The Connectionist Dual Processing Model of Reading and Visuo-Spatial Attention: The Dissociation of Two Visual Processing Factors on the Model’s Underlying Mechanisms

Abstract

In recent years there has been a shift towards more sophisticated computational models in typical reading. One such model is the Connectionist Dual Process (CDP++) model of reading aloud. There are arguments about the extent to which and how visuo-spatial attention is used when reading. Whilst many factors have been shown to correlate with reading performance, this thesis restricts its focus to two visual processing factors: Visual Attention Span (VAS) and focused visuo-spatial attention. Overall, we are interested in whether two types of attention can be dissociated on the lexical and sublexical pathways of the CDP++ model. This was tested through three tasks: a simple 250 word reading task, a VAS task, and a Posner task. Quantitative statistical methods were used to determine the relationships between the tasks. Whilst not all of our predictions were confirmed we did find a number of useful insights into reading. A wide attentional window allows pronunciations to be generated correctly by the lexical route. It was found that the Posner task may suffer from validity problems in terms of it measuring the desired attentional process. In the future, tasks likely to measure attention in reading with higher validity may help us better understand the dissociation that has been reported in the literature. Elucidating this is important because the way factors affect how reading works may be used to help understand dyslexia, as well as predict outcomes from training programs.