Chang, Y. 1 , Furber, S. 2 & Welbourne, S. 1
1 Neuroscience and Aphasia Research Unit (NARU), University of Manchester
2 Advanced Processor Technologies (APT) Group, University of Manchester
During reading, visual information from print is transformed into sound and meaning. This transformation process involves a series of processing stages including mappings between visual, orthographic, phonological and semantic representations. Most existing computational models of single word reading have been focused on only the last three (orthographic, phonological and semantic) stages, with little attention devoted to the role of visual processing. To rectify this, we incorporated a visual processing stage into several parallel models of reading. Specifically, we investigated to what extent visual processing is involved in the "serial" phenomena observed in normal reading (e.g., the length effect and the position of irregularity effect). These effects were previously thought by many (e.g., Coltheart et al., 2001) to imply the existence of a serial processing mechanism operating alongside the parallel processing mechanism that subserves human reading. However, the simulations reported here show that purely parallel models, which include a visual processing stage, can account for the length by lexicality effect found in normal reading (Weekes, 1997). Further analyses revealed that the length effect arises as an emergent property of mappings between length-dependent visual representations, centred around an optimal viewing position, and length-independent phonological representations. A critical factor is that the model must be allowed freely to develop its own internal orthographic representations. In addition to simulating length effects, the model could also simulate the position of irregularity effect (Rastle& Coltheart 1999; Roberts et al. 2003) and the "whammy effect" (Rastle& Coltheart, 1998). Overall these findings suggest that understanding the nature of the mapping between visual forms and phonological representations is vital to a proper explanation of these controversial ?serial? phenomena.