A computational model of neglect dyslexia.

This paper presents a straightforward theory of neglect (a pre-attentive salience system selects objects' centers for the focus of an attentional "spotlight") as a computational model. This construction permits simulations of model "lesions" and allows checking unequivocally the model's implications. The current model can account for some of the common patterns of observations of neglect dyslexia: that errors increase with word length and that short words may be read as "too long". Most importantly, it also accounts for a dissociation between neglect dyslexia and neglect on line bisection tasks. A two stage model can account for the patterns of performance seen in many neglect subjects, but more important is the demonstration that a theory's implementation as a computational model permits unambiguous evaluation of the theory's implications.