1. Recent studies of the monkey superior colliculus (SC) have identified several types of cells in the intermediate layers (including burst, buildup, and fixation neurons) and the sequence of changes in their activity during the generation of saccadic eye movements. On the basis of these observations, several hypotheses about the organization of the SC leading to saccade generation have placed the SC in a feedback loop controlling the amplitude and direction of the impending saccade. We tested these hypotheses about the organization of the SC by perturbing the system while recording the activity of neurons within the SC. 2. We applied a brief high-frequency train of electrical stimulation among the fixation cells in the rostral pole of the SC. This momentarily interrupted the saccade in midflight: after the initial eye acceleration, the eye velocity decreased (frequently to 0) and then again accelerated. Despite the break in the saccade, these interrupted saccades were of about the same amplitude as normal saccades. The postinterruption saccades were usually initiated immediately after the termination of stimulation and occurred regardless of whether the saccade target was visible or not. The velocity-amplitude relationship of the preinterruption component of the saccade fell slightly above the main sequence for control saccades of that amplitude, whereas postinterruption saccades fell near the main sequence. 3. Collicular burst neurons are silent during fixation and discharge a robust burst of action potentials for saccades to a restricted region of the visual field that define a closed movement field. During the stimulation-induced saccadic interruption, these burst neurons all showed a pause in their high-frequency discharge. During an interrupted saccade to a visual target, the typical saccade-related burst was broken into two parts: the first part of the burst began before the initial preinterruption saccade; the second burst began before the postinterruption saccade. 4. We quantified three aspects of the resumption of activity of burst neurons following saccade interruption: 1) the total number of spikes in the pre- and postinterruption bursts, was very similar to the total number of spikes in the control saccade burst; 2) the increase in total duration of the burst (preinterruption period + interruption + postinterruption period) was highly correlated with the increase in total saccade duration (preinterruption saccade + interruption + postinterruption saccade); and 3) the time course of the postinterruption saccade and the resumed cell discharge both followed the same monotonic trajectory as the control saccade in most cells. 5. The same population of burst neurons was active for both the preinterruption and the postinterruption saccades, provided that the stimulation was brief enough to allow the postinterruption saccade to occur immediately. If the postinterruption saccade was delayed by > 100 ms, then burst neurons at a new and more rostral locus related to such smaller saccades became active in association with the smaller remaining saccade. We interpret this shift in active locations within the SC as a termination of the initial saccadic error command and the triggering of a new one. 6. Buildup neurons usually had two aspects to their discharge: a high-frequency burst for saccades of the optimal amplitude and direction (similar to burst neurons), and a low-frequency discharge for saccades of optimal direction whose amplitudes were equal to or greater than the optimal (different from burst neurons). The stimulation-induced interruption in saccade trajectory differentially affected these two components of buildup neuron discharge. The high-frequency burst component was affected in a manner very similar to the burst neurons.(ABSTRACT TRUNCATED)