In animals, self-administration procedures have been used to assess reinforcing properties of drugs and predict abuse liability. In humans, subjective effect have been used. In the present studies, there two approaches were combined in human subjects. During the first four sessions of a nine-session experiment, subjects were given opportunities to ingest two drugs. After ingestion, subjects were free to leave. Mood questionnaires were filled out before ingestion and 1, 3, and 6 h later. During the last five sessions, subjects chose which drug to ingest. In study I, subjects chose 5 mg d-amphetamine over placebo. Subjective effects were typical of amphetamines but were not correlated with amount of amphetamine preference. In study II this same comparison was repeated three times. The preference for amphetamine decreased but subjective effects did not change. In study III, subjects tended to prefer 5 mg d-amphetamine to 20 mg fenfluramine. Subjective effects after amphetamine were similar to those in study I but those after fenfluramine did not differ from placebo. In study IV, subjects preferred placebo to 5 or 10 mg diazepam. The subjective effects of diazepam were typical of sedatives. These results indicate that drug self-administration procedures with humans can be useful for testing abuse liability and that subjective measures do not always correlate with drug preference.