This paper represents a comprehensive review of laboratory studies investigating the abuse liability of anxiolytic and hypnotic drugs in humans. The subjective effects of these drugs, most of which are either barbiturates or benzodiazepines, have been measured using various self-report questionnaires and their reinforcing effects have been studied using self-administration and choice procedures. Studies using both subjective and reinforcing effects reveal orderly relations between the two main chemical classes of anxiolytic/hypnotics (e.g. barbiturates are associated with higher abuse liability than benzodiazepines), between different doses of the drugs (e.g. higher doses are usually associated with higher abuse liability) and among different compounds within a class. The subjective and reinforcing effects of barbiturates and benzodiazepines depend critically upon the subject populations that are tested and it is argued that individuals with histories of drug abuse provide a more sensitive indicator of abuse liability than healthy volunteers. Several principles of abuse liability testing are discussed, including the selection of an appropriate subject population, the use of blind drug administration procedures, the comparison of a test compound to an appropriate standard, the inclusion of a placebo and a wide range of doses of the test drug and the use of multiple measures of likelihood of abuse.