Abstract:
Semantic memory can be selectively impaired following various types of brain damage. For example, cases have been reported where knowledge of living things (animals, fruit, vegetables) is impaired, while knowledge of nonliving things (utensils, vehicles, tools, etc) is relatively spared and vice versa. One theory to account for these apparently categorical losses of semantic knowledge is the sensory-functional theory (SFT), in which it is proposed that knowledge of living things depends to a greater extent on sensory/visual attributes and knowledge of nonliving things depends to a greater extent on functional attributes. Knowledge of visual and functional attributes is thought to be represented across modality-specific regions, and damage to regions representing visual or functional attributes can cause categorical deficits for living and nonliving things respectively. In Study 1, a stimulus set was designed to investigate the SFT in which attribute (visual or functional) was crossed orthogonally with category (living or nonliving). Twenty participants defined 120 living and nonliving items and the relative proportion of visual to functional attributes given for item definitions was calculated. A 68-item visual-functional (VF) set was selected, comprising 34 living (17 visual, 17 functional) and 34 nonliving (17 visual, 17 functional) items, matched on word frequency, familiarity and complexity. In Study 2, the VF set was used in a speeded picture-naming task with 14 left temporal lobectomy (TL), 17 right TL and 19 control participants. The left TL group made significantly more errors naming functional compared to visual items while there was no difference in accuracy for right TL and control groups. There was no effect of category (living or nonliving) for any group. These results support the SFT because attribute type had a significant effect on accuracy, but category did not. At a neural level, results indicated that functional representations may be iateralized to the left hemisphere. In Study 3, a single case (BML), with apparently categorical semantic impairments for living things following a left inferior temporooccipital stroke, was tested with the VF set. BML's performance on 3 out of 4 tasks (picturenaming, naming-to-general-definitions and an item definition task) was significantly poorer when items were visual, compared to functional, while category had no effect. In addition, BML's performance on 4 tasks assessing knowledge of musical instruments was similar to his performance on living things but significantly poorer than his performance on other nonliving items. Thus the prediction that deficits of living things reflect an underlying impairment in visual knowledge was supported. Finally the location of BML's lesion was consistent with the claim that visual representations are distributed across modality-specific cortex.