Evaluation of a technique for training new speech contrasts: generalization across voices, but not word-position or task.

We used the perceptual fading technique (Jamieson & Morosan, 1986) to teach unilingual adult Canadian francophones to identify the voiceless and voiced linguadental fricatives, /0/ and /o/. Training began with the identification of synthetic consonant-vowel (CV) exemplars that contained exaggerated amounts of frication (140 ms), with feedback given to identify errors and correct responses. Subsequently, stimuli with progressively shorter fricative durations were added to the identification set. After just 90 min of such training, francophone adults were better able to identify both the training stimuli and an untrained set of natural CVs produced by four different speakers, two men and two women. These results replicate and extend those reported by Jamieson and Morosan, and establish that such training generalizes to the identification of natural CVs produced by a variety of voices, both male and female. Thus, the effects of training are not restricted to a small range of acoustic cues. However, such learning was still strongly sensitive both to other aspects of the acoustic context and to the testing situation: Training failed to improve identifications of /0/ and /o/ phonemes when these were presented in word-medial and word-final positions, and identifications of synthetic and natural examples of /o/ vs. /d/ did not improve, even though identifications of /0/ vs. /o/ were very good.