Low, nonreinforcing doses of heroin have been shown to shift the dose-response function of cocaine leftward in rhesus monkeys trained under a progressive-ratio schedule of i.v. drug injection. Our study sought to determine 1) whether a reciprocal enhancement of heroin self-administration would be observed when heroin was combined with low, nonreinforcing doses of cocaine, and 2) whether self-administration of cocaine-heroin combinations could be antagonized by the opioid antagonist naltrexone. Rhesus monkeys (n = 4) were prepared with i.v. catheters and trained to self-administer cocaine under a progressive-ratio schedule. The initial response requirement of this schedule was fixed-ratio 120, which doubled across the session to a maximum of 1920. Injections were separated by a 30-min time out. Cocaine dose-response functions (6.4-100 micrograms/kg/injection) for injections/session and breakpoints were monophasic, i.e., increased with dose until responding reached a maximum. Heroin dose-response functions (1.6-25 micrograms/kg/ injection) either increased to a peak and then decreased or reached an asymptote. When nonreinforcing doses of cocaine (3.2-25 micrograms/kg/injection) were combined with heroin, the heroin dose-response function was shifted to the left, without change in maximum injections/session. Pressession treatments with naltrexone (3.2-1600 micrograms/kg, i.m., 10-min presession) antagonized self-administration of heroin and heroin + cocaine combinations in a dose-dependent fashion. However, naltrexone treatment had no effect on cocaine self-administration. Antagonism by naltrexone of self-administration of heroin and heroin + cocaine was surmounted by increasing the dose of heroin either alone or in the heroin + cocaine combination. In vivo apparent pA2 and pKB analyses of these data revealed values of approximately 8.0, consistent with a role for mu opioid receptors in the self-administration of heroin and cocaine-heroin (i.e., "speedball") combinations.