Acoustic correlates of perceived rhythm in spoken English

Abstract

Two experiments examined the relationship between speech timing and perceived rhythm for a corpus of approximately 900 spoken sentences. In experiment 1, trained listeners applied an annotation system for perceptual isochrony to the corpus. For each sentence, listeners assigned beats to syllables and judged whether the intervals between successive beats were equal or unequal, permitting the identification of perceptually isochronous speech fragments (or beat chains). For each such beat chain, the intervals between vowel onsets of successive beat syllables were determined. Overall, there was good agreement among labelers about what constituted a beat chain, with the average inter-beat-interval equal to approximately 550 ms. In experiment 2, naive listeners rated the rhythmicity of each sentence on a 6-point scale, ranging from 1 (very nonrhythmical) to 6 (very rhythmical). Sentences were judged as more rhythmic when the longest identified beat chain in experiment 1 for that sentence contained more beats and had an average IBI closer to 550 ms. IBI variability was not a significant predictor of perceived rhythmicity. The results of both experiments will be discussed in terms of preferred tempi in perceived rhythm and entrainment models of auditory event timing (McAuley and Jones, 2003).