Ten patients with mild to moderate Parkinson's disease were compared with ten age-matched normal controls in a series of saccadic paradigms in order to test various hypotheses relating to the origin of the Parkinsonian saccadic defect. The paradigms comprised a reflex saccade paradigm, a standard remembered saccade paradigm, a remembered saccade paradigm with delayed centre-offset, and a remembered saccade paradigm with a second target flash immediately prior to saccade execution. Finally, subjects executed both reflex and remembered saccades in a standard remembered paradigm (the "two-saccade" paradigm). As has been reported previously, Parkinsonian subjects demonstrated hypometria on all remembered saccade paradigms, particularly the "two-saccade" paradigm. There was, however, no significant difference between the first three remembered saccade paradigms. These studies serve to refute a simple attentional capture hypothesis, and a hypothesis that suggests that the abnormality of remembered saccades is due to concurrent reflex saccade suppression. On the basis of the results, further hypotheses are advanced in an attempt to explain all published work on Parkinsonian saccades.