Word Cognate Status and Concreteness: ERPs in L1 and L2 processing

Abstract

Event Abstract Back to Event Word Cognate Status and Concreteness: ERPs in L1 and L2 processing Ivo Popivanov1*, A. Janyan1 and E. Andonova1, 2 1 New Bulgarian University, Bulgaria 2 Bremen University, Germany To address the issue of bilingual language representation and processing we crossed a semantic variable (concreteness) and the cognate status of words in a translation verification task. Cognates are words that have a similar form and meaning in both languages (L1 & L2) while non-cognates in a translation task have similarity in meaning but not in word form. The general finding is that cognates are recognized and produced faster than non-cognates (e.g., Kroll & Stewart, 1994) which are usually attributed to a common set of form-based representations across the two languages. Models of bilingual language processing disagree on the issue whether cognate processing is affected by overlapping representations at the lexical level only or the semantics beyond (Kroll & De Groot, 1997). We test the models’ predictions and examine the extent to which semantics (more specifically, concreteness, i.e., concrete vs. abstract words) is involved in cognate vs. non-cognate processing. 11 young Bulgarian-English bilinguals performed a single word translation verification task from L2 (English) into L1 (Bulgarian). The words were presented for 300 ms with a 500-ms interstimulus interval and with an intertrial interval that varied between 1700 and 2200 ms. We used a go/no go task (to press a button if a non-equivalent translation is detected and do not press a button if the second word is a translation equivalent). Grand average ERPs were analysed during both L2 and L1 processing in two time windows on mean amplitude. A repeated-measure ANOVA showed that during L2 word processing the semantics of non-cognates was activated starting from 140 ms after stimulus onset while for cognates it was only activated later, starting at word offset. In addition, the magnitude of the cognate effect was larger during the second time interval, i.e., at later stages of processing. Furthermore, the analyses suggested that during the phase of translation recognition (L1 word processing) cognate processing did not involve semantics in either time window while non-cognate processing did activate it starting from 200 ms after word onset till the end of the second window (600 ms). This study shows for the first time that cognate processing in L2 activates semantics very early which is consistent with studies on monolinguals (see Pulvermüller, 2007). Overall, the results offer support to word processing in the bilingual’s two languages being semantically mediated and affected by the cognate status of words, and it thus contributes to our understanding of language representation in bilinguals. Findings are discussed in terms of the models of bilingual language representaion (Kroll & De Groot, 1997).