It can be seen that a number of proposals have been made with regard to the organizing principles of semantic memory. Various authors have suggested that our knowledge is organized by input modality, type of information or taxonomic category. However, where more than one of these organizing principles have been incorporated into models of semantic processing, the relationship between the principles has not always been made clear. In this section, an attempt will be made to summarize the way in which these principles are proposed to relate to one another within the various models of semantic memory. However, it should be noted that in some cases the necessary information has not been supplied.
First, there are those models which propose that there is only a single semantic memory system. Such models include the Cascade model, the Identification Procedures theory and OUCH.
According to the Cascade model, the same semantic system is accessed regardless of input modality. Information about the visual form of objects is represented within structural descriptions, while functional/associative semantic information is contained in the amodal semantic store. Category-specific deficits can arise from impairments at a number of different levels of the Cascade system. However, it is argued that when category-specific
deficits are the result of damage to stored semantic representations, the specific categories affected will be determined by the type of information impaired. Hence there may be a degree of organization by information type (e.g. into sensory and functional knowledge) within the unitary semantic system.
Within the Identification Procedures theory, input modality is an important factor in that pictures are considered to have privileged access to a subset of semantic information (i.e. that which is contained in identification procedures). Information type is also, to some extent, a relevant factor, in that the information which is necessary for the identification of visual instances of concepts is stored in identification procedures, while other types of information are stored in the amodal system. The authors of this theory do not consider how category-specific deficits may be accounted for by this hypothesis.
According to OUCH, pictorial stimuli have privileged access to certain semantic predicates in a unitary semantic system, because individual perceptual features of objects may be semantically interpreted. Hence input modality does have some influence, but semantic memory itself is not organized in terms of this factor. Category-specific deficits are suggested to be the result of damage to related semantic predicates within the unitary semantic system. However, these predicates are not reducible to a functional/sensory distinction, and correspond quite closely to the boundaries of taxonomic categories.
The relationship between the various organizing principles in multiple semantics models is less clear. For example, the relationship between organization by input modality and organization by information type has been understood in a number of different ways by different authors. The formulations proposed by the main exponents of this hypothesis will be discussed here.
McCarthy and Warrington (1988) have argued that visual and verbal stimuli initially access functionally separable semantic subsystems, and that within each o f these modality-specific subsystems living and non-living things are represented independently (perhaps in terms of sensory and functional attributes respectively). This interpretation of their model is
supported by their statement that "the differential salience of information from different sensory and motor channels in the acquisition of knowledge is at the basis of the categorical organization of meaning" and that "this basic categorical organization of meaning provides a blueprint, which is to some extent duplicated in functionally independent modality-specific meaning systems" (pp. 429-430).
As argued above, Shallice's (1988a) description of the multiple semantics hypothesis corresponds in many ways to the Modality-Specific Content hypothesis. Within the latter hypothesis, both input modality and information type are important factors. Information type is relevant in that the visual and verbal semantic subsystems are thought to contain visual and non-visual information respectively, while input modality is relevant because each of these modality-specific systems is considered to be directly accessible only from modality- congruent input. Warrington and Shallice (1984) have argued that category-specific disorders may be an emergent property of a system that is organized by information type (sensory or functional). However, within a simple version of the Modality-Specific Content hypothesis, the distinction between sensory and functional information appears to be synonymous with that between the visual and verbal semantic subsystems.
Within Shallice's later description of the multiple semantics hypothesis in terms of a distributed network (1988b; 1993), regions of the net are considered to be specialized for difierent types of process (including the representation of sensory and functional features). Shallice had argued that different input modalities could have stronger and more direct links with some of these processes than others. In addition, he has suggested that category- specific deficits could arise from damage to sub-regions of the specialized processing areas. However, he has not stated whether organization by input modality or organization by type of process would be primary.