A Semantic Subsystem of Visual Attributes

Abstract

We propose that the many different forms of selective semantic impairment that have been reported over the past 20 years may be classified into three general classes: semantic-category selective, modality-of-input selective, and semantic-attribute selective. Particular patients may exhibit more than one form of selectivity, i.e. there can be doubly and perhaps even triply selective semantic impairments. We then describe a patient with a singly selective semantic impairment of a form not previously described: he was unable to access visual semantic attributes in semantic memory, whereas he could access semantic attributes relevant to other sensory modalities, and could also access non-perceptual semantic attributes. This pattern of results was independent both of modality of input and of semantic category of probed item. We infer from these data the existence of a semantic subsystem specific to the storage of information about visual attributes of animate and inanimate objects. An ERP study of semantic processing in normal subjects provided further evidence in support of this claim about a particular component of semantic memory. We conclude by proposing that semantic memory is organized into subsystems of perceptual-attribute knowledge, one subsystem for each of the different perceptual modalities, plus a subsystem in which non-perceptual knowledge is represented. That subsystem is in turn categorically organized into semantic categories such as ‘animate’.