See also Brandom (1994: xiii, xxii, 7, 61, 155, 190) for some other significant statements.
It seems plausible to assume that these attitudes can only have non-conceptual content, despite the fact that Davidson arguably doesn't allow for (literal) non-conceptual content. As I suggest below that this may also be the case of Brandom's practical deontic attitudes, the analogy seems warranted.
A view which he shares with Davidson, and generally, with all those who hold that thought depends on language.
In other words, I submit that X depends on Y if and only if it is impossible to account for X without either invoking Y or being in a position to provide an account of Y as well.
And we have agreed above to understand the claim that non-conceptual content depends on conceptual content as implying just that.
Which is not to say that all systems capable of non-conceptual intentionality would therefore count as derivatively intentional systems, for among these, some would be capable conceptual intentionality as well. Hence, the claim that some system is originally or derivatively intentional should not be confused with the claim that conceptual (non-conceptual) intentionality is original (derivative) with respect to non-conceptual (conceptual) intentionality.
But it must be admitted that Brandom doesn't give any clear indication that he would embrace this position, and gives instead the impression (see, e.g 1994 : xiii) that the practical deontic attitudes that are supposed to institute deontic statuses are to be taken as non-intentional at all, which is the source of some puzzlement. On either view, however, it could be granted that not any practical deontic attitude is to count as a practical attribution of a conceptually contentful state, or even of a contentful state in general, for not all practical deontic attitudes are part of practices exhibiting the right kind of structure and complexity. Brandom's account would therefore be immune to at least one kind of circularity.
Even though it is perhaps less clearly untenable than (8).
It should also be pointed out that (11)-(12) are incompatible with the assumption (made in section 6) that non-conceptual intentionality doesn't depend on conceptual intentionality.
However, this would require committing oneself to the claim that even non-conceptual content is essentially normative; and this would need some justification.
Part of this paper has been read at the annual meeting of the CPA, at the Université Laval, on May 26th 2001. I am grateful to Don Ross for his stimulating comments.
Donald Davidson, `What Metaphors Mean', Critical Inquiry 5 (1978): 31-47; reprinted in Sheldon Sacks (ed.) On Metaphor (Chicago: University of Chicago Press, 1979). 29-45, especially 44-45. Page references are to Sacks.
Monroe Beardsley, `Metaphorical senses', Noûs, 12 (1978), 3-16.
John R. Searle, `Metaphor', in Andrew Ortony (ed.) Metaphor and Thought (Cambridge: Cambridge University Press, 1979), 92-123. See 93.
Max Black, `How metaphors work: a reply to Donald Davidson', in Sacks (op.cit.), 181-192. See 192.
Susan Haack, `Surprising Noises: Rorty and Hesse on Metaphor', Aristotelian Society Proceedings, New Series 88 (1987/1988), 293-301. See 299.
Susan Haack, `Dry Truth and Real Knowledge', in Hintikka (op.cit.), 1-22. See 18.
See, for example, Frege 1950, par. 88.
This can be found in Dummett 1973b, p. 297 and also in Dummett 1991, pp. 175-6.
See Terquem & Gerono 1851, pp. 106-119 for the French translation of Euler's paper (1736).
One attempt to dismiss Dummett's problem was made by Susan Haack (1982, pp. 225-227). According to her, the tension between the requirements of validity and fruitfulness can be relieved if we distinguish between two senses of deduction. The first is that of deductive implication, which she characterizes as the relation of logical consequence between propositions. The second sense of deduction is that which corresponds to the intentional act of making an inference, which she calls `deductive inference'. Haack's claim is that validity is a property of deductive implications whereas fruitfulness applies only to deductive inferences. This strikes me as incorrect, since deductive inferences can also be evaluated as valid or not; neither can I see why the pair of predicates `fruitful/non-fruitful' could not be attributed to deductive implications.
See Quine 1951, sections 5 and 6.
Dummett 1973a, pp. 4-5.
As someone who accepts Frege's context principle (which makes its first appearance in Frege's writings in Frege 1950, introduction) must admit.
For instance, in Dummett 1991.
Wittgenstein 1922: 3.263, 3.3.
The terminology is not Dummett's. I took it from a paper on meaning holism by Eric Lormand (1996).
See, for example, his celebrated Two Dogmas of Empiricism (1951).
I was convinced of this by his paper on malapropisms (Davidson 1986).
Crispin Wright reminded me of this weakness of Dummett's distinction.
See, for instance, Dummett 1973b, pp. 300-305.
Dummett 1991, p. 252.
I- and E-rules, for short.
Dummett 1991, p. 229.
For example, in Dummett 1991, pp. 218-9.
This is in Prior 1960, p. 130.
In Pinto 1998, chapter 7.
We discussed this property of statements in the last section in connection with the verificationist notion of conservative extension.
For example, in Dummett 1973b, pp. 314-16.
See, for instance, Dummett 1973b, p. 314. There are, according to Dummett, two reasons why the moderate verificationist notion of truth must have such a convoluted expression. On the one hand, truth cannot be equated with verification by a direct means because this, as our discussion of the radical verificationist shows, would make it impossible to explain the usefulness of deduction. On the other hand, truth must be related to the direct method of verifying statements otherwise one would fall into some version of verificational holism. This last contention is far from straightforward unless one adds the further assumption that there must be a close connection between meaning and truth. Such a connection was first rendered explicit by Frege and Dummett is prepared to endorse it provided that truth is understood according to the moderate verificationist canon (see, for example, Dummett 1991, chapter 6).
This is corroborated, for example, by Dummett 1973b, p. 297 and Dummett 1991, p. 176.
Dummett 1991, pp. 197-199.
Dummett 1994.
Throughout this paper I shall intend first or literal meaning by `meaning', i.e., that notion which is constrained by (TM).
See, e.g., Etchemendy (1988). Davidson's (1990a, §1) response to Etchemendy is, in essence, that we shall not necessarily fall into inconsistency if we forgo the explicit definition and read a definition's clauses as substantive axioms. But, if Tarski is to be read as showing us that truth is inherently compositional, then it can only be via satisfaction, yet truth and satisfaction are related in Tarski's method only via explicit introductions of truth predicates in terms of satisfaction definientia. Tarski did not recursively define truth; he recursively defined satisfaction, from which a truth predicate may be explicitly introduced. If we introduce a primitive truth predicate to occur in axioms (as presented in §2), then, clearly, one cannot say that Tarski has shown us how truth is essentially tied to the interpretation of sub-sentential structure, for, by such a method, truth is not defined in terms of satisfaction. As for explicit definitions, as Davidson well recognises, they are not expressive of compositionality: they amount to conditions on membership of a set TRUE. Suffice it to say, none of this is inimical to the method advertised in §2.
There is an interesting contrast here between early and later Davidson. The early papers look upon Tarski's work as offering one way to get at compositionality (see especially Davidson (1984, p. 61)). That is, there is no direct conceptual relation between Convention T and compositionality. This thought is clearly correct: Convention T is a criterion of adequacy that does not imply how a theory or definition should satisfy it. Likewise, a theory of truth, on Davidson's construal, should issue in interpretive instances of (T), but this demand does not in itself tell us just how this should be achieved. In his more recent writings, however, Davidson appears to argue that the structure of truth is conceptually necessary to thought or meaning (especially see Davidson (1990a, p. 296; 2000, p. 72)). I do think (see below) that we haven't got a good idea how to account for compositionality apart from a satisfaction-like relation; even so, if we were offered a non-truth theoretic explanation of compositionality that was not obviously mistaken, it would be dogmatic to say that there must be something wrong with it because it failed to deal in reference and truth.
The main texts are Fodor and Lepore (1991; 1992) and Fodor (1998a; 1998b, Chps. 4 and 5). Fodor, it should be said, is principally concerned with mental content rather than linguistic meaning, although, clearly, if we cannot account for the compositionality of thought with theory T, then we should not expect T to work with meaning.
Horwich (1998, pp.22-7) does identify what he terms the «constitution fallacy»: the idea that any analysis of a complex property must preserve its componential structure in its analysans. Thus, Horwich's claim is that the uniform notion of semantic value simply issues from the dogma that there must be some property to realise the uniform relation in each fact of the form `x means y'. I suppose this is a fallacy, but it has nothing to do with the issue at hand, for (i) a truth theory does not analyse meaning properties and (ii) the uniformity of semantic value is not based upon any general principle of constitution, fallacious or otherwise, but on what appears to be demanded by the phenomenon at hand.
The problem was especially vivid for Russell because his propositions (in 1903, at least) were made up out of particulars and universals, the very stuff our words are about, not mental representations of that stuff. Thus, realism vs. idealism turned on whether we can have an understanding of a thing independent of grasping the set of relations into which it enters. If not, then relations are internal, and we are landed with the One, as it were.
Davidson takes this kind of constituency to entail holism. This thought is correct on an interpretive understanding of the aim of a theory of meaning, for what is to be interpreted are sentences not words, words get interpreted simply in terms of their sentential role. I do not, however, share Davidson's assumption. If a theory of meaning is something a speaker genuinely knows, then sentential structure may be understood as projection from the lexical entries as represented in the speaker's mind. See §5.
This kind of complaint was made by Davidson (1984, pp.17-8) against Frege's notion of unsaturatedness. Higginbotham (1999) deploys the complaint against Horwich, who sees his schemata solution as being essentially Fregean. I was, however, unaware of the latter paper at the time of formulating my current ideas. Besides which, neither Davidson nor Higginbotham trace the issue back to Russell, where it belongs. The points I am making were influenced by Sainsbury (1997a), who relates Russell to Davidson.
A tempting line for the disquotationist is the attempt to generalise (T) in some manner so that it is no longer tied to particular languages. Such an attempt, however, appears forlorn (see Gupta (1993), also Davidson (1990a, p. 295-6)). A more promising line is one, very roughly, where truth is taken to be primitive, with assent to instances of (T) (or a propositional analogue) being taken to be necessary and sufficient for possession of the concept of truth (see Horwich (1998) and Soames (1999)). This is promising, however, only in the sense of avoiding the immediate generalisation problem; serious difficulties remain concerning compositionality and other factors (see Collins, J. (forthcoming). `Truth or Meaning? A Question of priority', Philosophy and Phenomenological Research).
This kind of approach to the theory of meaning (as truth theory) is set out in Higginbotham (1985, 1986) and Larson and Segal (1995).
It might seem here that a traditional worry resurfaces, viz., that, for many constructions (famously, context variable ones), the right flanks of instances of (T) will be ampliative in a way which either obviates the `absolute' reading of truth or vitiates the instances' intuitive acceptability. The worry is misplaced. There is no demand for homophonicity. The requirement, rather, is that the instances of (T) record intuitively acceptable paraphrases on their right flanks of the sentences described to their left. The paraphrases will be as ampliative as the lexical axioms from which the theorems are derived. That a truth theory has such consequences while cleaving to compositionality is what, in part, corroborates the hypothesis that a speaker's semantic competence is constituted by knowledge of the entailing theory. I think that there is good evidence that this demand can be met without a retreat to model theory. The issue, though, I should say, is wholly empirical.
Curiously, Davidson (1997b, pp. 20-1) claims to have no particular problem with language acquisition being largely governed by innate principles, and he cites Pinker's (1995) survey of the evidence in support of this nativism. He then, however, goes on to offer what is essentially an ostensive learning by negative evidence model (triangulation) of conceptuality as such, including lexical acquisition (also see Davidson (1997a, 2000)). But, famously, negative data appears to play no role at all in language acquisition; certainly the ostensive model does not square with the lexical `explosion' that takes place around 24-30 months. Elsewhere, Davidson (1989, p. 164) is more conciliatory: «Of course very many words and sentences are not learned [by ostension and correction], but it is those that are that anchor language to the world». I know of no data, however, to support the claim that a certain class of words is learnt by ostension. Besides which, the assumption that language is anchored to the world begs the question in favour of the social externalism which is presently in question.
One might argue that there are syntactic constraints, what is social is `meaning'. It is difficult to take such arguments seriously without being given a principled cognitive distinction between syntax and semantics (see below).
It does not follow, of course, that grammatical information is consciously accessible. Although sometimes it is, and when it is not, it clearly shapes our choices and judgements rather than simply causes what we say.
Quine, W.V.O., `Two Dogmas of Empiricism' in From a Logical Point of View, Harper Torchbooks, Harper and Row, New York. Second Edition. 1963. Pages 20-46.
ibid. Pages 22-23. The italics are Quine's.
ibid. Page 43.
ibid.
ibid. Page 24.
See, for example, Kripke, S., `Identity and Necessity', reprinted in Moore, A. (ed.) Meaning and Reference, Oxford Readings in Philosophy, Oxford. 1993. Pages 162-191.
See, for example, Evans, G. `The Causal Theory of Names', reprinted in Moore, A. (ed.) Meaning and Reference, Oxford Readings in Philosophy, Oxford. 1993. Pages 208-227.
More robustly descriptivist analyses are possible here. But these may come at a price. On a Russellean analysis, for example, it would seem to follow that there might be worlds in which such terms failed to refer, i.e. worlds in which orthographic identities might be false. To adopt a descriptivist position does not entail accepting every detail of Russell's analysis of descriptions, however. We may, for example, sympathise with Strawson's critique of Russell's theory (see, for example, Strawson, P.F., `On Referring' reprinted in Moore, A. (ed.) Meaning and Reference, Oxford Readings in Philosophy, Oxford. 1993. Pages 56-79). According to Strawson, reference is presupposed rather than entailed. Thus, where referent is lacking so too is truth-value, i.e. the relevant class of orthographic identities is neither true nor false. Orthographic identities containing referring terms are therefore true while orthographic identities containing vacuous terms are neither true nor false. It follows that no orthographic identity is ever false. In other words, there is no possible world in which any orthographic identity is false. Given the familiar interpretation of the notions of possibility and negation, orthographic identities are necessarily true, i.e. no orthographic identity is false in any possible world. Of course, this analysis assumes the correctness of Strawson's position; a precarious assumption given Stephen Neale's critique of Strawsonian analyses in his Descriptions, Cambridge MA, MIT Press. 1990.
Re Kripke, for example, see reference 6 above. Re Putnam, see, for example, Putnam, H., `Meaning and Reference', reprinted in Moore, A. (ed.) Meaning and Reference, Oxford Readings in Philosophy, Oxford. 1993. Pages 150-161.
I am indebted to Timothy Kenyon for many of the points made in this section of the paper which, moreover, shape the paper as whole.
Kripke, S., `Identity and Necessity', reprinted in Moore, A. (ed.) Meaning and Reference, Oxford Readings in Philosophy, Oxford. 1993. Page 180.
ibid.
Putnam, H., `Meaning and Reference', reprinted in Moore, A. (ed.) Meaning and Reference, Oxford Readings in Philosophy, Oxford. 1993. Page 159.
ibid.
Laudan, L., `A Confutation of Convergent Realism', reprinted in Papineau, D. (ed.) The Philosophy of Science, Oxford Readings in Philosophy, Oxford. 1996. Pages 107-138.
Kuhn, T.S., The Structure of Scientific Revolutions, The University of Chicago Press, Chicago, 1970. Second Edition. Chapter X, see especially pages 129-135.
I am again indebted to Timothy Kenyon for these points.
There are some reasons due to which private language is logically impossible. For instance, both language and consciousness are social constructs, which do not allow of any elements of logical privacy. But this is not a part of Wittgenstein's original argument.
In making references to the first part of Philosophical Investigations, I shall just indicate section numbers; e.g. (23) means section 23. To the second part, I shall give page numbers. E.g., (226e) means page 226 of Anscombe's English translation (Basil Blackwell, Oxford 1963). Italics mostly are mine.
Not, as alleged by Kripke, «the set of responses in which we agree, and the way they interweave with our activities» (Kripke 1982, p.96).
In Prolegomena section 39: «Of the System of the Categories» Kant tells us that the discovery of transcendental categories «presupposes neither greater reflection nor deeper insight, than to detect in a language the rules of the actual use of words generally, and thus to collect elements for a grammar. In fact both researches [grammar and epistemology] are very nearly related, even though we are not able to give a reason why each language has just this and no other formal constitution, and still less why an exact number of such formal determinations in general are found in it».
This is a question Wittgenstein raised at the beginning of his later period. «Is meaning then really only the use of a word? Isn't it the way this use meshes with our life?» Wittgenstein 1974, section 29, p.65.
I'm definitely not the first one who uses the term «transcendental» in this fashion. K-O. Apel, for example, used it in a similar way. See his Towards a Transformation of Philosophy, Routledg & Kegan Paul Ltd., 1980.
See Note in Editing, Wittgenstein 1974, p.487.
See also 92 and 97. «We ask: 'What is language?', 'What is a proposition?' And the answer to these questions is to be given once for all; and independently of any future experience»(92). «Thought is... the a priori order of the world: that is, the order of possibilities, which must be common to both world and thought ... It is prior to all experience, must run through all experience»(97). These are (roughly) Kantian and Tractarian views.
This is a thesis that cannot be explored here. Ayer once declared: «[I do not] seek to deny that, as a matter of fact, one's references to one's private experiences are made within the framework of a public language. What I am querying is Wittgenstein's assumption that this is a logical necessity»(Ayer 1985, p.74). The private language argument is a transcendental argument, not a logical argument in the ordinary sense. A short answer to Ayer's above challenge is, the only thing that is given and has to be accepted is form of life. Cf. the above note 1.
Wittgenstein 1975, sections 16 and 19, pp.60, 61. Numbering is mine.
Only in this connection Ayer is correct in saying that the distinction between public and private is idle.
Kripke 1982, pp. 60, 80, 103.
Ibid., note 83, italics are original, p.103.
Ayer 1959, p. 78.
See Ayer 1988, p.16. Ayer died in 1989.
Ayer 1973, p.98.
Ibid.
See above note 15. Kripke in his influential book on private language devoted two full pages to endorse this view, see Kripke1982, pp.60-2. E.g., «If I really were in doubt as to whether I could identify any sensations correctly, how would a connection of my sensations with external behavior, or confirmation by others, be of any help? Surely I can identify that the relevant external behavior has taken place, or that others are confirming that I do indeed have the sensation in question, only because I can identify relevant sensory impressions (of the behavior, or of others confirming that I have identified the sensation correctly). My ability to make any identification of any external phenomenon rests on my ability to identify relevant sensory (especially visual) impressions». Ibid., p. 61.
The first formulation appears in Ayer 1954. My quotation is from Ayer 1985 p.76 and 1982 pp.151-2. It is not surprising that Kripke's comment (on the 1954 version) is «the objection seems cogent». See Kripke 1982, p. 62.
Kripke 1982, p. 80.
Ayer 1985, p. 74.
It's true that individuals always contribute novel meanings and ideas to a language but this is not a part of the problem.
To solve this problem, a constructive theory of language formation is needed. Such a constructive theory can be found, e.g. in Transcendental Idealism and Pragmatism.
In my development of the views on Wittgenstein's philosophy I owe very much to Dr. Arthur Falk of Western Michigan University for his seminars during 1998 and 2000.