1. To evaluate the various hypotheses that the basal ganglia preferentially control one mode of movement to the exclusion of others, we recorded the discharge of single neurons in the globus pallidus (GP) in rhesus monkeys during their performance of five trained wrist-movement tasks. The tasks were designed to dissociate several modes and parameters of movement to see whether pallidal neurons would discharge in relation to one and not the others. All tasks were performed by flexing and extending the wrist with opposing or assisting torque loads (0.2 Nm). The five tasks included 1) VisStep, a visually cued step tracking task; 2) VisRamp, a visually guided hold-ramp-hold tracking task; 3) VisSine, a visually guided rapid sinusoidal tracking task; 4) SelfRamp, a self-paced hold-ramp-hold task with delayed alternation, trained velocity, and no visual feedback of wrist position; and 5) SelfSine, a self-paced rapid sinusoidal movement without visual feedback of wrist position. Wrist position and velocity were monitored during all recordings; and wrist, arm, shoulder, and back electromyographs (EMGs) were monitored periodically. Unit discharge was recorded extracellularly from both segments of the GP. The results were similar in the present analysis and are considered together. As a control, units were also recorded in the dentate nucleus of the lateral cerebellum, and the EMGs of many muscles were recorded in limbs, neck and trunk. 2. For 100 GP neurons [41 in the internal segment (GPi) and 59 in the external segment (GPe)], the activity of which changed with task performance, the discharge patterns varied greatly across tasks. The discharge of 96/97 neurons (99%) changed during VisStep, 66/91 neurons (73%) changed during VisRamp, 41/81 neurons (51%) changed during VisSine, 7/34 neurons (21%) changed during SelfRamp, and 25/80 neurons (31%) changed during SelfSine. Of 74 neurons that were fully tested in four or more tasks, 16 (21%) were related only to one task; only 17 cells (23%) were related to all tasks; and, for the remaining 41 (55%) neurons, the relation of the discharge of a given neuron to one task did not predict its relation to other tasks. These task-dependent differences in the discharge of pallidal neurons were not correlated with differences in wrist position, velocity, load, or muscle activity (see also the following paper--Mink and Thach, 1991a). 3. From these data, we conclude that no one task engaged all pallidal neurons to the exclusion of other tasks.(ABSTRACT TRUNCATED AT 400 WORDS)