Ten healthy adult male research volunteers reporting smoked and i.v. cocaine use resided on a Clinical Research Center for 2 weeks and participated in nine daily sessions. A session consisted of seven choice trials. The first two trials were sampling trials in which subjects received one dose each of i.v. cocaine hydrochloride (0, 16 and 32 mg) and smoked cocaine base (0, 25 and 50 mg). Each of the remaining five trials was a choice trial in which subjects could choose to self-administer either of the doses from the initial two trials. Subjects 1) reliably chose active doses of cocaine compared to placebo, 2) chose to self-administer the low-smoked cocaine dose about as often as either the low or high i.v. cocaine doses and 3) reliably chose the high-smoked cocaine dose when compared to either active i.v. dose. With few exceptions, both low doses and high doses produced similar subjective and cardiovascular effects after the initial dose, regardless of the route of administration. This suggests that initial effects were not predictive of subsequent choice. Cumulative doses of smoked cocaine increased scores on a number of subjective-effects measures that were not similarly increased by cumulative doses of i.v. cocaine. These differences were predictive of smoked cocaine self-administration. After session ratings of drug "Liking" and "Quality" differentiated smoked from i.v. cocaine, reflecting route choice. However, there were no significant differences between these ratings for low and high doses. These results provide information about the relationship between subjective drug effects and drug self-administration, and demonstrate the utility of a choice procedure in analyzing these relationships.