ABSTRACT
Both theoretical and factual reasons make it difficult to choose among the “sensory/functional”, the “domains of knowledge”, and the “interconnections among semantic features” hypotheses. The first theoretical reason concerns the fact that there are different versions of each of these models and each version can be supported or falsified by a different set of data. Thus, there are at least three versions of the “sensory/functional” hypothesis. The first and most influential version is consistent with the distributed models of semantic representation proposed by Allport (1985), Shallice (1988), and Damasio (1990). These models assume that each object concept can be represented as an autoassociated pattern of activation distributed across different sensory and motor attribute domains, located in different cortical regions. A local lesion of this network should, therefore, preferentially disrupt semantic categories for which the damage falls at the centre of gravity of their pattern of interconnected attributes; nevertheless, other concepts having a similar pattern of autoassociated properties should also be concomitantly damaged. According to this version, category-specific defects for living things should be the result of a lesion impinging upon areas processing and storing visualperceptual information, whereas defects selectively concerning artefacts should result from lesions encroaching upon areas mainly dealing with functional knowledge.
