This study investigated whether alcohol use by public school students differs in its relationship to other variables when analyzed on the individual level versus the aggregated group level. The study presents a model and a set of analytic techniques for studying the ways that school and individual variables relate to adolescent alcohol abuse. Recently developed Hierarchical Linear Model procedures were used to estimate and test the school-level contextual effects, controlling for the student-level variables. The analyses are based on 61 schools, with 10,344 students from 6th, 8th, 10th, and 11th grades. The student outcome, abuse, is the level of alcohol abuse determined by multiplying the frequency of alcohol use by the extent to which the student gets "high" when using alcohol. The two student-level determinants in the within-school model are social, reflecting how much the student socializes with friends, and good, the degree to which the student fits the traditional stereotype of good by getting good grades, staying out of trouble, attending church, and talking with parents about problems. The estimated social effect in this model is positive, indicating that more socializing is associated with more abuse, while a negative coefficient for good suggests that increasing goodness is associated with decreasing abuse. The predicted effect of social on abuse was generally positive, and the predicted effect of good was generally negative. (The effects are represented graphically.) (LLL)