The forecasts had all been produced by September 1988. Definitional problems and the lack of crucial studies contributed to an underrepresentation of the epidemic in the national AIDS surveillance data and in most forecasts. A more realistic estimate is that from 300,000 to 480,000 cases will have occurred by the end of 1991, compared to the current best estimates of 120,000 to 400,000 cases. Each forecasting model has problems in comprehensiveness, empirical basis, or assumptions. These problems mean that only about two-thirds of all cases of AIDS and other fatal HIV-related illnesses were captured in the data underlying existing national forecasts. These forecasts should be adjusted upwards by an estimated 50 percent, but the most recent forecast of the Centers for Disease Control adjusted the data by only 19 percent. Figures, tables, footnotes, appended methodological information, and 88 references.