Evaluating a Model of Naming

We evaluate the computational model of lexical access proposed (1997). They argue that ts of their model to naming data obtained from normals and brain-damaged patients support assumptions regarding interactivity in the lexicon, global damage in aphasia, and continuity between normal and aphasic naming behavior. Our investigation reveals that the model ts the empirical data poorly and that the claims Dell et al. make on the basis of the model's performance would not follow even if the model were accurate. Although we improve the model's t using a novel automatic regression procedure, it cannot account for ve of Dell et al.'s twenty-one patients (24%), and we show that its limitations are inherent in its design. We argue that claims such as those made by Dell et al. can only be addressed by considering evidence from multiple related tasks and by comparing multiple computational models. When one names an object, the semantic speciication of the name of the object is mapped to the sequence of phonemes that is that word's pronunciation. This cognitive process, known as lexical access, is quick, automatic, and usually successful, but very occasionally it can result in slips of the tongue: groups of phonemes representing words that are We would like to give special thanks to Gary Dell for his helpful comments and suggestions regarding this work, and for providing us with the source code for his implementation of his model. We also thank Deborah Gagnon for sending us detailed patient naming data, and Trevor Harley and an anonymous reviewer for their useful comments. We are indebted to Michele Miozzo, the Harvard Cognitive Neuropsychology Laboratory, Stuart Shieber, and the Harvard AI Research Group for insightful suggestions and stimulating discussions regarding this research.