Previous studies of single neurons in the substantia nigra reticulata (SNr) have shown that many of them respond to similar events. These results, as well as anatomical studies, suggest that SNr neurons share inputs and thus may have correlated activity. Different types of correlation can exist between pairs of neurons. These are traditionally classified as either spike-count ("signal" and "noise") or spike-timing (spike-to-spike and joint peristimulus time histograms) correlations. These measures of neuronal correlation are partially independent and have different implications. Our purpose was to probe the computational characteristics of the basal ganglia output nuclei through an analysis of these different types of correlation in the SNr. We carried out simultaneous multiple-electrode single-unit recordings in the SNr of two monkeys performing a probabilistic delayed visuomotor response task. A total of 113 neurons (yielding 355 simultaneously recorded pairs) were studied. Most SNr neurons responded to one or more task-related events, with instruction cue (69%) and reward (63%) predominating. Response-match analysis, comparing peristimulus time histograms, revealed a significant overlap between response vectors. However, no measure of average correlation differed significantly from zero. The lack of significant SNr spike-count population correlations appears to be an exceptional phenomenon in the brain, perhaps indicating unique event-related processing by basal ganglia output neurons to achieve better information transfer. The lack of spike-timing correlations suggests that the basal high-frequency discharge of SNr neurons is not driven by the common inputs and is probably intrinsic.