A neural net model of normal and dyslexic spelling†

In this paper we describe a connectionist model of the development of alphabetic spelling, and show that its performance resembles that of dyslexic spellers if its computational resources are restricted during learning. We then describe the results of an experiment which tests the predictions of the model on normal and dyslexic spellers. Our results suggest that dyslexic children show a similar pattern of results to normal children of equivalent spelling age in the spelling of single words. Broad agreement is found between the behaviour of the simulation and the performance of human subjects. The fact that the neural network model provides a good characterisation of the spelling process in dyslexic children suggests that a dyslexic’s difficulty can be simply characterised as one of difficulty in mastering the mappings from sound to spelling in English, rather than in terms of qualitatively distinct processing.