2009/12/30

Swerts, M. (1997) Prosodic features at discourse boundaries of different strength. Journal of the Acoustical Society of America, 101 (1), 514–521.

Presentation: Shelly

Summary: Sarah

In view of the fact that discourse boundaries cannot be reliably distinguished in past studies, studying the phonetic correlates of discourse boundaries might seem to be problematic. In this regard, this study aims to determine different discourse boundaries by utilizing experimental designs. The idea was borrowed from Rotondo (1984), in which he regarded the occurrence of one boundary indicates the end of a “complete thought”. With this definition, in the experiment the subjects were asked to read a text and mark the end of every complete thought. The other group of subjects was also doing the same thing, but their boundary marking task was facilitated with speech audio input. The rationale of the authors was that the more people agree on the existence of one particular boundary, the stronger the boundary is. Results showed that the rating patterns of the two subject groups were highly correlated, with the text-and-speech group reached even higher agreement than the text-only group. The author further measured a number of prosodic cues, including pause, pitch reset, and boundary tone. Regressional analyses showed that all three correlates together could successfully predict boundary strengths, which in turn verified the effectiveness and reliability of the experiment. Via this method, discourse boundaries can therefore be more objectively determined, and their prosodic correlates can be more carefully studied.