A corpus-based study of creaky voice production in English and Mandarin
Published in Proceedings of Interspeech, 2026
Creaky voice has been extensively examined across varieties of English, but cross-linguistic sociophonetic research on it remains limited. This study presents an acoustic investigation of creaky voice production in Mandarin Chinese and US English, using read-speech corpora comprising 66 Mandarin speakers (34F) and 61 English speakers (28F). Audios were processed via a pipeline of segmentation, annotation, cleaning, and acoustic analysis. 5,874 vowel tokens were measured for F0, H1–H2c, HNR35, and CPP over the 70% mid-portion. Linear mixed-effects analyses revealed men produced creakier speech than women in both languages. Findings contradict the notion that creaky voice is predominantly a feature of young women’s speech in US English but align with findings of more creakiness in men’s speech. We discuss the Mandarin results in relation to recent work on Mandarin listeners’ social perception of creaky voice.
| doi: 10.21437/Interspeech.2026-3235 | materials: osf.io/xxx | data: osf.io/yyy |
Recommended citation: Li, M., Yao, Y., & Chang, C. B. (2026). A corpus-based study of creaky voice production in English and Mandarin. In B. Ahmed, M. Proctor, V. Sethu, & F. Cox (Eds.), Proceedings of Interspeech 2026 (pp. 1779–1783). Sydney, Australia: International Speech Communication Association.
Download Paper