Previous studies have pointed out the question of effective training and information to health professionals on pharmacovigilance. The lack of training is known to induce inadequate use of drugs and noncompliance of patients. Pharmacology teaching is performed in the third year of medical studies at the Toulouse Faculty of Medicine. The aim of the study was to investigate the perception of risk of adverse drug reactions (ADRs) by medical students at the end of the one year pharmacology course and two years later, after clinical training period. Sixty-seven students were interviewed in May 2005 and in October 2007. Visual analogue scales were used to define a score of perceived risk of ADRs associated with each drug class (ranking from 0 to 10) before and after pharmacology training. The drug classes evaluated were antibiotics, anticoagulants, antidepressants, aspirin, contraceptive pill, corticosteroids, drugs for arterial hypertension, drugs for diabetes (other than insulin), hypnotics, hypocholesterolaemic drugs, nonsteroidal anti-inflammatory drugs (NSAIDs), postmenopausal hormone replacement therapy and tranquilisers. After pharmacology courses (May 2005), antidepressants were ranked as the most dangerous drugs by medical students [median score (25th–75th centiles): 7.7 (6.3–8.6)], followed by anticoagulants [7.6 (6.6–8.4)] and hypnotics [7.4 (6.1–8.1)]. Contraceptive pills was listed in the last position [median score [4.7 (2.2–6.7)]. Two years later (October 2007), anticoagulants moved into the first position [8.3 (7.3–9.2)], followed by NSAIDs [6.9 (5.0–8.4)] and aspirin [6.8 (5.8–8.4)]. Contraceptive pills remained in the last position. No change was observed for NSAIDs and aspirin, still ranked as dangerous drugs by medical students after clinical training. Values of perceived risk were significantly increased for anticoagulant (+9.2%, P < 0.01) and hypoglycemiant drugs (+27.7%, P < 0.0001). The perceived risk significantly decreased for hypocholesterolaemic (−14.3%, P < 0.0001) and antidepressant drugs (−19.5%, P < 0.0001), but not for hypnotics. The study shows that the perception of risk of ADRs by medical students was modified after clinical training. They were still aware of potentially serious ADRs associated with anticoagulants, aspirin or NSAIDs, but they remained less cautious for drugs such as antidepressants. Additional pharmacology training at the end of medical studies will be useful.