The cortical connections of the dorsal (PMd) and ventral (PMv) subdivisions of the premotor area (PM, lateral area 6) were studied in four monkeys (Macaca fascicularis) through the use of retrograde tracers. In two animals, tracer was injected ventral to the arcuate sulcus (PMv), in a region from which forelimb movements could be elicited by intracortical microstimulation (ICMS). Tracer injections dorsal to the arcuate sulcus (PMd) were made in two locations. In one animal, tracer was injected caudal to the genu of the arcuate sulcus (in caudal PMd [cPMd], where ICMS was effective in eliciting forelimb movements); in another animal, it was injected rostral to the genu of the arcuate sulcus (in rostral PMd [rPMd], where ICMS was ineffective in eliciting movements). Retrogradely labeled neurons were counted in the ipsilateral hemisphere and located in cytoarchitectonically identified areas of the frontal and parietal lobes. Although both PMv and PMd were found to receive inputs from other motor areas, the prefrontal cortex, and the parietal cortex, there were differences in the topography and the relative strength of projections from these areas. There were few common inputs to PMv and PMd; only the supplementary eye fields projected to all three areas studied. Interconnections within PMd or PMv appeared to link hindlimb and forelimb representations, and forelimb and face representations; however, connections between PMd and PMv were sparse. Areas cPMd and PMv were found to receive inputs from other motor areas--the primary motor area, the supplementary motor area, and the cingulate motor area--but the topography and strength of projections from these areas varied. Area rPMd was found to receive sparse inputs, if any, from these motor areas. The frontal eye field (area 8a) was found to project to PMv and rPMd, and area 46 was labeled substantially only from rPMd. Parietal projections to PMv were found to originate from a variety of somatosensory and visual areas, including the second somatosensory cortex and related areas in the parietal operculum of the lateral sulcus, as well as areas 5, 7a, and 7b, and the anterior intraparietal area. By contrast, projections to cPMd arose only from area 5. Visual areas 7m and the medial intraparietal area were labeled from rPMd. Relatively more parietal neurons were labeled after tracer injections in PMv than in PMd. Thus, PMv and PMd appear to be parts of separate, parallel networks for movement control.