This model, introduced by McClelland & Rumelhart (1981) is one of the earliest computational models of visual word recognition, and remains influential to this day. The model assumes three levels of representation: features, letters and words. Interactions between these levels take place in parallel. Perception is assumed to be an interactive process involving both bottom-up and top-down processing. Another important feature of the model is lateral inhibition among the nodes within a layer (in practice, this was only implemented for the word level).