Abstract
The principles that behavior is strongly influenced by the behavior of friends, and that people select each other to be friends on the basis of common characteristics, are based in large part on the many studies of adolescents that found the behavior of friends to be similar. In most of these studies, subjects were asked to describe, the behavior of their friends, and this information was then related to subjects' reports of their own behavior. The research on adolescent smoking and drinking described in this paper departs from that tradition by including friend reports of friend behavior so that results based on the different sources for measuring friend behavior can be compared. Subject report of friend behavior is a stronger correlate, of adolescent behavior than friend report of friend behavior, but the two sources for measuring friend behavior produce the same correlations with subject behavior when using longitudinal designs and controlling for prior subject behavior. The findings are discussed in the context of theoretical and methodological considerations of peer influence and selection.