Connectionist modeling of relearning and generalization in acquired dyslexic patients

Connectionist models implement cognitive processes in terms of cooperative and competitive interactions among large numbers of simple, neuron-like pro cessing units. Such models provide a useful computational framework in whic h to explore the nature of normal and impaired cognitive processes. The cur rent work extends the relevance of connectionist modeling in neuropsycholog y to address issues in cognitive rehabilitation: the degree and speed of re covery through retraining, the extent to which improvement on treated items generalizes to untreated items, and how treated items are selected to maxi mize this generalization. A network previously shown to model impairments i n mapping orthography to semantics was retrained after damage. The degree o f relearning and generalization depended on the location of the lesion and had interesting implications for understanding the nature and variability o f recovery in patients. In a second simulation, retraining on words whose s emantics are atypical of their category yielded more generalization than re training on more typical words, suggesting a counterintuitive strategy for selecting items in patient therapy to maximize recovery. Taken together, th e findings demonstrate that the nature of relearning in damaged connectioni st networks can make important contributions to a theory of rehabilitation in patients.