Computer-based hazard perception test scores are associated with the frequency of heavy braking in everyday driving

Abstract

Computer-based hazard perception tests are used in a number of countries as part of the driver licensing processes, and hence evaluating the validity of such tests is crucial. One strategy for assessing the validity of the scores generated by a hazard perception test is to determine whether they can predict on-road driving performance. Only a few prior studies have attempted this, all relying on the subjective ratings of an examiner who was present during a single brief drive and was not blind to the driver’s demographic characteristics, potentially contaminating the outcomes. Additionally, only one such study focused on the most relevant participant group with respect to the validity of tests used in licencing processes, namely young drivers. We sought to remedy this situation in the present project by measuring young drivers’ performance over an extended period of everyday driving via g-force triggered video cameras (“dashcams”) installed in their own vehicles. As a precursor to the dashcam study itself, we developed a new computerized hazard perception test and assessed the validity of its scores by more traditional means (Study 1). As expected, test scores distinguished between high-risk and lower-risk driver groups, and correlated with scores on an established hazard perception test previously shown to predict crash risk. In the subsequent dashcam study (Study 2), the frequency of heavy-braking events (controlling for distance driven) was used as a more objective measure of driving performance. Results indicated that drivers with higher rates of heavy braking had slower hazard perception response times, further supporting the use of these scores as a valid measure of drivers’ ability to exercise hazard perception skill during real driving. More generally, this study also demonstrates the viability of using low-cost off-the-shelf dashcams to measure real-world driving behaviour.