In this study, the execution of delayed saccades in 15 DSM-III-R-schizophrenic patients and 15 normal subjects was investigated. While looking at a central fixation cross, a peripheral target was randomly presented at 10 degrees eccentricity. Subjects were instructed to saccade to the target when the fixation cross was switched off after 500 ms. Two experiments were conducted: (a) a delayed-saccade task and, (b) a memory-guided saccade task, that is, the peripheral target was switched off together with the fixation cross. In the delayed-saccade task, amplitudes of regular saccades did not differ between schizophrenic patients and normals. In the memory-guided saccade task, schizophrenic subjects showed marked hypometric saccades. Incorrect delayed saccades (while the fixation cross was on) were also hypometric in schizophrenics, but not in normal controls. The final eye position, i.e., the position reached after the execution of correction saccades, however, did not differ between patients and controls. This means that schizophrenics show a deficit in the programming of primary saccades, if the fixation point and the peripheral target are (a) both visually presented or (b) both memorized. The results support the hypothesis that these saccades are the result of an averaging effect between the fixation point and the peripheral target. It is further hypothesized that these deficits might be explained by a lack of prefrontal inhibition of ocular fixation areas.