The acoustic and perceptual nature of tone in Vietnamese

This study presents a description of Vietnamese tones in the three major dialects -Northern, Central and Southern - in their various aspects and at different levels of analysis. Acoustic phonetic materials were gathered from voice recordings of 34 informants, analyzed instrumentally and measured. The data obtained were treated statistically and normalized to yield typical values for the parameters of Fo, Intensity, Duration and Laryngealization, which serve to characterize all Vietnamese tones. The results established six non-stopped tones (level, falling, rising, drop, curve and broken) and two stopped tones (stopped rising and stopped drop) for Northern Vietnamese, which in this respect also represents Modern Standard Vietnamese. The Central and Southern dialects have only five non-stopped tones and two stopped tones,since there are no differences in the phonetic realizations of the standard curve and broken tones in these dialects. Four types of Fo contour (level, falling, rising and concave) and three levels of average Fo (high, mid and low) were established. Useful data on intensity, duration and laryngealization occurring on different tones were also obtained. Analysis of the variation of tone parameters in their relationships with segmental and suprasegmental environments points to the complex interaction various factors in the production of tone. The phonetic nature of Vietnamese tones can thus be understood to have acoustic properties related to the laryngeal mechanisms for the production of voice, with Fo contour, relative Fo level and voice quality as the basis for differentiating the tones. Perceptual tests on cross-dialect tone recognition with natural and synthetic tones, involving 84 Northern, Central and Southern Vietnamese subjects, showed that natural tones of the three dialects in meaningful contexts were readily identified by all subjects, while isolated real speech tones and synthetic tones presented varying degrees of difficulty in recognition according to individual tones, dialects and subjects. Analysis of tone features involved in misperceptions suggested that contour features played a dominant role in tone identification. A model of tone perception is proposed, including conversion processes and interpretation rules for translating the physical phonetic parameters into phonetic features (pitch targets and degrees of laryngealization) then into phonological features (contour, concave, high, low, falling, rising, creaky, stopped). The model is supposed to reflect the dynamic process of tone perception in which the phonetic properties of tones are perceived in terms of the native speakers' phonological structures