The records of drivers, selected from the file of licensed drivers in Ontario, were reviewed to study the relationship between demerit points, other driver characteristics, and the frequency or risk of future collisions and traffic convictions. A stratified sample of 500-600 drivers from each of five levels of demerit points was selected. Low-point drivers differed significantly from high-point drivers in age, sex, and class of licence; estimates of risk of collision or conviction in each demerit point group had to take account of these differences. Discriminant analysis was used to identify drivers likely to be involved in collisions or to be given traffic convictions, and to identify accidents involving injury or fatality. Of the traits considered (demerit points, age, sex, class of licence, history of previous accidents), demerit points represented the only variable of importance in predicting future collision involvement. Since it is the only one of these variables which can be altered by driver behaviour it offers an opportunity to prevent accidents.