The ocular following responses elicited by brief unexpected movements of the visual scene were studied in 10 rhesus monkeys. Test patterns were either random dots or sine-wave gratings [spatial frequency (Fs) 0.046-1.06 cycles per degree (c/degree)]. Test stimuli were velocity steps [speed (V) 5-400 degrees/s] of 100-ms duration, applied 50 ms after spontaneous saccades to avoid saccadic intrusions. Eye velocity response profiles were nonmonotonic and idiosyncratic, but consistent and closely time-locked to stimulus onset. Two measures of response amplitude were used: initial peak in eye velocity (ei), and average final eye velocity over the period of 110-140 ms measured from stimulus onset (ef). Using random dot patterns, response latencies were short, e.g., when the criterion for onset was an eye acceleration of 100 degrees/s2, mean latency (+/- SE) for eight monkeys with a 40 degrees/s test ramp was 51.5 +/- 0.6 ms. Using gratings of low spatial frequency (Fs less than 0.5 c/degree), latency was inversely related to, and solely a function of, contrast and temporal frequency, Ft (where Ft = V X Fs). We conclude from the latter that ocular following is triggered by local changes in luminance, and propose a model of the detection mechanism that reproduces all the essential features of these data. Moderate low-pass spatial filtering ("blurring") of the random dot pattern, by interposing a sheet of ground glass between the animal and the scene, progressively increased the response latency and decreased ef, but ei was either little affected or increased. When used with gratings, the ground glass simply reduced the contrast (range: 0.5-0.003), with very similar consequences for ocular following: latency increased and ef decreased, but ei changed little over the first decade of contrast reduction, increased over the second, and began to show attenuation (often pronounced) only at the lowest contrast. We suggest that these anomalous increases in ei with reductions in contrast are secondary to the delay in response onset and might be explained if the motion detectors responsible for triggering ocular following act as a gate for integrated retinal slip inputs to the tracking system proper: the delay in detection causes a buildup in the error signal driving the tracking response. En masse movement of the visual field was not the optimal stimulus for ocular following.(ABSTRACT TRUNCATED AT 400 WORDS)