Predictors of nonword reading and dyslexia/ 1 PREDICTORS OF NONWORD READING AND DYSLEXIA Predictors of Exception Word and Nonword Reading in Dyslexic Children: The Severity Hypothesis

The classification of dyslexic children into discrete subtypes yields a poor description of the dyslexic population at large. Multiple regression methods were used to examine continuous variation in component reading subskills (nonword and exception word reading) and their underlying cognitive skills within a group of 59 9-15 year-old dyslexic children. Two measures of phonological skills contributed unique variance to nonword reading: phonological processing and verbal short-term memory skills. In contrast, the only unique predictor of exception word reading was reading experience. The results are discussed within a connectionist framework that views the decoding deficit in dyslexia as stemming from poorly specified phonological representations. The extent of the nonword reading deficit is determined by the severity of the underlying phonological impairment. In contrast, exception word reading is influenced more by print exposure. Predictors of nonword reading and dyslexia/ 3 Predictors of Nonword Reading in Dyslexic Children. The strong developmental association between phonological skills and learning to read (Fowler, 1991; Goswami & Bryant, 1990; Share, 1995) provides a back-drop to the theory that the proximal cause of developmental dyslexia is a phonological processing deficit (Brady & Shankweiler, 1991; Morton & Frith, 1995; Snowling, 2000; Stanovich & Siegel, 1994). Within this view, literacy development is affected in dyslexic children who come to the task of reading with poorly specified phonological representations (Elbro, Borstrom & Petersen, 1998; Hulme & Snowling, 1992). However, cases of children with dyslexia who do not have a phonological processing impairment(Castles & Coltheart, 1996; Goulandris & Snowling, 1991; Hanley, Hastie & Kay, 1992) pose a problem for the phonological deficit hypothesis. In contrast to dyslexic children with poor nonword reading (phonological dyslexia; Campbell & Butterworth, 1985; Hulme & Snowling, 1992; Temple & Marshall, 1983), developmental surface dyslexic children place extensive reliance on phonological strategies for reading and spelling (Coltheart, Masterson, Byng, Prior & Riddoch, 1983). Individual differences in dyslexia have been conceptualised using dual-route (Castles & Coltheart, 1993) and connectionist (Seidenberg & McClelland, 1989) models of reading. According to the dual-route model, reading can be Predictors of nonword reading and dyslexia/ 4 accomplished either by a direct reading system involving mappings between printed words and their meanings, or by a phonological system incorporating grapheme-phoneme correspondences. The direct route is used for reading exception words and familiar regular words, whereas nonwords have to be read using the phonological reading system. In contrast, within connectionist models, exception word and nonword reading is accomplished using a single mechanism operating over distributed representations of orthographic and phonological units. Such models gradually abstract the statistical relationships between orthographic inputs and phonological outputs, allowing novel words to be read through generalisation of this knowledge (Seidenberg & McClelland, 1989). Generalisation within connectionist models depends upon the structure of orthographic and phonological representations (Plaut, McClelland, Seidenberg & Patterson,