Smooth pursuit eye movements allow primates to fixate and track small, slowly moving objects. Pursuit usually requires visual targets; our aim was to determine the properties of the visual signals transmitted to the pursuit motor system. Rhesus monkeys were rewarded for tracking spots of light that underwent discreet changes in velocity under a variety of visual conditions. We measured the resulting smooth eye acceleration in a 100-msec interval that began with the initiation of pursuit and ended before there had been time for visual feedback. This approach allowed us to vary the parameters of visual stimulation and measure eye movement responses in a way that provides estimates of the properties of signals transmitted by visual pathways. The initiation of pursuit showed different properties early and late in the interval we studied. In the first 20 msec of pursuit, eye acceleration was in the correct direction, but was independent of the initial position of the moving images, the velocity of the stimulus, or the presence or absence of background illumination. Thereafter, the initiation of pursuit depended strongly on all of the above parameters. Eye acceleration was highest when the moving images fell close to the fovea and decreased sharply as eccentricity was increased up to 21 degrees. When the background was diffusely illuminated, eye acceleration showed velocity selectivity; it was highest for a middle range of velocities (30 to 60 degrees/sec) and decreased for higher velocities. When the background was dark, eye acceleration increased as a function of target velocity up to 150 degrees/sec. We conclude that the initiation of pursuit has at least 2 visual components. The two components have different latencies and show quite different relationships to the visual properties of the stimulus, suggesting two cell populations that could provide the visual inputs for pursuit.