Two macaques and three humans fixated luminous targets in a dark field. All subjects had greater dispersion of eye position from trial-to-trial (between-trial variability) than would be predicted from sampling error and within-trial variability. Monkeys had greater between-trial dispersion on the vertical meridian than humans because of less precise control of saccades. Mean vertical eye position of the monkeys varied idiosyncratically with the fixation task (spot-dim or line-tilt). Between-trial fixation variability of both monkeys and humans was large enough to affect the interpretation of experiments relating visual performance to retinal anatomy or to neurophysiology.