. The hypothesis that there is a positive relationship between violence and game attendance in the National Hockey League is tested; and the implications of the results for policies aimed at rectifying the violence considered. The empirical analysis, which is based principally on game by game data for the 1989/90 season, confirms the positive relationship, although it occurs mainly in American, not Canadian, cities. Since there is an incentive for teams to promote violence and because the legal system normally only prosecutes players not teams, the standard methods used to control violence—self regulation and the judicial process—are unlikely to work. Consideration should, therefore, be given to more direct control of violence.