H. P. Grice, answered from the texts and cited to the page.
Conventional implicature is that portion of a word's or expression's total signification which is part of its conventional force — its established meaning — yet which falls outside what is strictly said, in the favored sense, by an utterance of it.1 Let me put the distinction a little more carefully. When I speak of the total signification of an utterance as divisible, I have in mind two cross-cutting distinctions: first, between what is said and what is implicated; second, between what is part of the conventional force of the utterance and what is not.2
These two distinctions yield three possible elements in the total signification: what is said, what is conventionally implicated, and what is nonconventionally implicated — though in any given case one or more of these elements may be absent.3 The canonical illustration is but. When I say "He is poor but honest," the truth-conditional content — what would have to obtain for the utterance to be true — is simply the conjunction: he is poor and he is honest.
Yet something further is conveyed by but that is not conveyed by and: a contrast, a suggestion that the two conditions stand in some tension with one another. That contrast is conventionally implicated; it is carried by the word itself, not derived by any inference from the Cooperative Principle or the maxims. And this is what distinguishes conventional from conversational implicature: a conversational implicature is calculable from what is said together with the assumption that the speaker is being cooperative, whereas a conventional implicature attaches to the expression as part of its conventional meaning.
One diagnostic consequence of this is that conventional implicature is not cancellable in the way conversational implicature is. A generalised conversational implicature can be cancelled explicitly — by adding a clause that states or implies that the speaker has opted out — or contextually, if the form of utterance is used in a setting that makes clear the speaker is not following the usual cooperative path.4
No such cancellation is available for the contrast carried by but; you cannot say "He is poor but honest — I don't mean to suggest any contrast" without a kind of incoherence. And yet the contrast is not part of the truth conditions: you can deny what is said by "He is poor but honest" simply by denying that he is poor or that he is honest, without thereby denying the contrast.
A further diagnostic is nondetachability — and here the difference between conventional and conversational implicature is again instructive. A generalised conversational implicature, insofar as its calculation requires only knowledge of what has been said and not the particular manner of expression, will typically be nondetachable: you cannot find a synonymous expression that simply lacks the implicature.5
Conventional implicature, by contrast, is detachable — since it attaches to the specific lexical item, you can convey the same truth-conditional content with a different word (and instead of but) and lose the implicature. The two diagnostics thus cut in opposite directions for the two categories. Now, there is a complication I want to be honest about.
The relation between conventional implicature and what is said requires careful handling, and I noted in Further Notes on Logic and Conversation that the nature of conventional implicature needs to be examined before any free use of it for explanatory purposes can be indulged in.6 In particular, when a strong theorist of, say, disjunction wants to claim that the exclusive reading of or is conventionally implicated rather than part of what is said, he is open to the following pressure: if but is the model case of a word carrying a conventional implicature, then the negative form It is not the case that A but B is uncomfortable to say — and by parity, It is not the case that A or B, if or in the strong sense is the operative reading, should be equally uncomfortable.7
The fact that it is not — that negating A or B simply seems to amount to asserting that neither A nor B — is evidence that the exclusive reading is better handled as a conversational implicature, cancellable and nondetachable, rather than as a conventional one. The deeper point is this: conversational implicata are, initially at least, not part of the meaning of the expressions to whose employment they attach.8
They are calculable, cancellable, and nondetachable. Conventional implicata are part of the conventional meaning — but they are not part of what is said in the truth-conditional sense. The three-way distinction — said, conventionally implicated, nonconventionally (conversationally) implicated — is what gives the framework its analytical purchase, and it is the distinction that Modified Occam's Razor is designed to protect: do not multiply senses beyond necessity when a pragmatic account, whether conversational or conventional, will do the work.
one may distinguish, within the total signification, between what is said (in a favored sense) and what is implicated; and second, one may distinguish between what is part of the conventional force (or meaning) of the utterance and what is not.Studies in the Way of Words, pp. 47–49
the total signification of an utterance may be regarded as divisible in two different ways. First, one may distinguish, within the total signification, between what is said (in a favored sense) and what is implicated; and second, one may distinguish between what is part of the conventional force (or meaning) of the utterance and what is not.Studies in the Way of Words, pp. 47–49
This yields three possible elements—what is said, what is conventionally implicated, and what is nonconventionally implicated—though in a given case one or more of these elements may be lacking.Studies in the Way of Words, pp. 47–49
a generalized conversational implicature can be canceled in a particular case. It may be explicitly canceled, by the addition of a clause that states or implies that the speaker has opted out, or it may be contextually canceled, if the form of utterance that usually carries it is used in a context that makes it clear that the speaker is opting out.Studies in the Way of Words, pp. 46–47
it will not be possible to find another way of saying the same thing, which simply lacks the implicature in question, except where some special feature of the substituted version is itself relevant to the determination of an implicature.Studies in the Way of Words, pp. 46–47
the nature of conventional implicature needs to be examined before any free use of it, for explanatory purposes, can be indulged in.Studies in the Way of Words, pp. 53–54
if a model case for a word which carries a conventional implicature is but, then the negative form It is not the case that A or B, if to be thought of as involving or in the strong sense, should be an uncomfortable thing to say, since It is not the case that A but B is uncomfortable.Studies in the Way of Words, pp. 53–54
conversational implicata are not part of the meaning of the expressions to the employment of which they attach.Studies in the Way of Words, pp. 47–49