Individual variability in phonetic convergence of vowels and rhythm

Abstract

Phonetic convergence (PC) has been demonstrated for segmental (vowels, voice onset time) and suprasegmental (stress, intonation) properties [Nielsen (2008), Delvaux and Soquet (2007), Phillips and Clopper (2010), Rao, Smiljanic, and Diehl (2011)]. Since PC is subject to large individual differences [Ni Chiosáin (2007)], the current study examined individual variability in PC in both segmental and suprasegmental domains for native speakers of American English. Six female and six male pairs read CVC syllables and a short paragraph before and after an interactive map task. For each dyad, convergence in vowels was measured using formants and the cosine similarity metric for individual vowels and for the entire vowel space. Convergence in rhythm was measured using the centroid of the envelope modulation spectrum [EMS + centroid, Rao and Smiljanic (2011)]. Overall, speaker pairs converged to different extents in both measures. Vowel type, dialect background, and gender were found to influence the degree of convergence. In general, men were more likely to converge in rhythm whereas women were more likely to converge in vowels. This supports the findings that gender-based differences in convergence are due to perceptual sensitivity to indexical features [Nami et al. (2002), Babel (2009)] and particular sound features in spoken utterances.