SummaryTimely estimates of HIV incidence are needed to monitor the epidemic and target primary prevention but have been difficult to obtain. We applied a sensitive/less-sensitive (S/LS) enzyme immunoassay (EIA) testing strategy to stored HIV-positive sera (N= 452) to identify early infections, estimate incidence, and characterize correlates of recent seroconversion among persons seeking anonymous HIV testing in San Francisco from 1996 to 1998 (N= 21,292). Sera positive on a sensitive EIA but negative on a less-sensitive EIA were classified as early HIV infections; sera positive on both EIA were classified as long standing. Seventy-nine sera were from people with early HIV infection. Estimated HIV incidence was 1.1% per year (95% confidence interval [CI], 0.68%–1.6%) overall and 1.9% per year (95% CI, 1.2%–3.0%) among men who have sex with men (MSM). Early HIV infection among MSM was associated with injection drug use, unprotected receptive anal sex, and multiple sex partners in the previous year. No temporal trend in HIV incidence was noted over the study period. The S/LS strategy provides a practical public health tool to identify early HIV infection and estimate HIV incidence in a variety of study designs and settings.