Antisaccades have not only longer latencies but also lower peak velocities than prosaccades. It is not known whether these latency and velocity differences are related. Studies of non-human primates suggest that prosaccade peak velocity declines as latency from target appearance increases. We examined whether a similar relationship between peak velocity and latency existed in human saccades, whether it accounted for the difference in peak velocity between antisaccades and prosaccades, and whether it was affected by schizophrenia, a condition that affects antisaccade performance. Sixteen control and 21 schizophrenia subjects performed prosaccade and antisaccade trials in the same test session. In both groups antisaccades had lower peak velocities than prosaccades. Latency did not influence the peak velocities of antisaccades in either subject group. At short latencies, the peak velocities of prosaccades were also similar in the two groups. However, while prosaccade peak velocities declined minimally with increasing latency in control subjects, those in the schizophrenia group declined significantly until they reached a value similar to antisaccade peak velocities. We conclude that, in normal subjects, the effect of latency on prosaccade peak velocity is minimal and cannot account for the lower velocity of antisaccades. In schizophrenia, we hypothesize that the latency-related decline in prosaccade peak velocity may reflect either an increased rate of decay of the effect of the transient visual signal at the saccadic goal, or a failure of the continuing presence of the target to sustain neural activity in the saccadic system.