Seton, B. 1 , Loerts, H. 2 , Stowe, L. 1, 2 & Schmid, M. 1
1 University of Groningen, Center for Language and Cognition Groningen, The Netherlands
2 University of Groningen, School of Behavioral and Cognitive Neuroscience, Groningen, The Netherlands
In this study, we used event-related potentials (ERPs) to compare early bilingual Turkish-Dutch adults between 18 and 27 (AoA < 3 yrs) to a native Dutch control group when listening to Dutch sentences containing manipulations of gender concord. While Turkish does not have grammatical gender, the Dutch language has a rather opaque system of a common and a neuter gender that is visible on definite determiners and on adjective endings after indefinite determiners. Previous ERP studies on grammatical gender processing in bilinguals looked at late L2 aquirers and used visual presentation of stimuli. In the present study, we were interested in early bilinguals and auditory processing of gender violations. The two groups were matched on age and level of education. Participants were presented with auditory sentences in Dutch, either containing correct or incorrect grammatical gender concord, followed by a grammaticality judgement task (GJT). The Turkish-Dutch bilinguals who scored highest on the GJT showed an ERP pattern which was similar to that of the Dutch control group, namely a negative peak around 400 ms post-stimulus (N400) as well as a positive peak in activation around 600 ms after the target onset (P600) for the gender violations. The bilingual participants with the lowest accuracy rates on the GJT showed no difference between correct and incorrect sentences. The latter group only showed a P600 on a control condition with tense violations. These results imply that early acquisition and balanced use of the two languages are not necessarily a predictor of native-like processing.