Ten cocaine-dependent participants were trained to discriminate between intravenous saline and 20 mg/70 kg cocaine. During the first session, saline and cocaine injections were alternated twice, with each separated by 1 hr. The injections were identified by letter codes. During the next 3 sessions, 12 trials were conducted, with saline and cocaine administered 6 times each in pseudorandom order. Thirty minutes following each injection, participants were asked to identify the injection by letter code. Seven of the 10 learned the discrimination (at least 10 trials correct). To evaluate sensitivity, the investigators tested participants with different doses of cocaine in test sessions. In the next phase, methamphetamine (5 and 10 mg/70 kg) and pentobarbital (50 and 100 mg/70 kg) were given intravenously during test sessions to determine whether the discrimination exhibited pharmacological class selectivity. During the evaluation of sensitivity and selectivity, training sessions were interspersed. As dose of cocaine increased, the number of participants identifying the test dose as cocaine increased, demonstrating sensitivity. The higher doses of methamphetamine and pentobarbital substituted for cocaine. The physiological and subjective effects of cocaine and methamphetamine were stimulant-like and dose related. Pentobarbital produced no physiological changes but increased Visual Analog Scale ratings of Sedation, Good Drug Effect, and High. This failure to demonstrate pharmacological selectivity may be related to participants' learning a drug-vs.-no-drug discrimination, and thus it may be necessary to alter training procedures in future studies.