| dc.description.abstract | The complex tone sandhi patterns of Chinese dialects present analytical challenges to theoretical phonology, and productivity tests can help us address the issue from another perspective. Previous studies have shown that sandhi productivity is negatively affected by phonological opacity, positively affected by clear phonetic motivations, and positively correlated with lexical frequency of the sandhi patterns. It is further argued that the phonological grammar of tone sandhi patterns includes both grammatical constraints and lexical listing. We complement this research endeavor with a sandhi pattern whose productivity has not been previously studied: pattern substitution in Wuxi Chinese. Pattern substitution in Wuxi is left-dominant, whereby the base tone of the first syllable is first replaced by another tone before being spread to the sandhi domain. As a first step towards understanding the productivity of the pattern, we focus on disyllabic combinations between the three Yin tones on non-checked syllables T1, T3, and T5. Twenty native Wuxi Chinese speakers produced four sets of stimuli, including one set of Actual-Occurring real words (AO-AO), two sets of novel words made up of Actual-Occurring morphemes (*AO-AO1, *AO-AO2), and one set of novel words composed of an Accidental-Gap syllable and an Actual-Occurring morpheme (AG-AO). The difference between *AO-AO1 and *AO-AO2 was that the first AO morpheme of *AO-AO1 occurs in the initial position of real disyllable words, while that of *AO-AO2 does not. Both acoustic and statistical analyses were conducted. The results show that speakers had no difficulty producing real words with the expected sandhi, but pattern substitution is not fully productive in novel words. AG-AO showed the lowest productivity, while there was no significant difference between *AO-AO1 and *AO-AO2. This indicates that speakers may have tonal allomorph listings for morphemes as well as for syllables of morphemes (*AO-AO2). When they could not find the syllables in real syllable listing (AG-AO), they tend to spread the base tone of the first syllable or do nothing. Moreover, T3 showed the highest substitution productivity, and the similarity between T3, a low rising tone, and its substitution, a high rising tone, is the highest. It suggests that speakers may rely more on phonetic similarity rather than lexical frequency in applying tone sandhi to novel words. | |
