Objective To examine the prevalence and predictors of repeat HIV testing. Design, setting and participants Cross-sectional data from two random household- based and two bar-based samples of gay/bisexual men in two medium-size cities (Tucson, Arizona and Portland, Oregon) with substantial numbers of AIDS cases, in 1992 (n = 2602). Main outcome measure The prevalence and predictors of repeat testing among men who reported being HIV-tested at least once but not being HIV-positive (n = 1583). Results In total, 51% of the sample had been tested three or more times, and 15% were tested more than once every 6 months. Men with higher risk were more likely to be repeatedly tested, although oral but not anal risk was a significant predictor of repeat testing in regression analyses. Men who did not know the HIV status of their primary partner were less likely to be repeatedly tested. Men who perceived that social norms favored secondary prevention, specifically adherence to medical recommendations for the treatment of HIV infection, and who communicated more often about testing were more likely to be repeatedly tested. Conclusions Policy and clinical recommendations for repeat testing must be based on consideration of the complexity and multi-faceted nature of repeat testing. For some individuals, repeat testing may play a legitimate role in HIV prevention by reinforcing safe behavior and providing confirmation of HIV-negative status. However, for others repeat testing may indicate a need for different or more intensive interventions to encourage safe sex.