Michele A. Cox¹ (), Ashley M. Clark¹, Paul Jolly¹, Sanjana Kapisthalam¹, Yuanhao H. Li¹, Ruitao Lin¹, Soma Mizobuchi¹, T. Scott Murdison², Alina Neverodska¹, Jie Z. Wang¹, Zhetuo Zhao¹, Michele Rucci¹; ¹University of Rochester, ²Reality Labs

Recent studies have shown that humans exhibit remarkable precision in controlling eye movements during fixation. However, because of technical challenges in measuring eye movements, most of these studies were conducted as observers examined stimuli rendered on fixed-distance displays with their heads immobilized. Thus, little is known about the precision of natural, head-free fixation on real, three-dimensional objects. To overcome these limitations, we developed a system capable of measuring eye movements at high resolution during normal head movements. This device consists of a specifically-designed magnetic induction eye-tracker integrated with a motion capture system, together enabling simultaneous measurements of head and eye movements with arcminute (1/60th degree) resolution. Using this apparatus, we examined oculomotor control during fine discrimination of real objects at nearby working distances (30-60 cm). Targets were designed for either shape or a depth judgement and 3D printed from parametric models to scale angularly (1 degree) with viewing distance. The shape judgement consisted of reporting the position (left or right) of a 0.5 arcminute gap in a Landolt C. The depth judgement consisted of reporting whether the central prong of a 3-pronged fork was closer or farther than the two flankers (0.5 arcminute disparity; horizontal prong spacing and diameter 0.2 degrees). Normal-sighted observers (N=10) performed these tasks while seated and were allowed to freely move their heads. Trials were self-paced and ended with verbal reports. Our data show that head-eye control differs in shape and depth judgements. Head rotations and translations were attenuated and fixations considerably longer in the stereo judgements. The eye-in-head speed of intersaccadic fixation was also reduced during the stereo judgement, an effect that persisted even after controlling for head motion differences across tasks. Computational modeling is used to explore the benefits of longer, slower fixations for fine depth discrimination.

Acknowledgements: Supported by Reality Labs and NIH grants R01EY018363 and P30EY001319.