1. Models of the saccadic system propose that there is an integration of the pulse signal, and there is good evidence that the integrator is reset gradually (Nichols and Sparks 1994, 1995). Other studies of the superior collicular contribution to the saccadic system have proposed a sensory, not motor, nature for its signal. 2. To test experimentally the resetting of the integrator and the nature of the collicular signal, we electrically stimulated the superior colliculus during periods of fixation and during the course of visually guided saccades. Trains of stimuli which were presented during periods of fixation evoked saccades with fixed vectors. Identical stimulation at the beginning of a visually guided saccade evoked saccades whose direction was rotated and amplitude extended from the fixed vector. The direction of the rotation was opposite that of the visually guided saccade, and the magnitude of this rotation could be as large as 80 degrees. 3. Stimulation which was applied at progressively later times during the visually guided saccade, evoked saccades with progressively smaller rotations and progressively less elongations. The time period during which saccades were modified persisted beyond the end of the visually guided saccade, when the eyes were stationary. Thus, we confirm the previous findings (Nichols and Sparks 1994, 1995; Robinson, 1972), that the end of the saccade is not a period of quiescence within the oculomotor pathways. 4. Our results confirm that the resetting of the integration of the saccade signal is gradual rather than abrupt. Furthermore, these data suggest that the superior colliculus signals a motor error.