ABSTRACT
Given that correlated features play a key role in semantic processing, and that their role differs quantitatively for living versus non-living things, it has been hypothesised that they are important for explaining category-specific deficits. For example, they were a key part of Devlin et al.’s (1998) account of category-specific deficits in patients with Alzheimer’s disease (AD). Category-specific deficits are most often explained in terms of differential damage to a specific type of feature: perceptual in the case of living things, functional in the case of non-living things (Farah & McClelland, 1991). Because perceptual and functional features are assumed to be stored in separate areas of the brain (temporolimbic and frontoparietal, respectively), lesions localised to either region may preferentially damage one type of information, resulting in a category-specific impairment. Localised damage is compatible with some forms of neuropathology, such as herpes encephalitis. However, category-specific deficits have also been observed in patients with AD, a disease that causes more widespread, patchy damage in areas that typically include both temporolimbic and frontoparietal regions (Garrard, Patterson, Watson, & Hodges, 1998; Gonnerman et al., 1997; Silveri & Gainotti, 1988). Therefore, it is difficult to imagine how widespread damage could result in category-specific deficits without implicating other factors. Devlin et al. (1998) provided an account of category-specific impairments in AD using a model similar to that of McRae et al. (1997). To explain the simulation results, Devlin et al. appealed to differences in the density of correlated features for living versus non-living things, and the patterns of distinguishing features in the two domains.
