The Journal of Neurophysiology Vol. 84 No. 6 December 2000, pp. 2725-2738
Copyright ©2000 by the American Physiological Society

Task-Dependent Modulation of the Sensorimotor Transformation for Smooth Pursuit Eye Movements

Department of Psychiatry and Center for Neurobiology and Behavior, Columbia University, New York, New York 10032

Ferrera, Vincent P. Task-Dependent Modulation of the Sensorimotor Transformation for Smooth Pursuit Eye Movements. J. Neurophysiol. 84: 2725-2738, 2000. To investigate the transformation of retinal image velocity into smooth pursuit eye velocity, eye movements were measured in the presence of two moving targets. In the first experiment, the targets were identical in all respects except for direction of motion, and the monkey was not cued to attend to either target. In this experiment, smooth pursuit eye velocity elicited by two targets was the vector average of the response evoked by each target alone. In subsequent experiments, we examined the effects of stimulus and task parameters on the selectivity of pursuit. When the targets were made different colors and monkeys were cued for the color of the rewarded target, their pursuit eye movements were biased in the direction of the rewarded target, but still showed a substantial influence of the nonrewarded target (distractor). It did not matter whether the same target color was used for an entire session or whether the color was randomized from trial to trial. Reducing uncertainty about the axis of motion of the rewarded target also had little effect. However, the pattern of image motion appeared to have a substantial effect; radial image motion favored averaging, and winner-take-all pursuit was found only with nonradial image motion. We conclude that the sensorimotor interface for pursuit uses a flexible decision rule that can vary continuously from vector averaging to winner-take-all. We present a simple recurrent network model that reflects this range of behavior. The model has allowed us to identify three computational elements (selection bias, competitive inhibition, and response normalization) that should be taken into consideration in future models of smooth pursuit.