Optimal prosody for different proficiency Chinese talkers of English: Pitch and speech rate

This paper investigates native listeners' perceived accentedness for pitch- and duration-manipulated speech of Chinese L2 talkers and compared it with that of nonnative Chinese and Korean listeners. It attempts to find out actual numeric values of pitch and speech rate where low proficiency talkers are perceived as optimally improved proficiency and where high talkers are judged as the least accented. Results showed that Chinese low talkers were judged as a higher level of proficiency, intermediate, when the pitch ranges were synthetically expanded twice. High talkers were, on the other hand, degraded into intermediate proficiency when their pitch ranges was manipulated to be almost zero, but they were rated as optimally accented, close to the score of 1, when the pitch ranges were extensively widened two or three times. Speech rate did not, however, affect native listeners' perception of low talkers' stimuli, but the rating scores showed a curvelinear shape along with the changes of speech rate for high talkers. The optimal speech rate, where high talkers were perceived as the least accented, emerged as 6~6.5 s/s. It was 1.2 to 1.5 times faster than high talkers' original speech. These results cast a few pedagogical implications which can readily apply to teaching English. Prosodic factors such as pitch range or speech rate are sufficient attributes to altering the perceptual categories of L2 talkers' proficiency (except for speech rate for low talkers). The optimal numeric values of pitch range and speech rate will be solicited practically for low and high proficiency talkers in the pedagogical environment. In addition, nonnative Chinese and Korean listeners did not show statistically significant correlations with native listeners in the perception of foreign accent regardless of their L1 backgrounds or L2 proficiency. This might be plausibly due to the fact that prosodic features of nonnatives' interlanguages were considerably dissimilar to those native listeners. This might induce inconsistent ratings between native and nonnative listeners.