Professionally Speaking - The prosodic and discourse structure of professional monologue in English

Abstract: This thesis combines corpus-based, experimental and instrumental methods to investigate how professional speakers of English use prosody to signal important textual boundaries above the sentence level. In an initial study, prosodic features included in the transcriptions of the Spoken English Corpus, or readily derivable from these, are investigated. These include minor and major tone-unit length, speech-rate, pauses and nuclei. While the examination of tone-unit lengths and speech-rate allows some interesting conclusions to be drawn about the prosodic style of professional monologue, they do not immediately provide indicators of textual structure. Pauses and certain nuclei, on the other hand, show a strong propensity to indicate finality although there is an overlap, in this respect, between normal sentence boundaries and those which involve structuring at a higher level. By means of an informant study, it is shown quite clearly that speakers are able to signal their intentions regarding the locations of discourse topic boundaries and that prosody provides a vital tool in accomplishing this. Also considered are the degree to which informant responses reflect the possible hierarchical structuring of texts and correspondences with paragraphing. Having thus established an independent method whereby discourse topic boundaries can be located. The boundaries perceived by informants are analysed for prosodic and textual correlates. This is done using an instrumentally analysed version of the corpus aligned with the written transcriptions. Pauses and F0 values at perceived discourse topic boundaries and unmarked sentence boundaries are measured and analysed statistically. Using a simple neural network it is possible to make a highly accurate distinction between sentence boundaries left unmarked by the informants and the discourse topic boundaries on which all informants are agreed. In analysing textual correlates of anomalous perceived boundaries it can be observed that certain textual conditions restrict the resetting of F0 following the boundary or cause the onset to be delayed.