The most prominent explanation for specific reading difficulties is the phonological deficit hypothesis; that reading difficulties are caused by difficulties in accessing the phonological, or sound structure of speech. This makes it difficult for children to 'sound out' words in reading and spelling, and difficult to link spoken and written words in memory.

On the basis of this theory, it might be expected that children with early speech difficulties would be likely to show later literacy difficulties. In fact, many children with speech difficulties show good reading progress, and only a subgroup have literacy difficulties. Researchers have suggested various hypotheses to explain this fact. Most prominently, researchers have argued that only children with both speech and language difficulties are likely to go on to have literacy difficulties, and that therefore language difficulties are a more important risk factor than speech difficulties. This seems counter-intuitive given that speech is, on the face of it, more closely related to phonological processing. On this basis it seems that we need to consider carefully what we are measuring with standard language assessments.

Other researchers have argued that only children who show unusual or disordered speech errors are likely to go on to have literacy difficulties, or that only children who continue to have speech difficulties at the point at which they begin literacy instruction will be impaired. All of these alternative hypotheses can be examined using this dataset.

Previous studies have focused on group differences, and have generally used clinical samples referred through speech therapists or specialist schools. These methods will be subject to referral bias, and may underestimate the role of individual differences - for example, children with both speech and language difficulties tend to have more severe speech impairments than children with pure speech problems. A recent study of my own (Carroll & Myers, in preparation) showed that while children with both speech and language difficulties were more likely to have literacy difficulties, as previously suggested, language skills were not a significant predictor of growth in reading over time. This suggests that the link between the two could be explained by another third variable.

The ALSPAC study provides a way to investigate this possibility. Extensive background information is available on the families and children involved. There are repeated language assessments of a representative sample of children throughout the preschool years, together with clinical assessments of a smaller group at 24 months, 61 months and 8 years. These could be used to predict reading and spelling at 7, 9 and 12 years, and phonological skills such as nonword repetition, phoneme deletion and spoonerisms at the same time points. While the sample is considerably reduced by using measures from the 'children in focus' assessments, it remains a large sample in terms of clinical studies of this type.

The analyses will focus on using structural equation modelling and if necessary multiple regression to illustrate the relationships between the variables over time. Some group difference analyses examining those children in the sample showing significant difficulties will also be included. The measures requested for this analysis are shown on the following page