Pupil dilation is a useful, noninvasive technique for measuring the change in cognitive load. Since it is implicit and nonverbal, it is particularly useful with preverbal or nonverbal participants. In cognitive psychology, pupil dilation is most often measured by corneal reflection eye-tracking devices. The present study investigates the effect of gaze position on pupil size estimation by three common eye-tracking systems. The task consisted of a simple object pursuit situation, as a sphere rotated around the display screen. Systematic errors of pupil size estimation were found with all three systems. Implications for task-elicited pupillometry, especially for gaze-contingent studies such as object tracking or reading, are discussed.