Two experiments in which the effect of average movement velocity on reaction time is examined in relation to spatial accuracy are reported. Experiment 1 tested the hypothesis that fast inaccurate movements are more easily accessed during response selection by varying the compatibility of the stimulus-response relation. The second experiment employed spatially accurate movements in combination with the task variables S-R compatibility and foreperiod duration. The results are consistent with a two-stage motor-preparation notion consisting of a motor-programming stage and a program-loading stage.