Fengping Hu¹ (), Joyce Yi Xin Chen², Denis G Pelli¹, Jonathan Winawer¹; ¹New York University, ²University of Michigan - Ann Arbor

Online vision testing allows efficient collection of data from a diverse participant pool. However, many vision science studies depend on participants looking at a fixation mark. Traditionally, vision scientists check fixation with a camera-based gaze tracking device. However, current online implementations of gaze tracking using webcams are not sufficiently accurate. A recent alternative strategy took advantage of the natural tendency to coordinate hand and eye movements, encouraging accurate fixation by asking participants to track a moving fixation mark with a mouse-controlled cursor (Kurzawski, Pombo, et al., 2023). This dynamic fixation task greatly reduces peeking at a peripheral target, but does not eliminate it. Here, to further improve fixation, we exploit the “crowding” phenomenon by adding clutter around the fixation mark. We call this “crowded dynamic fixation”. We assessed fixation accuracy while measuring a peripheral threshold. Compared to stationary fixation, dynamic fixation without clutter reduced RMS fixation error by 39%. Crowded dynamic fixation reduced RMS error even more, by 53%. With a 1.5º tolerance, this corresponds to “peeking” on 9% of trials with a stationary fixation, 4% with a dynamic fixation, and 0% with crowded dynamic fixation. This improvement in fixation eliminated implausibly low thresholds on the peripheral task, presumably by eliminating peeking. We conclude that, by preventing peeking, crowded dynamic fixation brings accurate gaze control to online testing.

Acknowledgements: Funding: NIH grants R01-EY027401 (JW), R01-EY027964 (DP) and R01-EY033628(JW).