A comparison of training effects on non-native tone sandhi production between American English and Cantonese speakers

Previous studies suggest that productions of Mandarin tone sandhi by both American English speakers and Cantonese speakers were perceived more native-like after a laboratory perceptual training, whereas little is known about the effects of tonal or non-tonal backgrounds. Ten Cantonese-speaking trainees and ten American English-speaking trainees matched in age and Mandarin proficiency were recruited to the pre- and post-training recording sessions. Elicited with audio and visual stimuli, participants naturally produced disyllabic real and wug words where the two Mandarin tone sandhi rules (T3 + T1/T2/T4 sandhi and T3 + T3 sandhi rules) should be applied. In total, 7680 sandhi syllables obtained from two sessions were perceptually evaluated by two phonetically trained Mandarin-speaking raters on a 101-point scale. Statistical results indicated that native tonal/non-tonal backgrounds influence Mandarin learners’ improvement in the two sandhi rules differently. The Cantonese trainees outperformed the English trainees in the sandhi of T3 + T1/T2/T4 before training, and the two groups had statistically comparable performance after training, although both groups exhibited significant improvement. For the sandhi in T3 + T3, improvement occurred for the Cantonese trainees while not for the American trainees after training, suggesting that the successful learning of phonological T3 sandhi rule may require a tonal background.