Perception and production of non-native prosodic categories

Cross-language phonological variation is manifested at the levels of segmental and prosodic structure. Modeling the interactions of two segmental systems during non-native speech perception and production has been highly successful (e.g. Best et al., 2001; Flege 1995; Kuhl and Iverson, 1995). However, these models lack an account for cross-language perception and production of prosodic structure. The perception and production of non-native prosodic contrasts was investigated by testing native English speakers' perception and production of Mandarin. Mandarin is a lexical tone language in which pitch changes over syllables signal differences in lexical meaning whereas in English pitch patterns over whole phrases express pragmatic meanings.;Native English and Mandarin listeners' perceptual sensitivity to Mandarin lexical tones was tested in monosyllabic and tri-syllabic utterances. The English listeners exhibited overall lower and more variable sensitivity compared to the Mandarin listeners. In particular, the English listeners' sensitivity to the Mandarin tone contrasts varied depending on the acoustic similarity of the pitch contours for contrasting tones and on whether the tones were presented in monosyllabic or tri-syllabic utterances. In contrast, the Mandarin listeners were highly sensitive to all contrasts. Furthermore, the English listeners appeared to have been influenced by assimilation to English prosodic categories. Finally, while Mandarin listeners mostly attended to lexical tone targets, English listeners attended mostly to global aspects of the stimuli.;In a tone production task, the English participants imitated native Mandarin tone productions. These imitations were then evaluated based on native Mandarin listener judgments and acoustic analysis. The English participants accurately imitated the contour shape but deviated from the Mandarin model in ways that were shifted towards English production norms. Listeners' perception and production abilities were not significantly correlated.;In sum, listeners may perceive non-native prosodic contrasts on the basis of a combination of language-independent auditory processing strategies and assimilation to native prosodic categories. Models of cross-language perception of prosody must incorporate contextual variation into their predictions and account for listeners' use of both auditory and linguistic processing. These studies serve as the basis for a long-term research program aimed at extending current models of cross-language speech perception to the prosodic level.