(p.305) B. Reply to Quine on Events
(p.305) B. Reply to Quine on Events
Abstract and Keywords
Davidson starts off by contrasting his semantic approach to Quine's: whereas he reads off ontic commitments from natural languages as they are, Quine is willing to revise them in favour of a ‘language of science’. Davidson points out how that difference in approach has led them to diverging accounts of belief sentences (intensional contexts). He then retracts the individuation conditions he offered for events (in Essay 8) in terms of their causes and effects, conceding the circularity Quine spots in them. Recognizing that ordinary enduring objects with vague boundaries fare no better (explosions and mountains lack clear individuation conditions too), Davidson rests satisfied that events present not only a comparably permissible but also an indispensable ontological category to explain the workings of natural language.
It's a pleasure to be replying to a paper of Quine's, a pleasure for several reasons. One reason is that it has almost always been the other way around, and this is my chance to return the compliment; another is, of course, that there is so much to learn from Van Quine: he always has fresh points to make, and he makes us see familiar points anew. Some of the novelty in replying to a paper of his is reduced by the fact that his paper comments from time to time on work of mine; but then I can console myself by noting that some of his comments concern earlier comments of mine on his work. That is the end of this particular regress; but we can continue the sequence of comment on comment on comment ad libitum if not ad infinitum and I certainly hope we do.
Quine is right, of course, about the reasons I became interested in events and adverbial modification, though his remarks on how the connections between sentences can offset the apparent failure of single sentences to convey all we want put some old problems in a new light for me. Along with all the agreement, though, I also sense a difference in emphasis. Like Quine, I am interested in how English and languages like it (i.e. all languages) work, but, unlike Quine, I am not concerned to improve on it or change it. (I am the conservative and he is the Marxist here.) I see the language of science not as a substitute but as a suburb of our common language. Science can add mightily to our linguistic and conceptual resources, but it can't subtract much. I don't believe in alternative conceptual schemes, and so I attach a good deal of importance to whatever we can learn about how we put the world together from how we talk about it. I think that by and large how we put the world together is how it is put together, there being no way, error aside, to distinguish between these constructions. As Quine pointed out, adverbial modification happened to be the problem that first led me to think about events, and part of the reason for this in turn was an interest in actions, and (p.306) in particular what it could mean to say that two actions are one, especially since the sentences that many of us were thinking of as ‘describing’ actions apparently contained nothing like a singular term or description referring to an action. In any case, a welter of closely related problems soon became apparent. Obviously there had to be some close connection between verbs and their nominalizations, between ‘die’ and ‘death’, and hence between sentences like ‘Mozart died in 1791’ and ‘Mozart's death occurred in 1791’. (These do not have the same truth conditions.) There was also the fact that we use the definite and indefinite articles with verb nominalizations, say that one event is identical with another, quantify over events, and count events under sortals. Causal talk too turned out to be involved. Then there was the problem of causal verbs like ‘kill’ and the transitive form of ‘break’, along with the question how the transitive and intransitive forms of certain verbs are related. (This is an obvious problem only when the intransitive form is not elliptical for the transitive form.) I mention all these problems that are related to the semantics of certain verbs and adverbs only because I want to stress the importance, if one is interested in the semantics of natural languages, of looking for a theory that fits all the relevant phenomena. We have to take the problems one at a time, but we shouldn't be happy ending up with solutions that can't be embedded in a unified theory.
One important consequence of holding that our natural languages embrace our major conceptual tools (they are ‘universal’, as Tarski said) is that we must be able to explain a language by using that same language (well, almost). If we add this idea to Tarski's demand that a theory of truth end up able to prove, for every sentence of the language under study, a biconditional of the form ‘s is true iff p’ where ‘p’ is replaced by a translation of s, we end up requiring that our biconditionals, our T‐sentences, be, in Quine's happy word, ‘disquotational’. Disquotational T‐sentences are an ideal which the vagaries of grammar make hard to achieve. But the ideal has its value nevertheless.
One application of the ideal is this: as long as we do our best to explain a language in (almost) the same language, we cannot make sense of the idea that the logic of the object language differs from the logic of the theory of that language. The merits I find in predicate logic are not just the ones Quine mentions: that it is what science (‘for better or worse’, Quine says) uses, and that we know how to (p.307) produce a theory of truth for a language with the structure of predicate logic. The further merit that interests me is that if we can recognize this structure in a language we can give a disquotational semantics for it: we can use it to explain itself.
These remarks have some applications to the semantics of verbs and adverbs. Anthony Kenny long ago raised the problem of the variable polyadicity (as he called it) of verbs of action, the fact that ‘Eve ate’, ‘Eve ate an apple’, ‘Eve ate an apple in the Garden of Eden’ seem to treat ‘ate’ as a one‐place, two‐place, and three‐place predicate in turn—with no end in sight.1 If this were all there were to the problem, then as Quine points out it would be possible to treat ‘ate’ as a permanently two‐place predicate, true of persons and sequences. Quine explains that ‘slowly’ might then have to be taken as naming some entity; but I think this is hasty. For as Quine says, ‘slowly’ is to adverbs what ‘slow’ is to adjectives: it cannot stand alone. It is a syncategorematic adverb. A slow runner may be a fast 5‐year‐old, and slowly crossing the Atlantic may be rapidly rowing the Atlantic. It all depends on the comparison class. Mention of the comparison class, however, provides the reference needed to turn ‘slowly’ into ‘slow in the class of Atlantic crossings’, thus giving the multigrade predicate idea another lease on truth, though also making overt reference to events (crossings). But this way of treating ‘slow’ and ‘slowly’ already bothers me. For suppose a slow runner is a runner slow in the class of runners. Then a T‐sentence for a sentence containing the phrase ‘slow runner’ will have to iterate ‘runner’ on the righthand side of the biconditional, once to ensure that we can infer that what holds of the slow runner holds of the runner, and once to help describe the class in which the runner is slow. But if the word ‘runner’ requires reference to the class of runners here, it is hard to think why it does not do so elsewhere. Yet if the semantics of any predicate regularly requires reference to the class it defines, and we stick to the disquotational mode, an endless regress starts: each attempt to give the semantics of a simple predicate leads to a complex predicate containing the simple as part. Perhaps we must settle for reference to classes when it comes to the syncategorematic adjectives and adverbs, but I hope not.
Of course turning verbs like ‘runs’ into two‐place predicates true of
runners and sequences won't work anyway, for reasons Quine made clear. The trouble isn't only the non‐existence of appropriate entities, it's the work done by all those little prepositions and their cohorts. Thus there would be no telling, from the fact that Eve ate the sequence
Quine mentions that I once criticized one of his treatments of sentences about propositional attitudes on the ground that it called for an infinite basic vocabulary; and he has replied once before. In ‘Intensions Revisited’ he wrote, ‘Critics of that paper [“Quantifiers and Propositional Attitudes”2] reveal that I have to explain—what I thought went without saying—that the adoption of a multigrade predicate involves no logical anomaly or any infinite lexicon . . . As for the use of quotation, it of course is reducible by inductive definition to the concatenation functor and names of signs.’3 He named no names in that article. But I was one of those who had criticized Quine's suggestion on quotation, and for the fun of carrying on the War of the Quotes (‘You quote me and I'll double‐quote you’), I think there is still the same trouble with the present suggestion. For if we add the name of a sign to our basic vocabulary, we must add a name of that name in case we ever need to refer to it, and then the name of that name . . . This is really a trivial matter, however, since the difficulty could be overcome by using descriptions of the items in the lexicon (e.g. ‘the thirty‐first item in the lexicon’) instead of names.
Indeed, is there anything wrong with treating ‘believes’ as a multigrade predicate true of people, sentences (open and closed), and sequences of objects believed to satisfy the sentences? As Quine pointed out, this would make ‘believes’ much like Tarski's satisfaction, which relates languages, sentences (open and closed), and sequences of objects geared to the free variables in those sentences if any. Perhaps too like. For ‘satisfies’, as applied to a language L, is a predicate that can't, on pain of contradiction, belong to L, while (p.309) ‘believes’ must, to do its dubious work, be in both metalanguage and object language. The threat of paradox seems clear, especially if we take ‘knows’ instead of ‘believes’.
There may be natural ways around this threat. Something else worries me more, and will bring me back to an earlier and more directly relevant point. If the T‐sentence for ‘Ralph believes of Ortcutt that he is a spy’ says this sentence is true if Ralph believes ‘x is a spy’ is satisfied by the sequence
I turn now to Quine's comments on what sorts of thing events are and how they are to be individuated. He says my suggested criterion for individuating events is radically unsatisfactory, and I agree. I accepted the criterion only tentatively, but stressed that I thought it was about as good a criterion as we have for physical objects. Quine has made clearer to me what was wrong with my original suggestion, and I hereby abandon it.4 In the essay in which I put my suggestion forward I considered Quine's alternative: events, like physical objects, are identical if they occupy the same places at the same times. I originally rejected this suggestion (again tentatively) because I thought one might want to hold that two different events used up the same portion of space‐time; and also because I thought the boundaries of events were even less clear than the boundaries of objects. The first concern is perhaps overdone; I speculated that we might identify the rotation of the sphere with its heating up by realizing both events were identical with the history, during that period, (p.310) of the constituent particles. Quine's criterion is neater, and better, since it does not need a scientific theory to back it up.
On the matter of vagueness of location, Quine notes, as I did, that the center of an explosion locates it much as the summit of a mountain locates it, the outer edges remaining vague. But Quine goes on to point out that the vagueness in the spatiotemporal boundaries of objects and events as normally sorted does not compromise the clarity of his principle of individuation. This is true, and it brings home a mild paradox. One might have thought that where counting has a clear application individuation has to be at its best, since we must be able to tell one from two. Yet Quine's clear principle for individuating objects yields no way of counting, while the different and undoubtedly less clear principle that individuates tables does fairly well when it comes to counting. Perhaps it is obvious that individuating items in a grand category like events or objects is quite different from individuating kinds within those categories, such as desks or people; I had not fully appreciated this.
I may also have made the mistake of thinking that, if objects and events are both individuated by spatiotemporal location, we must identify events with objects. But Quine makes us see that this is a separate matter. For events and objects may be related to locations in space‐time in different ways; it may be, for example, that events occur at a time in a place while objects occupy places at times. It is easy, though, to question the distinction. If a wave crosses an ocean, that is an event from the point of view, so to speak, of the ocean. But the wave is also an object in its own right, keeping to a general shape while rapidly exchanging waters. Examples like this are easy to multiply. A lenticular cloud, unlike other clouds, stands still relative to the surface of the earth while the flow of wind on which it depends carries newly condensed particles of water into its defining area while subtracting others by vaporization. From the point of view of the air which contains it, the lenticular cloud is an event; from the point of view of the mountains which caused it, the cloud is an object.
These difficulties in deciding between objects and events are, however, generated by identifying space‐time content with space‐time content. Grammar allows no such confusion. The undulations of the ocean cannot be identified with the wave or the sum of waves that cross the sweep of ocean, nor can the complex event composed of condensations and evaporations of endless water molecules be identified (p.311) with the lenticular cloud. Occupying the same portion of space‐time, event and object differ. One is an object which remains the same object through changes, the other a change in an object or objects. Spatiotemporal areas do not distinguish them, but our predicates, our basic grammar, our ways of sorting do. Given my interest in the metaphysics implicit in our language, this is a distinction I do not want to give up. (p.312)
(1) Anthony Kenny, Action, Emotion and Will. This view of Kenny's is also discussed in Essay 6.
(2) W. V. O. Quine, ‘Quantifies and Propositional Attitudes’.
(4) The proposal here retracted is in Essay 8, p. 179.