Rises all the way up: The interpretation of prosody, discourse attitudes and dialogue structure

This dissertation is about what prosody contributes to dialogue interpretation. The view of prosody developed in this account is based on detailed quantitative investigations of the prosodic forms and interpretations of cue word and declarative responses, specifically with respect to the distribution and interpretation of terminal pitch rises. Drawing on results from corpus, production and perception studies, I argue that the underlying contribution of terminal rises is to signal that the dialogue has not come to a viable stopping point with respect to the task at hand. This approach enables us to explain previously incongruent findings about the connection between rises and attitudes like uncertainty. From this perspective, the perception of such attitudes does not arise directly from prosodic form, but instead depends upon a range of contextual factors. The experimental results indicate that the most important of these is how an utterance relates to the current question under discussion, rather than sentence or dialogue act type. However, variation in prosodic form is also affected by higher level factors like dialect, task, and speaker role: rises become more frequent on non-questioning moves as the need to co-ordinate becomes greater. The experimental results allows us to make significant headway in clarifying the relationship between the prosodic, semantic and information structural properties of responses. This, in turn, sheds light on several outstanding questions about the contribution of the rise in fall-rise accents and its relationship to information structural categories like contrastive topic. Overall, we see that rises don\u27t act on the proposition that carries them, nor do they mark out specific IS categories. Instead they reveal the state of the discourse from the speaker\u27s perspective. From a methodological point of view, I show that to gain a robust understanding the contribution of prosody on a particular meaning dimension, we need to take into account the baseline induced by the discourse configuration itself. These studies show the utility of using functional data analysis techniques to give more direct view of prosodic variation in larger datasets without manual prosodic annotation