Three types of experimental studies are reviewed: (1) intravenous and oral caffeine self-administration by laboratory animals, (2) oral caffeine self-administration by humans, and (3) human subjective effects of caffeine relevant to reinforcing effects. These studies show that, under appropriate conditions, caffeine can serve as a reinforcer and can produce elevations in subjective drug liking and/or euphoria. In this regard, caffeine can be distinguished from a wide range of behaviorally active compounds, such as the amphetamine analog fenfluramine and the major tranquilizer chlorpromazine, which do not produce such effects. Caffeine can also be distinguished from classic drugs of abuse such as cocaine, d-amphetamine or pentobarbital which generally maintain high levels of self-administration (or liking) in contrast to caffeine which tends to maintain lower levels of self-administration (or liking) or maintain self-administration under a more narrow range of parametric conditions. Several human studies and one animal experiment suggest that physical dependence substantially potentiates the reinforcing effects of caffeine. Other human and animal studies indicate that there may be substantial differences between individual subjects in the reinforcing effects of caffeine. An important challenge for future human and animal drug self-administration research will be to delineate more precisely the conditions under which caffeine does and does not serve reliably as a reinforcer.