Tone sandhi in Xiangshan Wu Chinese and beyond: patterns of variation and diachronic development

This thesis investigates tone sandhi in Xiangshan Wu Chinese, an under-documented dialect within the Northern Wu group, with a focus on disyllabic and trisyllabic domains. Drawing on fieldwork data, it provides the first systematic acoustic documentation and analysis of Xiangshan tone sandhi. The findings contribute new empirical data, introduce innovative methodology for bottom-up prosodic research, and offer a new theoretical perspective on the development of tone sandhi systems in Wu dialects. Chapters 3 through 6 present the empirical findings. Chapter 3 establishes the Xiangshan monosyllabic tone system as a foundation for the sandhi analyses. Chapters 4 and 5 examine disyllabic tone sandhi across three syntactic structures, identifying general patterns aligned with typological observations across Northern Wu: lexical compounds and adjective-noun phrases (Chapter 4) share broadly left-dominant sandhi (LDS), while verb-object phrases (Chapter 5) typically exhibit right-dominant sandhi (RDS). Chapter 6 extends the analysis to trisyllabic domains, which involve more complex syntactic configurations built on, but not reducible to, the disyllabic types. Noun-headed trisyllables typically show a temporal expansion of LDS patterns, while verb-headed ones deviate from their disyllabic RDS counterparts, instead exhibiting a distinct LDS pattern. Crucially, though both adopt LDS, their output sandhi forms differ systematically, maintaining the syntactic contrast. Despite these general patterns, the Xiangshan system exhibits rich and structured variation, summarised along six aspects in Chapter 7: syntactic sensitivity, historical relevance, register split, domain independence, non-discrete dominance, and speaker variation. Existing synchronic frameworks, particularly edge-based and stress-based approaches, fail to account for this full range, especially those historically rooted. I propose a diachronic account tracing the emergence of LDS as a morphophonological marker of lexical cohesion, in contrast to the phonetic simplification underlying RDS. This account explains the Xiangshan and other Northern Wu sandhi patterns, offering a more general model for tone sandhi evolution. Methodologically, this study contributes an integrated approach combining machine-learning clustering with a vision-aided auditory categorisation paradigm. This method addresses challenges in categorising f0 objectively while maintaining perceptual grounding, offering a replicable framework for future research on under-documented languages.

Name dc:type.qualificationname

Doctor of Philosophy (PhD)

Level dc:type.qualificationlevel

Doctoral

Grantor dc:publisher.institution

University of Cambridge

Year dc:date.issued

2025

Author dc:creator

• Shi, Yibing

Advisors dc:contributor.advisor

• Post, Brechtje
• Nolan, Francis

• Chinese languages
• tone sandhi
• tones

Licence

• https://www.repository.cam.ac.uk/bitstreams/0f01041c-ea7b-4509-af29-c20d0584a1ab/download
• http://purl.org/NET/rdflicense/allrightsreserved

Language dc:language

eng

DOI dc:identifier.doi

https://doi.org/10.17863/CAM.130601

OAI identifier oai:identifier

oai:www.repository.cam.ac.uk:1810/403777