Event-related potentials were recorded in a reaction time (RT) paradigm, where the duration of a learned interval (either 0.7 s or 2.5 s) delimited by two brief button-presses was to be accurately controlled. A preparatory signal (PS) either did not give or gave prior information concerning the duration of the following response (neutral condition or primed conditions, respectively). In the latter case, the information was either validated (valid condition) or invalidated (invalid condition) by the response signal (RS). When duration was not known in advance (invalid and neutral conditions), RTs were longer before a response of short than long duration. This difference was not found under the valid condition. During the preparatory period (PP), the amplitude of the contingent negative variation (CNV) was larger when the duration was primed than when it was not. A larger CNV appeared when the PS primed a short rather than a long duration. This effect occurred in the early part of the PP over the supplementary motor area (SMA) and in its latest part over the primary motor area (MI). The RT and the electrophysiological pattern were interpreted as revealing the occurrence of programming operations regarding the temporal dimension of the response. The time course of the CNV over the SMA and MI suggested that these two areas were hierarchically organized. Between the RS and the onset of the response, differences probably related to programming effects were still found over MI: the activities were larger under the valid than under the neutral condition. However, no sign of deprogramming (expected in the invalid condition) was observed: similar amplitudes were found under the neutral and invalid conditions. Deprogramming operations seemed to be postponed during response execution where the invalid condition evoked larger activities than the two other conditions over the SMA. Finally, MI but not the SMA yielded a Bereitschaftpotential before the second press ending the response (i.e., during response execution). These results suggest that the duration of a motor response can be a part of the motor program and that the SMA plays a major role in programming processes but not in response execution, contrary to MI.