Evaluating the effects of perceptual training on application of Mandarin tone sandhi rules by English speakers

Abstract

Application of the tone sandhi rules is a widely studied topic in speech production, but the production data have rarely been perceptually evaluated. Meanwhile, whether perceptual training improves speech production at the suprasegmental level remains a question. Adopting the approach of perceptual evaluation, this study examines the effects of perceptual training on the Mandarin tone sandhi rule application by ten American English speakers. All the participants attended a pre-training recording session, a short-term laboratory training session, and a post-training recording session. Ten native speakers of Beijing Mandarin also participated in the recording as a control group. There were 192 target stimuli, with both real and wug words in each recording session. In the training session, the participants were presented with auditory and visual inputs of Mandarin tone sandhi rules, which were followed by an identification test to evaluate the training. Two trained Mandarin-speaking linguists first manually segmented the target syllables, normalized the intensity at 55 dB, anonymized the participants, and then performed the perceptual evaluation of each target syllable on a 101-point scale. A significant effect of perceptual training was found on the half-third sandhi rule, but not on the third tone sandhi. The roles of familiarity (trained versus untrained words) and context (real versus wug words) on the training effect were also investigated.