Variable f0 of toneless moras in Suzhou Chinese: a computational analysis

Toneless Tone Bearing Units as an analytical device have been widely used in tone and intonation studies, yet there is a lack of consensus on how toneless TBUs realize in f0. This study investigates the variable f0 realization of phonologically toneless moras in Suzhou Chinese (Northern Wu) with the help of Naive Bayes classification models. By embedding the toneless moras in different contexts (e.g., between two H tones, preceded by H and followed by L), I have found both intra- and inter-speaker variation across the tonal contexts. Speakers of Suzhou Chinese realized toneless moras in three main ways: insertion of a 'Default L' tone, linear pitch interpolation between the left and right tonal contexts, and spreading of the left tonal context. In addition, I found insufficient evidence for the proposal that toneless TBUs as weak elements in speech realize as a stable target in the mid pitch range (Y.Chen & Xu 2006). The categorical pitch variation of toneless moras is indicative of a phonological model of variable/optional processes, where a single underlying representation (here, the absence of tone) may map to distinctive surface forms (Coetzee & Pater 2011; Coetzee & Kawahara 2013).