Road safety education is considered essential to teach children to interact with traffic safely. Many programs, however, do not consider the separate component skills of the road-crossing task, the functional and behavioural factors that may put some children at increased risk, and the most beneficial methods to transfer knowledge to improved behaviour in real-world environments. A targeted and practical training program using a simulated road environment has been developed and evaluated amongst primary school children using a randomised controlled trial. Significant reductions in proportion of critically incorrect road-crossing responses were found immediately after training (56%) and one-month post-training (47%) by the case group compared with pre-training responses, and relative to any changes in responses of the control group. The beneficial effects were greater for younger children, females, children with less well developed perceptual, attentional and cogntive skills, and those with little traffic exposure. The effects of the training program on other outcome measures (proportion of missed opportunity responses, decision time and safety rating responses) were less clear but showed some beneficial effects. This paper discusses the use of the simulated training program, a novel and safe way, to improve road crossing decisions. It is suggested that improvements can be made to child pedestrian education by providing tailored and practical programs that target the component skills of road-crossing decisions and improve essential skills through intensive training and feedback on known risk factors.