Changes in L2 production variability associated with visual biofeedback training

Abstract

Previous research suggests that acoustic variability across repeated productions of a phoneme could index the robustness of speech motor plans. Variability at onset is thought to reflect robustness in feedforward control [1,2], while variability at midpoint may reflect the narrowness of sensory targets and/or speakers’ capacity for feedback correction [3]. In L2 production, speakers may show elevated variability at onset because of unfamiliar motor plans and at midpoint due to weak sensory targets [2]. This study investigated variability in L2 vowel production before and after a brief training incorporating visual biofeedback. We hypothesized that midpoint variability would decrease after training, reflecting refinement of the auditory target, whereas onset variability may remain unchanged because the limited training might not be sufficient to establish a robust feedforward plan. Sixty native English speakers received 1 h of biofeedback training to produce two Mandarin vowels (/y, u/). After training, both vowels showed decreased midpoint variability, but only /y/ showed reduced onset variability. In addition, for /u/ only, higher pre-training midpoint variability was correlated with greater improvement in accuracy (distance from native-speaker target). These results are discussed in connection with the differing relationships of Mandarin /y/ and /u/ to the English vowel inventory.