Recent findings from large corpus studies of spontaneous English speech show that the predictability of a word is negatively correlated with the duration of its pronunciation. This predictability effect extends to word repetition in discourse, where words shorten as they are re-used, to word frequency, with shorter pronunciations for high frequency than low frequency words, and to homophones, where the higher frequency meaning has a shorter duration than its lower frequency counterpart. Standard Mandarin, due to its relatively unmarked phonology and morphology, presents the potential for a large degree of homophony and resulting lexical ambiguity. We test whether predictability factors such as word frequency, repetition, and prosodic phrasal position influence the duration of Mandarin homophones in simple sentence reading. Native monolingual Mandarin speakers read phonologically identical sentence pairs, which differed only in semantics and orthographic representation. Each pair member contained either an ambiguous high or low frequency word; these ambiguous pairs were placed in one of four different prosodic locations throughout the stimuli. Sentences were read multiple times with durations measured for each ambiguous token. The results corroborate previous findings from corpus analyses showing duration reduction with repeated mention and effects of relative meaning frequency on homophone durations. In addition, we found differences in duration due to prosodic phrasal position, but no interaction of prosodic position with other forms of word predictability. Our findings indicate that speakers produce subtle durational cues to indicate a range of information about the words they use. Such information may be useful to listeners who may use this subtle information to minimize ambiguities.

Index Terms: speech prosody, word frequency, Mandarin Chinese, lexical ambiguity, word duration