In a second-year family medicine course taught using simulated patients the students commented most favourably on the clinical relevance of the topics, the enthusiasm of teachers, and the opportunity to interview simulated patients with their fellow students, in contrast to their basic science courses, which did not give them patient contact. They felt most confident about skills acquired in relation to diseases with a limited number of key symptoms, signs and treatments (meningitis, otitis) and less confident about diseases with many symptoms and treatments (diabetes, trauma, arrhythmias). They made few comments about alternative cases which might have been selected. Their adverse comments were about the workload. During tutor meetings over 4 years, a key concern which emerged was to find cases with a level of complexity suitable to the students. The tutors emphasized these principles of case selection: the cases should be based on real life and include most of the signs and symptoms of the disease; contain one or at most two foci; have nodal decision-making points; emphasize clinical reasoning; reinforce prior knowledge; permit the transfer of knowledge to other cases; and permit the assessment of associated technical skills.