The prosody of questions in natural discourse

Abstract

For this paper, we examined a corpus of 73 wh-questions and yes/no questions, both positive and negative, from natural discourse.We found that the locus of interrogation (the initial auxiliary in yes/no questions or the initial wh-word in whquestions) most frequently gets an L+H* pitch accent, especially in wh-questions and negative yes/no questions.Positive yes/no questions are more variable, and included 40% unstressed auxiliaries.Nuclear stress was primarily falling in wh-questions, as expected; but positive yes/no questions were almost twice as often falling or level as rising, contrary to expectation.Finally, the topic of the question turned out to be marked primarily with some version of an H* accent rather than an L+H* accent, and the focus with L+H* rather than some variant of H*, contrary to predictions in the literature.