Five computational models of lexical access during production are tested for their ability to account for the distribution of aphasic picture-naming errors. The naming profiles (N= 14) were chosen from the literature to represent patients who make a relatively large number of omission errors. The most successful models combined the damage assumptions of the semantic-phonological model of lexical access (Foygel & Dell, 2000) with a treatment of omission errors as largely independent from overt errors (Ruml, Caramazza, Shelton, &Chialant, 2000). An explanation for the occurrence of omission errors was provided by the addition of a lexical-threshold parameter (Laine, Tikkala, ocjuhola, 1998) to the model. Suggestions for further testing of these models are introduced, as is a new website that allows other researchers to make use of the models.