The frontal eye field (FEF) and superior colliculus (SC) are thought to form two parallel systems for generating saccadic eye movements. The SC is thought classically to mediate reflex-like orienting movements. Thus it can be hypothesized that the FEF exerts a higher level control on a visual grasp reflex. To test this hypothesis we have studied the saccades of patients who have had discrete unilateral removals of frontal lobe tissue for the relief of intractable epilepsy. The responses of these patients were compared to those of normal subjects and patients with unilateral temporal lobe removals. Two tasks were used. In the first task the subject was instructed to look in the direction of a visual cue that appeared unexpectedly 12 degrees to the left or right of a central fixation point (FP), in order to identify a patterned target that appeared 200 ms or more later. In the second "anti-saccade" task the subject was required to look not at the location of the cue but in the opposite direction, an equal distance from FP where after 200 ms or more the patterned target appeared. Three major observations have emerged from the present study. Most frontal patients, with lesions involving both the dorsolateral and mesial cortex had long term difficulties in suppressing disallowed glances to visual stimuli that suddenly appeared in peripheral vision. In such patients, saccades that were eventually directed away from the cue and towards the target were nearly always triggered by the appearance of the target itself irrespective of whether or not the "anti-saccade" was preceded by a disallowed glance. Those eye movements away from the cue were only rarely generated spontaneously across the blank screen during the cue-target time interval. The latency of these visually-triggered saccades was very short (80-140 ms) compared to that of the correct saccades (170-200 ms) to the cue when the cue and target were on the same side, thereby suggesting that the structures removed in these patients normally trigger saccades after considerable computations have already been performed. The results support the view that the frontal lobes, particularly the dorsolateral region which contains the FEF and possibly the supplementary motor area contribute to the generation of complex saccadic eye-movement behaviour. More specifically, they appear to aid in suppressing unwanted reflex-like oculomotor activity and in triggering the appropriate volitional movements when the goal for the movement is known but not yet visible.