The longitudinal Guttman simplex (LGS), a method of modeling stage sequences or levels over time, allows for individual differences in sequence progression or development. The LGS is particularly appropriate in modeling tobacco, alcohol, and marijuana use stages with adolescent subjects where relationships among drugs and degree of involvement may be explored. To examine adolescent substance use, data were obtained from junior high school students (N=541) in the Los Angeles, California area who were part of a 5-year substance abuse prevention program. Psychosocial constructs and drug use were measured for this phase while students were in grade 7 (1983) and grade 9 (1985). The results revealed that a relatively sizable percentage of variability in substance use as measured by a LGS could be accounted for by psychosocial scales measured one year prior to use assessment. Analyses suggest that a student's self-report of the number of friends who use, and the student's intention to use drugs, related to increased use one year later. The strongest, consistent predictor of substance use identified (friends who use substances) may not be easily modified by preventive interventions. Constructs identified as indirectly predicting use, such as prevalence estimates and perceived consequences of use may be amenable to change as a result of prevention programming. (NB)