BACKGROUND: The general proficiency in physical diagnostic skills seems to be declining in relation to the development of new technologies. The few studies that have examined this question have invariably used recordings of cardiac events obtained from patients. However, this type of evaluation may not correlate particularly well with bedside skills. Our objectives were 1) To compare the cardiac auscultatory skills of physicians in training with those of experienced cardiologists by using real patients to test bedside diagnostic skills. 2) To evaluate the impact of a five-month bedside cardiac auscultation training program. METHODS: 1) In an academic primary care center, 20 physicians (trainees in internal medicine and family practice) and two skilled academic cardiologists listened to 33 cardiac events in 13 patients directly at bedside and identified the cardiac events by completing an open questionnaire. Heart sounds, murmurs and diagnosis were determined beforehand by an independent skilled cardiologist and were validated by echocardiography. Thirteen primary cardiologic diagnoses were possible.2) Ten of the physicians agreed to participate in a course of 45-minute sessions once a week for 5 months. After the course they listened again to the same patients (pre/post-interventional study). RESULTS: 1) The experts were the most skillful, achieving 69% recognition of heart sounds and murmurs and correct diagnoses in 62% of cases. They also heard all of the diastolic murmurs. The residents heard only 40% of the extra heart sounds and made a correct diagnosis in 24% of cases. 2) After the weekly training sessions, their mean percentage for correct diagnosis was 35% [an increase of 66% (p < 0.05)]. CONCLUSIONS: The level of bedside diagnostic skills in this relatively small group of physicians in training is indeed low, but can be improved by a course focusing on realistic bedside teaching.