We use social learning, self-control, and social disorganization theories to develop hypotheses related to tobacco and alcohol use among a racially diverse sample of Chicago adolescents. Data from the Project on Human Development in Chicago Neighborhoods were used to test models that examine relationships between neighborhood-level and individual-level attributes on tobacco-use and alcohol-use frequencies. Our primary goal was to test hypotheses about the invariance of statistical relationships over three waves of data by using equality of regression coefficients tests. We found that the effect of deviant peers decreased over time for alcohol use but was invariant for tobacco use. The effect of self-control on alcohol use was invariant, yet its effect on tobacco use was not invariant and increased significantly by the third wave. Contextual variables were not, for the most part, significantly related to tobacco and alcohol use, and, contrary to our expectations, the effects were invariant over time.