The design of a longitudinal study related to marijuana decriminalization should yield reasonable estimates of the incidence and prevalence of marijuana use and data on aggregate 'growth curves' for marijuana use. The first studies should probably be sequential cross-sections rather than longitudinal repeated measures. This will permit determination of the proportion of the population currently using marijuana, the proportion which has ever used it, etc., at a relatively low cost of time and money. It may well be useful in these studies to experiment with ways of collecting retrospective data on use patterns. Samples restricted to relatively young people, even those of high school age and below, are appropriate and will generate more useful data than samples of a wider age range. With adequate data on rates of use and rates of aggregate change, researchers will be in a better position to design a good longitudinal study of the determinants of drug use patterns. The content of the survey should focus on the measurement of drug usage and information that would permit the testing of one or two relatively precise theories of the determinants of drug use. Seven notes are provided. (Author summary modified)