Peripheral and central vision in hazard localization with a gaze-contingent scotoma paradigm

Research on static scene perception using simulated scotoma paradigms suggests both central and peripheral vision are sufficient for scene gist perception (Larson & Loschky, 2009), but it is unknown whether this applies to dynamic scenes. This study investigated whether peripheral vision and foveal vision can independently support road hazard detection in dynamic real-world road scenes. Two types of gaze-contingent scotomas were simulated in separate blocks: a peripheral scotoma, where the video was only visible through a central circular aperture, or a central scotoma, in which a gray circle occluded the central visual field. Within a block, the size of the scotoma varied based on a 3-up/1-down staircase to estimate the maximum tolerable scotoma size while maintaining 80% hazard detection accuracy. Echoing the static scene literature, both peripheral and foveal vision are independently sufficient for this dynamic hazard detection task, demonstrating the utility of peripheral vision. Again, in line with data from static scenes, the central scotoma condition required larger visible area than the peripheral scotoma condition. Furthermore, a strong positive correlation was found between the two conditions, suggesting individuals’ hazard detection capabilities are generalized across the visual field. Our work broadens current accounts of peripheral vision to time-limited dynamic scenes, demonstrating that drivers can obtain sufficient information from non-foveal input. These results suggest that accounts centered only on where drivers fixate miss key components of the information acquisitive process.