1. One hundred twenty neurons were recorded in the central mesencephalic reticular formation (cMRF) of four rhesus monkeys, trained to make visually guided and targeted saccadic eye movements. Eye movements were recorded with the head fixed, using electrooculography (EOG) or subconjunctival scleral search coils. Seventy-six percent (92/120) of cells discharged before and during contraversive visually guided or targeted rapid eye movements, and 76% of these (70/92) responded during contraversive spontaneous saccades in the dark. cMRF neurons had large contraversive movement fields and either a high (> 10 spikes/s) or low background level of spontaneous activity in the dark. The optimal movement vectors (i.e., saccades with greatest response) were predominantly horizontal, although many had a vertical component. Cells with optimal movement vectors within +/- 25 degrees of pure vertical were more rostral in the MRF and were excluded from the analysis. 2. A subgroup of cMRF neurons (31 of 92) that discharged before and during visually guided saccades were examined for visual sensitivity. Slightly less than one-half of these cells (42%, 13/31) were visuomotor units, i.e., they responded to visual targets in the absence of eye movement. The other 58% (n = 18) did not discharge during the visual probe trial; they were movement-related cells. 3. Microstimulation (threshold 40-60 microA at 333 Hz) at the sites of many of these cMRF neurons produced contraversive saccadic eye movements at short latency (< 40 ms). The amplitude and direction of the elicited saccades were similar to the optimal movement vector determined from single-unit recording. This suggested that cMRF cells recorded at the same locus of electrical microstimulation participated in the network responsible for the production and control of rapid eye movements. 4. The 92 saccade-related neurons were divided into two groups on the basis of their background discharge rate. Firing rates for both low background (28%, n = 26) and high background (72%, n = 66) cells increased approximately 30 ms before contraversive saccades and reached a peak discharge just before saccade onset. The low background neurons had either no activity or generated a few spikes just before the end of ipsiversive saccades. The steady rate of discharge (> 10 spikes/s) of high background neurons was inhibited from approximately 20 ms before ipsiversive saccades until just before saccade end. 5. Cells were also subdivided on the basis of how their discharge rates fell at the end of saccades. Clipped cells (38%, n = 35) had activity that fell sharply with saccade offset. Partially clipped cells (62%, n = 57) had persistent firing in the 100 ms following the saccade that was > 20% higher than the firing during the 100 ms before the saccade. 6. Latencies between the 90% point on the rising edge of the peak discharge and the start of the saccade were < or = 5.3 ms for eye movement-related cells in two monkeys. Longer latencies (11-19 ms) were found when measured between the 10% point on the rising edge of the peak discharge and saccade onset. These latencies were equal to or shorter than those obtained for eye movement-related burst neurons in the intermediate and deep layers of the superior colliculus analyzed similarly. Delays between the peak discharge and peak eye velocity were 13.6-15.1 ms for the same group of cMRF eye movement-related cells. These were significantly shorter than the delays measured for eye movement neurons in the superior colliculus (SC) of one of the monkeys. These findings suggest that the buildup discharge of cMRF neurons occurs early enough before saccades to contribute to saccade triggering. The peak discharge, however, occurs with or after the burst in the SC, suggesting that this portion of the discharge serves a function other than saccade triggering. 7. The number of spikes in bursts associated with eye movement was correlated with saccade parameters.