Eye movement circuitry involved in saccade production offers a model for studying cognitive control: visually guided prosaccades are stimulus-directed responses, while goal-driven antisaccades rely upon more complex control processes to inhibit the prepotent tendency to look toward a cue, transform its spatial location, and generate a volitional saccade in the opposite direction. By manipulating the relative probability of these saccade types, we measured participants' behavioral responses to different levels of implicit trial-type probability and task-switching demands in conditions with relatively long inter-trial fixation and trial-type cue lengths. Results indicated that when prosaccades were less probable in a run, more prosaccade errors were generated; however, for antisaccades, trial-type probability had no effect on the percent of correct responses. For reaction times, specifically in runs with a larger probability of antisaccade trials, latencies increased for both anti- and pro-saccades. Furthermore, task switching resulted in a lower percentage of correct responses on switched trials, but a prior antisaccade trial led to slower reaction times for both trial types (i.e., a task switch cost for prosaccades and switch benefit for antisaccades). These findings indicate that cognitive control demands and residual inhibition from antisaccades alter performance relative to trial-type probability and task switching within a run, with the prosaccade task showing greater susceptibility to the influence of a large probability of cognitively complex antisaccades.