Express saccades are visually-guided saccades that are characterized by an extremely short latency of about 100 ms. The present experiments tested the hypothesis that a disengagement of visual attention is necessary for the generation of express saccades. All subjects produced large numbers of express saccades in the gap paradigm, in which the fixation stimulus is removed 200 ms before target onset (Exp. 1), but not in the overlap paradigm, in which the fixation stimulus remained on during the entire trial (Exp. 2). By means of peripheral cues (Exps. 3-5) and central cues (Exps. 6-7), visual attention was directed at the target location for the saccade before the actual appearance of the saccade target. In all experiments, the location cues facilitated rather than abolished express saccades. The generation of express saccades was facilitated even when the currently fixated visual stimulus was not removed before target onset (fixation-overlap; Exps. 5-7). The results are explained by the hypothesis that a disengagement of a separate fixation system is necessary for the generation of express saccades, a hypothesis that is in line with current neurobiological findings.