The syntax and phonology of non-compositional compounds in Yixing Chinese

This article explores the implications of the distribution of tone sandhi domains in Yixing Chinese (a largely undescribed Wu variety) for the syntactic analysis of a type of compound, labelled a non-compositional compound (NCC). Various diagnostics identify these compounds as a well-defined type in Yixing, including non-compositional/idiomatic semantics and opacity in coreference and coordination. NCCs also undergo a particular tone sandhi process, Pattern Substitution (PS). These diagnostics, we show, suggest an analysis whereby NCCs are formed by Merging two uncategorised roots. The root Merge is symmetric, leading to questions of how the resulting structure is to be linearised and labelled. The linearisation, we suggest, is determined post-syntactically by an Encyclopedia entry imposing an order due to the diachronic reanalysis of a compositional structure. This Encyclopedia entry is also the source of the non-compositional semantics of these structures, permitted by the encyclopedic search (en-search) of Borer (2013a). Symmetric root Merge cannot yield a compositionally derived meaning, which we show follows from the theory of labelling of Chomsky (2013, 2015). Our account explains proposals by authors such as Arad (2003) and Borer (2013a), who stipulate that a first-categorisation domain receives non-compositional, atomic content. The paper also includes a preliminary description of the Yixing tonal system in general.