Research studies on group versus individual bystander responses either have involved nonviolent emergencies or laboratory simulations. To investigate group versus individual bystander response to a violent assualt in a natural setting, 80 male college students (from an original pool of 393 white, male students) were observed either individually or in small (3 to 5 person) groups, as they witnessed an apparent (simulated) rape. Witness intervention modes (direct, indirect, no intervention) were coded by three hidden observers and a campus security guard placed on an alternate route from the rape scene. An analysis of the results showed that 85% of the subjects in the group condition intervened, compared to 65% of individual bystanders. Seventy percent of the groups intervened directly, while 50% of the individuals did so. The findings suggest that group processes which inhibit group response in one set of circumstances (as indicated by Latane's l970 laboratory research) can enhance group response in another set. The findings also highlight the importance of conducting such experiments in natural settings, despite the numerous methodological problems that must be overcome. (BL)