Three experiments examined saccade programming during short duration fixations between 50 ms and 150 ms. In experiment 1, subjects copy typed text, in experiment 2, subjects read and executed a letter detection task, and in experiment 3, subjects read for comprehension only. Fixation duration had no effect on the size of the departing saccade in the copy typing task; however, saccades leaving short duration fixations were larger than saccades leaving all other fixation durations in the letter detection task and smaller than saccades leaving long fixation durations in the standard reading task. Within Morrison's (1984) model, these results imply, first, that consecutive shifts of attention during a fixation can take different directions and, second, that successive shifts of attention during a fixation support different purposes. Within Fischer's (1986, in press) model, the results imply that the engagement/disengagement of attention and saccade programming do not constitute independent events.