Yelda Semizer¹ (), Kian Motahari¹, Benjamin Balas², Ruth Rosenholtz³; ¹New Jersey Institute of Technology, ²North Dakota State University, ³NVIDIA Research

Solving a maze often requires a series of eye movements. Peripheral information, limited by visual crowding, determines ability to perceive the path ahead, and informs the next fixation location. We examine to what extent peripheral vision is used for maze solving using a gaze-contingent paradigm by only rendering a portion of the maze within a circular window centered at the current gaze point, outside of which the maze is masked or degraded. If the visual system uses peripheral vision to solve a maze, performance should diminish when there is no peripheral information available. For a given maze design, performance should increase as a function of window size, up to the point at which the window size reaches that of the functional viewing field. If peripheral information is available but degraded outside the window, performance should be better but constrained by the usefulness of the remaining information in signaling the structure of the path ahead. If the information available to peripheral vision is not informative about the path ahead, performance should be poor. Observers solved mazes mentally within an eye-tracking setup. Only the central gaze region was presented in clear view while the peripheral regions were masked to varying degrees of visual angle (1, 2, 4, or 7 in radius). Three different masking conditions were applied, where the peripheral region was either blurred, fully masked, or masked by a grid-like structure matching the appearance of the maze paths and walls. Results showed that performance in the grid condition was worse than in the masked condition, which was worse than in the blur condition. Additionally, performance increased with window size, with no difference between 4 and 7 degrees. These findings suggest that the amount of peripheral information available, its relevance, and usefulness for the task are critical in mental maze solving.