Native speakers can identify second language learners based on both their inability to use grammatically correct constructions and their non-native prosody. This perception of foreign-ness influences native speakers? impression of the learners (i.e., the learners tend to be perceived as socially inferior to speakers of the target language group).
Researchers in second language acquisition have been investigating prosodic problems of Japanese learners; including mastery of long/short vowel distinctions, usage of accent patterns, and sentence intonation patterns. However, learners? productions often involve problems at multiple levels of the prosodic system. Levelt?s speech production model proposed the existence of a prosody generator that is integral to the phonological encoding process. The model assumes an integration of syntactic, morphological/metrical information, segmental and intentional information, in order to construct prosodic constituents. Using this model, we can assume that second language learners? speech sounds non-native-like because the prosody generator either lacks some information or that some of the information arrives at the generator more slowly.
Data from novice learners of Japanese were collected. Each data set contains production and perception of syntactic structures, mora-timing/word accent, and interrogatives. Also, free monologue production was collected in order to obtain a global estimate of foreign accented-ness. In order to collect spontaneous speech data, a picture elicitation task was used. All oral data, which was digitally recorded on a computer, were analyzed with computer speech analysis software to determine syllabic durations, as well as fundamental frequency and intensity contours.
The comparison of the various components of prosody, in both production and perception, together with ratings of global foreign accented-ness made it possible to identify the components of L2 acquisition that contribute to make the learner?s prosody more native-like or more non-native-like.