1. The activity of globus pallidus (GP) neurons (n = 1,117) was studied in two monkeys to reexamine the relation of neuronal activity to movement type (slow vs. fast) while they performed both a visually guided step and ramp wrist tracking task. To select neurons specifically related to wrist movements, we employed both a somatosensory examination of individual body parts and a statistical analysis of the strength of temporal coupling of neuronal discharges to active wrist movement. 2. Neuronal responses to somatosensory stimulation were studied in 1,000 high-frequency GP neurons, of which 686 exhibited clear responses to manipulation of body parts. Of the latter, 336 responded to passive manipulation of forelimb joints and 58 selectively to passive flexion or extension of the wrist. 3. In the external segment of GP (GPe), most neurons responding to passive wrist movement were found to be clustered in four to five adjacent, closely positioned (separated by 200 microns) tracks in single coronal planes. The clusters were irregular in shape with a maximal width of 800-1,000 microns. Separate clusters of neurons responsive to passive wrist movement were identified in planes 3 mm apart in one monkey and in planes 500 microns apart in the other. Multiple clusters of neurons were also found for neurons responsive to joints other than the wrist. These findings suggest a more discrete and complex representation of individual joints in the primate GP than previously conceived. 4. During the performance of the wrist flexion and extension task, 92 neurons showed clear and consistent changes in activity. For these neurons we measured, with a statistical method on a trial-by-trial basis, the strength of temporal coupling between the onset of active wrist movement and the onset of change in neuronal discharge rate. Fifteen neurons showed changes in activity time-locked to the onset of active wrist movement. 5. Twelve pallidal neurons were classified as "wrist-related" based on their movement-locked changes in discharge during task performance and their clear responses to passive wrist joint rotation on examination. All of these neurons exhibited statistically significant modulation of their discharge rate during both fast (peak velocity 97-205 degrees/s) and slow (peak velocity 20-62 degrees/s) wrist movements in the task. The amplitudes of modulation were larger during fast wrist movement than slow movement. These results suggest that the basal ganglia motor circuit plays a similar, rather than an exclusive, role in the control of slow and fast limb movements.