Extended encoding/decoding of embedded structures using connectionist networks

Central to both the symbolist and connectionist approaches to modelling cognition, is the focus of attention on (at least) two important issues; mental representations and mental processes. The choice of representation constrains the choice of process that can be applied to them. In the, so called, classical view on the cognitive architecture, the representations are characterised by the possession of a systematicity of representation. This allows the mental processes to enjoy a systematicity of inference. Connectionists on the other hand, have argued for a spatial basis of systematicity of representation and inference. The intention in this paper is to explore what the representational capacity of a commonly used generator for spatial systematicity is, and if it can be extended.