Towards a localist–connectionist model of word translation
Ton Dijkstra |
Donders Institute for Brain, Cognition, and Behaviour
Word translation is among the most sophisticated skills that bilinguals can perform. Brysbaert and Duyck (2010) have argued that the Revised Hierarchical Model (RHM; Kroll & Stewart, 1994), a verbal model for word translation in beginning and proficient bilinguals, should be abandoned in favor of connectionist models such as the Bilingual Interactive Activation Plus model (BIA+; Dijkstra & Van Heuven, 2002). However, the partially implemented BIA+ model for bilingual word recognition has neither been applied to bilinguals of different proficiency levels nor extended to the complex process of word translation. After considering a number of aspects of the RHM, a new localist-connectionist model, called Multilink, is formulated to account for the performance of bilinguals differing in their L2 proficiency in different tasks: lexical decision, language decision, and word translation.