Peer influence is generally believed to be a major cause of adolescent drug behavior. This paper reviews research findings on friend selection and projection to suggest that the magnitude of friend influence may be overestimated. This paper also observes that, although adolescent drug use is assumed to begin in response to peer group influence, peer groups have rarely been measured in studies of drug behavior. Social network analysis is identified as a promising method for measuring peer groups. The implications of this review for research and programs are considered.