Decriminalization is only one of a number of process variables that may impact upon variables associated with marijuana use. Other independent and mediating variables may be enforcement procedures, changes in court processing, mass communication themes, and dealer sales pitches. Consideration of all possible variables that might influence the dependent variable is necessary in order to isolate the explanatory power of the independent variable. Dependent variables should be directly related to drug use. The unit of analysis used in the independent variables is important. The unit of analysis may be individuals randomly sampled, high-risk groups, supply systems, or criminal justice components. The choice of the unit of analysis must match the level of behavioral explanation desired. The choice of intraindividual versus interindividual design for evaluating any chosen dependent variable is also crucial. Whatever specific approach is taken in theorizing and the consequent selection of independent and dependent variables, difficult problems of statistical control must be faced, because the researcher is operating in a nonexperimental context where there are many threats to validity. There should be a model for the process of change. Such a model would be used to attempt to explain all the data, including shifts and nonshifts in behavior. This would provide clues to possible threats to validity that should be controlled by statistical methods. Twenty-four notes are provided.