Clinical evaluation is the cornerstone of any cardiac diagnosis, although excessive over-specialisation often leads students to disregard the value of clinical skills, and to overemphasize the approach to instrumental cardiac diagnosis. Time restraints, low availability of "typical" cardiac patients on whom to perform effective bedside teaching, patients' respect and the underscoring of the value of clinical skills all lead to a progressive decay in teaching. Simulation-guided cardiac auscultation may improve clinical training in medical students and residents. Harvey(©) is a mannequin encompassing more than 50 cardiac diagnoses that was designed and developed at the University of Miami (Florida, USA). One of the advantages of Harvey(©) simulation resides in the possibility of listening, comparing and discussing "real" murmurs. To objectively assess its teaching performance, the capability to identify five different cardiac diagnoses (atrial septal defect, normal young subject, mitral stenosis with tricuspid regurgitation, chronic mitral regurgitation, and pericarditis) out of more than 50 diagnostic possibilities was assessed in 523 III-year medical students (i.e. at the very beginning of their clinical experience), in 92 VI-year students, and in 42 residents before and after a formal 10-h teaching session with Harvey(©). None of them had previously experienced simulation-based cardiac auscultation in addition to formal lecturing (all three groups) and bedside teaching (VI-year students and residents). In order to assess the "persistence" of the acquired knowledge over time, the test was repeated after 3 years in 85 students, who did not repeat the formal 10-h teaching session with Harvey(©) after the III year. As expected, the overall response was poor in the "beginners" who correctly identified 11.0 % of the administered cardiac murmurs. After simulation-guided training, the ability to recognise the correct cardiac diagnoses was much better (72.0 %; p < 0.001 vs. baseline). Rather unexpectedly, before the tutorial, the performance of VI-year students and of residents was not significantly different from their III-year colleagues, since the two groups correctly identified 14.2 and 16.2 % of the diagnoses, respectively. After the tutorial, the VI-year students and the residents also improved their overall performance (to 73.1 and 76.1 %, respectively; p < 0.001 for both when compared to before the tutorial). The persistence of this capability after 3 years was remarkable, since the 85 students who repeated the test without any further exposure to the 10-h teaching session with Harvey(©) correctly identified 68.4 % of the possible cardiac diagnoses (p < 0.001 vs. baseline). These data underscore the importance of clinical training in order to improve auscultation skills in our academic setting, prompting to redesign teaching curricula. Simulation-based cardiac auscultation should be considered as the "missing link" between formal lecturing and bedside teaching of heart sounds and murmurs.