Eye movement parameters are consistently investigated, and the distribution of these parametersare well known. Whereas saccade duration has been studied along with saccade amplitude , thedistribution of saccade duration has not yet been reported. We aim to investigate thedistribution of saccade duration in several eye movement datasets from the literature and fromour own data to confirm the common, but never reported, observation that the distribution ofsaccade duration is bimodal. We consistently observed the bimodality of saccade durations, nottask- or stimuli-dependent. We created two groups of saccades based on the saccade durationdistribution. Our results suggest that short duration saccades could be partly linked to bottom-up processes and long duration saccades to top-down processes. This study highlights theimportance of reporting the distribution of eye movement data, in addition to means, which donot allow a correct and representative analysis in the case of bimodal distributions.