The End of the Line for a Brain-Damaged Model of Unilateral Neglect

Abstract

For more than a century, it has been known that damage to the right hemisphere of the brain can cause patients to be unaware of the contralesional side of space. This condition, known as unilateral neglect, represents a collection of clinically related spatial disorders characterized by the failure in free vision to respond, explore, or orient to stimuli predominantly located on the side of space opposite the damaged hemisphere. Recent studies using the simple task of line bisection, a conventional diagnostic test, have proven surprisingly revealing with respect to the spatial and attentional impairments involved in neglect. In line bisection, the patient is asked to mark the midpoint of a thin horizontal lie on a sheet of paper. Neglect patients generally transect far to the right of the center. Extensive studies of line bisection have been conducted, manipulating-among other factors-line length, orientation, and position. We have simulated the pattern of results using an existing computational model of visual perception and selective attention called MORSEL (Mozer, 1991). MORSEL has already been used to model data in a related disorder, neglect dyslexia (Mozer & Behrmann, 1990). In this earlier work, MORSEL was “lesioned” in accordance with the damage we suppose to have occurred in the brains of neglect patients. The same model and lesion can simulate the detailed pattern of performance on line bisection, including the following observations: (1) no consistent across-subject bias is found in normals; (2) transection displacements are proportional to line length in neglect patients; (3) variability of displacements is proportional to line length, in both normals and patients; (4) position of the lines with respect to the body or the page on which they are drawn has little effect; and (5) for lines drawn at different orientations, displacements are proportional to the cosine of the orientation angle. MORSEL fails to account for one observation: across patients, the variability of displacements for a particular line length is roughly proportional to mean displacement. Nonetheless, the overall fit of the model is sufficiently good that we believe MORSEL can be used as a diagnostic tool to characterize the specific nature of a patient's deficit, and thereby has potential down the line in therapy.