During the fall of 1981, questionnaires were administered to 2,703 10th grade students in 9 high schools in a Southern city. A second questionnaire was administered to 2,258 11th grade students in the fall of 1982. A total of 1,625 had completed the previous questionnaire in 1981. Respondents estimated the number of times they had committed four offenses. For 11th graders who completed questionnaires in the 10th grade, offenses committed during that year were reported. The measure of perceived severity identical to that used by Grasmick and Bryjak asked each respondent to contemplate the likely punishment for an offense committed and to indicate how severe each perceives the punishment. Perception-of-severity questions not only covered penal sanctions but also perceptions of the severity of peer sanctions, parental sanctions, and educational sanctions derived from the application of formal penal sanctions. Perceived certainty of punishment was also measured. Although perceptions of punishment severity did act as a deterrent for criminal behavior, compared to perceived certainty of punishment and fear of informal penalties, penal sanctions had the weakest correlation with subsequent criminal involvement. The development of a model of informal social control offers the greatest promise for deterring juvenile criminal behavior. Tabular data and 43 references.