1. The present study tests several key aspects of the current model of the intrinsic circuitry of the basal ganglia, in particular the degree to which basal ganglia-thalamocortical circuits are functionally segregated at the level of the subthalamic nucleus (STN). To this end the responses of STN cells to somatosensory examination (n = 301 cells), the polarity and latencies of neuronal responses to passive and active movements (n = 223 cells), responses to microstimulation (n = 1589 sites), and cross-correlation functions of pairs of neighboring neurons (n = 72 pairs) were studied in STNs of three African green monkeys. 2. The activity of 55% of cells examined in STN was briskly modulated in response to passive movements of individual contralateral body parts. Of these, 86% responded to passive joint rotation of muscle palpation, but in some cases (25% of responding cells) responses were also elicited by light touch. In 91% of the responding cells responses were elicited by manipulations around a single joint only. 3. The caudoventral sector in STN was largely devoid of cells with responses to somatosensory stimulation. Within the rostrodorsal zone a lateral region containing neurons that responded to arm movements and a more medial region with neurons responding to leg movement were found. Cells responding to orofacial movements were located more dorsally and rostrally. Neurons with similar responses to active and passive movements of the limbs tended to be clustered within 'arm' and 'leg' zones. 4. Of identified arm cells in STN (n = 80), 36% responded to the application of torque pulses to the elbow (43 responses overall). Forty-eight percent of these cells responded to both extension and flexion torques. Ninety-three percent of the responses were initial increases in discharge, which characteristically occurred earlier and were shorter than initial decreases. Fifty-three percent of the responses were biphasic or multiphasic. 5. During active step tracking movements 40% of STN arm cells (n = 53 cells) responded with significant changes in activity. Thirty-six percent of these cells showed responses with both extension and flexion movements. Of the responses, 90% were increases in discharge. Only 14% of all responses were biphasic or multiphasic. Responses tended to occur around the time of movement onset (average latency 2 ms after movement onset). 6. Microstimulation (bipolar pulses, 40 μA, 200-500 ms train duration, 400 Hz) of the core of STN itself did not appear to produce movement. However, stimulation at the lateral borders of STN and of the adjacent white matter often led to limb or eye movement. 7. Cross-correlation analysis of simultaneously recorded pairs of neurons revealed significant synchronized activity in only 11% of pairs. 8. The somatotopic arrangement of neuronal responses and the paucity of neighboring cells discharging in synchrony strongly support the concept of functional segregation in the basal ganglia-thalamocortical pathways. The predominance of brisk increases in discharge in STN in response to movements most likely results from corticosubthalamic activation. The current model of basal ganglia anatomy predicts that this will lead to inhibition of movements. The inhibitory role of STN in motor control is further supported by the failure of electrical stimulation of the nucleus to induce movements. The late onset of responses of STN neurons in the step tracking task suggests that STN and the 'indirect' pathway are not involved in the selection or initiation of movements, but may rather have a role in the control of ongoing movements.