Factors related to plans to engage in sexual activity and unsafe sexual behavior in the future were investigated to provide data necessary to design the most appropriate AIDS risk reduction interventions for young adolescents. Data were collected using questionnaires administered to 303 seventh-grade students (145 male) at three schools in metropolitan New York City (43% White, 33% Black, 18% Hispanic, and 6% "other") in February 1990. The schools offered no AIDS prevention programs. The confidential questionnaires elicited information on intentions to become sexually active and to use a condom and on gender, ethnicity, AIDS knowledge, cigarette smoking and alcohol use, and decision-making skills. Data were analyzed using SPSS/PC. The dependent measures (sex activity and condom intentions) were submitted to multiple regression. Gender, two AIDS knowledge items, ethnicity, frequency of cigarette and alcohol use, and decision-making skills were simultaneously regressed on each of the dependent measures. It was found that the likelihood of planning to have sex before marriage was greater among boys and among students who used alcohol more frequently and was less among Hispanics and students who knew that refraining from sex is the most effective way to prevent AIDS. The intention to refuse sex with a longterm dating partner was influenced by gender (boys less likely to refuse), current alcohol use (users less likely to refuse), and decision-making (skilled decision-makers more likely to refuse). Those who intended to have sex before finishing high school were more likely to be boys, frequent smokers, and frequent alcohol users. Intention to use a condom was positively determined by knowledge that condom use reduces the risk of AIDS and by decision-making skills. These results indicate that social skills programs designed to prevent substance use and premature sexual activity may offer the best strategy in deterring problem behaviors and reducing the risk of AIDS. This study was limited by its small sample size, by its reliance on intended behavior, and by the cross-sectional rather than longitudinal nature of the data.