[Dry run for a 20-minute talk at the Cornell Prosody Workshop]
It has been accepted for some time that certain aspects of intonation involve what Bolinger called "gradience". Attempts have been made to incorporate gradient effects into formal models of pragmatic interpretation, most notably Ward and Hirschberg's proposal (1985) that gradient increases in pitch range on an English rising-falling-rising nuclear contour can bring about a quasi-categorical shift in the pragmatic force of an utterance from "uncertainty" to "incredulity". However, this involves the expression of speaker *attitude*, and Ward and Hirschberg's approach has never been applied to the intonational expression of essentially grammatical distinctions such as theme/rheme or question/statement. Instead, recent work on intonation in formal semantics and pragmatics (e.g. Schwarzschild, Steedman) tends to take categorically distinct ToBI labels at face value as correlates of categorically distinct semantic/pragmatic effects.
In this paper I review phonetic evidence from three recent PhD dissertations on such phenomena. Taken together, this work points strongly to the conclusion that the intonational distinctions underlying the pragmatic effects, though real, are phonetically gradient. This suggests that Ward and Hirschberg's approach to describing intonational pragmatics might profitably be generalized beyond speaker attitude, and used as the basis of a more general theory of how the gradient expression of degrees of uncertainty, emphasis, etc., can yield quasi-categorical pragmatic nuances related to information structure and speech act. On the phonological/phonetic side, such a theory would have the further advantage of eliminating a number of alleged distinctions of accent type (e.g. H* vs. L+H*) that are phonetically quesitionable and are known to cause problems for transcriber reliability.