Two ways of subtyping in dyslexia will be examined, namely visual/auditory (or 'dyseidetic'/'dysphonetic') and phonological/surface. The conventional view is that phonological dyslexics have difficulty with non-words but are relatively stronger at real words, whereas the surface dyslexic can manage non-words but has difficulty with irregular words. When errors were classified into 'visual' and 'auditory', it was found in the case of 68 dyslexic children that after a 2-year period of tuition there was a change in the ratio of 'visual' errors to 'auditory' errors-the proportion of 'auditory' errors decreasing markedly. Also, when suitable regular words, non-words and irregular words were given to 58 dyslexic children and to 58 younger children matched for reading age, in neither group did the pattern of results give any significant support for classifying dyslexic children into 'phonological dyslexics' or 'surface dyslexics'. A more economical hypothesis, which explains both sets of results, is that the observed changes reflect the instruction which the children had received. The alternative hypothesis, namely that there is some kind of 'cognitive substrate' which remains unchanged whatever the kind of teaching, was not supported by the data.