Models of Impaired Lexical Access in Speech Production

Two computational models of lexical access in aphasic speakers are compared. Both are derived from an interactive two-step theory of retrieval based on spreading activation. The weight-decay model (Dell et al., 1997) associates patients with deviant values of two parameters, global connection weight and global decay rate. The new semantic-phonological model characterizes patients by two different parameters, semantic weight and phonological weight. We provide a detailed analysis of the properties of both models and show that the semantic-phonological model is a serious competitor to the earlier model. Not only does it fit existing patient data as well as the weight-decay model does but it also has closer ties to other facts and theories in cognitive neuropsychology and psycholinguistics.