Stephanie Reeves¹, Jorge Otero-Millan^1,2; ¹Herbert Wertheim School of Optometry & Vision Science, University of California Berkeley, ²Department of Neurology, Johns Hopkins University

Humans make eye movements to explore the world around them. While saccades occur in all directions, horizontal saccades are more common than vertical saccades, and upward saccades more than downward saccades. We hypothesize that the higher frequency of upward saccades could be related to a downward bias in fixational drift, as previously shown for fixational eye movements. We also hypothesize that drift velocity depends on eye position by drawing the eye back to primary position. We analyzed four publicly available datasets (Reeves 2023, Reeves 2022, Chen 2021, and van der Linde 2009) to quantify the velocity of ocular drift as a function of eye position. After processing the sample data, there were 510,491 fixation periods across 73 subjects (more than 1,000 fixations per subject). For each subject, we used linear regression to measure the influence of horizontal eye position on horizontal drift velocity and vertical eye position on vertical drift velocity, respectively. The intercept served as a measure of overall bias and the slope as a measure of the effect of eye position on drift. Results revealed two main findings. First, the average slopes for both horizontal and vertical components were negative (horizontal: -0.013°/s per degree, p<0.001; vertical: -0.012°/s per degree, p<0.001), indicating that the eye tends to drift back to primary position during eccentric free viewing. Second, the average intercepts revealed an overall downward drift bias of -0.15°/s such that regardless of where participants looked vertically, the eye tended to drift down (p<0.001). There was also a rightward drift bias (0.08°/s, p<0.001) that was about 50% of the magnitude of the downward drift bias and only present in one of the four datasets (p>0.39). Taken together, these preliminary results show that the ocular motor system produces drift during fixation that is not completely random and is instead systematically biased.

Acknowledgements: R00 EY027846