations could about what be determined in principle independently the sentences in the system mean.
V. TRUTH

of any

What

a theory of meaning for a system of internal representation I and III I have hinted at one aspect of my views on be like? In Sections for a system of internal representation this question: a theory of meaning semantics of a more-or-less in part of a truth-theoretic must consist would Tarskian kind. If we do not give a theory of truth for the system of internal we cannot make sense of the idea that some of our beliefs representation, are true and others

are false; and I think we do want to be able to make the only sense of this idea (for reasons to be sketched shortly). Moreover, I have ever heard of which is not obviously kind of theory of truth that deficient is the Tarskian kind. The upshot, then, is that we need to give a Tarski-type semantic theory for the system of internal representation. In a

MENTAL recent article Harman

... no

REPRESENTATION in most

45 Gilbert

to which

I am this:

respects

very

sympathetic,

has questioned
reason has been of representation

system

for given we think

a compositional in, be it Mentalese

of meaning theory or English ([12],

for whatever p. 286).

if the notion of truth makes any sense, then truth however Presumably must be related in the following way: the truth of the and meaning sentence 'Caesar crossed the Rubicon' should follow from the meaning together with the fact that Caesar crossed the Rubicon. sense a theory of meaning must include a theory of truth-condi? In this has to be a tions. And as far as I can see, the theory of truth-conditions of the sentence theory of roughly the type that Tarski made famous. compositional Harman's semantics is based on an important critique of compositional insight. He points out in the article that there is a serious problem for those like Davidson, who philosophers, as what is essential in semantics :
Davidson The would (presumably) say that

regard knowledge

of truth-conditions

the speaker

understands

[the sentence

'Snow

is white'] by virtue of the fact that he knows it is true if and only if snow is white.
... is that [for the speaker to know any such thing he] needs some way difficulty to represent to himself uses If the relevant snow's the words being white. speaker to represent in the relevant that snow 'snow is white' is white,... Davidson's way a form of Mentalese if speakers have available be circular. in And, [theory] would they can represent is still the problem that snow of meaning so that the [theory is white, for Mentalese (p. 286). avoids] circularity,

which there

what

to be drawn is that knowledge of truth-conditions is not is important to the semantics of a system of internal representation. The theory must ascribe truth-conditions, not knowledge of truth-conditions, to the sentences of English or Mentalese; for if it doesn't ascribe truth But the moral conditions beliefs to these sentences it will not have given sense to talk of our as being true or false. I have assumed that we do want

to make sense of the idea that some of our beliefs are true and others are false; I would also assume that we want to make sense of the idea that some of our beliefs are about Julius Caesar and that other of our beliefs are about quarks. It seems to me however that there is a serious question as to why we should want to make sense of these ideas. Is our desire to do so based on anything other than a naive metaphysics that has no place in a properly scientific account of the world?

46

HARTRY

H.

FIELD

That this question is not a silly one can be seen from a fact noted end of the previous section. If the task of psychology is to state (i) (ii)

at the

the laws by which an organism's beliefs and desires evolve as he is subjected to sensory stimulations, and the laws by which those beliefs and desires affect his bodily
movements,

then

semantic

characterizations and desired

of beliefs

and desires

are

irrelevant

to

psychology: what the believed

saying anything at all about sentences mean, or what their truth-condi? tions are or what their subject matter is. For instance, we might imagine a that contained laws like the following: super-crude psychology '->' in the system of internal repre? sentation such that for all sentences Sx and S2 in the system, a person believes whenever rSx -> SJ and desires S2 then he also desires S?. there is some connective

one can state the laws without

this law might mean 'only if, that is, it the conditional; the fact that it obeys but might obey the truth-table for this truth-table is not something we need to say in stating the psychological it to be super-crude, since we are imagining laws.35a The psychology, The connective '->' that satisfied might also contain laws like this :

there is a privileged class of sentences in the system of repre? called the class of observation sentences, with the that each sentence in the class has associated with it property sentation, a particular stimulation believes Intuitively, associated a sensory type of sensory stimulation. Whenever the organism of the appropriate type occurs, sentence. the observation that sentence is if a particular observation of retinal stimulations that are

we might expect in this way with

the class

characteristically means something the psychological to semantically

sentence caused by nearby rabbits, then the observation like 'there are rabbits nearby' ; but even if this is true, theory need not say that it is true. Why then do we need characterize the sentences

in the system of inner repre? sentation ? Why not simply say that belief and desire are relations between sentences? Saying this would preclude us from people and meaningless

MENTAL of beliefs

REPRESENTATION

47

as being true or as being about rabbits, but would anything of scientific value be lost? Here is one answer to this question - not the whole story, I think, but the part of the story that is emphasized in the writings of such philosophers as Quine, Davidson, and Harman (and perhaps Lewis in [18]). Imagine speaking with a foreigner whose language we do not under? stand. A rabbit scurries by, in the foreigner's line of sight, and the foreigner raises his gun in its direction - by now we have rather overwhelming grounds for thinking that he believes that there is a rabbit nearby. But of the narrow psychology considered say so in the vocabulary heretofore? We can't say 'He believes the sentence...' (giving the name of the sentence): for we don't know his language. How about if we say instead associated evidence 'He believes some sentence which is an observation sentence with sensory stimulations serves as of type... (and which for syntactically related sentences according to laws..., and so The difficulty is clear: only someone with a great deal of very forth)'. detailed information about the psychology of our foreigner could fill in the blanks. A He third possibility believes some is much better: can we that we find ourselves

sentence of his language which plays ap? the role in his psychology that the sentence proximately 'There's a rabbit nearby' plays in mine. a semantic notion !For isn't

But this, it might be claimed, really involves it really just a long-winded way of saying He believes my And language some sentence

of his language that translates as 'There's a rabbit nearby' ?

into

a semantic notion? answer does not satisfy me, for though it does definitely motivate the introduction of a notion of translation, it does not motivate the intro? duction of any non-translational semantic notions; that is, itmotivates the isn't translation This

introduction of a more-or-less36 semantic notion about the relation of one language to another, but it does not motivate the introduction of any or 'refers' which relate language to the world. semantic notions like 'true' No reason has been offered, it seems to me, for regarding another person as having beliefs that are true or about rabbits. The Quinean reply, I would is that we need such notions as truth in connection with imagine,

48 our own

HARTRY

H.

FIELD

like 'Every sentence of language (e.g., to state generalizations the form "/? or not p" is true');37 and that we can then use the notion of on other grounds to carry over the that we have motivated translation truth concept to foreign languages. I do not want to discuss this Quinean

reply here; but it seems to me rather weak, and I would like to do better. I think that the reason why we need to be able to apply semantic notions like truth and reference to the sentences that people believe and desire is the theory that people's beliefs are, in many circumstances, reliable indicators about the world; and the only way to state this theory is to use the notion of truth (and probably the notion of reference as well). that could this theory is not a piece of gratuitous metaphysics Moreover, it is central to our getting information about the easily be dispensed with: about other people's for we are constantly world, using our opinions beliefs in forming opinions about the world. The fact that a child believes like (a fact that can often be inferred I to think he has done something from his behavior) gives good won't like; the fact that most physicists believe that there are gravitational is waves (a fact that can be inferred from reading a few physics books) that he has done something I won't reason good reason forth. These that there are gravitational waves ; and so inferences evidently proceed by means of certain reliability a person's beliefs that say under what conditions principles for me to believe that we hold

principles, about certain

things are likely to be true. The principles we need are not : after all, the fact that a child believes in Santa Claus is not easy to state good reason for me to believe in Santa Claus, and the fact that most of a certain religious cult believe that flying saucers will land on next month is not good enough reason for me farm in Arizona saucers will land there. (I don't think that the failure to believe that flying of these inferences is due entirely to independent evidence as to the falsity of the conclusion.) We do evidently have a stock of reliability principles, a certain

members

one can imagine that though we can not explicitly formulate them; and into an explicit theory. In a suitably they will someday be systematized even regard this 'reliability theory' as part of broad sense, we might psychology. What such a reliability is that it would have guess notion people look like I do not know. My theory would to include not only the notion of truth but the of reference: for we want to be able to say in the theory that some have very reliable beliefs about physics but very unreliable beliefs

MENTAL about

REPRESENTATION

49

the state of the economy, and so forth. We might imagine then that - and if we 'true', 'refers', etc. are primitives of the theory. If we do that theory is elaborated with such detail and imagine that the reliability

precision

that it is uniquely realized in each of the organisms to which the - then we could use this theory to give a theory is intended to apply functional account of truth and reference for systems of internal repre? sentation. Such a functional account the reason accounts

would of course be desirable, for are always desirable: that functional it would allow that the reference relation for the possibility is realized by different physical relations in different organisms.38 VI. MORE ON THE SEMANTICS OF INTERNAL REPRESENTATION Is there more to the semantics of a system of internal representation than I think that there is; and I think is given by truth-theoretic semantics? that this casts considerable doubt on the possible worlds analysis of

even independently of ontological In fact, considerations. propositions, on I think it casts some doubt (independent of ontological considerations) whether any notion of proposition is possible. In explaining these matters, I will begin by discussing the semantics of spoken and written languages, since I want to discuss some points that Quine has made in the context of them; but what I say will carry over to systems of internal representation. Many
A and

years ago Quine made

be or... concerned he may

the following
with synonymy be concerned how of

observation:
between

may lexicographer in another forms

the same sumed

It is an open question language. a single general under formulation

in one forms language between forms in synonymy the two cases can be sub? satisfactorily the synonymy concept... ([21] p. 56). with

The point

that Quine is making here is very relevant to the question of one is assuming a positive if one postulates propositions: propositions answer to Quine's For if sentences mean propositions, 'open question'. the same it would

two sentences are synonymous then apparently if they mean and this would be a general concept of synonymy, proposition; both intra-linguistically and inter-linguistically. apply Is Quine's intralinguistic synonymy. interspeaker

'open question' really open? Well, at least this much is true: seems a lot easier to define than interlinguistic synonymy (Similarly, synonymy, intraspeaker suggesting is easier to define than synonymy that there might be still further

50

HARTRY

H.

FIELD

in the synonymy concept.) That divergence easier to define than interlinguistic synonymy occasion

is synonymy intralinguistic at least for what Quine calls

sentences is a point that Quine argues in Sections 9 and 11 of : he points out that the difference in meaning 'Everest' and between [22] 'Gaurisanker' sentences) for a certain speaker is (construed as one-word revealed by the fact that different sensory stimulations would prompt him to assent

of to one than to the other; and that the difference in meaning is revealed by sentences in a given linguistic community these one-word for the fact that these sentences are intra-subjectively non-synonymous most members 'prompting it seems to me that this talk of of that community. Now, too behavioristic, into and leads Quine to assent' is much of second intention" worries, e.g., about "stimulations like "Assent

unnecessary (verbal stimulations

sentences that begin with to one-word - cf. it 'E' or I'll beat your brains out") [22] pp. 48-49. Nevertheless, seems to me that Quine's general point is correct: we can explain intra? I have de? of meaning by evidential considerations. linguistic differences this point

veloped

elsewhere,39 using a non-behavioristic (but, I admit, of evidence ; and in addition to solving such hum? idealized) conception drum difficulties with Quine's approach as the second-intention problem, it also obviates

the need for restricting the account to occasion sentences. I think is that it gives you dif? What is interesting about this approach ferences of meaning where you would intuitively expect them, but where you do not get them on the possible worlds approach about non-rigid designation). implausible assumptions following pair of sentences sible worlds approach: (10) comes (11) and (12) comes (13) But Temperature out equivalent Temperature is mean molecular to40 is temperature. clear, even without looking up the details of my energy Everest out equivalent Everest come out equivalent (without adopting For instance, the on the pos?

in meaning

is Gaurisanker to is Everest ;

it should be pretty

MENTAL

REPRESENTATION

51

account, that these equivalences will not hold on any sufficiently sophisti? even for speakers cated evidential criterion of intraspeaker synonymy, who believe the sentences (10) and (12). Such evidential tions natural account considerations if the latter are not (coupled with seem redundant41) truth-theoretic to me considera? a very to provide

however

of intra-speaker synonymy synonymy. For inter-speaker the situation is quite different, for here it is very hard to formu? late any evidential criteria for two words differing in meaning. The source of the difficulty is clear: you and I may disagree about what counts as some? for a certain sentence, not because that sentence means in different to you than it means to me, but because of differences thing our beliefs. One might to find some rule of the form 'If the rest of try there is an evidential difference of such and such a kind between your

evidence

sentence and mine, then they differ inmeaning' ; this is the task that Quine to strip away the effects of collateral refers to as "trying information". Quine has cast considerable doubts on the possibility of carrying out this task ([22], Section 9), and I think that if you look at the question in terms of my formal model in [6] you will find that Quine's doubts are reinforced. It seems to me that the criteria of inter-linguistic synonymy we actually the criteria provided by truth-theoretic are, almost exclusively, employ semantics.42 This

is not quite true; when there are two sentences Sx and S2 in one language that are equivalent from the point of view of truth and theoretic (or possible world) semantics, but which differ evidentially, when there is a sentence S3 in another language that is truth-theoretically to Si equivalent to both S? and S2 but is much more similar evidentially to S2, then we regard it as definitely a mistake to translate S3 as than - the translation Sx seems required. But except in such cases, little if S2 sameness of evidential role is required; if the Martians any inter-subjective have i.e., Gaurisanker, by their powerful singled out Everest, telescopes and have named it 'Schrdlu', we would translate their name 'Schrdlu' by 'Everest' or by 'Gaurisanker' indifferently, however much we might want to say that 'Everest' and 'Gaurisanker' differ inmeaning for us. One might

object to this argument by saying that translation is a loose and pragmatic notion: the true situation (one might say) is that 'Schrdlu' differs in mean? and from 'Gaurisanker' ;we translate the name 'Everest' or by 'Gaurisanker' because these approximate indifferently by 'Schrdlu' about equally well. I sympathize with this response, except for ing both from 'Everest'

52 one

HARTRY

H.

FIELD

thing: it assumes that there is some clear notion of inter-linguistic in meaning between words that refer to the same thing, and difference that is what I think needs to be established. semantics - i.e., possible world My view then is that truth-theoretic - is the possible worlds almost enough, but that there semantics without These are certain very fine-grained distinctions of meaning that it cannot explain. seem to be clearly drawable of meaning distinctions fine-grained and that fact appears to make the notion of only intra-linguistically, suspect. proposition It should be clear that these remarks about

synonymy between sentences in public languages apply also to synonymy between sentences in systems It should also be clear that to define the evi? of internal representation. dential differences between sentences in the same system of internal repre? one does not need to employ any concepts that go outside sentation, psychology in the narrow in motivating special problem siderations into a semantic sense. Consequently the introduction there of seems to be no con? such evidential

theory for the system of inner representation, the in a way that there did seem to be a special problem of motivating introduction of truth-theoretic concepts. I want to conclude this paper by raising what I regard as an open ques? are there between a theory of meaning for a tion: what connections for a spoken and a theory of meaning system of internal representation to the to one influential approach44 or written language?43 According the meaning of a for spoken and written of meaning languages, theory in such languages is to be explained in terms of beliefs (or desires, to correlated with the sentence. Roughly, etc.) that are conventionally is to know that this sen? of 'Caesar was egotistical' know the meaning correlated with the belief that Caesar was ego? tence is conventionally sentence tistical; and similarly for every other sentence in one's public language. correlation must be spelled out recursively, of course.) (The conventional that one can explain what it is to believe that This approach presupposes Caesar was egotistical without relying at any point on the semantic features in one's spoken or written language: of the sentence 'Caesar was egotistical' for if one relied on the semantic features of the spoken or written sentence of the sentence in one's account of the belief, then to explain the meaning in terms of the belief would is whether that presupposition involve a circularity. So a crucial question is correct.

MENTAL

REPRESENTATION

53 it is this: system of relying in

in terms of internal representation, Putting the presupposition that one can explain what it is for a sentence in the internal to mean that Caesar was egotistical without representation

one's explanation on the fact that certain words and phrases in the public language stand for Caesar or that certain other words and phrases in the public If this pre? language stand for the property of being egotistical. is correct, then the above approach to the theory of meaning is supposition about the if worked out, it would reduce all questions attractive: quite semantics of the spoken or written language to the corresponding ques? tions about those the semantics could questions I am inclined of the system of internal representation; and then be answered without further reference to to doubt however

that the presupposition is true. My guess is that in a typical case, part o/what makes a symbol inmy system a symbol that stands for Caesar of internal representation is that this as a result of my symbol acquired its role in my system of representation language. acquisition of a name of this sort that stands is true, for Caesar something for a public language. I believe however approach to a theory of meaning that the issues here are quite complicated, and deserve a great deal of further study. An it would in the public language. If appear to defeat45 the above

to the theory of meaning would be to try to opposite approach to the reduce the semantics of the system of internal representation semantics for public language. To do this would be to assume that al? important these systems of representation behavior, notions like meaning and truth applied: such semantic notions could be to the representational applicable systems of organisms with a only or written seems very im? To me, such a conclusion spoken language. plausible. It may nevertheless be that there is something to this approach, can be explained more be that certain aspects of meaning for public language than for systems of internal representation. directly can be explained The suggestion is not that these aspects of meaning for it may without reference to beliefs and desires, for two reasons. In the first place, in explaining the fact that the public word 'Caesar' referred to Caesar one could appeal freely to certain sorts of beliefs and desires, e.g., those beliefs and desires that a languageless for we organism could possess; though dogs, chimpanzees, sentation which played an etc. might have role systems of internal repre? in the explanation of their would not be ones to which

54 have that

HARTRY such beliefs

H.

FIELD can be accounted is an for in a

granted

and desires

way. In the second place, there language-independent important sense in which one could appeal to other beliefs and desires as well in our semantic theory for public language: we could appeal to beliefs and desires construed only restriction sentences we appealed I do not want as attitudes towards internal sentences all we liked; the would be on which semantic features of those internal to.

to draw any very definite conclusions from this last dis? to develop cussion. I merely want to say that it may well be necessary the semantic theory for internal languages and the semantic theory for one kind of semantics public languages together rather than developing independently of the other and reducing the other kind to it. The reduction? ist strategy (particularly the strategy that tries to reduce public semantics to internal semantics) is worth pursuing, but alternative strategies are worth pursuing as well.

University

of Southern

California

NOTES * I on these Janet Levin, Brian have from discussion issues with benefitted greatly The criticism of Schiffer. latter two especially, and Stephen by their relentless Loar, a very much to write I expressed ideas in conversation, have forced me the way my otherwise have written. than I would better paper to Ned Block, Keith Donellan and I am also grateful Devitt, Tyler B?rge, Michael on an earlier version. comments Hills for helpful David 1 See for instance Armstrong [1], Ch. 3, and Putnam [20]. 2 to Brentano 11 ; references in [2], Ch. is discussed Brentano's by Chisholm problem can be found there. 3 are scattered on functionalism See Stalnaker remarks throughout [25]. Lewis'

various writings; see [14], [15], [17], [18], and [19]. 4 See Quine [22], Section 45.
an can believe true: a foreign is thus different from believing speaker Believing* no idea what it means, true while sentence e.g., if he sees a headline having English a sentence that S is true) is, if (1) S true (i.e., to believe in the N. Y. Times. To believe true involves a sentence that S is true. Believing S* which means is right, to believe* the concept of truth; 6 The Banach-Tarski into put (For a finite back number believing* theorem does says not. be decomposed can then be 'pieces' one. size as the original can that a solid sphere roughly in such a way that the 'pieces' two (and solid a proof) the same spheres see [13], pp. 3-6.) 5

together a more precise

of disjoint to form again statement

MENTAL REPRESENTATION
7 For I'm of von Neumann set theory

55
including the axiom of

pedants:

speaking

to regard the introduction of the notion of core-believing stored-sentence from the absurdity of requir? approaches sentences to be stored: but the following will show I discussion ing infinitely many can be given think that the notion motivation. strong independent 10 To spell this out more role of fully, we'd need first to spell out the characteristic to say desires) beliefs in reasoning, and so on call this (as opposed deliberating, to save characteristic role Role B. Then the materialistic view of believing* that I am suggest? that in each believer there is some kind of physical of sentences ing is roughly storage such that the things stored in that way and their obvious are the things consequences that have Role B. 11 to be needed The 'almost' here because of the non-transitivity of 'is an appears I doubt obvious that there are any clear cases of believing of. that consequence either hairless that n hairs to assert numbers, to assert Russell then there was snow hairless is white; clear cases no is white but not believing was not that if Russell I regard or less analogous this as more to the fact n such that some of numbers with individuals less than n or greater with individual I would hairs is bald. not want but are no lead of such numbers, there are no such I would not want paradox. Similarly, are no clear cases of believing an ob? p and believing for that would lead by q, there are no such cases; to a Sorites that of one can't believe/? and yet not believe con? clear cases or snow

9 [4], p. 410. Dennett as an ad hoc device

choice; it is finitely axiomatized. 8 [16], pp. 182ff.

seems

are no

are bald that that because

but

because because

there there

that would

viously equivalent proposition to the conclusion Sorites-like reasoning the unobviously claim r. equivalent 12 The acceptance of indeterminacies to an instrumentalistic cession view

the kind

mentioned

here

is a minor

of ordinary belief talk, but only a minor one; for even if it does not always make sense a belief-ascription to ask whether is true or false, we have seen no reason to deny that a core-belief is always true ascription or false; and that is close enough. concession minor to an instrumentalistic (Another view of our ordinary belief talk will be made in Section III). 13 See for instance Fodor 2. [7] and Harman [11], Ch. 4, Section

14[5] 15 [3]
16 I have assumed however people we associate Russell' and uniquely; w if and a world thing world, rigidly so as desirable. The only a natural is hairless i.e., Some in this paragraph that proper names are always rigid designators. this: they believe that when we use the name 'Bertrand deny a property with this name we think picks out Russell which <f> that when we say 'Bertrand Russell is hairless', what we say is true at only in w.

quite to allow To

if there is exactly one thing in w with property ^ in w, and that in in (The thing with </> w need not be the thing with </> the actual it need not be Russell.) I find this doctrine can denote that names non I am making but the point in the paper could be generalized implausible; for see

property. we utter

names to be non-rigid for anyone who finds that designators a property associate with a name. this, reflect on how we would <j> with some expression that name way I can see to do that is to associate (in or in a system of internal which stands for that language representation) What the advocate of non-rigid is saying, is that when designation then, the sentence 'Bertrand Russell is hairless', that sentence serves to merely

56
abbreviate thing of the the sentence

HARTRY

H.

FIELD

form

'There

Truth-at-a-possible-world that we really mean, tively, a sentence if the associated S

by

that we really mean, is some? which (or sentence-analog) one is exactly ?/?(*), and x is hairless'. thing x such that can then be defined for the sentences (or sentence-analogs) in the text; and we can say that deriva? the process described the name (or 'Bertrand Russell' that we is true really at w mean if and when only we

containing sentence

sentence-analog)

utter S is true at w. 17 I don't to be assuming in the sentence that the primitive symbols really mean are names or predicates; I knows what that the account who analogs they are. All am suggesting assumes such is that whatever of (actual world) reference problems manner. in a materialistically raise are soluble acceptable symbols 18 an alternative account I should of a relation add that the difficulty of constructing to be acute if one takes propositions is especially between and propositions people fine-grained. 19 Lewis believes this, and certainly 20 This is true even of non-materialist of psychological properties, (Only an instrumentalistic entirely.) I suspect that Stalnaker does too.

they merely functionalist

functionalists; that allow would

realizations they too require be non-physical. the realizations of realizations the requirement deny

21

21a

[14], [15], and [17].

I do not sense psychological be a common that <j> Lewis's requirement impose to add that (to my mind if you want you may: requirement, pointless) of the paper will not be affected. of this section sense psychological to common : that we explicitly build At first blush into <l>. this would later on interfere with a pro? theories might the postulate of a system of internal seem to conflict be a with letting <j> to not appear to be committed would Lewis not is rather regard the fact in what he generous that a theory postu? the requirement that it say are at

theory. But the argument The posal

restriction to be made

representation sense common systems of

than our own as going worlds other against of internal sense. I believe that systems is common representation only what sense as are possible worlds. of common to being postulates least as close sense be a common that </> I have said that I think that the requirement it is not at all clear For one thing, I also think that it is harmful. pointless. of

regards lates possible

sense since common theory, internal However, representation. to :he does sense committed common

is theory that any

to being) are (or even come close sense of belief theories the common uniquely be to invoke non to get anything like unique realization the only way may realized; It may well be for instance that only a theory of belief common-sense suppositions. near can come anywhere a system of internal that explicitly representation postulates of to press this point here: the argument But I do not want realized. to being uniquely a system not postulate of internal even if <f>itself does this section goes through or in any other representation sense. common 22 own account of what Lewis' way (apart from propositions!) strain the bounds of

he does

are is strictly because inaccurate realizations speaking as his account I've recommended: the precaution something stands, at all times, of pain only if it realizes can serve as a realization pain in all organisms of its point, functionalism thus depriving [11], Ch. 3, Section 4). But it is (cf. Harman is to what Harman is equivalent I have that what clear suggests) (which suggested not take Lewis really had in mind.

what

MENTAL REPRESENTATION
23 Here I oversimplify: are needed

57

we might that Tu...,Tn include non-psychological imagine in the theory; and we might also that some psycho? imagine in the theory are not included terms needed their among Tu..., Tn, because logical reference is fixed of 4>. independently 24 not relevant to the present to imagine For reasons paper, Lewis finds it convenient terms that that have quite derive 25 "I ([14], mean the so that all the psychological terms are names. I rewritten theory has been of the next paragraph not to do this, and the features that do not look chosen in Lewis I will from this fact. The like anything result conclusion explicitly however take p. would to be have been of things at times" to a state he does not of course refers 'pain' be called a state-token, that is, an individual inner might it 'feels pain'). of the type that the organism feels when (an occurrence he mean what might be called a state type: a. sortal property of such inner attributes says that states on his procedure forthcoming of a special kind: attributes as well.

Lewis 165). When that it refers to what

occurrence Nor does

a property be expressed that might 'is a pain'. Rather, occurrences, by the predicate the attribute that is expressed pain is taken to be an attribute of organisms, by 'feels that the word is best reserved 'state' for pain', (cf. [14], footnote 1). I think myself what I've called state-tokens states and in his a confusion a disastrous into between state-types; sense and for Lewis' use of the and state-types in this section. in mine, term helps foster this can lead to

to be noted later confusion does not fall (Lewis himself as footnote 13 of [17] shows, the confusion, but I think his terminology has led to make others it.) 26 to leave out the word it might be better here: that Strictly speaking, 'physical' if materialism is only contingently in those way true, then we can allow organisms in which worlds is to have func? mental possible </> realized by irreducibly properties in the definition tional properties. This of functional in no would change property way affect materialism my properties own does the application to organisms of functional in those worlds where properties in particular, it would not affect is true; the application of functional in the actual world, if as I am assuming materialism true. Since is indeed interest lies wholly in the actual world, my addition of the word 'physical'

no harm.

I chose to put in the word to remind the reader of the materialist 'physical' partly as a convenient and partly to introduce of order. In order distinctions way premise, for the method of defining in (2) to make sense lead properties employed (and not to paradoxes), the functional cannot be one of the properties property being defined in the range If we use of property-quantifier. the notion of 'physical property' so that functional like those defined in (2) do not themselves narrowly, properties count as physical, then this condition is met. On this narrow use of the term physical, the materialist thesis is that all non-functional are physical. properties 27 The uniqueness I think that it must is Lewis'. be taken with a grain of requirement to explain my reservations the space here or to develop the salt, but I do not have to avoid needed in (2) cannot merely it. (The word be dropped.) machinery 'unique' 28 to Lewis, 'the property of pain' and refer to different According 'pain' proper? ties: 'the property to a functional of pain' refers and 'pain' refers (in the property, context a specific organism X at a specific of discussing time /) to the non-functional property neutral which on realizes the functional the question of whether way) property refers 'pain' to a realization of in X at t. I have tried property to remain or refers to the functional it.

(in a context-dependent

58 HARTRY
29 This is not intended

H. FIELD

relates

as an argument of believing that which using a notion against to propositions, but only as an argument which against using a notion to English relates people sentences of English A theory (or to utterances sentences). is exempt from the criticism, at since if there are propositions involving propositions people then not expressible there are propositions in English. presumably (However, if 'believes to propositions, it is not clear that' relates that sentences people particular may over case us could ever be to organisms very unlike there is no ontological gain in quantifying Such a person should read Putnam propositions. beliefs feel that over true.) literally over proper?

all even

attributing 29a Some makes (b) has out 30 that

people ties rather than a good

properties

temperature to stand for I have allowed

(a) that quantification are quite distinct from 210? C and '* has mean property even

[20]. Putnam in science, is needed and properties in that two predicates like '* meanings, can turn molecular 10"20 joules' energy

the same this

believed 31 Where

be genuine

neutrality in these cases it is not difficult to re? the neutrality; but even sacrificed frequently I say so as to accord with the position that I am temporarily phrase what excluding. 32 is formulated in such a way This assumes that the syntactic that all syn? theory are predicates To assume of tokens. this is not to assume that the tactic predicates can be given nominalistically, syntax which like sets, or even like sequences less the role of over abstract entities i.e., without quantifying can be intuitively as having more or regarded this is realized, Once the task of formulating the expression-types. no difficulty. in the way required presents

differ in meaning. though they clearly are for the possibility that the sentences which by allowing sentences in a genuine than sentence rather language, analogs. to achieve I have of formulation is difficult without verbosity,

syntactic

theory

can be formulated I should in such remark that the syntactic theory Incidentally, in the grammars a way as to allow wide diversity of the systems of representation of to apply. in which it might do to which is intended the theory (One way organisms a general of of syntactic Different sub-sets this is to postulate system categories. to which in different be instantiated the theory could this general organisms system and applies, be compatible in this way a wide with the general I have in mind.) syntax syntactic theory. diversity See Lewis in systems [16] for an of representation would illustration of the kind of

R could well be a disjunctive say of the form relation, perfectly in the left hemisphere of the brain and x bears y are both occurrences in the right hemisphere and x bears R2 to to y, or x and y are both occurrences Pi in the right and the one in the left in the left and the other y, or one is an occurrence are needed func? realizations for other bears R3 to the one in the right'. Disjunctive in a given organism that pain is realized it is perfectly theories tional e.g., possible or a like 'is a stimulation of the C-fibers time by a disjunctive at a given property to rule out the same flexibility in and there is no reason of the Z-neurons' firing reason I think that a main I make this point because the theory of belief and desire. is that they tend resist the inner representation hypothesis philosophers why many commitments. its neurophysiological to exaggerate 34 to be of the same in the same organism it is for two tokens This only defines what in in one organism and tokens tokens between of type-identity type; but the notion as a meaningless for psychological is not needed the other theory, and can be regarded notion. 35 I will elaborate on this point in the next section.

general 33 Incidentally, 'Either x and

CENTAL
36a The

REPRESENTATION

59

'? ' is of course to get any? crude: iaw' exceptionally involving proposed one probably in something like degrees at all plausible, has to bring thing that looks I am making the is unaffected: But the point and degrees of desirability. of belief can be regarded as attaching to sentences, of desirability of belief and degrees degrees and the tences At laws form Si can the degrees of a sentence of of belief and desirability be given relating ? sen? to the degrees and desirability of the component of belief SJ if B sentences and S2 and of other Si and S2. (E.g., containing compound r5i of the component as just less'. sentences or of the logical the notion ought of connectives

is degree of belief, one law might be that B(Si -+ S2) > max{B(S2), 1 - B(Si)}).
no point do the meanings to be mentioned. I think that loose there and is some

need 36

question

to whether explained,

in the very

semantic notion; 37 Cf. Quine [23], p. 11. 38 we can in a sense recognize a functional Without account, with in connection stand for a certain relation physical some that speak (say, organisms language 'refers' stand for a different relation physical we a functional account type; but without of do a certain not have for in connection

way pragmatic hence the 'more or

to count

used translation, as a genuinely

this: we organisms

can of

let 'refers' one kind

let and also general type), with organisms of another to explain the means what 'true'.) worlds ap?

39
40

these

two uses

of

'refers'

have

in common.

(Similarly

[6].
this Strictly, But we on the holds equivalence only could that a one-word imagine coarse-grained phrase 'glub' possible had been introduced

proach. mean on

into the language by the stipulation that it was to (rigidly) denote the property of
to (13) even then 'Temperature is glub' would molecular be equivalent energy; the fine-grained but they would worlds approach, clearly be non-synony? possible mous. 41 one treats evidential on whether Whether considera? they are redundant depends or a coarse-grained see [6], pp. 396-397. tions in a fine-grained manner; 42 are less clear: some? but intra-linguistic Our criteria for inter-speaker synonymy we use treat the like the inter-linguistic times we this case and sometimes case, common across to extend of considerations synonymy language intra-speaker different 43 Here with the truth-theoretic of meaning. aspects Also, primarily does not really presuppose the formulation in the what asking of internal that a person's is distinct from his system representation we can phrase the opposite If we adopt the question public language. presupposition, : in what ways as follows if any does one need to bring in the fact that a system of the question text suggests, I am internal conditions to bring theory 44 See is also used representation for the system of internal in the fact that a public for it. publicly, a complete in giving theory and in what ways does for thinking, in giving of one truth need speakers. I am concerned

representation; is used language

a complete

of meaning Schiffer in [19] is quite similar. Harman's [24]. Lewis' approach suggestion 1 theory of mean? in [9] that the "level 2 theory of meaning" the "level presupposes seems a similar to involve idea. ing" but not conversely 45 It might not defeat at the end of Section V we give a functional it, if as suggested of reference: in public for then we language could 'Caesar' to specify that we need grant a realization to specify in order for the use of the the reference

account word

60
relation

HARTRY

H. FIELD

need

as applied to the system of internal but hold that we do not representation, to mention it in the functional it does not seem to me at theory itself. However, use of words to mention all obvious that we do not need the public in a satisfactory functional for systems of internal it seems theory of truth-conditions representation; to me in anywhere near enough that no one has developed detail any such theory for us reference It may well be for instance that a fairly detailed account causal to be explicitly built into reliability other kind (or whatever theory or of truth-conditions it is in which the notion of reference psychological theory tell. has and that such a detailed causal refers to of of

broad

appears),

of a public word that acquisition can become which people causally having beliefs about the object.

will to explicitly have mention theory as one of the mechanisms to something by to an object in the way to related that is relevant

BIBLIOGRAPHY
A Materialist David, [1] Armstrong, & Paul, 1968. Kegan, Press, [3] Davidson, [4] Dennett, pp. 1957. Donald, Daniel, 'Truth 'Brain Writing and Meaning', and Mind Synthese Reading', 17 (1967), in [8]. pp. 304-323. Theory of the Mind. London: Routledge,

[2] Chisholm, Roderick, Perceiving: A Philosophical Study. Ithaca: Cornell U.

[5] Field, Hartry,

347-375. [6] Field, Hartry, 74 (1977), pp. [7] Fodor, Jerry,

'Tarski's Theory
'Logic, 379-409. Meaning,

of Truth', Journal of Philosophy 69 (1972),

and Conceptual New Role', Journal of Philosophy

The Language

of Thought. Mind, VII).

[8] Gunderson, Keith, the Philosophy of 1975. Gilbert, [9] Harman, pp. 590-602. [10] Harman, [11] Harman, [12] Harman, [13] Jech, Gilbert, Gilbert, Gilbert,

ed., Language, Vol. Science, 'Three Levels

1975. York: Crowell, and Knowledge. in Studies (Minnesota U. of Minnesota Press, Minneapolis: Journal of Philosophy 33-43. 1973. in [8]. 1973. 65 (1968),

of Meaning', Nous Princeton

'Language Thought.

Learning', Princeton:

4 (1970), pp. U. Press,

Thomas,

and Communication', Thought, 'Language, Amsterdam: The Axiom North-Holland, of Choice. the Mind-Body 'How 'General to Define Problem. Theoretical in Donald Dordrecht: Theoretical Englewood Terms', Davidson Reidel, Cliffs: Journal

[14] Lewis, David,

Materialism [15] Lewis, [16] Lewis, (eds.), [17] Lewis, [18] Lewis, David, David, Semantics David, David,

'AnArgument for the Identity Theory', inDavid Rosenthal

Prentice-Hall, of Philosophy

(ed.),
1971. 67

and

(1970), pp. 427-446.

Semantics', Language. of Natural and 'Psychophysical and Gilbert 1972. Australasian Harman

Identifications',

Journal of Philosophy 50 (1972), pp. 249-258.

23 (1974), pp. 331-334. Synthese Interpretation', in [8]. and Language', 'Languages in Putnam, 'On Properties', Papers, Philosophical Hilary, [20] Putnam, 1975. U. Press, Cambridge Cambridge: 'Radical [19] Lewis, David,

Vol.

1.

MENTAL

REPRESENTATION

61
in Quine, 1960. From a

[21] Quine, Logical [22] Quine, [24] Schiffer,

in Linguistics', of Meaning 'The Problem Willard, & Row, 1963. York: Point Harper of View. New M.I.T. and Object. Word Press, Cambridge: Willard, Stephen, Meaning. Oxford: Oxford in Alfred New U. Press, 1972.

[23] Quine, Willard, Philosophy of Logic. Englewood Cliffs: Prentice-Hall,

Robert, 'Propositions', [25] Stalnaker, in the Philosophy Issues of Language. Merrill and Daniel Mackay Haven: Yale Press, 1976.

1970.
(eds.),

submitted 16 August 1977 Manuscript Final version received 31 October 1977