Both linguistic intuitions and psychological evidence argue that intonation plays a role in the interpretation of spoken English utterances. The work presented here is a pilot study for exploration of this role. We suggest that intonation is one of several extrapositional linguistic features of an utterance which, along with extralinguistic information, signal that the content of an utterance is a question, assertion, request, greeting, or some other speech act. We present our model for recognition of speaker intentions. We show, using a forced-choice paradigm, that the request/question distinction can be intonationally disambiguated. We discuss the intonational features that may be responsible, and indicate the steps necessary to making use of these results in automated speech act recognition.