An Articulatory, Acoustic, and Auditory Study of Burmese Tone

Abstract

This dissertation investigates the complex tonal contrast in the modern Burmese language. The four Burmese tones are reported to make multi-way distinctions in pitch, phonation, duration, intensity, vowel quality, and syllable structure, but the literature is frequently conflicted about the association of these qualities with each tone. Bradley (1982) has described Burmese as a register, rather than tonal, language due to the bundling of prosodic qualities headlined by contrastive phonation types. However, contrastive phonation in Burmese has eluded detection in numerous instrumental phonetic studies (Thein Tun 1982, Ladefoged et al 1988, Watkins 1997). Production and perception experiments were conducted in order to first clarify the phonetic description, and in turn build a phonological model of the contrast.

Audio, electroglottographic, and aerodynamic recordings of ten native speakers were collected and analyzed for Duration, F0, and Phonation type (as measured by Open Quotient of the glottal waveform, rates of oral airflow, and spectral tilt - all measured dynamically over the syllable rhyme). The perception experiment studied the forced choice listener identifications of stimuli re-synthesized to controlled levels of each dimension.

Major findings include (a) the contrast in phonation type is two-way (two breathier lax tones vs. two creaky-voiced tones), (b) breathy and creaky phonation are temporally dynamic in Burmese - produced systematically only at the vowel offset, and (c) the phonation contrast is neutralized in juncture while pitch contrasts are maintained.

The results indicate that a purely Tone or Register analysis of Burmese is inappropriate, in the sense that pitch contours or phonation type cannot alone serve as the basis of contrast. Instead, the present state of affairs is argued to represent an intermediate stage of tonogenesis, where multiple phonetic properties exist side-by-side to reinforce suprasegmental categories. The analysis is cast in Optimality theoretic terms, which act to highlight the phonological dilemma introduced by a seemingly redundant distribution of properties. The study has implications for the typology of Tone and Register systems, as it portrays a language in a transitional state between the two - some intrinsic pitch distinctions have been phonologized, but a concomitant voice quality distinction remains integral to the grammar.