Riven, L. & de Almeida, R. G.
Concordia University
According to the 'good enough' approach, the language comprehension system often fails to complete a syntactic analysis of sentences when heuristic interpretations are available. For instance, when interpreting non-canonical sentences like 'The dog was bitten by the man', native speakers have been shown to rely on linearity cues rather than structure, often reporting that 'the dog' is the agent of the sentence. We investigated the attentional dynamics underlying such performance errors. English-French bilinguals reported the agentive noun of active and passive sentences in their L1 (English) and L2 (French) in single-task and dual-task blocks. The dual-task blocks required participants to retain six digits during sentence interpretation. Participants performed at ceiling for active sentences in all conditions, but we observed a crossover interaction between task demands and language of presentation for passive sentences (p=0.020). Without cognitive load, participants were less accurate in their L1 (M=60%) than their L2 (M=69%), but when extraneous task demands were introduced, accuracy increased in L1 (M=67%) and decreased in L2 (M=52%). These data suggest that variation in ?heuristic? and ?algorithmic? processing is not associated linearly with task difficulty, and that the nature of competition between these types of processes is different in native and non-native languages.