To determine whether the excess prevalence of human immunodeficiency virus type 1 (HIV-1) infection in US black and Hispanic homosexual men relative to white men can be explained by differences in sociodemographic factors, history of sexually transmitted diseases, or sexual and drug-use behaviors, the authors conducted a cross-sectional analysis of baseline HIV-1 seroprevalence and HIV-1 risk factors among 4,475 non-Hispanic white, 234 Hispanic white, and 194 black homosexual men from four centers in the United States (Baltimore/Washington, DC, Pittsburgh, Chicago, and Los Angeles). HIV-1 seroprevalence was significantly higher in Hispanic men (50%; odds ratio (OR) = 1.83,95% confidence interval (Cl) 1.41–2 39) and black men (47%; OR = 1 62,95% Cl 1 21–2.16) compared with white men (35%). Both Hispanic and black men more frequently reported a history of sexually transmitted diseases Overall, Hispanics had the highest risk profile and blacks the lowest risk profile with respect to certain high-risk sexual behaviors (e.g , receptive anal intercourse and use of anonymous sexual partners) and recreational drug use. After multivariate adjustment, black race remained a significant independent risk factor for HIV-1 seropositvrty (OR = 1.60, 95% Cl 1.13– 2.26), but Hispanic ethnicity was no longer statistically significant (OR = 1.17, 95% Cl 0.82=1.69). Most of the excess HIV-1 prevalent infection among Hispanics was explained by their predominant recruitment from Los Angeles—the study center with the highest HIV-1 seroprevalence—and their greater prevalence of a history of sexually transmitted diseases and certain high-risk sexual practices. By contrast, adjustment for these same risk behaviors failed to explain the observed black-white differences in HIV-1 seroprevalence, and further studies are needed to elucidate the reasons for these unexplained racial differences. HIV-1 educational programs for homosexual men should take into account the behavioral differences that exist between white and minority racial/ethnic groups. Am J Epidemiol 1993;138:415–29 .