Grainger, J.
In my past research I have described and tested a functional architecture for word recognition in bilingual persons, the Bilingual Interactive-Activation (BIA) model, that implements the principle that lexical access is initially language non-selective. In this theoretical approach, information about which language a word belongs to provides top-down feedback in order to limit the interference caused by co-active representations in the non-target language. I will summarize this approach and some key findings that support it. However, in this work we chose to ignore how the lexical representations of a second language are learned in the first place, and how the principles implemented in the BIA-model might emerge through prolonged exposure to the L2. Here I will provide a tentative integrative account of the process of second language vocabulary acquisition and its evolution up to the stage of high proficiency in the L2. The integrative approach combines elements of the Revised Hierarchical Model (RHM) and the BIA-model, while implementing standard neural net learning algorithms.