OBJECTIVES: Avoidable breaches of patient confidentiality due to inadvertence or ignorance occur frequently. The aim of this study was to explore whether participants were able to identify violations of confidentiality and how serious these violations were. DESIGN: After a 2-hour theoretical education session, participants filled out a standardised questionnaire presenting 6 hypothetical cases (e.g. politician's illness mentioned to the doctor's family, violence inflicted by the police mentioned to a lawyer at a dinner party, both without patient consent) and asking whether confidentiality had been violated. Answers were given on a score of 0-3, where 0 = no violation and 3=serious violation. SETTING: University of Geneva, Switzerland. PARTICIPANTS: Three professors of criminal law (representing the gold standard) and convenience samples of 311 law, medical and dental students took part. Main outcome measures Students' answers were compared to the gold standard (answers of criminal law professors). RESULTS: According to our gold standard, a violation of confidentiality took place in all 6 cases (scores: 3, 2, 3, 2, 3, 3; means of students: 1.7, 0.6, 1.0, 0.4, 1.5, 2.6). Most students (88%) mistakenly believed in 1-3 cases that no violation took place. In only 2 cases did more than 90% of participants recognise the violation. Law students attributed significantly lower scores than medical or dental students in 3 cases and higher scores in 1 case. CONCLUSIONS: Despite theoretical education, participants did not fully understand obligations towards patient confidentiality when it came to practical situations, especially when colleagues and authorities (police, those in a judicial context) asked for information. Teaching should address the risk of patient identification and the seriousness of violations in order to motivate doctors to avoid unjustified violations of confidentiality.