Let’s start by taking a look at a classic computational model of visual word recognition – the interactive activation (IA) model introduced by McClelland and Rumelhart in 1981. To load the model choose Load model from the main menu:
select the file titled ia.eNm. After a few seconds you will see the architecture of the model displayed on the right-hand side of the screen:
The IA model is sometimes referred to as a simple model, but this is misleading. It’s true that the architecture of the model is simple – it consists of just three layers, corresponding to letter features, letters and words. Furthermore, this is a localist model, and so each node within the model corresponds to a meaningful entity (i.e., a feature, letter or word). Nevertheless, the model can give rise to complex behaviour. This complexity arises from the nonlinearity of the model’s activation equation and the feedback within and between layers.