Standing out in context: Prominence in the production and perception of public speech

This study investigates the relationship between signal-based factors and discourse context in the production and perception of prominence, extending prior work on prominence in American English by examining speech from complete TED Talks as examples of a public speech style. First, we conduct a production study. The TED Talk speech samples are analyzed for relationships between (1) phonological pitch accent, (2) acoustic prosodic features, and (3) information structure, modeled here in terms of referential and lexical givenness as well as focus related to semantic alternatives. Second, we conduct a perception study. Word-level prominence ratings obtained from untrained listeners are analyzed for the effects of these (1) phonological, (2) acoustic, and (3) information structural factors. The aim of this study is to evaluate the interaction of the three kinds of factors on prominence ratings with speech that is fully contextualized, allowing for information about the speaker, the situational context, and the discourse context to interact with and possibly mediate signal-based cues to prominence. Results from the production analysis show a weak probabilistic relation between pitch accent assignment and information structure for these TED Talk speakers, but only in the presence versus absence of pitch accent with differences between lexical and referential information status. No systematic relationship is found between a word’s phonological pitch accent type and information structure category, though we do find some evidence of the expected gradient variation in the acoustic expression of prominence: Words that are more informative tend to have higher pitch and a wider pitch range. Despite the weak association between signal-based prominence and information structure in production, listeners perceive and rate prominence in a manner that is, to varying degrees, in accordance with the Accentual Prominence and Givenness hierarchies. In sum, the findings contribute new evidence of the mediating effects of contextual factors relating to the phonological status of pitch accents, information structure, and speech style on the interpretation of acoustic cues to prominence.