Abstract
Homonymous hemianopia, unilateral blindness caused by damage to early visual cortex, causes deficits in many every-day visually-guided activities, including driving, walking and reading. Previous work has reported abnormal fixation distribution in such individuals when in visual search tasks. However, little is known about the functional consequences of these abnormal fixation patterns in complex, naturalistic settings. We measured eye position while three homonymous hemianopes (HH) and four visually-intact controls performed a model-building task. Subjects were seated at a real table and asked to assemble 10 models consisting of four slats attached together with three bolts and nuts from a wooden construction set (Baufix - Mennie et al., 2003). Overall performance in hemianopes was remarkably similar to that of controls. Hemianopes were not impaired in their speed or ability to complete the task, nor were their fixations less accurate than controls. This is remarkable because many of the saccades were made into the blind visual field. Such saccades must be programmed on the basis of spatial memory, since visual input is absent. There were some subtle differences in performance: hemianopes exhibited significantly more look-ahead fixations (LAF) while also spending a greater proportion of time fixating task-irrelevant parts of the table than controls. State transition matrices and pathfinder associative networks found the sequence of eye movements in hemianopes less predictable than in controls. Since LAFs might provide spatial information for planning subsequent movements, this data is consistent with a heavier reliance on visual memory in programming accurate eye and hand movements . This might be associated with a greater need for frequent updating of the internal representation using LAFs.
The authors would like to thank Brian Sullivan and Kelly Chajka for help in data collection.