The organisation of the bilingual lexicon: a PET study

In the literature on bilingualism, cognate relatedness has been shown to interact with proficiency in the foreign language such that cognate items are a measure of higher mastery than non-cognate ones. Systematic variation of these items in an activation study thus allows inter-subject control of different proficiency levels in the same bilingual subject. A positron emission tomography (PET) experiment was conducted with 11 Belgian subjects who were native speakers of Flemish/Dutch and had good proficiency in French. The experimental task consisted of internally naming pictures of objects in French and Dutch. For each language, there were two conditions. In one condition, the picture names were cognate across both languages, i.e. the translation equivalents shared phonological/orthographic similarity. In a second condition, picture names were non-cognate items without phonological/orthographic similarity in the French-Dutch translation equivalents. The items in both subsets were matched for visual complexity, object familiarity and familiarity of the French word. The results of our PET experiment on bilingual production showed hardly any difference between L2 cognates, L1 cognates and L1 non-cognates. The few significant differences of activation between L1 and L2 were almost exclusively due to increased activation for French non-cognate items in left inferior frontal and temporo-parietal areas. Thus, non-cognate foreign language naming recruited additional brain regions compared to native language non-cognate and native as well as foreign language cognate items. In our study, these areas of additional activation in L2 noncognates were restricted to left hemispheric inferior frontal and temporal structures in or at the periphery of the known language areas. In conclusion, the bilingual production study presented here supports an important role of proficiency in activation differences between L1 and L2. The results suggest a relation between activation in left BA 20, BA 44 and BA 47 and effortful lexical retrieval. Our results are not in line with the hypothesis of entirely distinct neural substrates for the different languages of a bilingual individual. (C) 2003 Elsevier Science Ltd. All rights reserved.