Self-injection of three barbiturates, six benzodiazepines, and chlorpromazine was examined in baboons. Intravenous injections of drug were dependent upon completion of 160 lever presses (a 160-response fixed-ratio schedule). A 3-h time-out period followed each injection, permitting a maximum of eight injections per day. Prior to testing each dose of drug, self-injection performance was established with cocaine. Subsequently, a test dose was substituted for cocaine. Amobarbital, pentobarbital, and secobarbital maintained the highest levels of self-injection, which were similar to those maintained by cocaine. Clonazepam, clorazepate, diazepam, flurazepam, medazepam, and midazolam maintained relatively modest levels of self-injection, while chlorpromazine maintained only low levels, which were in the range of vehicle control. Of the six benzodiazepines, midazolam produced the highest levels of self-injection. At the highest self-injected doses, the barbiturates produced anesthesia in contrast to the benzodiazepines, which produced only sedation. None of the drugs affected food intake except for chlorpromazine, which produced dose-related decreases. The differences among the drug classes (i.e., barbiturate, benzodiazepine, phenothiazine) with respect to the maintenance of self-injection correspond well with the results of previous animal and human drug self-administration studies.