Discriminant analysis of acoustic cues to three Green Mong tones

Abstract

Green Mong is a tone language with an inventory of seven contrastive tones. The tonal system of Green Mong incorporates both pitch and phonation-type distinctions, and appears to make use of many of the acoustic parameters which have been shown to correlate with phonation-type cross-linguistically. This study looks at three Green Mong tones that have similar pitch contours, but are distinguished by the distinctive use of breathy, creaky, and modal phonation. Acoustically, the three tones are distinguished to some degree on the basis of pitch, relative amplitude of lower and higher harmonics (H1–H2), duration, vowel quality, and voice onset time. Discriminant analysis is used to estimate the value of these different cues for classifying tokens into tone categories based on observed characteristics in tone production. Discriminant functions were generated using measurements from 270 CV words produced by six speakers, for which every combination of three vowels (/i/, /u/, /a/) five consonants, and three tones (breathy, creaky, modal) was attested. Results of the analyses will be presented, as well as results for classification of a separate, test set of 102 CV tokens produced by the same six speakers.