The authors approximated human immunodeficiency virus (HIV) seroincidence in a population of men who have sex with men and who sought repeated anonymous HIV testing in San Francisco in 1995. The number of seroconversions and person-years of observation were estimated using the date and result of the current test and the self-reported date and result of the previous test. Estimates for HIV seroincidence (2.8 per 100 person-years, 95% confidence interval 2.3-3.4) and predictors of seroconversion were similar to those estimated from a prospective study of men who have sex with men conducted in San Francisco at the same time. While the limitations of self-reported data in a self-selected population are recognized, data from repeat testers may provide a practical surveillance tool.