Beginning in the summer of 1987, the New York State Department of Health undertook serosurveillance for antibody to HIV by the blind testing of available blood samples. Six populations were selected as "windows" on the epidemic: newborns; homeless adolescents; prisoners; and clients of family-planning, sexually-transmitted-disease, and drug-treatment clinics. For each of these windows the objectives were to establish the baseline prevalence and geographic distribution of infection, monitor trends of infection overall and in various populations, and use the information to target and design preventive activities. Testing of all newborns in New York State began in December 1987, and all six study windows were operational by June 1988. As of March 1990, blind testing for HIV antibody had been performed on more than 725,000 blood specimens in these six studies. The chief limitation in the study is that the groups available for study are determined by the availability of blood specimens. Some population groups, such as adult males, may be difficult to assess, and the study group is not representative of the entire population. Two years of testing, however, have yielded sufficient data for reliable analysis and projection of trends. 1 table and 9 references