AN EMPIRICAL METHOD FOR DISTINGUISHING PHYSIOLOGICAL PUPILLARY LIGHT RESPONSE FROM GEOMETRIC DISTORTION OF THE PUPIL IMAGE DURING HORIZONTAL MOVEMENTS OF ARTIFICIAL AND HUMAN EYES

Pupillary Foreshortening Error (PFE) is a systematic distortion in video-based eye trackers that occurs because the pupil is a circular opening in a flat iris; as the eye rotates, the pupil's image on camera changes from a circle to an increasingly thin ellipse. This geometric distortion displaces the calculated pupil centre relative to the optical axis, and results in an inaccurate pupil-size estimation. During saccades, rapid rotations induce sharp, non-linear artifacts often mistaken for physiological pupillary responses; similar, albeit slower, distortions occur during smooth pursuit. While correction typically requires complex 3D models or polynomial regression, we introduce a pragmatic empirical method using artificial eyes to isolate true physiological responses. Our data demonstrates that as the angle between the camera’s optical axis and the pupillary axis increases during horizontal smooth pursuit, the observed pupil size decreases. This relationship is reliably modelled by a truncated Gaussian function. Notably, constriction dynamics were independent of eye-velocity but varied with initial pupil-size. In human subjects, the constriction rate exceeded that predicted by artificial eyes, confirming a concurrent physiological response. We estimated constriction at different eccentricities using the aforesaid fit obtained from artificial eyes and contrasted that with the data obtained from saccade by artificial eyes and humans at that eccentricity. Estimated constriction matched observed constriction for artificial eyes; subtracting the same from the observed constriction of physical eyes provided true constriction of pupils. This subtraction method enables the estimation of pupil-size modulation by cognition during fixation, mid-saccade "flight," and immediately after the gaze shift.