Abstract

It seems unlikely that there will ever be consensus about the extent to which we can reliably distinguish semantic phenomena from pragmatic phenomena. But there is now broad agreement that a sentence's meaning can be given in full only when it is studied in its natural habitat: as part of an utterance by an agent who intends it to communicate a message. Here, we document some of the interactions that such study has uncovered. In every case, to achieve even a basic description, it is necessary to pool semantic information, contextual information, speaker intentions, and general pragmatic pressures.