The effects of visual attention and fixation upon the distribution of saccadic latencies: express (E-), fast regular (FR-), and slow regular (SR-) saccades were investigated. Extinguishing a fixation or an attention point 200-300 ms before target onset increases the incidence of E-saccades while concurrently decreasing the proportion of SR-saccades. Since this extinction forces a disengaging of attention, these changes in relative proportions of saccades reflect the elimination of one of the steps involved in programming saccades. It is shown that a previously attended stimulus has a favored status relative to other stimuli in the visual field. If, after being turned off, the previously attended fixation point or a peripheral attention stimulus is turned on near the time of the target's appearance, the occurrence of the E-saccades is greatly reduced. However, the appearance of any other stimulus in the visual field at or near the time of the target onset does not inhibit E-saccades. Contrary to the conclusions reached by Posner and Cohen (1984), a stimulus presented at the formerly attended location can attract attention more efficiently than a stimulus presented at another, new location.