Brekelmans, G. 1 , Evans, B. G. 2 & Wonnacott, E. . 1
1 Department of Language & Cognition, University College London
2 Department of Speech, Hearing and Phonetic Sciences, University College London
High variability phonetic training (HVPT), where speech contrasts are exemplified in multiple talker input, has become a standard paradigm in teaching novel L2 contrasts. From a statistical learning perspective, more varied input should aid generalization by focusing the learner on critical acoustic features, whilst de-associating irrelevant cues. However few studies have directly compared high and low variability input, and there is little work with child learners. Children may benefit less from varied input due to lesser phonological working memory (Martin et al, 1989). This study compares HV versus LV phonetic training in adults and children in two experiments. In Experiment 1 adult (N=48) and child (n=48) native speakers of English were trained to map words containing novel (Dutch) vowels to shapes representing letters using HV (4 talkers) or LV (1 talker) training. In Experiment 2 adult (N=48) and child (N=48) participants were trained to map similar words to pictures depicting their meaning. In both experiments, both adults and children showed significant learning of the mappings, however critically, there was no evidence of a high variability benefit in learning or generalization. Learning was stronger in Experiment 2, particularly for children, suggesting a benefit for a more meaningful learning context.