Background: Ability to drive was an important factor of quality of life for subjects with spinal cord injuries (SCI). However, the effect of virtual reality (VR) environment on driving ability and simulation-based driving training of people with SCI has not yet been investigated in any systematic, objective study. The purpose of this study was to examine the effect of the virtual reality created by a driving simulator, and determine the number of simulator sessions necessary for patients with spinal cord injuries to reach maximum driving competence. Methods: This was a longitudinal, prospective before-after trial. It was comprised of 12 spinal cord injury patients who attended driving rehabilitation between July and December 2005. At their initial and subsequent evaluations, the participants' driving skills were measured as they drove along a simulated 6 km two- and three-lane urban road with traffic signals, overpass, underpass, obstacles, and a number of straight and curved stretches of road. The primary outcome measures consisted of total driving time, average speed, center-line violation, stop-line violation, collisions, and steering/braking stability, with a sampling rate of 16 Hz. Each training session lasted for 30 minutes and was carried out twice a week for about 1.5 months. Results: After 5 sessions of simulator driving training, there was a significant increase in the average speed and/or total driving time. The participants could stop their cars more precisely at the stop-line in traffic signal testing, and there was significantly less speed variation and center line violation in overpass testing. Conclusion: This study shows the significant effect of a virtual environment on the progress of driving rehabilitation, and suggests that incorporating virtual reality into rehabilitation programs will accelerate the maximal recovery of the patient's driving competence.