Executive functions allow us to respond flexibly rather than stereotypically to the environment. We examined two such functions, task switching and inhibition in the antisaccade paradigm, in two studies. One study involved 18 normal subjects; the other, 21 schizophrenic patients and 16 age-matched controls. Subjects performed blocks of randomly mixed prosaccades and antisaccades. Repeated trials were preceded by the same type of trial (i.e., an antisaccade following an antisaccade), and switched trials were preceded by a trial of the opposite type. We measured accuracy rate and latency as indices of processing costs. Whereas schizophrenic patients had a threefold increase in error rate for antisaccades compared to normals, the effect of task switching on their accuracy did not differ from that in normal subjects. Moreover, the accuracy rate of trials combining antisaccade and task switching was equivalent to a multiplication of the accuracy rates from trials in which each was done alone. Schizophrenic latencies were disproportionately increased for antisaccades, but again they were no different from normal subjects in the effect of task switching. In both groups the effect of task switching on antisaccades was a paradoxical latency reduction. We conclude that the executive dysfunction in schizophrenia is not generalized but selective, sparing task switching from exogenous cues, in which the switch is limited to a stimulus-response remapping. The accuracy data in both groups support independence of antisaccade and task-switching functions. The paradoxical task-switching benefit in antisaccadic latency effects challenges current models of task switching. It suggests either carryover inhibition by antisaccadic performance in the prior trial or facilitation of antisaccades by simultaneous performance of other cognitive operations.