Experience

Ernest LePore, who must count by now as a leading entrepreneur of analytical philosophy, has edited the proceedings of the 1984 Rutgers conference on the philosophy of Donald Davidson. The scale of that conference is reflected in the size of this volume, which contains 28 substantial papers. And this is but half the story: a companion volume of similar size, drawn from the same conference, and dealing with Davidson’s essays on actions and events, was published simultaneously with Truth and Interpretation. Despite the size, many of the papers in the volume under review repay careful study: the book is a fitting tribute to the distinguished writings it discusses.

The editor has usefully reprinted two relatively inaccessible papers by Davidson, and included a new one. After an editorial introduction locating the subsequent essays in relation to Davidson’s work, the volume is divided into five sections: on truth and meaning; on the application of Davidson’s ideas in semantics; on radical interpretation; on epistemology; and on ‘the limits of the literal’. Instead of commenting on every paper, I will consider a few in slightly more detail. But within the many successful papers not mentioned below, the contributions of Akeel Bilgrami, Tyler Burge, Michael Dummett and James Higginbotham include some particularly penetrating points.

A major question arises for every type of content a belief may have. What is the relation between theories of those contents which individuate them by reference to truth conditions, and theories which claim to individuate them without any such reference? Are such theories in competition, or not? And if not, how should theories of each sort relate to one another? In many areas, our current understanding is not sufficiently far advanced to allow very precise theorising on these issues. But our understanding in the special case of logic is better than elsewhere, and these questions are ones which arise in connection with Gilbert Harman’s paper ‘The Meanings of Logical Constants’.

Harman argues that the meaning of a logical constant such as ‘and’, ‘not’, ‘for all’, is determined by the relations of immediate implication in which it stands. Immediate implication is a psychological notion: an immediate implication is one that can be immediately recognised. So – to take the easiest case – conjunction can be defined as that operator which applied to two sentences P and Q yields a sentence with these three properties: it immediately implies P, it immediately implies Q, and is itself immediately implied by the set consisting of P and Q. Harman displays considerable ingenuity in developing the account for harder cases. He says that the meanings of logical constants are not given by the contribution they make to the truth conditions of sentences containing them. He is led to the conclusion that truth conditions are no more relevant to the meaning of logical words than they are to the meaning of non-logical predicates such as ‘horse’.

Details aside, is Harman offering something incompatible with a truth-conditional theory? A view contrary to Harman’s would be that what he is doing is giving a substantive account of what is involved in understanding constants whose sense is specified by a theory of truth conditions. In fact, precisely the classical assignment of contributions to truth conditions would validate the transitions Harman classifies as immediately obvious. A truth-condition theorist should not say that one can simply read off the correct substantive account of understanding a particular expression from its truth-theoretic clause. His position should rather be that the correct substantive account – an account of what Harman would call its conceptual role – determines its contribution to truth conditions.

Harman does offer a specific objection to giving the meaning of logical constants in truth-conditional terms. His objection is that a constant which operates on P and Q to produce a sentence synonymous with P and Q means something different from one which operates to produce a sentence synonymous with not-( not-P or not-Q). Surely they do have different meanings: but an alert truth-condition theorist should be asserting this point, not denying it. Harman seems to be assuming that logically equivalent sentences must be counted as having the same truth condition. But this would not be so on the theory of that great originator of truth-conditional theories, Frege. Nor would the clauses for the two constants in a Davidsonian truth-theoretic implementation of the conception say the same thing. For one constant, the relevant clause would say that a sentence in which it operates on P and Q is true just in case P is true and Q is true. For the other constant, the clause would say that the corresponding sentence is true if it’s not the case that either P is not true or Q is not true.

If truth conditions are determined by conceptual or inferential role at least in the logical case, are they a theoretical epiphenomenon – a varnish we can add after the real depiction of meaning has been given using other materials? They are not mere varnish, because anyone who cares about the distinction between a mere impression of validity and genuine validity will need to appeal to the notion of truth. Harman writes, naturally and intuitively, of recognising an implication. I do not think the distinction between a genuine implication, there to be recognised or not as the case may be, and a mere impression of an implication, can be elucidated without using the notion of truth: one sentence really implies another only if the second must be true when the first is. The relations between truth conditions and understanding in the logical case are of course subtle, and deserve a more extended discussion. Moreover, if a claim of determination of truth conditions by inferential role can be sustained, it is at least an open question whether a comparable determination claim about the conceptual role of the non-logical concept ‘horse’ may not equally be true. But precisely because Harman’s account may after all be consistent with a truth-conditional conception, some of his ideas may have a wider appeal than his own claims allow.

The longest and most sharply focused section of the volume is entitled ‘Language and Reality’. Almost all the papers in it address the arguments in Davidson’s paper ‘A Coherence Theory of Truth and Knowledge’. Davidson argues that there is a presumption in favour of the truth of a belief that coheres with a significant mass of beliefs. If he is right, massively false belief is impossible.

Davidson argues from the identity of the content of beliefs. As he formulates it, the argument invokes two considerations. First, an interpreter should, roughly, take the conditions under which a speaker assents to a sentence to be the truth conditions for the sentence. Second, there is an argument that the interpreter’s beliefs must be largely true. For we can consider an interpreter whose beliefs about the non-intentional world are all true. This second interpreter will make the first, fallible interpreter largely right. So in turn the original interpretee will be largely right. Thus there is a presumption in favour of the truth of a belief which coheres with a significant mass of beliefs, because this mass of beliefs must be largely true. For Davidson, coherence yields correspondence.

This is a slightly circuitous argument, but the detour is avoidable. If the argument is sound, then correct interpretation is, roughly, given by correlating what is held true with what is really the case. We can cut out the first, fallible interpreter; while the knowledgeability of the second interpreter was simply to ensure that what is held true is correlated with what is true. Cutting out the detour, we still have a claim of substance: that correct interpretation involves assigning beliefs largely true, but in any case intelligible in the light of the way the world really is.

Davidson’s is a theory about the truth of beliefs: it is not meant to be a definition of truth. He explicitly takes truth as primitive. Of definitions of truth in terms of justified assertibility, or anything else guaranteed to be epistemically accessible, he remarks that they are shotgun weddings ‘which do not marry the original mates’. Davidson is also a coherence theorist in that he holds that only another belief can count as a reason for holding a belief. He devotes some pages to arguing that perceptual experiences cannot justify beliefs.

In his contribution, Richard Rorty categorises Davidson as significantly close to the American pragmatists. Rorty lists four theses he says are common to Davidson and a pragmatist such as William James. The fourth thesis states that there is no point to debates between realism and anti-realism, because these debates presuppose the empty notion of beliefs being made true by something. In other writings, Davidson has indeed done as much as anyone could to scotch the idea that there is something relevant to the truth of sentences besides the meanings of words and the way the world is. There is no third intermediary, no ‘content’ of a sort to which a scheme/content dualism can be applied. But what is the connection with realism? Realism is, with refinements, the doctrine that a sentence can be true though unverifiable. Certainly this is the central doctrine in dispute in the recent literature Rorty cites. Realism in this sense does not imply that there is some third thing which makes sentences true; nor does anti-realism. In accordance with his self-description, Davidson falls squarely on the realist side of this dispute. If he did not, he would not take truth as primitive for the reasons he does. He would regard a wedding of truth with warranted assertibility as made in heaven. Part of the reason he does not is that he is committed to distinguishing the ontological issue which exercises Rorty from the issues about the nature of semantical properties at stake in the dispute over realism.

Is Davidson right to assign perceptual experience so small a role in his account of the determination of the content of attitudes? Two fundamental classes of content-constituents are individuated in part by their links with perceptual experience. One is the class of singular demonstrative perceptual ways of thinking of an object, ways of thinking made available to thought by a perceptual experience, and naturally expressed in such phrases as ‘that cat’. The other is the class of relatively observational concepts, mastery of which requires a certain sensitivity in their use to perceptual experiences as of objects falling under them. Without such concepts, it is not clear how a subject could think about the world around him. These links with experience also need to be exploited if we are to conform to two Davidsonian requirements: those of explaining intentional action and exhibiting the rationality of belief formation. Because the subject perceives a glass to be in a particular direction from him, he reaches out in that direction and not any other when he intends to drink from it. Because the subject perceives something as square, it is rational in suitable circumstances for him to believe it is square; there are circumstances in which it would not be rational without the perceptual experience.

Davidson fires a battery of arguments against the claim that perceptual experience can justify beliefs. Undoubtedly much further work remains for someone who wants to defend that claim. But in my judgment, none of Davidson’s arguments destroys his target. He asks us to consider whether someone would be justified in believing that a green light is flashing if, even though he has an experience as of a green light flashing, he believes he does not have such an experience. But it seems that, in the relevant cases, this counterfactual has an impossible antecedent. A case can be made that it is required for possession of a concept of experience of a given type that the subject be disposed to judge that he has such an experience when in fact he does so. Again, Davidson implies – rightly, in my view – that only content-possessing states can justify. But experiences do have a representational content concerning the world, and meet this requirement. Davidson brings several other arguments, which would have to be considered one by one. Underlying many of them is a desire to avoid epistemic intermediaries which make a thinker’s contact with the world problematic. But his arguments leave room for a conception of experience on which it is not an intermediary, but a state which makes legitimate a description of the subject as related to the world itself. This is the conception implicit in our pre-theoretical talk of perceiving things around us, and of perceiving that the door is open.

It might be said that Davidson’s procedure for radical interpretation already adequately accommodates experience. ‘If an interpretation procedure generates the correct beliefs and desires for a subject, and some fundamental beliefs have constitutive connections with experience, the same procedure must implicitly determine that the relevant links with experience are fulfilled; we do not need to consider them explicitly.’ But there must be something wrong with the last step in this speech, because a parallel argument would allow us to leave desire out of explicit account. ‘If an interpretation procedure generates the right beliefs for a subject, and belief has various constitutive links with desire, the same procedure must implicitly determine that the relevant links with desire are fulfilled; we do not need to consider them explicitly.’ The significant question is not whether there is implicit determination, but whether any procedure yielding detailed attitudes of one type must also determine a range of detailed attitudes of some other given type. Davidson would say that this is so for belief vis-à-vis desire; I would say that it is so for both belief and desire vis-à-vis experience. What we ought to be explicit about are the simultaneous equations to be solved which link experience, belief, desire, intention and meaning with action, just as Davidson aims to be explicit about these equations when experience is not on the list.

In his lucid paper, Colin McGinn also urges that experience be considered explicitly in an interpretation procedure. But I doubt that it is possible to incorporate it in the way McGinn envisages. McGinn’s view is that a radical interpreter can assign detailed contents to a subject’s perceptual experiences in advance of solving for his detailed beliefs and desires. ‘Having identified the subject’s sense-organs and their receptive condition, we then make a suitable ascription of experience. Thus if, for example, the subject’s eyes are open and a running rabbit is in his line of vision, then we suppose that he has an experience as of a running rabbit.’

But there is no such thing as the state of affairs in the line of vision as this conception requires. A person unfamiliar with modern culture may see an object as an upright plank with a board attached; another may see it as a noticeboard; a third, a Russian familiar with English letters but unable to understand English, may see it as a notice in English; a fourth will perceive it as a sign saying: ‘Danger! Do not cross this line.’ Their actions will differ accordingly. It is hard to see how to fix the right level of conceptual richness of the content of the experience without drawing on its impact on intentional action, via belief and desire. McGinn says that Fodor’s modularity thesis about the perceptual systems shows that there is no dependence of the content of experience on the content of beliefs and desires. But we ought to distinguish causal independence from constitutive independence. The fact, if it is a fact, that a subject’s beliefs and desires do not causally influence the content of his perceptions does not imply the following: that what makes his perceptions have the content they do is independent of the judgments they cause when taken at face value.

An alternative to McGinn’s view would be an interpretational account. On this alternative, the content of an experience is one which makes intelligible the actions the experience causes, and would make intelligible the actions it would cause in counterfactual circumstances. This would have to be restricted to cases of causation in which the experience functions, in the context of other attitudes, as the reason for the beliefs which in turn cause the actions in the appropriate way. On such an interpretational account, variation in conceptual richness of experiential content is only to be expected. An interpretational account is just a beginning. It is an over-arching constraint: there will be many more specific constraints on the ascription of particular contents to experience. But it is only when we have a theory of the determination of the content of experience that we will be in a position to describe the limit on intelligible forms of radical scepticism.