Driver gaze behaviour when monitoring for road hazards: Effects of visual distraction
Poster Presentation 43.412: Monday, May 18, 2026, 8:30 am – 12:30 pm, Pavilion
Session: Eye Movements: Natural, complex tasks
Schedule of Events | Search Abstracts | Symposia | Talk Sessions | Poster Sessions
Ginnie Wee1 (ginnie.wee@mail.utoronto.ca), Jiali Song1, Benjamin Wolfe1; 1University of Toronto Mississauga
Safe driving requires detecting hazards in dynamic road scenes. Dividing attention between the road and a secondary task increases collision risk, and eye tracking can reveal how monitoring strategies change under distraction. During visual search, divided attention increases the incidence of looked-but-failed-to-see (LBFTS) errors, in which observers miss targets despite having looked directly at them. This study examines whether these errors occur during dynamic road hazard monitoring (Experiment 1), and whether eyes-off-road (EOR) warnings reduce distraction’s impacts (Experiment 2). In Experiment 1, twenty licensed drivers performed two tasks – a hazard detection task in which drivers reported the locations of hazards in 320 dashcam videos (2 to 8s), and a secondary visual task requiring continuous off-road monitoring. Tasks were performed separately under focused attention, or simultaneously under divided attention. Hazard detection accuracy was lower under divided attention compared to focused attention regardless of whether drivers fixated on the hazard (t(19)<-3.81, p<0.001). Critically, visual distraction increased both LBFTS and didn’t-look-and-didn’t-see (DLDS) errors, in which drivers miss hazards because they did not look at them. Under distraction, approximately two-thirds of missed hazards were DLDS errors. As most misses were DLDS errors, Experiment 2 investigated whether real-time gaze monitoring with EOR warnings reduced miss rates. Eight drivers completed the divided attention condition from Experiment 1 in an EOR warning condition and a warning-absent condition. In the warning condition, a visual alert was given when drivers looked away from the road video for a set duration (1 or 1.5s, in separate conditions). EOR events rarely elicited warnings in the 1.5s condition, suggesting that this threshold is less likely to be helpful, but frequent warnings in the 1s condition may reduce LBFTS and DLDS errors. Together, these results suggest that distraction increases both kinds of errors and point to interventions that may reduce these errors.
Acknowledgements: This work was funded by Transport Canada Enhanced Road Safety Transfer Payment Program.