Training foreign speech contrasts

Abstract

The present work investigated the efficiency and generalizability of a technique for training francophone Canadians to identify and distinguish the phonemes / e I and I ?? /. The technique was developed in a pilot experiment, based on the fading technique used by Terrace (1963) to train perceptual contrasts with few subject errors. The initial technique involved training, beginning with the presentation of consonant-vowel (CV) exemplars, beginning with exaggerated amounts (140 ms) of the appropriate type of frication, in an identification training task with feedback. Subsequently, stimuli with progressively shorter fricative durations were added to the stimulus set for identification. In Experiment 1, just 90 minutes of such training was found to enhance the ability of francophone adults to discriminate between these sounds and to identify both the training stimuli and natural CVs produced by a single male speaker. Control subjects showed no such gains. Experiment 2 was conducted to test the generality of this technique. The results of the identification of the synthetic / e / - / ?? / training stimuli replicated the results of Experiment 1. The improved performance of subjects on the identification of natural CV s produced by two additional male and two female speakers demonstrated that the effect of training extended across voices. Training failed to improve identifications of /0 /and/ ??/phonemes in word-medial and word-final positions. Identifications of synthetic and natural examples of/??/ and /d/, in word-initial position also failed to show any effect of training. Experiment 3 attempted to train / 0 / -/ ??/discrimination and identification abilities using single synthetic exemplars of each phoneme. Identification training with feedback replicated the improved identifications of natural tokens that had been demonstrated in the results of Experiment 1, and showed moderate improvement in identification performances of the synthetic continuum of sounds. Together, the results indicate that laboratory training using intact speech exemplars can efficiently enhance listeners' abilities to percieve a foreign speech contrast. The utility of this training has been shown using the simple presentation of prototypes and using the more sophisticated fading procedure. Evidence to suggest the superiority of the fading technique to train the voiceless/voiced "th" contrast is weak. However, the flexibility and sophistication of this technique may prove to be necessary in the training of speech contrasts which differ more dynamically, along multiple speech cue dimensions.