SUMMARYEvaluation of teaching by the use of questionnaires to students is now commonplace. If it is to be useful the data obtained must be reliable and valid. One criterion of reliability is the response rate, and a low response rate may indicate low validity, i.e. that the questions asked do not reflect the students’ real concerns. In order to inform questionnaire design a critical incident study was undertaken. A 20% random sample of students in each of the 5 years of the course were asked to describe one piece of good teaching and one piece of bad teaching, and say why they were good or bad. There was a 65% response rate, and replies were independently categorized by three people. The factors identified fell into three ‘domains’: interpersonal behaviour of teachers; planning and preparation; and the ability to run the session well. There was no evidence that teaching which ‘played to the gallery’ or was very examination‐oriented would earn high ratings fr