The interplay between melodic and durational contours in prosodic prominence and phrasing

Prosodic functions such as prominence and phrasing can be expressed in speech through durational and melodic modifications. However, research in this area has often focused on either duration or melody independently, with limited attention to their interaction as a functional unit. This is especially true for research on Brazilian Portuguese (BP), which, as a Romance language, is generally assumed to rely primarily on durational cues for the expression of prosodic functions. While this seems to hold for sentence-reading data, some studies indicate that it may not hold for more spontaneous forms of speech. This reveals two open questions in our understanding of the prosodic realization of prominence and phrasing: (1) how different prosodic parameters may interact; and (2) to what extent laboratory speech represents naturalistic speech patterns. The present study addresses these issues by analyzing the interplay between the melodic and durational parameters in the realization of terminal and non-terminal boundaries (as instances of prosodic phrasing) and narrow informational focus (as an instance of prosodic prominence), in a function-oriented experimental phonetic approach. The corpus used for this study consists of recordings of five male and five female native BP speakers from the state of São Paulo in an informal interview speech style that offers more spontaneous speech data to the analyses. To identify occurrences of the target prosodic functions, the recordings were divided into excerpts that were given to lay listeners trained to mark the words they perceived as realizations of (non-)terminal boundaries and narrow informational focus in their corresponding orthographic transcriptions. Each excerpt was annotated by five listeners and a marked word was considered an effective realization of a target function if at least four agreed on its marking. Acoustic measurements were also extracted for parameters related to the melodic and durational contours. The interplay between these parameters was examined for the effective realizations of the target functions through cross-correlation analyses. The analyses reveal consistent correlations between the prosodic functions, duration peaks and an increase in fundamental frequency range. Moreover, terminal boundaries show a pre-stress melodic rise, followed by a fall on the stress, while non-terminal boundaries and occurrences of narrow informational focus show similar pre-stress lengthening patterns and melodic rises on the stress and just before it, respectively. Additionally, foci also appear to show a significant fall after this rise. These findings suggest an interplay between melody and duration in the expression of prosodic prominence and phrasing. They also highlight the need to complement results obtained for sentence-reading data with those obtained for more spontaneous speech data, to assess the extent to which laboratory speech reflects naturalistic speech patterns and to observe these patterns when it does not fully reflect them.