| The development of models for the bilingual lexicon began in the 1950s. Its history is characterized by a continual process of making the models more complex in order to account for the data. Still today, no one model exists that satisfactorily represents the evidence because each current model only accounts for one aspect of processing.; Two main factors determine how the bilingual lexicon is organized: relative fluency in the two languages and their linguistic structures. Kroll's Revised Hierarchical Model accounts for fluency by describing the asymmetry of the lexical systems in the beginning bilingual. De Groot's Distributive Representation accounts for semantic relationships across language in the more fluent bilingual, while Grainger and Dijkstra's Bilingual Interactive Activation Model accounts for orthographic relationships across language.; The Overlapping Distributive Model combines elements from these three models and adds the phonological component. It proposes that the bilingual lexicon consists of two intersecting networks of information that share common representations and store language specific representations separately. While each lexicon is an integral unit capable of operating independently, cross-language links exist between any items that are phonologically, orthographically or semantically related.; Two experiments were conducted to test one prediction of the Overlapping Distributive Model: the more form related stimuli are, the greater effects they should produce. These experiments used highly fluent Spanish-English bilinguals as subjects, and stimuli which represented three levels of form relationship. The lexical decision task produced a pattern of results paralleling those found in monolingual studies, in which some form relationship produces priming while dose form relationships produce inhibition. While this is not the trend that was predicted, it still indicates that the more highly related forms have the strongest link.; The translation recognition task produced unexpected results in which there was inhibition for the positive responses, but a trend in the negative responses: it was increasingly difficult to reject nontranslation equivalents the more form information they shared. Furthermore, in both experiments, the negative condition with the greatest amount of form relation produced the greatest number of errors, also supporting the notion that they share a strong form link. |
