Moore's paradox

Abstract

G. E. Moore famously noted that saying ‘I went to the movies, but I don't believe it’ is absurd, while saying ‘I went to the movies, but he doesn't believe it’ is not in the least absurd. The problem is to explain this fact without supposing that the semantic contribution of ‘believes’ changes across first-person and third-person uses, and without making the absurdity out to be merely pragmatic. We offer a new solution to the paradox. Our solution is that the truth conditions of any moorean utterance contradict its accuracy conditions. Thus we diagnose a contradiction in how the moorean utterance represents things as being; so we can do justice to the intuition that a Moore-paradoxical utterance is in some way senseless, even if we know what proposition it expresses.

Notes

1 As he says, ‘to say such a thing as “I went to the pictures last Tuesday, but I don't believe that I did” is a perfectly absurd thing to say’[Moore 1952: 543].

2 For the concepts of reflexive truth conditions and incremental truth conditions, see Perry [2001].

3 Note that accuracy demands sameness of content, narrowly individuated. Suppose that Kay has a belief b that Elwood is a fool, and Elwood happens also to be president of the bowling league. If Kay says ‘The president of the bowling league is a fool’ that is not an accurate expression of her belief b, although it may be an accurate expression of some other, closely related belief.

4 That the work is done by accuracy and not sincerity conditions is crucial—this fact sets our solution apart from the solutions of speech act theorists such as Searle. See below. Note that both the truth and accuracy conditions of an utterance are derived semantically. No attention is paid to the conversational purposes of the agent, or the states of mind of his audience. Our solution applies, for example, to Moore's uttering (2) alone in his study. In this sense, our solution is semantic and not pragmatic.

5 Moreover, note that this is not so for utterances of the form (4). If x makes the utterance (4), sincerely or not, accuracy demands that x believes that p and N doesn't believe that p. There's no contradiction here with (4)'s reflexive truth conditions. A question here arises: what if it should happen that N is another name for the speaker, x—will there be in that case no way to satisfy both truth and accuracy conditions together? Were that so, our solution would overgeneralize to all cases where N is another name for the speaker. In response, we note that accuracy demands sameness of content, narrowly individuated. Consequently, even in cases where N and ‘I’ corefer, the content that p, but N doesn't believe that p, differs from the content that p, but I don't believe that p. Thanks to an anonymous referee for raising this issue.

6 In fact Wittgenstein [1953] in the Investigations IIx speaks only of the sentence ‘It may be raining, but it isn't’ as an Unsatz, but he does so in the context of discussion of moorean assertion more generally.

7 Some approaches try hard to capture semantic senselessness, but in the effort abandon the idea that ‘believes’ means the same throughout. See Malcolm [1991].

8 Where and how to draw a line distinguishing semantic and pragmatic features of language is of course disputed. Fortunately, we don't have to try for current purposes.

9 Searle writes: ‘This law, incidentally, provides the solution to Moore's paradox: the paradox that I cannot assert both that p and that I do not believe p, even though the proposition that p is not inconsistent with the proposition that I do not believe p’[Searle 1970: 65 n. 1 (emphasis added)].

10 It's worth stressing how pragmatic accounts differ from our proposal: on a pragmatic, or speech act approach, resolving the paradox involves identifying a conflict at the level of linguistic activity, not at the level of semantics. On our approach, resolving the paradox involves identifying a conflict at the level of semantics or representation. Accuracy conditions are not conditions on a speech act’s being successful, but are conditions on a representation's being correct. Thus our approach also differs from accounts that focus on warranted assertibility conditions (for example, Koethe [1978]). Assertibility conditions are conditions under which an assertion is appropriate or justified. Accuracy conditions are conditions under which a token utterance correctly represents a state of mind. The two conditions may in some circumstances be extensionally equivalent, but the concepts of accuracy and assertibility are very different. Thanks to an anonymous referee for raising this issue.

11 Albritton [1995] nicely motivates understanding the paradox this way.

12 Many adequacy conditions on a proper solution have been suggested (see Sorensen [1988] for a discussion of adequacy conditions on a solution); here we take up just the idea that an adequate solution must address ‘Moore-paradoxical belief’, as most pressing for our account.

13 Shoemaker [1995: n 1] suggests as a general principle, ‘If you have an explanation of why a putative content could not be coherently believed, you thereby have an explanation of why it cannot be coherently asserted’. Shoemaker's principle may be correct, but, as our example shows, there is a residue of paradoxicality about a moorean assertion that needs further explanation.

14 Notice that in utterances that report one's desires, for instance, no such possibility for conflict opens up. When one asserts one's belief about one's desire, ‘I want ice cream’, what accuracy demands of one's assertion is one thing, namely, that one have the relevant belief, and what truth demands is another, namely, that one have the relevant desire.