Integrating the application and realization of Mandarin 3rd tone sandhi in the resolution of sentence ambiguity

Chinese third tone sandhi (T3S) covaries with the prosodic hierarchy both in the probability of application and in the realization of pitch slope. This paper evaluates whether Mandarin-speaking listeners integrate the covariation between T3S and prosody to resolute sentence ambiguity. Twenty-seven structurally ambiguous sentences were designed, each containing two consecutive T3 syllables situated across a word boundary, and the strength of the T3-intervening boundary crucially differentiates different interpretations of the sentence. The first T3 was manipulated to bear either a low, a shallow-rising, or a sharp-rising pitch. Sixty native Mandarin-speaking listeners heard each of these sentences and chose from two written interpretations the one that was consistent with what they heard. The results show that listeners are more likely to report a major-juncture interpretation when T3S does not apply (low) than when it applies (rising), and in the latter case, when the T3S variant has a sharper rather than shallower slope. Post-hoc analyses show that the T3S application is a more robust parsing cue for short sentences (4-5 syllables long), whereas the pitch shape of T3S is a more efficient parsing cue for longer sentences, indicating that listeners make sophisticated use of tonal variation to facilitate sentence processing. © 2020 ISCA.