Second-language (L2) speakers have an observable accent, but L2 listeners may also have difficulty hearing differences between L2 speech sounds that signal a change in meaning (e.g., “rock” vs. “lock”). Classic studies have shown that perception can be improved through training, and that the benefit persists over time, but discrimination never reaches the accuracy level of a native listener. Recently Best & Tyler (2007) extended the Perceptual Assimilation Model (PAM; Best, 1995) to predict the likelihood of acquiring an L2 distinction. A number of PAM-L2’s predictions mirror those of the Speech Learning Model (SLM; Flege, 1995), but the models diverge in two key aspects: 1) SLM considers only similarity between an L2 category and a single native category whereas PAM-L2’s predictions are modulated by the assimilation pattern of phonetically neighbouring L2 sounds; 2) For PAM-L2, successful perceptual learning is more likely for learners with small L2 vocabularies, than those with large vocabularies, because the perceptual system must settle on phonological categories for vocabulary expansion to occur. Incomplete perceptual learning at the time of vocabulary expansion may lead to fossilisation and a reduced likelihood of successful training interventions.
The implication for language learning and instruction is that a learner’s perceptual assimilations of L2 sounds should be assessed before formal instruction begins, and a tailored perceptual training regime put into place before plasticity is reduced through increases in L2 vocabulary size. Such an approach would be best implemented in an automated computer program.
Authors: Michael D. Tyler
Event: SF08: Speed Papers