The present study was designed to assess the preferred methods of treatment of breast cancer by American oncologists, and the impact of clinical trials on their practice. We mailed 465 questionnaires to surgical, radiation, or medical oncologists. The questionnaire described five hypothetic patients with breast cancer, and respondents were asked to select their preferred treatment for each patient. For primary breast cancer, most physicians would offer the choice of local excision followed by radiation therapy or modified radical mastectomy. About 80% of physicians would prescribe adjuvant chemotherapy for a premenopausal woman with estrogen receptor-negative, axillary node negative breast cancer, and for a postmenopausal woman with estrogen receptor-negative, node-positive disease. This policy was favored by male and female physicians of each specialty. Almost all respondents would treat a young woman with inflammatory breast cancer with initial chemotherapy followed by radiation and/or surgery, and about 60% would recommend chemotherapy to a postmenopausal patient with estrogen receptor-negative disease and minimally symptomatic bone metastases. Clinical trials have compared treatment strategies that could be applied to patients described in our questionnaire. Preferred treatments for primary breast cancer, and for inflammatory breast cancer are supported by the results of clinical trials. Recommendation of adjuvant chemotherapy for node-negative breast cancer is not based on a consistent demonstration of improvement in survival, although randomized trials with short follow-up have shown delay to recurrence. Recommendation of adjuvant chemotherapy for a postmenopausal woman with node-positive breast cancer is contrary to the results of large randomized controlled trials (and to a meta-analysis), which have shown that this policy does not lead to improved survival. Our report suggests that even large randomized clinical trials may have a minimal impact on practice if their results run counter to belief in the value of the treatment.