Rubbernecking at a roadside accident: An exploratory, simulated eye-tracking study of young adult drivers in Abu Dhabi

Roadside accidents can trigger secondary risks outside the blocked lane, but this is poorly described in mixed traffic with human-driven vehicles (HVs) and automated vehicles (AVs). This simulator study examined 38 drivers aged 18–29 in a 4-km, 8-lane Abu Dhabi-like urban corridor containing a Lane 1 roadside accident between 2255 and 2330 m. STISIM outputs and Tobii fixation data were summarized in 13 consecutive 10-m stages before the accident center. Because later stages were missing only after crashes, primary inference was limited to Stages 1–8 (130–50 m before the accident center). Stage-wise outcomes were analyzed with generalized estimating equations, and gaze was summarized using 9 predefined areas of interest (AOIs). Crash occurrence differed by lane (Lane 1: 2/8; Lane 2: 13/22; Lane 3: 0/8). Significant lane-by-stage effects were found for speed, longitudinal acceleration, braking deceleration, steering wheel angle, minimum time to collision with the AV, and minimum range. Variability analyses showed stronger instability in Lanes 1 and 2 than in Lane 3. AOI analysis showed lane-dependent differences in fixation allocation: a larger share of fixations fell within the driver-side windshield AOI in Lanes 1 and 2, whereas Lane 3 showed a more distributed pattern. In Lane 2, crash trials showed later-stage compression of minimum range and minimum time to collision and narrower fixation allocation. These findings are exploratory and scenario specific. They point to candidate monitoring signals for mixed-traffic warning research, not validated deployment thresholds.