A Distributed Model of the Representational States in Classical Conditioning

Abstract A distributed neural network model is presented which is designed to replicate behavioural data from classical conditioning experiments. Other models in this field have tended to be feedforward systems with localised representations which seem inimical to the distributed, feedback approach from which more “cognitive-like” properties emerge. Classical conditioning is treated here as the learning of biologically relevant sequences, and achieves this through the use of feedback, and through the inclusion of an input evaluative function which instigates the learning process. This model may be regarded as a pointer in the search for structural correlates, and as an extension of attentional theories of classical conditioning.