Seven male research volunteers reporting intravenous cocaine and morphine use participated in 12 cocaine self-administration sessions. A session consisted of sublingual buprenorphine pretreatment (0, 2, 4mg) 50min before seven self-administration trials. The first three trials were sampling trials, in which subjects were given each of the three reinforcers available for that session; the remaining four trials were choice trials in which subjects could choose between two doses of cocaine and tokens exchangeable for special privileges. Each buprenorphine dose was tested for four consecutive days under different cocaine dose/token combinations (e.g. 8mg vs 16mg vs two tokens). Following placebo, under all choice conditions, subjects reliably chose higher doses of cocaine over lower doses, and tokens about as often as high cocaine doses. Buprenorphine (4mg) significantly decreased high-dose cocaine choice and increased token choice, when the available cocaine doses were 16 and 32mg/70kg. Both buprenorphine doses significantly increased ratings of "sedated" and opiate symptoms and decreased ratings of "bad drug effect" following all cocaine doses. Both buprenorphine doses, in combination with cocaine doses greater than 8mg, significantly increased self-reported "high" and occasionally increased "stimulated" scores above values seen following cocaine alone. The results demonstrate that in the presence of non-drug alternative to cocaine (tokens), buprenorphine decreased high-dose cocaine choice. Despite the decrease, subjects were clearly intoxicated from a drug combination that mimicked a cocaine-heroin ("speedball") effect. Although buprenorphine may be useful in reducing cocaine use, the complex relationship between this reduction in use, behavioral alternatives, and the subjective effects of the drug combination remains unclear.