Two formats of case presentation are traditionally used for teaching problem-solving skills: clinical vignette or chief complaint formats. While the first one is more commonly used, it does not completely reflect the actual problem-solving process during a real encounter, which may hamper the learners to integrate separately acquired data gathering skills into their reasoning process and affect their diagnostic performance in practice. The present study compared diagnostic accuracy when the reasoning stimulus was a case vignette containing all diagnostic information versus the patient's chief complaint only. Forty-two medical students, 53 residents and 60 general internists participated in the study. Diagnostic accuracy was significantly lower for the chief complaint format at the student, resident, and practitioner levels. Analysis of the data gathered in the chief-complaint format revealed that faulty diagnostic decisions resulted from a failure to gather critical data. The results suggest that data gathering techniques, semiology, and medical reasoning should be trained in association and that this effort should be pursued beyond medical school.