A computational model of bilingual inhibitory control in a lexical decision task

Results from empirical studies in bilingual research demonstrate that “code-switching” in bilinguals, i.e., the change from one language to another, comes at a cost. What is still unclear is the exact nature of the cost. In this paper, we develop a connectionist model of Green’s Inhibitory Control (IC) hypothesis for a bilingual lexical decision task (Green, 1998, 74), which hypothesizes that the cost of switching between two languages in this task is due to inhibitory forces among different task schema and word forms. We were able to fit the model to the empirical data from Von Studnitz and Green (1997, Experiment 1), thus providing a proof-of-concept model that the IC hypothesis can account for the observed language switching costs and for the L1 language advantage effect in a lexical decision task. More extensive, systematic explorations of the model and its parameters in related but different tasks are, however, required to determine the extent of the IC applicability. Although the IC hypothesis is simple, it advanced earlier hypothesis by suggesting that the major source of language switching costs lies not in the lexico-semantic system but in a task/decision system; this notion is incorporated by Dijkstra and Van Heuven (2002) in their BIA+ hypothesis. By modelling the IC hypothesis, we suggest that not only is it a plausible explanation for the locus of control but also the reasonableness of the BIA+ hypothesis as the basis for further research in code switching and language learning.