The literature on vehicle crash reconstruction provides a number of empirical or classical theoretical models for the distance pedestrians are thrown in impacts with various types of vehicles and impact speeds. The aim of this research was to compare the predictions offered by computer simulation to those obtained using the empirical and classical theoretical models traditionally utilised in vehicle-pedestrian accident reconstruction. Particular attention was paid to the pedestrian throw distance versus vehicle impact speed relationship and the determination of pedestrian injury patterns and associated severity. It was discovered that computer simulation offered improved pedestrian kinematic prediction in comparison to traditional vehicle-pedestrian accident reconstruction techniques. The superior kinematic prediction was found to result in a more reliable pedestrian throw distance versus vehicle impact speed relationship, particularly in regard to varying vehicle and pedestrian parameters such as shape, size and orientation. The pedestrian injury prediction capability of computer simulation was found to be very good for head and lower extremity injury determination. Such injury prediction capabilities were noted to be useful in providing additional correlation of vehicle impact speed predictions, whether these predictions were made using computer simulation, traditional vehicle-pedestrian accident reconstruction methods or a combination of both. A generalised approach to the use of computer simulation for the reconstruction of vehicle-pedestrian accidents was also offered. It is hoped that this approach is developed and improved by other researchers so that over time guidelines for a standardised approach to the simulation of vehicle-pedestrian accidents might evolve. Thoracic injury prediction, particularly for frontal impacts, was found to be less than ideal. It is suspected that the relatively poor thoracic biofidelity stems from the development of pedestrian mathematical models from occupant mathematical models, which were in turn developed from cadaver and dummy tests. It is hoped that future research will result in improved thoracic biofidelity in human mathematical models.