OBJECTIVE: To determine factors affecting compliance with screening mammography prescribed by family physicians. DESIGN: Secondary analysis of a nonrandomized trial. SETTING: University-affiliated family medicine clinic in Montreal. PATIENTS: Women aged 50 to 69 years who were given a written prescription for a screening mammography during their visit at the clinic between Oct. 12, 1991, and May 31, 1992, and who had not undergone mammography in the preceding 2 years and had never been treated for breast cancer. Information on the potential factors was obtained through a telephone questionnaire 2 months after the visit. OUTCOME MEASURES: Indicator of compliance presence of result of screening mammography in patient chart, potential factors influencing compliance: age, level of education, marital status, socioeconomic level, smoking status, perceived health status, perceived psychological well-being, risk factors for breast cancer, use of health services including frequency of Papanicolaou test, Health Belief Model variables. RESULTS: Of the 171 eligible women, 113 (66.1%) underwent the prescribed mammography within 2 months after the visit to the clinic, and 149 (87.1%) responded to the questionnaire. The patients' socioeconomic characteristics, perceived health status, health utilization indices and risk factors for breast cancer were not found to be predictors of compliance. The strongest predictor of compliance was the number of previous mammograms. Women who had undergone mammography previously were less likely to be noncompliant than those who had not (odds ratio [OR] 0.11, 95% confidence interval [CI] 0.02 to 0.51; p = 0.005). Women who did not comply were less likely than those who did to believe that a prescription from their physician would convince them to undergo mammography (OR 0.21, 95% CI 0.007 to 0.60; p = 0.004). Other factors associated with noncompliance were the expression of fear of mammography (OR 2.09, 95% CI 1.08 to 4.02; p = 0.03) and the lack of time to take the test (OR 3.07, 95% CI 1.21 to 7.80 p = 0.02). Being a smoker was negatively associated with compliance (OR 0.43; 95% CI 0.22 to 0.86; p = 0.02). The stepwise logistic regression model accounted for 87.5% of the outcome (chi2 for goodness of fit = 164.4; p = 0.0001). CONCLUSION: Family physicians who prescribe screening mammography, even to women who consult for other reasons, are likely to overcome some of the barriers observed in association with population screening rates. However, physician-oriented approaches are not likely to reach the 30% to 40% of reluctant women who appear to hold negative views toward physicians' recommendations. Further study is necessary to determine how better to reach these women.