In this paper I attempt a new analysis of utterances that (implicitly) convey typicality or nontypicality, which have been regarded as cases of conversational implicatures since Grice's classical analysis of them. Leading accounts of such utterances―Horn's and Levinson's―appear inadequate for a kind of them. I take it that utterances conveying typicality split up into two groups and they are different in their ways of generating a typicality implicature. And I regard utterances conveying nontypicality as more closely related with metalinguistic negation than Horn and Levinson assume.