Three experiments explored whether each saccade of a sequence of saccades is programmed prior to the onset of a sequence of saccades. Using a choice reaction paradigm that has been used to study the programming of sequences of manual responses, subjects were required to perform either one, two, or three saccades under blocked (experiment 1) and randomized (experiment 2) sequence length conditions. Replicating manual performance (Rosenbaum, Saltzman and Kingman 1984a) the time to initiate a sequence of saccades (T1) increased with the number of forthcoming responses in experiment 1 but not in experiment 2. Experiment 3 was analogous to Rosenbaum, Inhoff and Gordon's (1984b) manual choice reaction task in that subjects were instructed to choose between two sequences of saccades that shared either the sequence initial or sequence terminal fixation position. In contrast to manual choice performance, which shows shorter keypress responses for shared response elements, saccadic latencies were unaffected by the sharing of fixation positions. The results are discussed in the framework of exogenous and endogenous motor control systems that are interfaced with structures controlling oculomotor response execution.