Results of previous laboratory studies in humans suggest that the benzodiazepine diazepam has greater abuse liability than the benzodiazepine oxazepam. The validity of these laboratory-based experimental data were examined by analyzing Swedish data on drug abuse. Sales and prescription data showed that use of diazepam was somewhat lower than oxazepam, but of the same general order of magnitude (0.8:1). Prescription data showed that the drugs were prescribed for the same diagnostic indications. After adjustment for differences in use, 'prescription forgeries' and 'mentions in theft and loss reports' were found to be more frequent for diazepam than for oxazepam (2.3:1 and 2.5:1 for forgeries and theft/loss reports, respectively). This effect was consistent for each year examined (1982, 1983, 1984) and occurred when the data were recalculated to exclude Valium, the original and most widely known brand of diazepam. Finally, this pattern with prescription forgeries occurred across different geographical regions in Sweden (1982, 1983).