Talker variability versus variability of vowel context in training naïve learners on an unfamiliar class of foreign language contrasts

Numerous studies showed that learners can improve their ability to discriminate and/or identify non-native L2 contrasts through phonetic training, and that encountering sufficiently varied stimuli during training leads to effective generalization of the learning gains. However, previous studies often conflate talker and phonetic-context variability, and tend to prioritize talker variability. The current study investigated the relative importance of each source of variability in training naïve learners on a contrasting class of L2 sounds, plain vs. emphatic Arabic coronals, which enabled us to test generalization more rigorously than is possible with a single L2 contrasting pair (e.g., English /l/-/ɹ/). All trained participants showed significantly better identification of the contrast after training. For trained items, variability in the training materials did not matter. However, when participants had to generalize to unfamiliar contexts, variability of training materials did matter, but there was no benefit of one type of variability over the other. These findings highlight the importance of both talker and vowel context variability to effectively generalize learning of non-native sound contrasts.