Abuse of cocaine-opiate combinations ("speedballs") reported clinically was simulated in a rhesus monkey model of simultaneous cocaine and heroin self-administration. The reinforcing effects of nine cocaine-heroin combinations (cocaine [0.001, 0.01 and 0.10 mg/kg per injection i.v.] and heroin [0.0001, 0.001 and 0.01 mg/kg per injection i.v.]) were evaluated for 10 days on a second-order fixed ratio of 4 (variable ratio of 16:S) schedule and compared with self-administration of cocaine and heroin alone. Dose-effect curves for cocaine-heroin combinations usually were similar to those for cocaine and heroin alone, and intermediate doses maintained equivalent high levels of drug self-administration. No toxic effects were observed. In drug discrimination studies, cocaine (0.004-1.3 mg/kg) substitution resulted in a dose-dependent generalization to the training dose (0.4 mg/kg i.m.) in all five monkeys. Heroin substitution (0.01-1.0 mg/kg i.m.) resulted in dose-dependent and complete generalization to cocaine in three of five monkeys. Heroin pretreatment (0.1-0.18 mg/kg i.m.) did not change the cocaine discrimination dose-effect curve. Pretreatment with an opiate antagonist, quadazocine (0.1 mg/kg i.m.), had no effect on the discriminative stimulus effects of cocaine but antagonized the cocaine-like discriminative stimulus effects of heroin. Pretreatment with a dopamine antagonist, flupenthixol (0.018 mg/kg), antagonized the discriminative stimulus effects of cocaine but did not affect the cocaine-like effects of heroin. These findings demonstrate the feasibility of maintaining self-administration of cocaine-heroin combinations and suggest that cocaine and heroin have similar reinforcing and discriminative stimulus effects in rhesus monkeys.