“Perceived” talker variability in high-variability perceptual training

Abstract

It has been suggested that higher talker and acoustic variabilities in training materials are the driving factors impacting learning advantages, such as better generalization and longer retention of training effects. To investigate whether such advantages are diminished when trainees do not perceive talker variability in training stimuli, we gave three days of training on identifying Korean stop contrasts to three groups of English speakers: a group with three distinctive-voice stimuli (DV), a group with three similar-voice stimuli (SV), and a group with one-voice stimuli repeated three times (OV). We compared their performances twice: once right after the last training session and again one week later. Results showed the tendency that the DV group was better than the other two groups at generalizing their training effects to a novel talker, and this remained the same one week after their training. Although some participants in the SV group claimed they heard multiple voices during their training, their performance was similar to that of the OV group. Taken together, the results support the previously reported high-variability training benefits, and suggest that in training stimuli, the “perceived” talker variability, not just the talker and acoustic variability, is a driving factor for such benefits.