1 The classic statement of this point of view may be found in Sellars's ‘Empiricism and the Philosophy of Mind,’ in Sellars, W. Science, Perception, and Reality (London: Routledge and Kegan Paul 1963).Google Scholar

2 For Grice's psychological theory of speaker's meaning, see his paper ‘Meaning,' Philosophical Review, 66 (1957) 377-88; see also his ‘Utterer's Meaning and Intentions,' Philosophical Review, 78 (1969) 147-77. For Grice's program of explicating semantic concepts in terms of speaker's meaning, see his ‘Utterer's Meaning, Sentence-Meaning, and Word-Meaning,’ in Searle, J. ed., The Philosophy of Language (Oxford: Oxford University Press 1971).Google Scholar

3 See for instance the remarks by Chisholm in the Chisholm-Sellars Correspondence on Intentionality, Feigl, H. Scriven, M. and Maxwell, G. eds., Minnesota Studies in the Philosophy of Science, Vol. II, (Minneapolis: University of Minnesota Press 1958) 521-39.Google Scholar

4 Strictly speaking, a definition of denotation that is based on an enumeration of the types of singular terms and their respective denotation-conditions would have to be recursive, and so would not look exactly like (1). I neglect this complication here, since it does not affect my basic point.

5 Kripke seems to be suggesting this at several points in ‘Naming and Necessity,’ in Davidson, D. and Harman, G. eds., Semantics of Natural Language (Dordrecht: D. Reidel 1972).Google Scholar For instance, he seems to be saying (pp. 302-3) that he has not given a theory of the reference-conditions for names, because the picture he presents cannot be used to eliminate the notion of reference. This certainly suggests that a real theory of names could be so used. See also p. 285, and note especially the remark about Bishop Butler's dictum on p. 301.

6 For instance, consider the imaginary token·reflexive term ‘toof’ of an imaginary language L. The following, suppose, is the correct theory of reference for ‘toof’ in L:

Surely, it is possible for a term like ‘toof’ to exist. But (as far as I know) no actual language contains a term of the same type as ‘toof,’ and so the denotation-condition for ‘toof’ would not occur in a definition that has the form of (1). Of course, an infinite number of possible but non-existent types of token-reflexive singular terms can be generated in this way; Just replace ‘“toof”’ in (i) by the name of some other syntactic type, such as ‘ “threef”,’ and replace ‘two’ in (i) by the name of some other natural number, like ‘three,’ and so on.

7 Kripke suggested this idea in his ‘Speaker's Reference and Semantic Reference,' Midwest Studies in Philosophy, 2 (1977) 255-76. For further discussion of the relation between speaker's meaning and speaker's reference, see my paper, ‘Causes and Intentions: a Reply,’ Philosophical Review, 90 (1981).

8 See note 24 below for discussion of this point.

9 Gail Stine suggested that the notion of speaker's reference could be used to construct a Gricean theory of denotation in ‘Meaning Other Than What We Say and Referring,’ Philosophical Studies, 33 (1978) 319-37.

10 Below I will propose that φ should be understood to contain at most free variables y and α.

11 Here and below, I use the expression ‘to follow a rule’ in such a sense that it is possible to follow a rule without obeying it. This may be abnormal usage, but I need a word for the psychological relation that needs to hold between a person and a rule in order for the rule to apply to his behavior, and I am using ‘follows' to express this relation. So in my terminology, following a rule in uttering an expression is something like thinking of one's utterance as subject to the rule, and is not the same as actually obeying the rule.

12 Here I am considering only terms for which there is Just one rule of the form (2) in the language in question. In other words, I am restricting my attention to unambiguous terms.

13 Here and below I assume that rules of the form (2) are all either of the form mentioned in clause (b) of (G1) or of the form mentioned in clause (d) of (G1). In other words, from here on I assume that φ is a formula containing y free and containing at most free variables y and α.

14 By saying that (3) is ‘the correct theory of reference for Win L,’ I intend both to rule out the possibility that (3) merely happens to be true, and to suggest instead that (3) is true by virtue of the meaning that W has in L. In other words, (3) contains what I've been calling the ‘denotation-condition’ for W in L. Hence, this same condition must occur in a denotation-rule of L, and for a proponent of (G1) this implies that the condition occurs in a rule of the form (2).

15 Roughly, one rule or imperative logically implies another Just in case a person could not endorse the former without committing himself to endorsing the latter. A clear account can only be achieved by constructing a logic of imperatives. For an example of such a logic that is adequate for my purposes here, see Hector-Neri Castañeda's Thinking and Doing (Dordrecht: D. Reidel 1975), Chapter Four.

16 The principle I am invoking here is

(P) If R is a rule of l, and R logically implies R', then R’ is a rule of L.

Since several of my arguments here and below against various Gricean theories of denotation depend on (P), a Gricean might hope to avoid these arguments by denying (P). However, unless (P) is true, a Gricean theory of denotation cannot hope to give an adequate account of syntactically complex terms. let l be a language that contains terms of the form ˹α + β ˺. Since l will contain an infinite number of such terms, there cannot be a finite list of basic denotation-rules, each covering an instance of ˹α + β ˺. So there must be a single basic rule in l for all terms of this form (e.g., ‘refer to x with a term of the form ˹α + β ˺ only if x is the sum of the denotations of α and β ’). But then, the denotation-condition for an instance of ˹α + β ˺, say ‘2 + 1’, can only be obtained from a denotation-rule that is derived from the basic rule (e.g., ‘refer to x with “2 + 1” only if x is the sum of the denotations of “2” and “1”). So to be adequate, a Gricean theory will have to treat derived rules as rules of language, and this requires the truth of (P).

17 Here and below, when I speak of a theory of denotation's being ‘based on’ a given form of rule, I mean that the theory is Just like (G1), only with expressions for the form of rule in question replacing clauses (iii)-(b) and (iii)-(d) of (G1).

18 Here and below, forms of rules will be expressed without explicit universal quantifiers to bind the variables ‘s’ and ‘x'. This is Just for the sake of brevity; these quantifiers should be assumed to be implicit wherever they are necessary to bind ‘s’ and ‘x'.

19 The theories based on those rules in the list whose major connective is ‘only if’ imply like (G1) that complex terms are inevitably ambiguous. The theories based on the other rules in the list whose major connective is ‘if’ imply that any term which means the same as a description of the form ˹(ly) (y - a)˺ is inevitably ambiguous. As an example of the latter sort of theory, consider the theory based on (8). Let W be a term of L that means the same as a description ˹(ly) (y - a)˺. If (8) is an adequate concept of a denotation-rule, then L must contain the rule

(i) s is permitted to refer to x with a token α of W, if x - (ly) (y - a).

But, for any property F, the rule (i) logically implies the rule

(ii) s is permitted to refer to x with a token α of W, if x - (ly) (y - a & Fy).

So according to the theory in question, W must also mean the same as ˹(ly) (y - a & Fy)˺ since (ii) must be a rule of L if (i) is.

The arguments against the theories based on (7), (9) and (10) are exactly analogous to the argument against (G1), and the argument against the theory based on (11) is exactly analogous to the one Just given against the theory based on (8). The arguments against the theories based on (12).(15) are similar to those already described, but they require use of an additional principle about inten· ding to the effect that ˹s intends to refer to (ly) (y - a & Fy)˺ implies ˹ s intends to refer to (ly) (y - a)˺. The plausibility of this latter principle rests on that of the principle: ˹ s intends to do both A and B˺ implies ˹s intends to do A˺.

20 See, for instance, Burge, Tyler ‘Demonstrative Constructions, Reference, and Truth,’ Journal of Philosophy, 71 (1974) 205-23;CrossRefGoogle Scholar also, Devitt, Michael Designation (New York: Columbia University Press 1981) 42–56.CrossRefGoogle Scholar

21 In my paper ‘The Ambiguity of Definite Descriptions,’ Theoria, 45 (1979) 78-89, I argue that every definite description ˹the ϕ˺ is ambiguous as between its standard Russellian interpretation and an interpretation on which it means ˹that ϕ˺. From this it follows that every English description of the form ˹the unique ϕ˺ is, on one of its meanings, a demonstrative description. David Kaplan discusses demonstrative definite descriptions in his paper'Dthat,’ in P., French Uehling, T. and Wettstein, H. eds., Contemporary Perspectives in the Philosophy of Language (Minneapolis: University of Minnesota Press 1979) 383–400Google Scholar). Kaplan leaves aside the question of whether there are any such terms in actual languages, but he clearly believes that they are possible.

22 Strictly speaking, (20) is not the right form to be an instance of the form (16). In note 13 above, I stipulated that in the expression of such forms, ‘ ϕ ‘ is to range over formulae that contain at most free variables ‘y’ and ‘a'; in (20), however, the formula following ‘they such that’ also contains a free occurrence of the variable ‘s'. This formal defect is easily remedied by replacing this occurrence of ‘s’ by ‘the speaker of a', to obtain a rule equivalent to (20) that satisfies my restriction. Here and in similar contexts below I use's’ in place of ‘the speaker of a’ merely for the sake of brevity.

23 Here it is again important to note that the speaker is following the rule only in the sense that the rule applies to his utterance, not in the sense that the speaker is obeying the rule. See note 11 above.

24 My argument against (G2) depends upon the fact that (G2) allows a token to denote an object that its speaker is not referring to. ([G1] and the other theories so far discussed also have this feature.) A defender of Grice might want to reject this possibility, and might endorse a variant of (G2) that has an additional clause requiring that a token a denotes an object x only if a's speaker refers to x with a. This theory would certainly avoid my objection based on demonstrative descriptions. But it does so only by requiring that every term must be a demonstrative description; and surely, any theory that requires this is false.

For instance, definite descriptions like ‘the inventor of bifocals’ denote whichever objects uniquely satisfy their matrices, regardless of what their speakers are referring to. Even if Donnellan is right that this is true only of descriptions that are ‘used attributively,’ it is still nevertheless clear that there are such attributively used terms: see Donnellan, Keith ‘Reference and Definite Descriptions,' Philosophical Review, 75 (1966) 281–304.CrossRefGoogle Scholar Kripke has persuasively argued that proper names also may denote objects that their speakers are not referring to: see his ‘Speaker's Reference and Semantic Reference,’ op. cit. But even if these views about the descriptions and names of actual languages are wrong, it is surely at least possible for there to be languages whose terms behave in the way these views describe. And this mere possibility is enough to refute the idea that a term can only denote an object that its speaker is referring to.

25 Surely, any natural three-place relation M is expressible by a (possible) predicate M* that in turn could occur as part of a (possible) definite description of the form ˹(ly) (y - a & b bears M* toy and c)˺, where a, b, and care context-independent terms. But then the relation M would be part of the denotation-condition for this possible definite description. And it is very implausible to suppose that M could be part of the denotation-condition for such a definite description, and yet could not be part of the denotation-condition for a token-reflexive term that conforms to a principle of the form expressed by (25).

26 The argument against the theory based on (26) is exactly analogous to the argument against (G2). The argument against the theory based on (27) depends on the additional principle that ˹s intends to refer with a to they such that y - a and s refers toy with a˺ is equivalent to ˹s intends to refer with a to they such that y - a˺. The plausibility of the latter principle rests in turn on the plausibility of the principle that ˹s intends to both do A and do A˺ is equivalent to ˹s intends to do A˺. This is because ˹s intends to refer with a to they such that y - a and 5 refers toy with a˺ Just says in effect that5 intends to both refer to a with a and refer to a with a.

27 Here it is worthwhile contrasting self-conditioning rules with tautological rules (such as ˹s is permitted to do A only if either p or not-p˺). These are easy to confuse, since tautological rules are also impossible to disobey. But selfconditioning rules are not tautological, since tautological rules are logically implied by every rule and self-conditioning rules are not. For instance, we’ve seen that the self-conditioning rule (33) is not implied by (30). See note 15.

28 In a way, tautological rules are pointless too, but we can't help adopting them, since they are logically implied by every system of rules. But since selfconditioning rules are not tautological (see note 27), we can avoid adopting them, and since they are pointless, we would avoid adopting them.

29 See note 25 above.

30 Searle, John Speech Acts (Cambridge: Cambridge University Press 1969) 33Google Scholar

31 Brian Loar sketches a Gricean theory of meaning that makes this assumption in his paper ‘Two Theories of Meaning,’ in Evans, G. and McDowell, J. eds., Truth and Meaning: Essays in Semantics (Oxford: Clarendon Press 1976);Google Scholar see pp. 153-5.

32 I was originally stimulated to think about these matters by Hector-Neri Castaf'leda, when he was serving as director of my Ph. D. dissertation ('The Reference of Proper Names,’ Indiana University, 1976). The ideas in section 1 originally appeared in the Introduction of my dissertation. An earlier version of this paper was presented to the Pacific Division of the American Philosophical Association (March, 1981). Brian F. Chellas served as commentator on that occasion, and I am grateful to him for his lucid and valuable remarks. For helpful comments and suggestions, I am also indebted to Hector Castañeda, Carl Ginet, Richard Grandy, Lawrence Lombard, Lawrence Powers, Neil Wilson, and a referee for the Canadian Journal of Philosophy.