Fixed genetic differences between marijuana and nonpsychoactive hemp were identified, demonstrating the ability of AFLP to separate cultivars. Analysis of molecular variance in four cannabis populations indicated that 27.2 percent of the genetic variation is attributable to differences between marijuana and hemp lines, 20.9 percent to differences among hemp varieties, and 51.9 percent to variation within populations. Some of the fixed genetic differences between drug and fiber lines detected in this study could be linked to THC production, which should be examined in future research. Genetic markers that distinguish hemp and marijuana have practical use for drug enforcement in Canada and Europe, where hemp cultivation is permitted but marijuana cultivation is not. In the United States, where all cannabis plants are controlled substances, the technique has the potential to connect marijuana samples with source populations. For the purposes of this study, seed of the drug variety Skunk #1 and the hemp variety Carmen were imported under permit from the U.S. Drug Enforcement Administration. Thirteen Carmen plants from stock seed, 6 Carmen plants after a generation of inbreeding, and 12 Skunk #1 plants after a generation of inbreeding were screened for AFLP variation. In addition, seeds from two naturalized hemp populations were collected in Shakopee and Minneapolis, MN. Seeds were germinated in the laboratory, and 14 plants were sampled from each population. The repeatability of AFLP profiles was examined by estimating error associated with restriction-ligation, preselective amplification, and selective amplification from cannabis DNA accessions fingerprinted with a single primer pair. 4 tables, 1 figure, and 29 references