Computational modelling of phonological dyslexia: How does the DRC model fare?

Abstract

This paper investigates the patterns of reading impairment in phonological dyslexia using computational modelling with the dual-route cascaded model of reading (DRC, Coltheart, Rastle, Perry, Langdon, & Ziegler, 2001). Systematic lesioning of nonlexical and phonological processes in DRC demonstrates that different lesions and severity of those lesions can reproduce features of phonological dyslexia including impaired reading of nonwords, relatively spared reading of words, an advantage for reading pseudohomophones. Using the same stimuli for model and for patients, lesions to DRC were also used to simulate the reading accuracy shown by three individuals with acquired phonological dyslexia. No single lesion could replicate the reading performance of all three individuals. In order to simulate reading accuracy for one individual a phonological impairment was necessary (addition of noise to the phoneme units), and for the remaining two individuals an impairment to nonlexical reading procedures (increasing the time interval between each new letter being processed) was necessary. We argue that no single locus of impairment (neither phonological nor nonlexical) can account for the reading impairments of all individuals with phonological dyslexia. Instead, different individuals have different impairments (and combinations of impairments) that together provide the spectrum of patterns found in phonological dyslexia.