Perception of Coarticulated Tones

Yi Xu

Journal of the Acoustical Society of America, Pt 2, 90: p. 2362. 1991

Tones, like segments, coarticulate with one another in speech. Tonal coarticulation reduces the Fo excursion of a dynamic tone when it is in a mismatched context, that is, when the two ends of that tone do not have the same underlying pitch register as the adjacent parts of the surrounding tones, e.g., in case a rising tone is surrounded by two rising tones. To examine the perceptual effect of tonal coarticulation, trisyllabic Mandarin words were altered by changing one of their initial consonants so that they became nonsense words. These triplets were used as stimuli for identification of the tones on their middle syllables. Overall identification was well above chance, indicating listeners' use of the information provided by the tonal context. However, with mismatched contexts, identification was slightly poorer than with matched contexts, indicating that actual Fo contours is also important for tonal identification for connected speech. To further test the perceptual effect of tonal coarticulation, the first syllable and the last syllable within each triplets were swapped, and the resulting nonsense triplets were again used as stimuli for identification of the tones in the middle. When the first and the last syllable had the same tone, identification was unaffected. When the two differed, however, a rising tone was mostly heard as a falling one and vice versa when the swapping had changed a previously mismatched context into a matched one, whereas identification was not much affected if the swapping changed a matched context into a mismatched one. This indicates that listeners factor out, or compensate for, variations due to the tonal context when trying to identify a target tone.[Work supported by NIH.]

See other publications