The activity of subthalamic nucleus neurons related to motor performance was studied in three unrestrained cats operantly conditioned to perform a lever-release movement. The movement was initiated either rapidly after the trigger stimulus (a brief sound) in a simple reaction-time paradigm or after a delay in trials identified by a tone cue. These paradigms were randomly presented. The activity of 171 neurons was recorded in the contralateral and in the ipsilateral subthalamic nucleus, with respect to the performing limb. The mean spontaneous activity of cells in the ipsilateral side (18.5 +/- 13.8 imp/s, mean +/- SD) was higher than that in the contralateral side (8.5 +/- 8.1 imp/s). A total of 145 cells (85%) presented significant changes in activity in relation to the lever-release movement (task-related cells). The remaining 26 cells were either related to other events of the task (n = 15; lever-press or reinforcement occurrence) or not related at all to the task performance (n = 11). The majority of changes of activity of task-related cells were initial increases in discharge, which started on average, 127 ms before movement onset and lasted several hundreds of milliseconds. These increases in discharge were more frequent in the contralateral side (75 of 80 task-related cells, 94%) than in the ipsilateral side (43 of 65 task-related cells, 66%). The changes in activity, either increases or decreases, occurred early after the trigger stimulus, since 62% of them had a latency of less than 100 ms. Although the mean latency of initial increases was rather similar in both sides (97 ms contralateral versus 104 ms ipsilateral), the contralateral side was characterized by a high proportion of very early responses (less than 20 ms). For most neurons, the early changes in activity described above were absent after the trigger stimulus in the delayed condition. For certain neurons, the changes in activity prior to movement were different in reaction-time condition and in delayed condition, showing that the pattern of activity preceding movement might depend on the temporal requirements for motor initiation. The results suggest that a significant proportion of subthalamic cells are involved in the preparation and the initiation phases of the lever-release movement studied, although other hypotheses (e.g. stimulus-related responses) cannot be definitely ruled out. The timings and patterns of the changes in activity observed in the subthalamic nucleus in the present study, and in the pallidal complex previously, cannot be explained easily by the classical scheme where the external pallidum inhibits the subthalamic nucleus. The results suggest rather that the subthalamic nucleus, driven by a yet-to-be-determined excitatory input, exerts an excitatory influence on the pallidum and plays a crucial role in the control of the basal ganglia output neurons.