62

Consider an example. While still on Earth, Peter goes hiking in the mountains of northern New Zealand. Here he comes across Lake Taupo and is startled to see the famous tenor Luciano Pavarotti floating on its pristine waters. They talk amiably for a while and Peter goes off flushed with excitement. Understandably enough, this experience of Peter’s gives rise to many subsequent memories on his part, and to beliefs based upon them. Consider the ones that are tokened while he is still on Earth, many years prior to his trip to Twin Earth. There can be no question about their content. They are about Pavarotti and Lake Taupo and water.12 In fact, I take it, they would be about Pavarotti even in a possible world in which Pavarotti has a twin living in Los Angeles who Peter doesn’t know about but whom he subsequently meets. Even after he does so, the memories of the encounter on Lake Taupo, and the beliefs based upon them, would continue to be about Pavarotti, not Twin Pavarotti. Well, some years go by and Peter is moved surreptitiously to Twin Earth and becomes happily ensconced there. Eventually, some of the tokens of his mental names come to refer to the twin counterparts of the familiar Earthly figures, and some of the tokens of his general terms come to express twearthly properties: some tokens of Pavorotti' will come to refer to twin Pavarotti and some tokens ofwater’ will come to mean twater.13 One day, perhaps while reading a schedule of upcoming events, Peter is moved to reminisce about the occasion when he saw Pavorotti swimming in the waters of Lake Taupo. He calls up vivid and accurate representations of the scene. Of course, he takes himself to be remembering scenes involving the singer he is reading about now. But he isn’t. His memories, intuitively, are about the Earthly Pavarotti, the Earthly Lake Taupo, and Earthly water, previous perceptions of which are the sustaining cause of his later ability to recall what the scene looked like and how it felt. In the situation described, Peter’s externally individuated thought tokens are not epistemically transparent to him. In particular, Peter’s language of thought contains token expressions that possess different semantic values, despite being of the same syntactic type. And yet, clearly, Peter does not know that they do. Tokens of Pavarotti',water’, and Lake Taupo', in sentences expressing memories and beliefs about that memorable occasion, will mean Pavarotti, water and Lake Taupo, respectively; whereas other tokens of that type, in sentences expressing beliefs about his current environment, or current desires, will intuitively mean twater and twin Pavarotti. From the inside, however, there will be no indication of this: as far as Peter is concerned, they will appear to express precisely the same contents. 14 Externalism, then, would appear to be robustly inconsistent with the epistemic transparency of thought contents: thoughts that have the same content may look to introspection to have distinct contents (as in Pierre and Paul), and thoughts that have distinct contents may look to introspection to have the same content (as in Peter). Externalist contents fail, in other words, to possess the feature that Dummett proclaimed "undeniable." What problem, if any, does this pose? I shall argue that the problem is this. We don't just ascribe thoughts to a person in order to say something descriptively true of him. We use such ascriptions for two related purposes: on the one hand, to enable assessments of his rationality and, on the other, to explain his behavior. As these matters are currently conceived, a thought must be epistemically transparent if it is to play these roles. Without transparency, our conceptions of rationality and of rational explanation yield absurd results. We manifest our recognition of this fact by barring de re thoughts-thoughts which intuitively lack epistemic transparency-from figuring in assessments of rationality and psychological explanation. However, if we abandon transparency even for de dicto thoughts, and hence in effect altogether, then we must either jettison the notion of rationality and with it the practice of psychological explanation that it underwrites, or we must show these notions can be refashioned so as not to yield absurd results. The problem is that the first suggestion is wild and there appears to be no obviously satisfactory way of implementing the second. THE APRIORITY OF LOGICAL PROPERTIES AND DE RE THOUGHTS We may usefully begin with a discussion of de re beliefs. By a We re belief' I shall mean, by stipulation, a belief that is individuated by the objects it is about. Such beliefs are typically reported with the use of anof-clause’, rather than a that-clause', as in, Jane believes of the piano that it is ugly, and their content is given by a Millian proposition. Clearly, and in contrast with fully conceptualized, referentially opaque de dicto beliefs, it is both necessary and sufficient for the distinctness of two de re beliefs applying the same predicate that they concern distinct objects. I should emphasize here that I am not presupposing, for the purposes of this discussion, either that this account explicates theintuitive’ notion of de re belief, if there is such a thing, or even that there actually are de re beliefs in this stipulated sense. I am only interested in the question what would be true of such beliefs, if there were any. Well, one claim that is often made in connection with such thoughts is nicely expressed by Burge: individual entities referred to by [referentially] transparently occurring expressions, and, more generally, entities (however referred to or characterized) of which a person holds his beliefs do not in general play a direct role in characterizing the nature of the person’s mental state or event. The difference [between such entities] does not bear on Alfred’s mind in an sense that would immediately affect explanations of’Alfred’s behavior or assessment of the rationality of his mental activity…. Moreover, it seems unexceptionable to claim that the obliquely occurring expressions in propositional attitude attributions are critical for characterizing a given person’s mental state. Such occurrences are the stuff of which explanations of his actions and assessments of his rationality are made.15 Burge’s striking claim here is that de re beliefs, in contrast with fully conceptualized, referentially opaque, de dicto beliefs, don’t enter into assessments of a subject’s rationality or psychological explanations of his behavior. Here, for the purposes of further illustration, is Jerry Fodor making a similar point (Fodor’s emphasis is on psychological explanation): Suppose I know that John wants to meet the girl who lives next door, and suppose I know that this is true when “wants to” is construed opaquely. Then, given even rough-and-ready generalizations about how people’s behaviors are contingent upon their utilities, I can make some reasonable predictions (guesses) about what John is likely to do: he’s likely to say (viz., utter), “I want to meet the girl next door.” He’s likely to call upon his neighbor… On the other hand, suppose that all I know is that John wants to meet the girl next door where “wants to” is construed [referentially] transparently; i.e., all I know is that it’s true of the girl next door that John wants to meet her. Then there is little or nothing that I can predict about how John is likely to proceed. And this is not just because rough-and-ready psychological generalizations want ceteris paribus clauses to fill them in … .16 What reasons do Burge and Fodor offer in support of their respective claims? Burge, actually, has very little to say on the matter (though in fairness to him, I should point out that the essay from which the citation is drawn is largely concerned with other questions). And the exact interpretation of what Fodor has to say would take us too far afield. So without concerning myself overmuch with why these authors believe that de re beliefs are unfit for the purposes of content-based psychology, let me offer my own explanation. Let’s begin with the question about rationality. Suppose that Jane sees a wholesome-looking apple. She thinks de re of the apple that it is wholesome. That is, she comes to believe the Millian proposition . She subsequently sees the same apple with its blemished side exposed. She thinks de re of this apple that it is not wholesome. That is, she comes to believe the Millian proposition . Two things are true of Jane in this case. First, her de re beliefs about the apple logically contradict each other: the (Millian) proposition subtended by the one is p and the one subtended by the other is not-p. And, second, she cannot recover from this condition on an a priori basis; to discover that the beliefs contradict each other she would have to learn an empirical fact, namely, that the apple involved in the first thought is identical to the apple involved in the second. That the two thoughts logically contradict each other is not introspectively accessible to her. 17 Clearly, a similar case can be described to illustrate the fact that the logical consistency of Millian propositions is also not necessarily introspectively accessible. In general, then, the point is that the logical properties of de re propositions are not knowable a priori. The question is: How might this fact help explain why de re beliefs are unfit for the purpose of assessments of rationality? THE APRIORITY OF LOGICAL PROPERTIES AND NORMS OF GOOD REASONING The answer derives from our conception of the nature of rationality and, in particular, of what it is for someone to be a good reasoner. What does a person have to do in order to count a good reasoner? Clearly, it is not at all a question of knowing empirical facts, of having lots of justified true beliefs about the external world. Rather, it is a matter of being able, and of being disposed, to make one’s thoughts conform to the principles of logic on an a priori basis.” A surreptitiously envatted brain-transplanted from its normal adult body into a vat and attached to a computer that seamlessly duplicated and continued its previous course of experience-could be as good a reasoner as it ever was, despite the sharp escalation in the number of its false beliefs about the external world. Or so, at any rate, our conception of rationality requires us to think. So, rationality is a function of a person’s ability and disposition to conform to the norms of rationality on an a priori basis; and the norms of rationality are the norms of logic. We may, if we wish, put matters in a far less committal way: let’s say that being minimally rational is a matter of being able to avoid obvious violations of the principles of logic, given enough time to reflect on the matter and so on. But even relative to this very minimal notion of rationality we would appear to have stumbled onto a problem. For according to this view, our Jane, who innocently believed of one and the same apple that it is both wholesome and not wholesome, would appear to stand convicted of irrationality: she believes a pair of Millian propositions that contradict each other, but she is unable to recover from this predicament on an a priori basis, no matter how long she may be given to reflect on the matter. But, intuitively, there is nothing irrational about her. Therefore, either our conception of rationality is mistaken or we have to find some other systematic and non-arbitrary way of absolving Jane’s cognitive behavior. Obvious conservative solution: Bar an agent’s de re thoughts from entering into assessments of her rationality. The ban on contradictory belief and invalid inference is preserved, but only in application to de dicto beliefs, just as Burge says. THE APRIORITY OF LOGICAL PROPERTIES AND PSYCHOLOGICAL EXPLANATION Precisely parallel considerations explain why de re thoughts are also unfit for the purposes of psychological explanation. Since rationality is taken to consist in the ability and disposition to conform to the principles of logic on an a priori basis, any rational subject, regardless of his external conditions, may be expected to obey certain laws (or counterfactual-supporting generalizations): namely, those generalizations that mirror the introspectively obvious logical consequences of a person’s propositional attitudes. Thus, our ordinary psychological practice of explaining and predicting behavior is built upon appeal to such laws as this: If S occurrently believes p and occurrently intends to F if p, and if S has no independent reason for not F’ing, then S will intend to F or, at the very least, will be disposed to intend to F. If S intends to F iff p, but does not believe p, but merely q instead, (where p and q are logically independent propositions), then S will not intend to F. The trouble is that perfectly rational subjects will not obey these generalizations, when they are construed as quantifying over de re thoughts. Thus, suppose that Jack intends to call the FBI whenever he is within 50 feet of a spy. And let’s suppose that he believes of the dean of his College that he is a spy. The dean in fact is currently seated next to him at the beach, disguised as a lifeguard. So, de re, he believes of the `lifeguard’ seated next to him, that he is a spy. Yet he doesn’t budge. Yet Jack would appear to be a perfectly rational person. The elementary generalizations upon which the practice of psychological explanation depends fail to hold for de re thoughts. Therefore, either there is something wrong with our conception of rationality and, hence, with the generalizations that it underwrites, or we must find some way of justifying our belief in these generalizations, consistent with their failing to hold in this case. Obvious conservative solution: Except a subject’s de re thoughts from psychological explanations of his behavior. The elementary generalizations on which psychological practice depends are preserved, but only in application to de ditto beliefs, just as Burge and Fodor say. It seems to me, then, that we have before us the general answer to the question: What considerations might underwrite Burge’s claim that de re thoughts are unfit for the purposes of assessments of rationality and psychological explanation? The answer is that both these enterprises require that the logical properties of the propositional attitudes they manipulate be knowable a priori; and the logical properties of de re thoughts aren’t. (Again, I don’t claim that this is the answer that Burge himself would give.)