The nature of variation in tone sandhi patterns of Shanghai and Wuxi Wu

The primary goal of this dissertation is to understand the variation patterns in suprasegmental processes and what factors influence the patterns. To answer the questions, we investigated the variation patterns of tone sandhi in the Shanghai and Wuxi Wu dialects of Chinese. Shanghai disyllables and trisyllables have been documented to have two different sandhi patterns: tonal extension and tonal reduction. Some items can only undergo tonal extension, some items can only undergo tonal reduction, and some can variably undergo either type of sandhi. Previous works have indicated that the syntactic structure, semantic transparency, and lexical frequency of the items all play a role in the sandhi application. Additionally, the morpheme length of trisyllabic items (1+2, 2+1) is also expected to affect their sandhi application. A variant forms’ goodness rating experiment, together with a lexical frequency rating experiment and a semantic transparency rating experiment, showed that syntactic structure has a primary effect on sandhi application in general. It overrides the effect of semantic transparency, especially in modifier-noun items. The nature of the lexical frequency effect in Shanghai is related to the syntactic structure of the lexical item. Morpheme length effect is not found. Wuxi disyllables and trisyllables also have been observed to have two different sandhi patterns: tonal substitution and no sandhi. Some items can only apply tonal substitution, and some can apply either form variably. Syntactic structure and semantic transparency have been reported to affect Wuxi sandhi application, and morpheme length is also expected to have an effect in trisyllabic sandhi application. The three rating experiments conducted in Wuxi found that due to the lexical listedness of the opaque substitution pattern, frequency influences both modifier-noun and verb-noun positively, although modifier-noun prefers tonal substitution form more. Semantic transparency effect is only apparent for verb-noun disyllables. Moreover, morpheme length also distinguishes sandhi application between 1+2 and 2+1 modifier-noun items. In all, by using quantitative rating experiments, the present study shows that tone sandhi variation is regulated by both grammatical factors, such as syntactic structure, morpheme length, phonological opacity, and nongrammatical factors, such as lexical frequency