In addition to the usual measures of screening-test performance, it is important to consider testing frequency when evaluating a screening program. Data on which to base recommendations for the timing of screening tests are urgently needed. For example, in the cases of cervical and colon cancer, when the target is a precursor lesion, research indicates that less frequent screening may be appropriate. This finding may not apply, however, to screening for breast cancer by mammography, which requires currently recommended intervals for the early detection of malignancies. Resources now allocated to breast cancer might more effectively be applied to the construction of tests that would permit longer intervals between screenings. To achieve the National Cancer Institute's goal of reducing cancer mortality in the United States by the year 2000, it will be important to review the balance between population coverage and individual screening for each cancer and to emphasize prevention strategies that maximize population coverage while minimizing expenditures.