The effect of training structure on perceptual learning of accented speech

Abstract

Although previous research suggests that high variability training facilitates perceptual learning of systematic variation in speech, the extent to which the organization of training materials affects this learning remains relatively unexplored. The present study examined the role of training structure on the perceptual learning of speaker-independent properties of Spanish-accented speech. During training, native adult speakers of American English transcribed sentences spoken in English by four native Spanish-speaking adults. Training stimuli were presented to listeners either grouped by speaker, sentence, or randomized with respect to sentence and speaker. At test, listeners transcribed novel sentences produced by four unfamiliar Spanish-accented speakers. Transcription performance at test was found to vary as a function of the organization of training materials. Listeners transcribed test sentences more accurately when training sentences were randomized relative to when training sentences were grouped by speaker or when listeners received no training. Test performance when training sentences were grouped by sentence was intermediate. These findings suggest that variable training structure may direct listeners' attention to accent-general properties of speech, allowing for comparison across speakers' voices and linguistic content, and generalization of learning to unfamiliar accented speakers.