This project focuses on developing a computational model of naming deficit in aphasia in bilinguals, validating it with patient data, and using it to draw predictions for effective treatments. These predictions will be tested in a clinical trial with actual patients---to our knowledge the first time when a computational cognitive model is serving in this role. More specifically, a self-organizing map model of the lexicon will be built based on the existing DISLEX model, and trained with semantic representations and a large phonetic vocabulary similar to those of patients. It will be lesioned and retrained, simulating stroke damage and rehabilitation in individual patients. Predictions will be drawn about what the most efficient treatment protocol would be for each patient. It will then be extended to include lateral inhibition, demonstrating a possible mechanism underlying the observed cross-language transfer and interference effects. The University of Texas team will collaborate closely with the Boston University team, in particular building the model based on their semantic and phonological data, patient performance data, and treatment protocols. The results of the model will also be interpreted in terms of protocols available in their lab, resulting in recommendations in the clinical trial, and for the future. This project is supported by NIH under grant 1U01DC014922.