Voice quality as an indicator of speaker's social information and as a cue for sound change

Abstract

Voice quality reflects a variety of information in speech. The present study examines how voice quality may indicate a speaker's social information. We examine the ongoing /yø∼y/ merger in Shanghainese Wu. The first vowel formant (F1) and Harmonics-to-noise ratio (HNR) are extracted from 80 speakers. We also collect social information including age, gender, and immigrant background, family language, and working language. Mixed-effects linear regression models are built with the F1/HNR difference between /yø∼y/ as the dependent variable. Significant age effects are found on both F1 (β = 0.043, p = 0.006) and HNR (β = −0.055, p = 0.024); F1 is also more merged for female speakers than males (β = −0.521, p = 0.001); HNR is more distinct for speakers from non-immigrant families than speakers from immigrant families (β = 1.858, p = 0.008). The study shows a tendency for /yø∼y/ to merge in terms of both vowel quality and voice quality. It is noteworthy that speakers of native origin participate in the vowel-quality merging (cued by F1) but not in the voice-quality merging (cued by HNR). Taken together, voice quality can function as an independent indicator of a speaker’s social attributes.