Different responses to small visual errors during initiation and maintenance of smooth-pursuit eye movements in...

1. We have investigated the role of retinal and extraretinal signals in the initiation and maintenance of smooth-pursuit eye movements in trained rhesus monkeys. Visual targets were presented in open-loop conditions by using electronic feedback of eye position to form the command for target position. This allowed us to present stimuli that were stabilized with respect to the moving eye or to provide small, precisely controlled retinal position or velocity errors. 2. Pursuit was maintained with only small decreases in eye velocity if retinal errors were eliminated by stabilizing the tracking target in front of the fovea during pursuit at 15 degrees/s. This argues that the pursuit system employs "velocity memory" to maintain pursuit. We suggest that velocity memory is effected by an extraretinal signal derived from positive feedback of eye-velocity commands. 3. Small retinal position errors caused smooth eye accelerations if imposed during pursuit, but were ineffective for initiating the transition from steady fixation to pursuit. Small retinal velocity errors were effective both for initiating pursuit from steady fixation and for altering eye velocity during pursuit. 4. Retinal position errors were effective at changing smooth eye velocity in a variety of conditions that required prior activation of the pursuit system. These include pursuit with or without a stationary background, pursuit with a background that was stabilized with respect to the eye, pursuit with combined eye and head motion (cancellation of the vestibuloocular reflex), and use of pursuit to suppress optokinetic nystagmus. Position errors were ineffective during fixation of stationary targets, even if head motion was provided to evoke the smooth eye velocity of the vestibuloocular reflex. 5. We conclude that retinal position errors are effective only after the pursuit system has been activated. It follows that pursuit initiation involves an active transition from steady fixation and that this transition is normally triggered by retinal velocity errors but not by retinal position errors.