1 This is a somewhat recent development; the general attitude in formal semantics towards such sentences was exemplified by Carnap who, in 1937, simply excluded such sentences from the languages he investigated. See The Logical Syntax of Language (London, 1937), p. 168.

2 Some might prefer this description to the first, or vice versa; I see no real issue here, but if others do, they may simply choose the description they prefer - the claims I make later will not be affected.

3 In each case, it is possible to give a specification of what is said that applies to any speaker. For instance, anyone who utters ‘I am hot’ says of himself that he is hot; less informatively, any speaker who utters ‘John will be late’ says of some specific person named John that John will be late, and so on. The sense of ‘say’ I am concerned with is the one involved in giving indirect speech reports, e.g., the sense involved when we say that Nixon said that Richard Milhouse Nixon is president when he uttered ‘I am the president'. To say that the speaker said of some person named Nixon, perhaps contextually indicated, that he is the president, does not quite provide such a report. Nor does the ambiguous remark 'Nixon said that he is the president'.

4 The claim that what is said, or truth or falsity, is determined by the context of utterance can be found in the pronouncements of various writers. Usually, though, it is by no means clear that this claim is to be taken literally (or, what all can be included in the context). I have singled out Ziff because it is clear that he is committed to this claim, and a fairly strict construal of ‘context'.

5 I take this to be one consequence of Donnellan's comments in “Reference and Definite Descriptions” (The Philosophical Review, 75 (1966), pp. 281-304): someone might by uttering ‘The man with the martini is a famous philosopher' under certain conditions say that the man drinking water is a famous philosopher. If so, then what the description means does not even constrain what can be said in the way we might have initially thought. See “The Semantic Significance of Donnellan's Distinction”, Philosophical Studies, 37 (1 980), pp. 281-88.

6 I would include ambiguous sentences and sentences containing quantifiers (whose scope may vary) in this class, though this strains our ordinary notion of in· dexicality somewhat. More contentiously, I would also include definite descriptions, most of which are not definite at all. It seems quite clear that English expressions of the form ‘the ϕ’ function much more like expressions of the form 'that ϕ’ than Russell would have been willing to allow in framing his notorious theory of descriptions. I think, moreover, that this claim can be reconciled with Russell's theory in a way that Russell himself suggested much later, but that is a rather long story which I am not prepared to tell here.

7 One could begin to make a case for the claim that for certain sentences, including perhaps ‘I am hot', what they mean does by itself determine what the speaker says. But it is clear that this is not sufficient; we must at least add that the speaker is using the sentence ‘seriously and literally'; but then we need to know what is demanded by those adverbs. Saying that the speaker must mean what the sentence means may be true, but it does not really help much -sooner or later we have to get out of this circle if we are to understand what is at work.

8 Ziff, “What is Said”, in Davidson, D. and Harman, G. (eds.), Semantics of Natural Language (Dordrecht, 1972), pp. 709-21;CrossRefGoogle Scholar seep. 709.

9 Ibid., p. 713. Ziff is not entirely consistent in his use of ‘verbatim reporting'; perhaps he would respond that there isn't much to be consistent about. What is said is, as he points out, slippery stuff . .If George says ‘I want a banana’ and Josef says ‘He wants a banana', “where the reference is to George, then each has said the same thing, namely that George wants a banana” (710). But, “Suppose different persons offer the following reports about a UFO: The object had at least thirteen faces', The object had sixteen faces', The object had many sides',…; in a sense they are all saying the same thing, namely, that the object was a polyhedron“ (710). Ziff goes on to describe cases in which the sense of ‘what is said' can become even further attenuated. But indirect speech reports seem to be what are at issue in the examples I discuss, and I shall assume that in the UFO case the sense in which what is said is that the object is a polyhedron is a sense involving an abstraction from or generalization of the particular things various speakers say.

10 Ibid., p. 712. Notice that this is not obviously true. The first step in understanding how we understand that something is water, for instance, is not to investigate what determines that it is water. We did not have to develop chemistry, let alone physics, prior to carrying out this task. And, it is by no means clear that coming to grips with the factors which determine what is said would be the first step toward understanding how we understand what is said. The factors involved in interpreting utterances may be, and I shall argue are, quite different from those which determine what gets said. (Cf. note 18).

11 Ibid., p. 713. Later (716) he notes that given our current understanding, “Such a supposition is best seen as at best a heuristic maxim”.

12 Cf. Grice's theory of implicature in “Logic and Conversation”, in Cole, P. and Morgan, J. (eds.), Syntax and Semantics, V.lll (New York, 1975), pp. 41–58.Google Scholar

13 Ziff, op. cit., pp. 714-15.

14 lbid.,p.715.

15 I might, after all, fall over dead in mid-sentence. This might leave you permanently in the dark as to what I have said, but it would not leave what I said undetermined.

16 Ziff, op. cit., p. 715. Note that morpheme identification can only be the task of the hearer, who is trying to interpret an utterance.

17 Ibid., p. 716.

18 In a certain way, Ziff's confusion is fortuitous. His general worry is how we understand what is said, to the extent that we do, and what my criticisms show is that he has been investigating that question all along, rather than what he claims to be investigating. Notice, however, that this suggests that Ziff's methodological remark - that the first step in understanding how we understand what is said would be to gain some insight into the character and operation of the factors that serve to determine what is said -should be ignored. Ziff's methodological advice is undermined by the rest of his paper, as we might have expected (Cf. note 10).

19 See, for example: Montague, R., “Pragmatics and Intensional Logic”, in Thomason (ed.), Formal Philosophy (New Haven, 1974), pp. 95–118;Google ScholarWeinstein, S., “Truth and Demonstratives”, Nous, 8 (1974), pp. 179-85;CrossRefGoogle ScholarKamp, H., “Formal Properties of ‘Now'”, Theoria, 37 (1971), pp. 227-73;CrossRefGoogle Scholar R. Stalnaker, “Pragmatics”, in Davidson and Harman (op. cit.), pp. 380-97. The list is hardly complete.

20 Montague, op. cit., p. 98.

21 Weinstein, op. cit., p. 184.

22 Ziff, op. cit., p. 709. He makes the point with respect to accents: someone who utters ‘I vant a banana’ has said that he wants a banana.

23 In the first four chapters of my doctoral dissertation.

24 Perhaps this will only hold for subject-predicate sentences. The introduction of quantifiers, for example, produces problems; it is not clear that someone who asserts ‘All Americans are bourgeois fools’ refers to all Americans. For my part, I'm not convinced that such an abuse of ordinary speech is not simply to be tolerated.

25 The objection is due to Martin Huntley, who made a related point in a different context; the response is due to Dennis W. Stampe (both in conversation). I wish to thank Stampe, and Lilly-Marlene Russow, for useful discussion and advice.

26 Even if we waive the ‘serious and literal’ rider, this assumption is not all that safe. For what if the marriage ceremony was performed by an imposter, the couple has not lived together long enough for common law to apply, etc.?

27 Strictly, we should speak of the phonetic contour of the utterance, but I will ignore that complication here.