In order to characterize the discriminative stimulus properties of cyclazocine squirrel monkeys were trained to discriminate between intramuscular injections of 0.1 mg/kg of cyclazocine and drug vehicle in a discrete-trial avoidance paradigm in which a response on one of two levers would prevent or terminate the delivery of a mild electric shock to the tail. Behavior was considered to be under stimulus control when the monkeys completed at least 22 trials of a 25-trial session on the appropriate choice lever after they received cyclazocine or vehicle. Stimulus generalization (i.e., dose-response) curves were then determined for a variety of drugs over a broad range of doses. The following analgesics with narcotic antagonist properties produced stimulus control of behavior comparable to that produced by 0.1 mg/kg of cyclazocine: ketocyclazocine (0.1 mg/kg), butorphanol (1.0 mg/kg), oxilorphan (3.0 mg/kg) and levallorphan (3.0 mg/kg). Time course experiments revealed that the duration of complete stimulus control was relatively short: 1 hour after 0.1 mg/kg of cyclazocine and 3.0 mg/kg of oxilorphan and 0.5 hour after 0.1 mg/kg of ketocyclazocine. Naloxone produced a dose-related antagonism of the stimulus control by cyclazocine and butorphanol which was complete at 1.0 mg/kg. In contrast, as much as 10 mg/kg of naloxone only partially blocked stimulus control by ketocyclazocine and failed to modify stimulus control by oxilorphan and levallorphan. The monkeys showed partial generalization to other drugs with activity as narcotic agonists, antagonists or both (e.g., morphine, naloxone, pentazocine, nalbuphine, nalmexone and nalorphine), but little or no generalization to the nonopioid psychoactive drugs, d-amphetamine, mescaline, pentobarbital and scopolamine. This discrimination paradigm appears to have potential as a model for the quantitative assessment of the discriminative stimulus properties of narcotic antagonist analgesics.