Time-to-Collision Estimation in a Simulated Driving Task

Three experiments are reported that examine the role of visual information in the estimation of time to collision. Participants were shown video clips taken from within a car as it approached a stationary target. In the first two experiments the videos were occluded either at a constant time before collision or at a constant distance from the collision. Approach velocity and viewing time were also manipulated. The third experiment again varied approach velocity but did not occlude the video prior to collision with the target. In all three experiments the participants were required to press a pedal when they estimated that collision occurred. Results from all experiments indicated that approach velocity had the greatest influence on the accuracy of collision estimation, and accuracy improved with increasing velocity. Varying viewing time from 1 to 8 s had little effect on estimation accuracy. These results emphasize the role of global optic flow rate in the judgment of time of collision.