1. A dichoptic viewing arrangement was used to study the initial vergence eye movements elicited by brief horizontal disparity steps applied to large textured patterns in three rhesus monkeys. Disconjugate steps (range, 0.2-10.9 degrees) were applied to the patterns at selected times (range, 13-303 ms) after 10 degrees leftward saccades into the center of the pattern. The horizontal and vertical positions of both eyes were recorded with the electromagnetic search coil technique. 2. Without training or reinforcement, disparity steps of suitable amplitude consistently elicited vergence responses at short latencies. For example, with 1.8 degrees crossed-disparity steps applied 26 ms after the centering saccade, the mean latency of onset of convergence for each of the three monkeys was 52.2 +/- 3.8 (SD) ms, 52.3 +/- 5.2 ms, and 53.4 +/- 4.1 ms. 3. Experiments in which the disparity step was confined to only one eye indicated that each eye was not simply tracking the apparent motion that is saw. For example, when crossed-disparity steps were confined to the right eye (which saw leftward steps), the result was (binocular) convergence in which the left eye moved to the right even though that eye had seen only a stationary scene. This movement of the left eye cannot have resulted from independent monocular tracking and indicates that the vergences here derived from the binocular misalignment of the two retinal images. 4. The initial vergence responses to crossed-disparity steps had the following main features. 1) They were always in the correct (i.e., convergent) direction over the full range of stimuli tested, the initial vergence acceleration increasing progressively with increases in disparity until reaching a peak with steps of 1.4-2.4 degrees and declining thereafter to a nonzero asymptote as steps exceeded 5-7 degrees. 2) They showed transient postsaccadic enhancement whereby steps applied in the immediate wake of a saccadic eye movement resulted in much higher initial vergence accelerations than the same steps applied some time later. The response decline in the wake of a saccade was roughly exponential with time constants of 67 +/- 5 (SD) ms, 35 +/- 2 ms, and 54 +/- 4 ms for the three animals. 3) That the postsaccadic enhancement might have resulted in part from the visual stimulation associated with the prior saccade was suggested by the finding that enhancement could also be observed when the disparity steps were applied in the wake of (conjugate) saccadelike shifts of the textured pattern. However, this visual enhancement did not reach a peak unit 17-37 ms after the end of the "simulated" saccade, and the peak enhancement averaged only 45% of that after a "real" saccade. 4) Qualitatively similar transient enhancements in the wake of real and simulated saccades have also been reported for initial ocular following responses elicited by conjugate drifts of the visual scene. We replicated the enhancement effects on ocular following to allow a direct comparison with the enhancement effects on disparity vergence using the same animals and visual stimulus patterns and, despite some clear quantitative differences, we suggest that the enhancement effects share a similar etiology. 5. Initial vergence responses to uncrossed-disparity steps had the following main features. 1) They were in the correct (i.e., divergent) direction only for very small steps (< 1.5-2.5 degrees), and then only when postsaccadic delays were small; when the magnitude of the steps was increased beyond these levels, responses declined to zero and thereafter reversed direction, eventually reaching a nonzero (convergent) asymptote similar to that seen with large crossed-disparity steps; convergent responses were also seen with larger vertical disparity steps, suggesting that they represent default responses to any disparity exceeding a few degrees. 2) As the postsaccadic delay was increased, responses to small steps (1.8 degrees) declined to zero and thereafter re