In an antisaccade task, where saccades in the direction opposite of a suddenly presented stimulus are required, certain numbers of prosaccades can occur. The hypothesis is put forward that poor fixation and poor voluntary saccade control constitute two independent sources for the errors. This possibility is investigated by including the corrections of the errors in the analysis. First, the eye movements of 346 normal subjects (group N) performing a gap antisaccade and an overlap prosaccade task were measured. For each subject the proportion of express saccades in the overlap prosaccade task and the proportion of prosaccades in the gap antisaccade task were determined. The data of 150 subjects with more than 20% proerrors were divided into two groups: group A with relatively many, group B with relatively few express saccades in the overlap prosaccade task. Group A subjects produced their errors after significantly shorter reaction times and they corrected their errors significantly faster and more often than group B subjects. Second, we analysed the data of three groups of subjects: the complete normal group N, a group D of dyslexic subjects (n=343), and a group T containing all subjects irrespective of their cognitive achievements (n=780). A highly significant negative correlation exists between the correction rates and the error rates. A factor analysis of the variables performed for each group separately results in only two factors, one describing prosaccade the other antisaccade control. Only the error rate contributes significantly to both factors indicating that high errors may have two independent reasons.