In order to allow for detailed analyses by race, black and Hispanic areas were oversampled to obtain a sufficient number of interviews for reliable estimates. This oversampling was then balanced by weighting to yield correct proportions in the final analysis. Schools were selected in proportion to the number of students in grades 4 through 12, and each school was randomly assigned to either the young or the teenage sample. Every interviewer received a computer-drawn map delineating boundaries of the area in which he or she was to conduct interviews. The map contained a randomly designated starting point at which the interviewer began to list the first household in the area. Interviewers and teachers could not look at any student's work so that each student could be assured of response confidentiality. Questionnaires were self-administered in households contacted for in-home interviewing. Interviewers were instructed to use convenient hours for initial contacts in order to maximize the likelihood of finding respondents at home. Interviews were conducted in May and June of 1996.