I have argued that sentence implicatures exist not because conversational implicatures are derivable from conversational principles, but on the contrary because there are conventional ways of conversationally implicating things. Implicature practices are arbitrary to some extent rather than completely determinate. Because the Gricean paradigm is incompatible with their existence, conversational implicature conventions have scarcely been noticed let alone studied in depth. In this chapter, we begin examining the distinctive nature of implicature conventions. We compare and contrast them with more familiar linguistic conventions. I suggest that instead of generating conversational implicatures in any way, the conversational principles of Grice and his followers tell us why implicature practices are socially useful. Although the implicatures of a sentence are not derivable from its meaning, I try to explain why they invariably do bear some relation to its meaning that makes them seem fitting or appropriate. I conclude with an alternative speculation about how some implicature practices came to be nearly universal, a surprising but not unprecedented feature given that they are conventional. We should look to historical rather than theoretical linguistics for the explanation.

FIRST-ORDER VERSUS SECOND-ORDER SEMANTIC CONVENTIONS

Let us say that the conventions giving words, phrases, and sentences their meanings are first-order semantic rules: they assign meanings and implications directly to sound sequences or letter strings. Then conversational implicature conventions are second-order semantic rules: they assign additional implications to linguistic forms only insofar as they have specified meanings by virtue of the first-order semantic rules.