62

The Transparency of Mental Content INTRODUCTION Michael Dummett once wrote: It is an undeniable feature of the notion of meaning-obscure as that notion is-that meaning is transparent in the sense that, if someone attaches a meaning to each of two words, he must know whether these meanings are the same. As the surrounding discussion makes clear, Dummett’s claim isn’t merely the fairly weak thesis that, for any two expressions in a person’s repertoire, it must be possible for that person to come to know whether or not they mean the same. Rather, the claim is that it must be possible for that person to come to know such a fact purely introspectively, without the benefit of further empirical investigation-a priori, as I shall also occasionally put it. Dummett doesn’t say much about the basis for his conviction that ‘transparency’ in this senseepistemic transparency as I shall sometimes call it, to distinguish it from the referential variety-is “an undeniable feature of the notion of meaning.” He gives the impression of finding the claim too obvious to need arguing. In this respect, I believe, he follows in the footsteps of Frege and Russell, both of whom gave epistemic transparency a pivotal, if unargued, role in their respective theories of linguistic and mental content. In another paper, I hope to discuss this historical point. In this paper, however, my concern will be purely philosophical and, then, only with mental content. I believe that the notion of epistemic transparency does play an important role in our ordinary conception of mental content and I want to say what that role is. Unfortunately, the task is a large one; here I am able only to begin on its outline. I shall proceed somewhat indirectly, beginning with a discussion of externalist conceptions of mental content. I shall show that such conceptions violate epistemic transparency to an extent that has not been fully appreciated. Subsequently, I shall look at the implications of this violation and at the reconstructive project that a rejection of transparency entails. I am inclined to think it unlikely that we will get a conception of propositional content that underwrites epistemic transparency. But I am concerned that we have not fully appreciated the role that transparency currently plays and the work that would need to be done were we to discard it. EXTERNALISM, SELF-KNOWLEDGE AND TRANSPARENCY The minimal intended significance of the Twin Earth thought experiments may be put like this: Either mental contents are not individuated individualistically, or they are not individuated in terms of their truth conditions. Thus, consider Oscar and twin Oscar; they are alike in all intrinsic, nonrelational respects. But whereas Oscar grows up in a physical environment which contains H2O, twin Oscar grows up in an environment which contains XYZ. And widespread intuition appears to have it that whereas Oscar’s utterance of Water is wet' expresses a thought that is true if and only if H2O is wet, twin Oscar's utterance of that sentence expresses a thought that is true if and only if XYZ is wet. Hence, either the contents of the two thoughts are not individuated individualistically or they are not individuated in terms of their truth conditions.2 Thus, consider also Bert and twin Bert; they are alike in all intrinsic, nonrelational respects. But whereas Bert grows up in, and defers to, a community in which the wordarthritis’ is used to apply only to a certain kind of inflammation of the joints, twin Bert grows up in, and defers to, a community in which arthritis' is used to apply to all rheumatoid ailments, even to those that might occur in the thigh. And widespread intuition has it that whereas Bert's utterance ofI have arthritis in my thigh’ expresses a thought that is true if and only if he has arthritis in his thigh (and, hence, can never be true), twin Bert’s utterance of that sentence expresses a thought that is true if and only if he has (some other disease) tharthritis in his thigh (and, hence, may occasionally be true). As before, the moral would appear to be: Either the mental contents of the two thoughts are not individuated individualistically or they are not individuated in terms of their truth conditions.3 Let us call the former view externalism, and the latter individualism. According to externalism, then, the psychologically relevant contents of a subject’s intentional states are individuated in part by certain sorts of facts about the physical and/or social environment in which he happens to be situ- ated.4 And although this would appear to undermine the claim that we can have purely introspective, non-empirical access to the contents of our thoughts, we are assured by an increasing number of philosophers that this appearance is illusory: there is in fact no conflict between an externalist conception of mental content and introspective self-knowledge. I want to begin with the question, On what is this confidence based? It is based on a thought that is succinctly expressed by Davidson: Showing that there is no conflict [between externalism and knowledge of content] is basically simple. It depends on realizing that whatever is responsible for the contents of our thoughts, whether known or not, is also responsible for the content of the thought that we have the thought.5 How exactly should we understand Davidson’s thought here? Let’s assume, without loss of generality, that we think in a language of thought (in fact, I’ll pretend that we think in English. This will make the exposition easier but is not required for anything that’s to follow: similar claims could have been made without the assumption that psychological states possess syntactic structure). With the convenient assumption in place, we may say that Peter believes that p just in case Peter has a token of a sentence S, which means that p, in his belief box; and Peter desires that q just in case Peter has a token of a sentence Q, which means that q, in his desire box; and so on for the other attitudes. It is now easy to give a completely general elaboration of Davidson’s thought. For since externalism is guaranteed to be entirely neutral on the matter, we may safely assume that there are syntactic mechanisms which ensure that when, and only when, a token of the sentence S enters Peter’s belief box, then ceteris paribus, so does a token of the sentence I believe S'. And so, provided that the embedded and unembedded tokens of S may always be counted upon to express the same content, Peter's second-order belief will always report correctly on the content of his firstorder belief, the externalist nature of content individuation notwithstanding. Hence, and in this sense, externalism is shown to be consistent with authoritative first-person knowledge of thought content. This is, I think, the best and simplest way to develop the idea that Davidson expresses.6 Now, there are, I think, a number of questions that might be raised about this proposal. For instance: can we in fact always count on the embedded and unembedded occurrences of S to express the same content? (As we shall presently see, there may be special reasons for doubting the reliability of this covariation under externalist assumptions.) Also: the envisaged proposal would appear at best to ensure the availability of reliable beliefs about content, not knowledge; and there are some tricky questions about justification that need answering. But I am not going to pursue either of these lines of questioning here; for present purposes, I wish simply to grant the claim that externalism is consistent with the sort of knowledge of content envisaged by this elaboration of Davidson's idea. My point will rather be that the kind of self-knowledge that is thereby secured falls short-way short-of the kind of self-knowledge we normally think of ourselves as possessing. Let us grant the claim that, for any sentence S in my belief box, I can form a correct secondorder belief about its content by prefixing the wordsI believe … ‘ to that very sentence. It is important to observe, however, that no other epistemic capacity is thereby grounded. For instance, suppose I have two sentences P and Q in my belief box that are identical in content; it needn’t follow that I will know, or be able to a priori come to know, that they are, and this despite my being said to know what the contents of P and Q are individually. Nor, similarly, does it follow that if I have two sentences Q and R in my belief box that are distinct in content, that I will know, or be able to a priori come to know, that they are; and this despite my being said to know what the contents of Q and R are individually. In fact, it doesn’t even follow from the sort of knowledge that is underwritten by the proposal, that I will know, or be able to a priori come to know, whether two distinct tokens of the same sentence type S express the same or distinct contents. To put the point concisely, the proposal does nothing to reconstruct the intuitive idea that we know the contents of our thoughts in the sense that we can introspectively discriminate between them, that we can tell them apart from each other: it doesn’t by itself underwrite the epistemic transparency of mental content. EXTERNALISM, TRANSPARENCY AND THE SEMANTICS OF TRAVEL The thesis of the epistemic transparency of content may be usefully broken up into two parts: (a) If two of a thinker’s token thoughts possess the same content, then the thinker must be able to know a priori that they do; and (b) If two of a thinker’s token thoughts possess distinct contents, then the thinker must be able to know a priori that they do. Call the first the thesis of the transparency of sameness and the second the thesis of the transparency of difference. Now, it is fairly easy to show that externalist contents do not satisfy the transparency of sameness. Kripke’s notorious Frenchman, Pierre, already shows this for the special case of Millian contents (themselves, of course, a species of externalist content). The details are well enough known that I need not rehearse them here. A plausible story culminates in Pierre having two expressions- Londres' andLondon’-which refer to the same city without, however, his realizing that they do. The thought he expresses with London is pretty is the same as the thought he expresses with Londres is pretty; but this fact is unavailable to introspection alone. We need a different sort of case, however, to illustrate the general claim that even a non-Millian, referentially opaque content would fail the transparency of sameness, if it were individuated externalistically. Brian Loar has described a case which may be adapted to just this purpose.’ Paul, an English speaker, has been raised by a French nanny in a sheltered way. She speaks English with Paul, but amuses herself by referring to the cats around them as “chats” and never as “cats.” Paul, satisfying the usual competence and deference constraints on concept possession, is to be credited with beliefs involving the concept cat. In particular, the belief he would express with “All chats have tails” would be the belief that all cats have tails. Now, Paul occasionally sees his parents, who speak of animals called “cats.” Because on those occasions, no cats or pictures of cats are ever present, Paul does not realize that cats are his familiar “chats.” However, Paul’s parents tell him quite a lot about cats and in particular that all cats have tails. On this basis, it is again true of Paul that the belief he would express with “All cats have tails” would be the belief that all cats have tails.9 Intuitively, however, it is quite clear that Paul will not be able to tell a priori that the belief he expresses with “All chats have tails” is the same belief that he expresses with “All cats have tails.” So much, then, by way of illustrating externalism’s violation of the transparency of sameness. What has not been generally appreciated, however, is that externalism also allows for a coherent description of the converse case: two tokens of a thinker’s language of thought belonging to the same syntactic type have distinct meanings, but the thinker is in principle not in a position to know a priori that they do. These converse cases, illustrating externalism’s violation of the transparency of difference, are, I believe, important even for issues beyond the ones presently at hand.10 To see one way in which such converse cases might arise we need to think-harder perhaps than is common in the literature- about the semantics of so-called one-world traveling cases'. Suppose that Earth and Twin Earth are part of the actual world and that Peter, a normal, competent adult Earthling, is suddenly and unwittingly transported to Twin Earth. An accidental tourist, Peter goes to sleep one night at home and wakes up in twin home in twin bed. He suffers no discernible disruption in the continuity of his mental life. Here on Twin Earth, Peter happily lives out the rest of his days, never discovering the relocation that he has been forced to undergo. How should we think about the semantics of Peter's thoughts? Well, one intuition that is shared by practically everyone who has thought about these cases is that, after a while (how long is unclear), some tokens of, say,water’ in Peter’s language of thought will shift from meaning water and will come to mean twater. Thus, to quote just one example, Tyler Burge writes: The thoughts would not switch as one is [quickly] switched from one actual situation to another twin actual situation. The thoughts would switch only if one remained long enough in the other situation to establish environmental relations necessary for new thoughts. So quick switching would not be a case in which thoughts switched … But slow switching could be such a case. i i Viewed from an externalist perspective, it’s hard to deny the force of the intuition that Burge is expressing here. We imagine Peter moving to Twin Earth and staying there for a while. Over time he thinks to himself a variety of thoughts that he expresses with sentences like: I won’t go swimming today; the water is too cold. Or: There’s too much water in this whiskey. It seems very plausible that these expressions of beliefs about one’s present environment, expressions of current desires and current intentions, are expressions of thoughts that involve the concepts twater, not water. So it seems right to say that some tokens of water' will shift from meaning water to meaning twater. But it seems to me equally compelling to say that certain other tokens ofwater’ won’t shift. Specifically, it seems to me that those tokens of `water’ occurring in memories, and in beliefs about the past based upon them, will retain their Earthly interpretations, despite being tokened on Twin Earth. Such thoughts, unlike, for instance, beliefs with undated general contents, or thoughts about one’s present surroundings, are caused and sustained by previous perceptions long gone. In the normal case, they owe little, if anything, to current perceptions and cognitive transactions with one’s environment. From a purely intuitive standpoint, they would be expected to retain their Earthly interpretations, despite the admitted shift in their syntactic cousins. After all, why should mere travel from one place to another obliterate one’s capacity to have certain memories. Normal tourism doesn’t have such an effect; why should travel to another planet with somewhat unusual geographical properties?