Young drivers, 17-24 years old, are at a greater risk of being involved in fatal crashes than any other age group. Hazard perception and speed choice have been repeatedly linked to the high crash risk seen in this age group. Video-based tests that measure hazard perception and speed choice are only informative if they predict driver safety. Past validation techniques have methodological limitations that could undermine current validity evidence for video-based tests. The current study aims to address these limitations to assess the validity of a hazard perception test and two tests of speed choice by observing on-road behaviour using g-force activated in-car cameras. It was hypothesised that speed choice would be a valid predictor of on-road behaviour, with higher preferred driving speeds and more positive attitudes towards speeding predicting a greater number of g-force events. It was also hypothesised that hazard perception ability would be a valid predictor of on-road behaviour, with better hazard perception and a greater sensitivity to hazards both predicting a lower number of g-force events. Participants (N = 47) had a g-force activated camera installed in their car for 5-8 weeks. The rate of safety-relevant events was established and categorised into braking, turning, illegal and other. A hazard perception test measured drivers’ hazard perception ability. A video speed test and implicit association test of speeding measured preferred driving speeds and implicit attitudes towards speeding, respectively, and both measured hazard sensitivity. No significant results were found when comparing speed choice or hazard sensitivity to g-force event rate. Hazard perception ability was significantly associated with on-road behaviours where slower response times to hazards were associated with a greater rate of g-force activations due to heavy braking. This offers partial support for the validity of the hazard perception test, revealing that it can describe heavy braking behaviour.