Research

Ohio State Dissertations in Linguistics (OSDL)

Shu-Hui Peng (1996)

Phonetic Implementation and Perception of Place Coarticulation and Tone Sandhi

Advisor: Mary Beckman

Download .pdf file [PDF]

Abstract:

This dissertation examines two cases of contextual variability for evidence concerning the nature of phonological and phonetic representations and the relationship between them. The first case is Mandarin tone sandhi whereby tone 3 becomes tone 2 when followed by another tone 3. The second case is place "assimilation" between Taiwanese coda and onset stops.

In the tone sandhi study, the overall F0 of the derived tone 2 was slightly lower than that of the underlying tone 2. A forced-choice identification test between /tone 2 + tone 3/ and /tone 3 + tone 3/ showed that the derived tone 2 and the underlying tone 2 were perceptually indistinguishable to Mandarin speakers. On the otherhand, concept formation test showed that most Mandarin speakers nonetheless categorized the derived tone 2 as tone 3, although they could be trained to categorize the derived tone 2 as tone 2 (with less consistent performance).

In the Taiwanese place "assimilation" study, EPG measurements showed a noncategorical gestural coarticulation. However, the latency of the second gesture with respect to the first decreased as speech rate increased, so that gestural 'deletion' eventually occurred at faster speech rates. The dental gesture was "deleted" more frequently than the velar gesture. A concept formation test showed that coda /t/ and coda /k/ had different patterns of categorization. As speech rate increased, the dental-velar sequence was categorized as /t/ lessfrequently, and as /k/ more frequently. By contrast, the velar-dental sequence was categorized as /k/ to about the same extent as for the velar-velar sequence. This difference in categorization may be attributed to different patterns of gestural coarticulation in the dental-velar sequence and the velar-dental sequence.

These results pose difficulties for all current models of phonological and phonetic representations. Models which make a hard and fast separation between (discrete categorical) phonological representationand (continuous) phonetic representation cannot explain the gradualness of the progression from coarticulation to deletion. Models which collapse these two representations into one articulatory representation cannot explain the categorization of the sandhi tone.