Stimulant drugs have been illicitly consumed for non-medical purposes for centuries. Within the past several decades procedures for assessing the abuse liability of a compound prior to marketing have been developed, which should limit the introduction of new stimulants with high abuse liability to the public. Drugs with high abuse liability are positive reinforcers in that they maintain behavior leading to their consumption. Assessment of abuse liability has traditionally involved a comparison of the profile of effects of a known drug of abuse to that of a drug with unknown abuse liability. With respect to stimulant drugs, amphetamine and cocaine may be considered prototypic stimulants with high abuse liability. Various procedures have been used to study abuse liability including the measuring of (1) subjective effects, (2) drug 'liking', (3) estimates of value, (4) the accuracy of drug identification, (5) the ability of trained subjects to discriminate one drug from another and (6) drug self-administration. In this paper the behavioral effects of a number of stimulant drugs will be compared along these dimensions and the effects along each dimension will be related to each drugs' known abuse liability. Dissociations among the various measures will also be described, and the implications of these dissociations discussed. We conclude that the best single assessment of abuse liability is obtained from drug self-administration studies, but that the most accurate assessment of abuse liability is obtained when the effects of a drug are evaluated along as many of these dimensions as possible.