Acoustic correlates of vocal alignment to accented speech

Abstract

Language users can vocally align to the speech characteristics of other talkers both within and across accent groups. Although perceptual assessments of vocal alignment indicate convergence to accented speech, the acoustic properties of the speech signal that contribute to these perceptual judgments are less clear. The present investigation examined the acoustic correlates of vocal alignment to foreign-accented and native speech and related acoustic to perceptual measures of convergence. Native speakers of American English shadowed utterances produced by four model talkers; male and female native speakers of English and Spanish. Baseline utterances were obtained from visual naming blocks and shadowed utterances were compared to both baseline and model talker utterances. Fundamental frequency (F0), utterance duration, and vowel spectra (F1, F2) served as acoustic indices of vocal alignment. The results showed that different patterns of alignment emerged across the three measures. Utterance duration, but not fundamental frequency or vowel spectra, demonstrated convergence. However, both duration and vowel spectra measures significantly predicted perceptual assessments of vocal alignment. The results suggest that perceptual judgments of alignment behavior depend on multiple acoustic properties and that the dimensions of acoustic change during vocal alignment appear similar across accented and native speech.