Shannon Barrios · Even highly proficient bilinguals have homophonous representations for certain minimally contrastive L2 words
Producing and perceiving sound contrasts in a second language (L2) can be hard, particularly for adult learners. This difficulty persists despite high proficiency and lengthy exposure to the target language (Flege, Bohn, & Jang, 1997; Pallier, Bosch & Sebastián-Gallés, 1997; Pallier, Colomé, & Sebastián-Gallés, 2001). A key issue in L2 speech perception research is whether these difficulties are due to nonnative-like perception, representation, or both. We report results from two studies that show that  even highly proficient Spanish-English bilinguals fail to distinguish some nonnative sounds, such as /i/-/I/. Thus,  English words that differ minimally in these contrasts (i.e. ‘sheep’-‘ship’) are represented as homophones by these bilinguals.
Experiment 1 was an AX discrimination task examining the ability of 28 English speakers and 28 Spanish bilinguals to discriminate English word pairs distinguished by common and English-specific vowel contrasts. Experiment 2 employed Pallier et al’s (2001) medium-term repetition priming paradigm to investigate word recognition processes, using the same participants and stimuli.
Experiment 1 showed effects of Group & Condition (p<.05), as well as a Group-Condition interaction (p<.05), suggesting that Spanish bilinguals fail to discriminate the English contrasts. Experiment 2 also showed effects of Group & Condition (p<.05). Planned comparisons revealed no difference in the facilitation effect observed for minimal pairs (‘sheep’-‘ship’) when compared to repetitions (‘sheep’-‘sheep’) for the Spanish group only, suggesting that these bilinguals have shared representations for word pairs distinguished by English-specific vowel contrasts.
Consistent with Pallier et al (2001), these results show that even highly proficient bilinguals may fail to accurately perceive and, thus, represent certain nonnative contrasts. Moreover, replicating these findings with new language groups allows us to exclude alternative explanations for past findings based on cross-linguistic similarity.