Listen-and-repeat training of non-native vowel quality

Abstract

This study investigated how listen-and-repeat training affects the production of non-native (Swedish) vowels /y/ and /u/ by speakers of different Namibian languages. Seventeen speakers, who did not have /y/ or /u/ categories in their L1, participated in the experiment. Training effects were measured with acoustic analysis and an identification task performed by 40 proficient Swedish speakers, to see whether the acoustic quality and the perceptual salience of the speakers’ non-native production evolved during training. We expected the speakers’ production to change as a function of training, and the change to be reflected on the vowel formant values and the identification task. The results showed that the speakers produced /u/ close to the trained target already in the first session, but changed their production away from the target after the first training. The speakers’ production of /y/ did not change significantly. The speakers did not reach a perceivable spectral difference between the two non-native vowels. The participants’ productions remained inconsistent throughout the experiment. There was great inter-speaker variation, which could not be accounted for by the speakers’ language backgrounds.