The current study investigated whether capture of the eyes by a salient onset distractor and the disengagement of the eyes from that distractor are driven by the same or by different underlying control modes. A variant of the classic oculomotor capture task was used. Observers had to make a saccade to the only gray circle among red background circles. On some trials, a green (novel color), red (placeholder color) or gray (target color) distractor square was presented with sudden onset. Results showed that when participants reacted fast, oculomotor capture was primarily driven by bottom-up pop-out: both types of distractors (green and gray) that popped out among the red background elements showed more capture than a red distractor that did not pop-out. In contrast to initial capture, disengagement of the eyes from the distractor was driven by top-down target-distractor similarity effects. We also examined the time-course of this effect. The distractor could change from green to either the target or placeholder color. When the color change was early in time (30-40 ms after its onset), dwell times were strongly affected by the change, whereas the effect on oculomotor capture was weak. Importantly, a change occurring as early as 60-80 ms after distractor onset did neither affect capture nor dwell times, corroborating the assumption of parallel programming of saccades. © 2010 The Author(s).