Tóth, D. & Csépe, V.
Institute of Cognitive Neuroscience and Psychology, Research Centre for Natural Sciences, Hungarian Academy of Sciences
Several models of reading development postulate that phonological skills, especially phonological awareness play an influential role in reading acquisition. This view is also supported by factor-analytic studies, in which the assumed phonological factor strongly (and often reciprocally) predicts the level of the reading factor. However, these models make the strong assumption that factors (latent cognitive skills) and specific subgroup of tasks have a close and exclusive relationship. This means that all covariances between the observed performance on distinct tasks (e.g. phoneme deletion, pseudoword reading, rapid naming of letters) can be explained by the covariance of distinct factors (e.g. phoneme awareness, phonological recoding, RAN). Based on the reanalysis of a large, international database (N>2000) of the 3DM reading battery, we suggest an alternative view: the common variance between all reading related tasks is directly attributable to a general (g) factor, and specific, independent factors account for additional covariances between specific subgroups of observed measures. (The model resembles the direct hierarchical model of general intelligence.) We argue that the g factor of reading reflects low-level cognitive abilities of strong biological origin, which determine how fast and effectively the subsystem of fast, automatic, phoneme-level access develops in the phonological system, enabling the mapping or recoding of phonological and orthographic information at an elementary level. However, g is not omnipotent: specific factors explain a remarkable portion of shared variance between reading fluency measures, between accuracy measures of various tasks, between speed measures of several tasks, etc. We reviewed various behavioural genetic studies, re-analysed published datasets of large-scale reading studies, and conducted an additional large-scale experiment; all of these provided strong, independent support for our model. The model is universal (in the sense that it is not language-specific) and general (in the sense that it relates to both typical and atypical reading development).