The ability to make predictive saccadic eye movements is dependent on neural signals that anticipate the onset of a visual target. We used a novel paradigm-based on the saccade-countermanding task-as a tool to investigate rhythm saccade pacing and to provide information on the mechanisms of predictive timing. In particular, we examined the ability of normal subjects to stop a sequence of periodically paced eye movements when cued by a stop signal that was presented at different times with respect to the last target of the sequence (stop signal delay, SSD). The timing of the stop signal affected the ability to stop the saccadic sequence (make a saccade to a central target rather than to the peripheral alternating targets) in different ways, depending on the preceding tracking behavior. For the same SSD, subjects cancelled fewer trials during predictive tracking (promoted by tracking targets alternating at a fast pacing rate, 1.0 Hz) than during reactive tracking (tracking alternating targets at a low pacing rate, 0.2 Hz). In addition, on non-canceled trials, there was an increase in the delay of the corrective saccade to the central target with increasing SSD for pacing at 0.2 Hz, but the timing of the corrective saccade remained near constant for 1.0 Hz pacing. In examining the timing between movements, we estimate that the repetitive GO process that drives the saccades during predictive tracking begins earlier and has a shorter duration than the repetitive GO process during reactive tracking. These behavioral results provide further insight into the initiation process of predictive responses. In particular, the reduced reaction time and the corresponding short duration of the predictive process may result from a faster accumulation of neuronal discharge to a relatively fixed threshold.