63

REFERENCE AND EPISTEMIC TRANSPARENCY Why, though, are the logical properties of referentially individuated thoughtswhether these be de re thoughts normally so-called, or thoughts involving Millian proper names-not knowable a priori? What makes these thoughts special in this regard? The answer, of course, is that the logical properties of referentially individuated thoughts fail to be a priori just because these thoughts fail to be epistemically transparent. It is precisely becauseand only because-a subject is unable to tell a priori whether the thought he expresses with Tully is bald is the same as the thought he would express with Cicero is bald that he is unable to tell whether the thought he would express with Tully is bald and Cicero isn’t is or is not a contradiction, on a Millian construal. Referential transparency entails epistemic opacity. It is worth noticing in this connection that violations of the transparency of sameness and violations of the transparency of difference induce different sorts of logical defect. A thinker for whom the transparency of sameness is false may well fail, as we have seen, to be able to tell a priori that a given proposition is in fact a logical consequence of other propositions that he believes. That it is such a consequence is a fact he will be able to discover only a posteriori, by learning some empirical facts. Thus, Pierre both believes that He lives in London and believes that If he lives in London (‘Londres’), he lives in the same city as Oscar Wilde lived. Yet he is unable to draw the conclusion that he lives in the same city as Oscar Wilde lived in. Such a thinker, however, need not be supposed ever to actually reason invalidly. It is consistent with the falsity of the transparency of sameness, in other words, that all the simple inferences that look a priori to such a thinker to be valid, are valid; what is falsified is the claim that all the simple inferences that are valid, will necessarily so look. On the other hand, a thinker who suffers from the converse defect-failure of the transparency of difference-will suffer from the converse failing. For such a thinker, certain inferences may well look valid, when they in fact are not. And that they are not is a fact he will be able to discover only a posteriori, by learning some empirical facts. Peter provides an appropriate example. Since his language of thought contains token expression that differ in semantic value despite being of the same syntactic type, he will be tempted to think that certain inferences are instances of, e.g., modus ponens, when they in fact aren’t. Thus, he might muse to himself as follows: Whoever floats on water, gets wet. This thought, by virtue of expressing a general quantified proposition, is to be regarded as having Twearthly content, i.e. as being about twin-floating and twin water. Now, however, he combines it with a memory belief that he would express with the sentence Pavarotti once floated on water to conclude Pavarotti once got wet. The inference will seem valid to him; but it arguably isn’t. The second premise, by virtue of expressing a memory belief that is rooted in an Earthly experience, will be about Earthly floating and Earthly water. True premises, aided by a failure of univocity that Peter is in principle not in a position to introspect, will combine to produce a false conclusion. Returning to the main line of argument, we see, then, that the. fundamental answer to the question: Why are de re thoughts unsuitable for the purposes of assessments of rationality and psychological explanations, is this: It’s because de re contents (and Millian propositions quite generally) are not epistemically transparent. By contrast, fully conceptualized de dicto thoughts, nowhere subject to co-referential substitution, are supposed to be, as Burge rightly says, “the stuff of which explanations of his actions and assessments of his rationality are made.” If the diagnosis on offer is correct, however, this contrast can hold up only if fully conceptualized de dicto thoughts are transparent. But as Loar’s example of Paul and my example of Peter show, fully conceptualized de dicto contents will themselves fail to be epistemically transparent-and hence will themselves fail to be suitable for the purposes of psychological explanation and assessments of rationality-if they are individuated externalistically in the manner that Burge advocates. On what basis, then, does Burge distinguish between them? Why are externalistically individuated, and, hence, non-transparent, de dicto thoughts held to be suitable for the purposes of rational psychology, when de re thoughts are conceded not to be? THE APRIORITY OF LOGICAL PROPERTIES AND THE INTRODUCTION OF SENSE It seems to me that once epistemic transparency is identified as a semantically significant thesis, its role in a variety of important disputes in the philosophy of language and mind becomes obvious. I shall discuss two of these: the role it has played in the canonical argument for the thesis that names have sense and not merely reference, and the role it plays in generating Kripke’s puzzle about belief. Beginning with the former, many philosophers would probably resist the claim that transparency plays a part in the canonical argument motivating a nonMillian view of names, because they would resist the claim that there is any such argument. Most philosophers write as if it’s merely obvious-and, hence, in need of no argument-that someone might be in a state truly described by (1) Mary believes that Ali was a champ but not thereby in a state truly described by (2) Mary believes that Clay was a champ. It’s worth seeing, however, that it isn’t merely obvious. It becomes compelling, as I shall now argue, only when one makes the assumption, left implicit by Frege, that beliefs involving proper names are fit for the purposes of assessments of rationality and hence must be epistemically transparent. It is only under the terms of this assumption that one gets an argument for the referential opacity of proper names in the first place. 19 For consider how a belief in referential opacity is typically motivated. We are given a case which goes like this: Mary sincerely asserts Ali was a champ.' She also sincerely assertsClay was not a champ.’ She asserts these sentences even though it is clear that, as she is using the names Ali' andClay’ they refer to one and the same legendary boxer. Now, given the following principle for reporting beliefs Jones’ sincere assertion of p' expresses his belief that p we may conclude that (1) Mary believes that Ali was a champ and that (3) Mary believes that Clay was not a champ. It is important to notice, however, that nothing so far bars us from supposing that beliefs involving names are referentially transparent, and, hence, that (1) is equivalent to (2) Mary believes that Clay was a champ. For all that this would entail is that Mary has contradictory beliefs, a state of affairs that is, presumably, perfectly possible. We need to be given a reason why an ascription of contradictory beliefs is unacceptable in the present instance. Otherwise, we would have no case illustrating, and consequently no argument for, the referential opacity of beliefs involving names. What is that reason? We get such a reason only if we insist that beliefs involving proper names must be fit for the purposes of assessments of rationality and psychological explanation, and hence must have logical properties that are knowable a pri- ori-must, that is, be transparent. Armed with such an assumption the argument for referential opacity is finally enabled. For the assumption insists that the attribution of a contradictory pair of beliefs involving proper names is acceptable only if the fact that they contradict each other is a priori available to the subject to whom they are attributed. Yet it seems perfectly clear in this case that no matter how much Mary might search her own mind, she won't discover that the belief she expresses withAli was a champ’ is referentially (and, hence, on a Millian construal, logically) inconsistent with the belief that she expresses with `Clay was not a champ.’ With the insistence in place, then, it follows, that (1) and (3) could not be attributing logically contradictory beliefs and, hence, that (1) is not equivalent to (2). If we ignore, as Frege himself evidently did, views that attempt to accommodate this result by analyzing belief in terms of a three-place relation, we get Frege’s conclusion: there must be a level of semantic description of beliefs involving names other than the referential. KRIPKE’S PUZZLING PIERRE We may observe the very same dialectic at work in Kripke’s famous Pierre case. In one scenario (I won’t discuss the other), Pierre assents both to Londres is pretty and to London is not pretty. Kripke argues that, in this case, there is no saying what it is that Pierre believes, no satisfactory belief ascription. His overall idea is to protect Million theories from the charge that they generate absurd belief ascriptions, by showing that parallel absurdities can be generated solely from principles constitutive of belief as such, and without reliance on distinctively Millian principles (for example, substitutivity): When we enter into the area exemplified by … Pierre, we enter into an area where our normal practices of interpretation and attribution of belief are subjected to the greatest possible strain, perhaps to the point of breakdown. So is the notion of the content of an assertion, the proposition it expresses.20 But is it really true that we are unable to say what it is that Pierre believes in the case as described? Why can’t we say, applying our disquotational principle, that Pierre believes both that London is pretty and that London is not pretty? Kripke writes: … there seem to be insuperable difficulties with this alternative as well. We may suppose that Pierre … is a leading philosopher and logician. He would never let contradictory beliefs pass. And surely anyone, leading logician or no, is in principle in a position to notice and correct contradictory beliefs if he has them… . [Pierre] cannot be convicted of inconsistency: to do so is incorrect.21 Here we see the assumption of epistemic transparency playing an explicit role: Pierre cannot be ascribed contradictory beliefs because “anyone … is in principle in a position to notice and correct contradictory beliefs if he has them”; and yet in this case it is clear that he will not be able to do so. But the blame ought not to be placed on the very idea of belief or propositional content as such, at least not in the first instance. Rather, the source of the problem lies in the fact that Kripke is working both with the requirement that content be transparent and with a notion of propositional content that falsifies that assumption. The impression of a puzzle is generated by keeping two conflicting elements at play at the same time. No wonder, then, that no satisfactory belief ascription to Pierre is forthcoming. A SIMPLE SOLUTION? Our story thus far has unfolded as follows. We have the view that rationality, or at any rate, good reasoning, is the disposition to conform to the principles of logic on an a priori basis. This view, in conjunction with a non-transparent conception of propositional content, yields highly counterintuitive results. Our willingness to exclude de re thoughts in particular, and Millian thoughts in general, from the province of rationality -based psychology manifests our recognition of this fact. It commits us to the assumption, unless we are to forego rationality-based psychology altogether, that de dicto contents are epistemically transparent. Yet most of us adhere to conceptions of propositional content, chief among them externalist conceptions of propositional content, that have it as a consequence that de dicto contents are not transparent. We cannot have it both ways. We must either reject such conceptions of propositional content, or we must show how to refashion the idea of reasoning so that it no longer consists in the disposition to conform to logic on an a priori basis. But isn’t there in fact a simple way of refashioning the idea of reasoning that will take care of the problem cases?22 Well, how would it go? We can’t just say: A person is absolved for believing a contradiction provided that he doesn’tor can’t-‘see’ that it’s a contradiction. Irrationality often consists in precisely such failure. We want in some sense to capture the fact that Pierre or Peter or Paul are blameless for not seeing the contradictions that the stories attribute to them. The question is how is that to be done? The only proposal I can think of is this: A thinker is to be absolved for believing a contradiction, provided that the contradictory character of the proposition he believes is inaccessible to mere a priori reflection on his part. The trouble is that, against the background of a non-transparent conception of propositional content, any contradictory proposition will satisfy that description. Since on a non-transparent conception, it is precisely not available to mere a priori reflection that a given belief is a belief in a contradiction, practically any contradictory belief will be absolvable under the terms of this proposal. The only exceptions will be those beliefs of which it is simultaneously true (i) that they are beliefs in contradictions, (ii) that the subject believes them to be beliefs in contradictions, and (iii) which he nevertheless refuses-mysteriously enough-to change his mind about. Any such subject would undoubtedly count as exemplifying a form of extreme irrationality; but he could hardly be considered the normal case. CONCLUSION There is a pervasive tension between our conception of rationality and the practice of psychological explanation it underwrites, on the one hand, and currently dominant conceptions of mental content, on the other. The former presuppose what the latter deny. One or the other conception must be reconsidered.23, 24