A semantic model to study neural organization of language during bilingualism

A neural network model of object semantic representation, developed in previous years, is used here to simulate the case of learning of new words from a foreign language. The network consists of some feature areas, devoted to description of the object properties, and a lexical area, devoted to memorization of words. Neurons in the feature areas are represented as Wilson-Cowan oscillators, to allow segmentation of different simultaneous objects via synchronization in the gamma-band. Excitatory synapses among neurons in the feature and lexical areas are learned, during a training phase, via a time-dependent Hebbian rule. In this work, we first assumes that some words and the corresponding object representations are initially learned during a preliminary training phase (“native words”). Subsequenly, a foreign word is learned by simultaneously presenting the new word together with the native one. Simulations show that, during the first learning period, stimulation of the foreign word alone allows retievial of the object properties via the intermediate activation of the native word. Conversely, after a prolonged training period, the foreign word becomes able to retrieve object representation per se. These results agree with recent data on second language representation.