Acoustic correlates of Burmese voiced and voiceless sonorants

Abstract

Abstract Voiceless sonorant consonants are typologically rare segments, appearing in only a few of the world’s languages, including Burmese. In this study, Burmese sonorants and their adjacent vowels are investigated in an attempt to (1) determine what acoustic correlates distinguish voiced and voiceless sonorants and (2) determine whether there are multiple realizations of voiceless sonorants and, if so, establish what acoustic correlates distinguish them. In order to pursue these questions, a production study was carried out and target words were analyzed, demonstrating that Burmese voiceless sonorants have a spread glottis period resulting in turbulent airflow 78 % of the time. Findings from linear mixed-effects models showed that voiced and voiceless sonorants are significantly different in terms of duration of the sonorant, F0 of the sonorant, and strength of excitation measured over the following vowel. A linear discriminant analysis was able to predict voicing category with 86.7 % accuracy, with the duration of the spread glottis period being the best indicator of voicelessness, followed by the cues that were significant in the linear mixed-effects models. In cases when the spread glottis period is absent from voiceless sonorants, the sonorant only has correlates that are associated with voicelessness (such as F0 and strength of excitation) but not correlates that are associated with the spread glottis gesture (such as duration and harmonics-to-noise ratio). These results have implications both for our understanding of the acoustics of Burmese sonorants and for our understanding of voiceless sonorants more generally.