How much visual field loss can you tolerate on the road? Impact of central and peripheral scotomas on road hazard localization
Poster Presentation: Tuesday, May 20, 2025, 2:45 – 6:45 pm, Pavilion
Session: Scene Perception: Spatiotemporal factors
Schedule of Events | Search Abstracts | Symposia | Talk Sessions | Poster Sessions
Ido Zivli1 (), Ginnie Wee1, Jiali Song1, Benjamin Wolfe1; 1University of Toronto Mississauga
Our ability to perceive the gist of a scene in a glance is well-established for static (Greene & Oliva, 2009) and dynamic scenes (Wolfe et al. 2019). Strong views of the Useful Field or Functional Visual Field argue that dynamic scene tasks require central vision whereas peripheral vision is less useful. However, little empirical evidence speaks to this claim. Here, we examined the extent to which observers could tolerate foveal and peripheral visual field loss in a road hazard localization task. Eight licensed drivers viewed 2s excerpts of 270 dashcam videos from Road Hazard Stimuli (Song et al. 2024; videos subtended 39 x 22 DVA). 66% of videos contained hazards, defined as situations which require immediate driver response to avoid a collision. Hazards occurred in the left or right of frame in equal proportion. The remaining 33% of videos contained no hazards. Participants indicated whether a hazard was on the left, on the right or absent. Using a gaze-contingent display, we simulated central and peripheral scotomas in two separate blocks. Central scotomas were simulated by removing a circular window at the gaze location, leaving the rest of the video visible. Peripheral scotomas only displayed the video inside a circular window at the gaze location. To determine the threshold scotoma size at 80% localization performance, the diameter of the circular window was controlled with a trialwise 3-up-1-down staircase. In the central scotoma condition, participants tolerated a 6.80 DVA central scotoma. In the peripheral scotoma condition, participants required an 8.34 DVA visible window. Our results suggest that detecting immediate hazards in road scenes is highly resistant to visual field loss. Moreover, central and peripheral vision both support this task, suggesting a more complex account of peripheral vision use in dynamic scene perception.
Acknowledgements: Supported by a NSERC DG, University of Toronto XSeed awards to BW