Network position ("isolate," "member," "liaison"), peer-group substance use, and their interaction were examined as predictors of cigarette, alcohol, and marijuana use in a sample of 163 urban sixth, seventh, and eighth graders. Two measures of peer substance use were compared: one based on social network analysis (SNA), the other on perceptions of use. Results varied by substance. For cigarettes, network position and the interaction between position and peer-group use predicted use in the model using SNA to measure peer use. Liaisons were most likely to smoke, but isolates' and members' smoking was significantly associated with peer smoking. For alcohol, perceptions of peer-group alcohol use predicted individual use. For marijuana, peer-group use and the interaction between position and peer use predicted use, regardless of measure. Liaisons' marijuana use was significantly associated with peer use. The importance of SNA for understanding peer factors in adolescent substance use is discussed. (Contains 3 tables and 1 figure.)