The research was conducted in a southeastern city with a population of 265,968. Youths in this area are particularly vulnerable to violence due to the incidence and intensity of violent crime around them. The participants included randomly selected 5th, 7th, 9th, and 11th graders from schools with high numbers of conduct code violations. The data were analyzed by means of logistic regression techniques. Seventy-nine percent of the participants were categorized as nonviolent, including 48 percent in the most nonviolent group. Among the remainder, 13 percent were in the most violent group. Violent behavior was predicted in youths who were exposed to violence, were gang members, had family or friends who are gang members, or had peer support. Factors that predicted nonviolent behavior included adult social support, an attitude that religion is important, younger, and female gender. Findings are limited by the use of self-report data, the preponderance of black participants, and the collection of data only from students who attended school that day. Nevertheless, findings supported the role of certain protective factors in reducing the risk of violent behavior among vulnerable youths. Therefore, violence prevention programs should use a framework that simultaneously reduces risk and promotes protection. Tables and 40 references (Author abstract modified)