This paper reports the results of a study designed to investigate the relation between vehicle handling and accident frequency. Because evaluation of the driver control process can only be made on a subjective basis, emphasis is placed on evaluating the role of the vehicle rather than the man-car combination.Deficiencies in handling are likely to be associated with accidents involving loss of control, and it is shown that over 80 percent of loss of control accidents involve single vehicles only. This means that the single vehicle accident rate can be used as a measure of proneness of loss of control. Single vehicle accident rates are determined by model of car and the effect on these rates of other factors, such as variations in driver age and ratio of male to female driver population between the different models, are assessed. This allows accident rates to be established which are, to a large extent, independent of driver effects, so that their relation with vehicle handling performance can be directly assessed.In selecting parameters with which to evaluate the role of the vehicle, emphasis is placed on simplicity of measurement to allow as many different models of car as possible to be included in the study. Accordingly, simple design parameters as well as more complex vehicle response parameters are considered.Multiple regression techniques are used to establish the relative importance of handling response parameters when compared to driver effects and also to decide which are the most appropriate parameters to consider. The results show that in explaining the variation in the accident rate between different models of car, driver effects account for as much as 70 percent; if driver effects are removed from the accident rate then handling parameters explain about 35-40 percent of the remaining variation between models of car. The important parameters appear to be weight, a measure of the change in understeer as a function of lateral acceleration and power to weight ratio.