start » program » presentations

Accuracy of fast, post-saccadic eye movements recorded with pupil-based eye trackers is influenced by relative motion between the pupil and the iris

To be able to accurately measure saccade offsets and fixation onsets is important when, e.g., calculating fixation durations, saccade amplitudes, and when controlling the stimulus presentation in a gaze-contingent experiment (Nyström & Holmqvist, 2010). A complicating factor when deciding where the saccade ends is a type of post-saccadic eye movement known as dynamic overshoot, which can be described as a oscillation in the eye-tracker signal that occurs directly after a saccade (Bahill et al., 1975). In data recorded with dual Purkinje (DPIs) eye-trackers, which contain large post-saccadic oscillations, dynamic overshoot has been attributed to motion of the lens relative to the eyeball (Deubel & Bridgeman, 1995). In contract, data recorded with scleral search coils contain smaller or no overshoots, which may reflect that coils modify the neural command signals that control saccades, or that the annulus of the coil slips in relation to the cornea (Frens & van der Geest, 2002). While dynamic overshoot can be found in data recorded with different techniques and across different participants, little is known of why they occur in data recorded from modern video-based eye trackers using the principle of pupil and corneal reflection tracking.

We collected eye movements from three participants making horizontal, abducting saccades from the right side of a computer monitor to its center. Eye movement data and eye images were recorded monocularly from the participants' left eyes at 500 Hz with the SMI Hi-Speed system. The movement of the pupil- and iris centers were extracted from the sequence of eye images and compared with the recorded eye movement data.

Results showed a high correlation between the eye movement data and the motion of the pupil center extracted from the eye image. There was however a large discrepancy between the movements of the pupil- and iris centers during the post-saccadic oscillation. This suggests that the pupil moves relative to the iris, and that this relative motion is reflected in the eye movement data. Consequently, pupil-based eye trackers do not accurately measure post-saccadic eyeball rotation, but rather a superimposed movement consisting of eyeball- and pupil motion (and possibly motion of the iris). These results are important for researchers using pupil-based eye trackers to answer fine-grained questions about the oculomotor system, in particular directly after saccade endings.


Bahill, A., Clark, M., & Stark, L. (1975). Dynamic overshoot in saccadic eye
movements is caused by neurological control signal reversals. Experimental
neurology, 48(1), 107–122.

Deubel, H., & Bridgeman, B. (1995a). Fourth purkinje image signals reveal
eye-lens deviations and retinal image distortions during saccades. Vision
Research, 35(4), 529–538.

Frens, M. A., & Van Der Geest, J. N. (2002). Scleral search coils influence
saccade dynamics. Journal of Neurophysiology, 88(2), 692–698.

Nyström, M., & Holmqvist, K. (2010). An adaptive algorithm for fixation,
saccade, and glissade detection in eye-tracking data. Behavior Research
Methods, 42(1), 188–204.