In the longitudinal models predicting alcohol and marijuana use, the constructs of early-onset substance use and friends’ support of substance use were so highly intercorrelated that multicollinearity disrupted the potential for testing the model. Creating a supraordinate construct of substance use lifestyle that was indicated by the two constructs resolved the statistical problem. The solution, however, is more than statistical; it is also substantive. It appears that in general, early- to middle-adolescent substance use is indeed a lifestyle process that is highly linked to the peer network and activities that support and promote adolescent substance use. If this social and developmental process is firmly in place by age 16–17, there is a strong likelihood of at least maintenance of substance use, if not escalation to dependence and abuse.
