On the case for multiple semantic systems: A reply to shallice

Abstract

What is the nature of our stored knowledge about objects? When asked to describe a bee, a person might reply that bees are insects that live in hives, sting, produce honey, and engage in complex social interactions; they tend also to be small and to have black and yellow stripes. We can ask whether all this information is stored within a single undifferentiable system, whether it is stored in a single system which can fractionate into separable sub-domains, whether it is stored in systems which are separated according to the modality of the input they receive, etc., A primary aim of the paper by Riddoch, Humphreys, Coltheart, and Funnel1 (this volume) was to clarify these questions, and the theoretical positions adopted in various models. We also made a distinction between representations involved in the perceptual identification of stimuli presented in different modalities (e.g. representations specifying the structural characteristics of objects or printed words), and those specifying functional and associative knowledge about the stimuli. Information about, say, the size and visual form of bees might then be represented separately from that specifying that bees make honey. We proposed that the former kind of knowledge is represented in a structural description system-though it is quite feasible that this system might represent other visual attributes of objects, such as their colour, and not just their structure-whilst functional and associative knowledge would be represented in a separate, semantic system. Different tasks might draw on these different representations to contrasting extents. To judge that a seen object is familiar might require only access to structural knowledge. To name it might require access to both structural and semantic knowledge. To define an object, given its name, might require access either to structural, semantic or possibly both types of knowledge (perhaps depending on the object). Jackendoff, 1987).