Resumo

This study aims to intercept the bilingual switching mechanism to verify the micromodularity of bilingual lexical access. Late Dutch(L1)-Portuguese(L2) bilinguals were tested in two ERP priming experiments. First, different crosslinguistic morphological pairs were tested in monomodal (written) and bimodal setting (auditive-written) (such as vis-VISSERIJ (L1toL1); peixe-PEIXARIA(L2toL2); vis-PEIXARIA(L1toL2) or peixe-VISSERIJ(L2toL1)). Second, similar pairs with image primes preceded by 3,2,1 countdown, in L1 or in L2. In the first experiment, faciliation was expected regardless of switching, based on the theoretical framework of Distributed Morphology claiming all words are derived compositionally. Also, we assume that morphosyntactic and phonological and logico-semantic modules are language specific, whereas semantic entries are shared. in the given priming scenario. Thus, in the given priming scenario, in which prime and target (e.g. peixe-VISSERIJ) contain roots that access the same encyclopedic entry, e.g. [√peix + -nom. (L2) ↔ ‘an aquatic animal typically with caudal fin, fins and gills, etc.’ ↔ √vis + -nom. (L1))], a priming effect was expected is expected regardless of language switching as (i) semantic entry is already activated by the prime; (ii) switching occurs at a ‘language neutral’ moment. ERP latency measures for N400 signatures, commonly correlated to lexical access, showed there is, indeed, no difference in access timing for the bilingual condition as there is no main effect for language except for a slight advantage for target recognition in L1 which did not reach significance across all derivations. N400 latencies of the Cz derivation, for example, did not yield a main effect for language a=0,263/F1(3,13)=1,41 with L1 to L1=428,57ms(13,02ms), L2 to L1=415,36ms(9,57ms), L2 to L2=428,93ms(9,48ms), L1 to L2=438,93ms(11,75ms). That indicates that neurological data tapped into early lexical access stages equally fast for L1 and L2, whereas behavioral measures showed a disadvantage for L1 to L2 switching (reaction times: L1toL1=670,76ms(13,07ms) x L2toL1=681,71ms(11,95ms), a= 0,139; and L2toL2=718,78ms(14,62ms) x L1toL2=753,49ms(13,99ms), a=0,00). This suggests that in later phases of L2 word recognition other aspects come into play, such as inhibition/activation processes and costly additional morphological L2 processing, and increased burden on memory and motor-executive interfaces. However, essentially bilingual lexical access seems to be compositional. In the second experiment, inhibition for language switch combinations was expected as we predicted the image prime would trigger the derivation of an associated word guided by the preceding countdown (either in L1 or L2), starting at the conceptual module going into the language-specific derivation at which point a written target was presented: e.g., [countdown in L2] três, dois, um→[prime] →(CONCEPT of [fish])→√peix+e (L2)→[target] VISSERIJ (L1). Monolingual and bilingual combinations were presented. Nor behavioral nor ERP measures confirmed our hypothesis categorically, with absence of main effect for language (a=0,434/F1(3,12)=0,869, for N400 latencies), possibly indicating that we were not successful in setting up a ‘switching’ paradigm, or that the image prime is not suitable for achieving morphological priming (which was confirmed by absence of priming effect in comparison to unrelated pairs).