Although surveillance for HIV infection has traditionally focused on the incidence of AIDS and the prevalence of HIV, new diagnostic technologies that allow the estimation of incident HIV infection have become available. Number and distribution of new cases of HIV infection, rather than old cases, are the data most relevant to guide rational application of HIV prevention programs. Historically, incident HIV infection has been measured in longitudinal cohort studies, diagnosed clinically or since 1993 by detection of seroconverting patients (during the window period before appearance of HIV antibody) who are viremic as measured by p24 antigen or RNA-PCR. The sensitive-less sensitive EIA test (or serologic testing algorithm for recent HIV seroconversion [STAHRS]) has now made the serologic diagnosis of incident HIV infection in individual patients as well as the estimation of HIV incidence in populations possible. Examples of the public health application of this are studies of HIV incidence in anonymous test site attendees, sexually transmitted disease clinic patients, and in-treatment injection drug users in San Francisco. These sorts of studies allow us not only to measure incidence cross-sectionally but also facilitate surveillance for HIV subtypes and primary antiretroviral resistance, targeting early antiretroviral therapy and partner notification, and understanding who is "failing" prevention. Having an HIV surveillance system that focuses on incident rather than prevalent infection should be our long-term goal.