The relations among spatial memory, Stroop-like colour-word subtests, and errors on antisaccade and memory-guided saccadic eye-movement trials for older and younger adults were tested. Two types of errors in the antisaccade task were identified: short latency prosaccade errors that were immediately corrected and longer latency uncorrected prosaccade errors. The age groups did not differ on percentages of either corrected or uncorrected errors, but the latency and time to correct prosaccade errors were shorter for younger than older adults. Uncorrected prosaccade errors correlated significantly with spatial memory accuracy and errors on the colour-word subtests, but neither of these neuropsychological indices correlated with corrected prosaccade errors. These findings suggest that uncorrected prosaccade errors may be a result of cognitive factors involving a failure to maintain the goal of the antisaccade task in working memory. In contrast, corrected errors may be a consequence of a fixation system involving an initial failure to inhibit a reflexive prosaccade but with active goal maintenance enabling correction to take place.