1. We have identified a visually driven output from the flocculus of the monkey by studying the simple-spike responses of Purkinje cells (P-cells) during the initiation of smooth-pursuit eye movements. We report on two groups of P-cells that appear to be the horizontal and vertical gaze-velocity P-cells (GVP-cells) studied previously during periodic target and head motion. 2. During pursuit of periodic target motion, one group of P-cells prefers downward motion (down GVP-cells), and the other prefers motion toward the side of recording (ipsi GVP-cells). The two groups have mean directional preferences that are nearly orthogonal, but their responses during pursuit of sinusoidal target motion and sinusoidal vestibular stimulation are in other respects quantitatively similar. 3. During the initiation of pursuit to step-ramp target motion, GVP-cells show a large transient change in simple-spike firing rate followed by a sustained change in firing that persists during steady-state pursuit. 4. The transient response is directionally selective, so that GVP-cells show a pulse of simple spikes for pursuit in the ON-direction and a dip in simple-spike firing for pursuit in the OFF-direction. The amplitude of the transient response is too large to be explained by the sensitivity of GVP-cells to eye velocity measured during pursuit of sinusoidal target motion. 5. To test whether the transient change in simple-spike firing was related to a visual input or to an eye-acceleration input to the flocculus, we recorded the firing of ipsi GVP-cells during a rapid eye acceleration caused by a transient vestibular stimulus in darkness. Most GVP-cells showed little or no transient response under these conditions, even though eye acceleration was greater than during the initiation of pursuit. We conclude that the transient response at the initiation of pursuit is probably caused by visual mossy-fiber inputs to the flocculus. 6. The sustained change in simple-spike firing is also directionally selective, with large increases in simple-spike firing for pursuit in the ON-direction and smaller decreases for pursuit in the OFF-direction. For pursuit in the ON-direction, the amplitude of the sustained response is well predicted by the sensitivity of GVP-cells to eye velocity measured during pursuit of sinusoidal target motion. 7. To determine whether the sustained response was driven by visual inputs, we recorded simple-spike firing when image motion was prevented by electronically stabilizing the target image on the fovea during steady-state pursuit.(ABSTRACT TRUNCATED AT 400 WORDS)