We adopted the estimate of the intraclass coefficient of reliability, R, to evaluate the reliability of saccadic eye movements quantitative analysis. At a one-week interval we recorded refixative saccadic eye movements twice from fifteen healthy subjects by means of the binocular electrooculographic technique. R was computed for the constants and the slopes of the amplitude/duration and the amplitude/peak velocity relationships, for the mean precision values and for the mean latency values adjusted for subject's age. Our data demonstrated that the reliability of saccade parameters is fairly good for the amplitude/peak velocity relationship, good for the precision and very good for the amplitude/duration relationship. Finally, we believe that the normal variability values we obtained can be usefully employed in neurophysiological longitudinal studies not only in normal subjects, but also in pathological condition provided the more reliable parameters and the more adequate strategies to compute normal variability values are chosen.