Two experiments were conducted to determine the spatial and temporal organization of the arm trajectory in human subjects as they pointed to single- and double-step target displacements. Subjects pointed either without (Experiment 1) or with (Experiment 2) vision of their moving hand throughout the trial. In both experiments, target perturbation occurring in double-step trials was clearly perceived by the subjects and was randomly introduced either at the onset or at peak velocity of hand movement. Regardless of whether or not visual reafference from the pointing hand was available, subjects corrected the trajectory of their moving hand to accommodate the double-step. Moreover, asymmetrical velocity profiles were observed for responses to both types of target, with or without vision of the moving hand. The acceleration phase was a fixed pattern independent of the type of step stimulation. However, a clear dissociation, both in the deceleration phase and accuracy of responses to double-step targets, emerged according to the timing of target perturbation. When targets were perturbed at the onset of hand movement, subjects modulated the deceleration phase of their response to compensate for 88 to 100% of the second target displacement. In contrast, when targets were perturbed at peak velocity of hand movement, subjects were unable to modulate the deceleration phase adequately and compensated for only 20 to 40% of the perturbation. These results suggest that motor error is dynamically evaluated during the acceleration phase of a movement toward a perturbed target, allowing amendments to the trajectory to be performed during the deceleration phase. This main corrective process appears to be basically independent of visual reafference from the moving hand.