Peggy M. Schmid
Grace H. Yeni-Komshian
Dept. of Hear. and Speech Sci., Univ. of Maryland, College Park, MD 20742
This study makes use of a listening for mispronunciations (LM) task utilizing English sentences produced by native and nonnative speakers. The test sentences included high (H) and low (L) predictability target words containing mispronunciations (MPs) that were constructed by changing the initial segmental phoneme in the target words along the dimensions of voicing, place, and manner. Sentences not containing MPs were judged by native listeners for overall accent and the speakers were classified as native, having a mild-to-moderate accent or having a heavy accent. Another group of native listeners were presented the LM task. The results showed that the listeners were faster and more accurate in detecting MPs produced by the native than non-native speakers. They were more accurate, but not faster, in detecting MPs produced by non-native speakers with milder accents, as compared to heavier accents. The results also indicate that the listeners were faster and/or more accurate n detecting MP in the H sentences, and this was more pronounced with the sentences produced by the non-native speakers. The findings suggest that listeners are less accurate in detecting mispronunciations produced by non-native speakers and that the accuracy of their responses decreases as the degree of accent increases.