The medical records of 243 asymptomatic women aged 50 years or older were reviewed at a community-based family practice center to determine the proportion who had been referred for a screening mammogram and to identify correlates of mammography referral. Patient demographic characteristics, breast cancer risk factors, and characteristics of past patient-physician encounters were considered. Between July 1, 1981, and July 1, 1987, 40 (16 percent) of the women had received a mammography referral from their currently assigned physician. All but two of the women had actually obtained the mammogram. The primary predictors of mammography referral were the known risk factors for breast cancer: a family history of breast cancer (prevalence rate ratio [PRR] = 9.3, P = .001) and a history of benign breast disease (PRR = 7.9, P = .002). Other predictors included having a Papanicolaou test performed by the current physician (PRR = 4.1, P = .03), having a test for stool occult blood returned by the patient (PRR = 10.2, P = .003), having been instructed in smoking cessation by the current physician (PRR = 10.0, P = .05), and, possibly, being a former smoker (PRR = 4.6, P = .09). Patient demographic characteristics, other known breast cancer risk factors (age, obesity, alcohol use, and pregnancy history), and the sex of the physician were not predictive.