Individual Differences in the Production and Perception of Prosodic Boundaries in American English

Abstract

Theoretical interest in the relation between speech production and perception has led to research on whether individual speaker-listeners’ production patterns are linked to the information they attend to in perception. However, for prosodic structure, the production-perception relation has received little attention. This dissertation investigates the hypothesis that individual participants vary in their production and perception of prosodic boundaries, and that the properties they use to signal prosodic contrasts are closely related to the properties used to perceive those contrasts. In an acoustic study, 32 native speakers read eight sentence pairs in which the type of prosodic boundary (word and Intonational Phrase boundary) differed. Phrase-final and initial temporal modulation, pause duration, and pitch reset at the boundaries were analyzed. Results showed that, as a group, speakers lengthened two phrase-final syllables, shortened the post-boundary syllable, and produced a pause and pitch reset when producing an IP boundary. However, individual speakers differed in both the phonetic features they used and the degree to which they used them to distinguish IP from word boundaries. Speakers differed in the onset and scope of phrase-final lengthening and presence of shortening (resulting in six different patterns), pause duration, and the degree of pitch reset at the IP boundary, including in ways that demonstrated a trading relation between these properties for some individuals and an enhancement relation for others. The results suggest that individuals differ in how they encode prosodic structure and offer insights into the complex mechanism of temporal modulation at IP boundaries. In an eye-tracking study that tested the perceptual use of these acoustic properties by 19 of these same participants, the productions of a model talker were manipulated to systematically vary the presence and degree of IP boundary cues. Twelve unique combinations of cues, based on the main patterns in the production study, were created from four phrase-final lengthening patterns, two pause durations (presence/absence of a pause), and three pitch reset values. Patterns of fixation on the target boundary image over time showed that, as a group, listeners attended to the information conveyed by pause duration and final lengthening as that information became available, with pause being the most salient cue for IP boundary perception. A clear pattern did not emerge for pitch reset. Adding to the body of research on weighting of the acoustic properties for IP boundary, these results characterize the time-course of the perceptual use of different combinations of IP boundary-related properties. To examine the production-perception relation, a series of perceptual models in which each participant’s average production values were entered as predictor variables tested whether the production patterns are reflected in the same individuals’ perception. The results did not provide statistically significant evidence of a production-perception relation, although a trend in the pause duration models across three different conditions was suggestive of a pattern in which individuals with longer pause durations were faster to fixate on the IP boundary target than those with shorter pause durations. The lack of evidence of a close production-perception relation for individual speaker-listeners is inconsistent with the main hypothesis but is in line with the results of several previous studies that have investigated this relation for segmental properties. Further investigation is needed to determine whether, despite the absence of a strong production-perception relation, specific individuals might nonetheless show the link predicted by some theoretical approaches.