Pronunciation Pedagogy and Speech Perception: Three Studies

This dissertation investigates how second language (L2) speakers perceive non-native speech and how language learners can be helped to better perceive differences between their L2 output and target language speech, thus facilitating improvements in pronunciation. Study 1 investigated which dimensions underlie the perception of L2 speech by L2 listeners. Fifteen L2 listeners (and 10 native English listeners who served as a baseline group) rated 30 L2 audio-recordings from controlled reading and interview tasks for dissimilarity, using pairwise comparisons. Multidimensional scaling analyses revealed that fluency and global aspects of the speakers’ pronunciation explained listener judgments but that there was little agreement across the L2 and native listener groups. Study 2 investigated the role of language background in comprehensibility judgments. English speakers from Mandarin, French, Hindi, and English language backgrounds (10 per group) rated the speech of 30 L2 speakers from the same language backgrounds for comprehensibility and provided verbal reports about each rating. They then rated the speakers for segmental and word stress errors, intonation, and speech rate. Correlations between the speech measures and comprehensibility ratings for each L2 listener-speaker group and hierarchical regressions carried out for each L2 listener group revealed that different speech measures contributed to comprehensibility for different listener-speaker groups, with language background accounting for an additional six percent of the variance in comprehensibility ratings for the Mandarin listeners after the linguistic variables were taken into account. Analysis of the verbal reports for whether the listeners attributed their ratings to the speakers’ language background showed only a moderate relationship to the quantitative data. Study 3 investigated the efficacy of shadowing, a common pronunciation practice technique. Sixteen learners practiced shadowing with iPods for eight weeks. Two language tasks (shadowing task, extemporaneous speaking task) were administered as pre-, mid-, and post-tests, and were rated by 21 L1 English listeners. The shadowing task was evaluated for the learners’ ability to imitate a speech model, and the extemporaneous speaking task was rated for comprehensibility, accentedness, and fluency. Results indicated that the learners improved significantly in all speaking measures apart from accentedness and were largely positive about the activities.