Eye movements reveal that drivers can predict the location of hazards in dynamic road scenes but gaze and awareness are dissociable

Parsing complex dynamic scenes is critical for navigating our visual world, and driving safely is a daily task that requires responding to hazardous events which unfold quickly. Theories of driver awareness suggest that drivers need to look at a hazard to detect them, particularly in anticipation of hazard onset. Indeed, scene context may guide eye movements to likely hazard locations. However, given the complexity of real road scenes, drivers may rely instead on explicit cues of an impending collision before identifying the correct location. In 2024, we recorded eye position while 30 licensed drivers localized hazards in annotated dashcam footage. On correct trials, drivers started to look at where the hazard will be 2s before onset, highlighting the importance of anticipatory processes in dynamic scene perception and safe driving. However, hazard-directed looking is not indicative of awareness, as 40% of missed hazards were foveated prior to response. Our results demonstrate early gaze guidance by scene context, which can help drivers anticipate hazards, an idea consistent with theories of driver awareness. However, looking directly at the hazard is not required to correctly detect them, and gaze alone should not be used as an index of awareness in driver monitoring systems.