Background Pedestrian injuries are among the leading causes of morbidity and mortality in middle childhood. One limitation to existing pedestrian safety interventions is that they do not provide children with repeated practice needed to develop the complex perceptual and cognitive skills required for safe street crossing. Virtual reality offers training through repeated unsupervised practice without risk, automated feedback on success of crossings, adjustment of traffic to match children's skill and a fun, appealing environment for training. Objective To test the efficacy of virtual reality to train child pedestrians in safe street crossing. Setting Birmingham, Alabama, USA. Methods A randomised controlled trial is underway with an expected sample of four groups of 60 children aged 7-8 years (total N=240). One group receives training in an interactive, immersive virtual pedestrian environment. A second receives pedestrian safety training via widely used video and computer strategies. The third group receives what is judged to be the most efficacious treatment currently available, individualised behavioural training at streetside locations. The fourth group serves as a no-contact control group. All participants are exposed to a range of field and laboratory-based measures of pedestrian skill during baseline and postintervention visits, as well as during a 6-month follow-up assessment. Outcome Measures Primary analyses will be conducted through linear mixed models testing change over time in the four intervention groups. Three pedestrian safety measures will serve as primary outcomes: temporal gap before initiating crossing, temporal gap remaining after crossing and attention to traffic while waiting to cross. Clinical Trial Registration This study is registered at the US government website, www.clinicaltrials.gov, under the title 'Using virtual reality to train children in pedestrian safety', registration number NCT00850759.