Audit procedures for mammography interpretation in a national breast cancer screening program are described. Agreement between radiologists at five screening centers and a single reference radiologist was evaluated based on mammograms at initial screen in 739 women not known to have breast cancer and 204 breast cancer cases detected at or within a year of initial screen. Kappa statistics, used to measure agreement in the interpretation of mammograms between the reference and center radiologists beyond that attributable to chance, were 0.409 in women not known to have breast cancer, 0.472 in women with diagnosed breast cancer, and 0.493 for all women with p less than .0001 for all three categories. To our knowledge, this is the first report presenting measures of agreement made systematically, simultaneously, and on a continuing basis, during a screening program for quality assurance.