UC Berkeley

In this dissertation, I describe three related experiments which explore the relationship between perceptual and articulatory learning during the process of second language phoneme acquisition. Novel phoneme acquisition is a well-documented challenge for adult learners which can persist even after extensive experience with the target language. The perceptual challenge typically manifests as an inability to distinguish between two or more target categories. Articulation performance often reveals a significant effect of the native language on pronunciation. While these problems have been extensively studied as independent phenomena, there is less work relating the joint development of articulatory and perceptual categories in second language acquisition. As a result, questions remain about the effects of cross-modal training - the extent to which learning in one domain can support development in the other.

This dissertation contributes to that body of literature with experiments that compare the effects of perceptual and articulatory training on the perception and production of Hindi coronal stop consonants by native English speakers. It focuses on adult learners who have no prior exposure to Hindi to explore patterns of learning at the earliest stages, before stable second language targets have formed. Of particular interest is the transfer of articulatory learning to perceptual categorization, a trajectory which has been explored in only a small handful of studies.

In experiment 1 (chapter 2), the joint contributions of perceptual and articulatory learning on category acquisition were assessed with a multi-day training experiment. Benefits were found for within-mode learning - perceptual training aided discrimination, and pronunciation was improved during articulatory training. However, cross-modal learning did not have an effect - pronunciation was not significantly improved by perceptual learning, and articulatory learning did not have an additional benefit on discrimination performance.

Experiment 2 (chapter 3) sought an explanation for the lack of a cross-modal effect in experiment 1. New learners who received only a single session of training showed discrimination improvement whether they received articulatory or perceptual training. This finding suggests that learners in experiment 1 failed to show an improvement from cross-modal learning not because of general inefficacy, but because they had already received prior training during perceptual learning. This result is taken as evidence that completely novice learners benefit from instruction about category targets regardless of the mode of training, but that within-mode learning may be more beneficial once some experience is gained.

In experiment 3 (chapter 4), the neural correlates of category learning were tested. The mismatch negativity response, a component of the electrophysiological response to auditory stimuli, was used to test pre-attentive categorization of selected target categories. The findings from this study indicate that learners are able to detect category differences both before and after training - a contrast to the behavioral results reported in experiments 1 and 2. This study suggests that pre-attentive auditory detection of non-native contrasts is possible even when behavior indicates an inability to categorize targets.

Taken together, these results provide some evidence for the efficacy of cross-modal training, but the effect is restricted to certain conditions. I argue that cross-modal learning may be most effective for purely novice learners; when a learner has not been introduced to a category paradigm before, it may be that any cue to category identity can be leveraged to begin to detect contrasts. This is provided that detection is undertaken as a conscious categorical task, and not a pre-attentive response. After some learning has taken place, within-mode training becomes more important to continued development of category representations. This work supports arguments that acoustic and perceptual processes may be the primary source of robust perceptual phonetic categories, while simultaneously supporting the involvement of motor representations as a secondary mechanism to support those categories. The dissertation concludes with predictions about the trajectory of learning beyond the novice stage, and suggestions for future avenues of behavioral and neural studies of non-native phonetic category representations.