Training segmental productions for second language intelligibility

Two experiments examine the effects of minimal pairs-based segmental intelligibility training on American English speech produced by Mandarin Chinese speakers. In each experiment, 15 hours of segmental production training is delivered by a computer-based training system that provides feedback on the recognition of minimal pair members. In Experiment 1, five Mandarin Chinese speakers of English as a second language train on production accuracy of target words representing six different English phonemic contrasts. Experiment 2 tests whether the same type of training can be expanded from six phonemic contrasts to all of the most frequently occurring contrastive vowel and consonant sounds in American English, whether word-level training improves sentence intelligibility, and whether improved sentence intelligibility is related to vowel versus consonant segment training.; Results show that computer-based minimal pairs-segmental intelligibility training produces significant improvement on the production of phonemic contrasts in isolated words and that these results generalize to phonologically similar words excluded from training. Vowel and consonant trainees make similar word level improvement. The effect of word level training on improvement in sentence intelligibility is significant when results from trainees in both vowel and consonant training groups are collapsed. Neither training group alone shows significant improvement in sentence intelligibility as a result of word training.