Prosodic structure in language understanding: Evidence from tone sandhi in Mandarin

Abstract

Two experiments show that listeners can use phrase-level prosodic information to disambiguate local lexical ambiguities that occur due to the operation of the tone sandhi rule in Standard Mandarin Chinese. In Chinese, each word is associated with a tone; changes in a word's tone that occur in relation to the tones of words nearby in phrasal context are referred to as “tone sandhi,” e.g., ([L] → [LH]/_[L]). In experiment 1, listeners identified lexical tones for ambiguous, unambiguous, and nonsense words in phrasal contexts where the tone sandhi rule might have applied. Comparable results in the lexical versus nonsense conditions indicate that judgments did not rely simply on lexically stored tonal information, but also made reference to prosodic structure. In experiment 2, subjects chose the correct written English translation for auditory sentences of Mandarin. Global prosodic information was manipulated to create different levels of “prosodic closeness” between two critical items in a tone sandhi environment, but the syntactic relation between these items was held constant. Results show that listeners relied on the prosodic structure of the phrases to determine whether or not the tone sandhi rule had applied, and consequently to identify individual lexical items. The evidence is taken to support the notion that prosodic structure influences auditory language comprehension processes. [Work supported by NICHD.]