Superadditive effects of multiple lesions in a connectionist architecture: Implications for the neuropsychology of optic aphasia

Mark Sitton, Michael C. Mozer, Martha J. Farah
2000 Psychological review  
Traditional accounts of neuropsychological disorders have been based on the assumption that brain damage results in a single focal lesion of the cognitive architecture. Although this strategy has been productive for neuropsychology, some syndromes defy explanation in this way, resulting in an unparsimonious architecture with many specialized components or connections. Optic aphasia is one such syndrome. Patients with optic aphasia have difficulty naming visually presented objects. However, the
more » ... eficit is not in visual recognition per se because patients can pantomime the appropriate use of objects, and the deficit is not in naming per se because they can name objects from auditory cues. To explain optic aphasia, past accounts have hypothesized multiple semantic systems or multiple functional pathways to visual naming. Farah (1990) instead sketched an account based on lesions to multiple pathways-one that maps visual input to semantics, and the other that maps semantics to naming responses. This account supposes that the effects of the two lesions are superadditive, meaning that a task requiring both damaged pathways (e.g., naming a visually presented object) would manifest a much higher error rate than expected based on the sum of error rates on two tasks requiring just one pathway or the other (e.g., gesturing the appropriate use of a visually presented object, and naming from auditory cues). We have explored this hypothesis by modeling superadditive effects of damage in a connectionist architecture. The resulting model explains other phenomena associated with optic aphasia, including the tendencies of patients to: produce a large number of naming errors that are semantically related to the target but few visually related errors, show response perseveration from one trial to the next, "home in" on the correct response over time, and make fewer errors on naming from a verbal description than on gesturing the use of an object from a visual presentation. More broadly, superadditive effects of damage provide a novel class of explanations for neuropsychological deficits that might previously have seemed to imply the existence of highly specialized processing components.
doi:10.1037//0033-295x.107.4.709 fatcat:rwqre7f7fzcydesnbcfduzqcq4