A new approach to estimating exposure is presented and applied to determining relations between car mass and driver fatality likelihood. The new approach considers two groups of fatal crashes in the FARS files. The first group contains crashes in which car drivers are killed in single car crashes or in crashes with trucks. These are both examples of non-two car crashes. It is hypothesized that the likelihood of a car driver fatality in such crashes depends on car mass. The second group of fatal crashes contains crashes in which either pedestrians or motorcyclists are killed in crashes with cars. It is hypothesized that the likelihood of the pedestrian or motorcyclist being killed in such crashes is independent of the mass of the car. The new exposure approach implies that the ratio of the number of people killed in the mass dependent crash to those killed in the mass independent crash gives an estimate of how car mass affects the likelihood of a driver fatality. The approach further implies that the estimate obtained is an estimate of the physical effect of mass, essentially independent of driver behavior. It is found that the new exposure approach yields relationships between driver fatality likelihood and car mass that are more precise and consistent than can normally be obtained in accident research. The effects found, which are attributed to the physical properties of the vehicle, essentially independent of driver behavior, are larger (for example, a driver of a 900 kg car is 2.6 times as likely to be killed as is a driver of a 1800kg car) than those based on fatalities per car.