To examine whether recent intravenous use of cocaine might be associated with increased risk of human immunodeficiency virus type 1 (HIV) infection, the authors studied 2,597 active intravenous drug users: 2,399 with recent cocaine injection and 198 with recent injection of heroin or other drugs but not cocaine. These subjects were adult residents of Baltimore City and the surrounding Maryland counties, recruited via outreach into the community between February 1988 and March 1989. In contrast to the first report on the cocaine-HIV association, the present study sample was not recruited solely from drug treatment programs. In the present study, estimated HIV seroprevalence was 26.4% for recent cocaine injectors as compared with 10.6% among all other recent intravenous drug users; the relative odds estimate was 3.03. In the untreated segment of the sample, HIV seroprevalence was 26.0% for recent cocaine injectors as compared with 8.9% among others (relative odds (RO) = 3.61). The estimated degree of association did not change appreciably when multiple logistic regression was used to hold constant potentially confounding and/or mediating variables such as receptive anal intercourse, number of sex partners, and use of injection equipment obtained at shooting galleries (RO = 2.64). Augmenting these cross-sectional data, preliminary prospective data showed excess risk of HIV seroconversion among recent cocaine injectors (estimated relative risk = 2.11). While other research has examined the cocaine-HIV association, the present study differs in that it has allowed a test for whether the association was a spurious artifact of studying drug users recruited solely from drug treatment programs, a broad array of alternative determinants of HIV infection have been held constant, and the association has been examined with seroconversion data. The results lend support to the abiding concern about the risk of HIV infection among cocaine users.