ABSTRACT
Connectionism consists of old ideas about representation, spreading activation, semantic decomposition, and associative memory all packaged in something that looks like a cartoon brain with cartoon neurons, connected by cartoon synapses. The grand claims of the advocates of connectionism must also be tempered by the realisation that connectionist modelling, so far, has been confined to “toy” problems, using highly tailored processing constraints. The most basic idea of self-organising systems is that of multiple, interacting forces, which can “settle into” a “preferred state”, known as an attractor. The crucial role of a uniquely structured language for mediating joint-attention and joint-action rapidly and smoothly has been described by a number of psychologists. A “vertical” nesting of hyperstructures may form the basis of the well-known distinctiveness, arising in the course of development, between “basic”, “subordinate” and “superordinate” levels of representation — often described as taxonomic structures.
