Abstract
Computation theory has been used extensively in recent decades to model cognitive behaviour. Of fundamental importance in this approach are the machines in which a functional distinction exists between program and memory. The processes within these machines can be described in terms of symbol manipulation. The distinction between program and memory can be implemented with neural networks that resemble structures found in the (visual) cortex. Networks of this kind can model selective attention, visual search and compositional representation (e.g. feature binding). The importance of a neural network implementation of symbol manipulation is briefly discussed.