An outline is provided of an experimental study of the classification of mental ill-health presented to general practitioners (GPs). Videotaped real-life consultations were assessed by 27 GPs for the purpose. A high degree of inter-observer variation was recorded. Some of the implications of the findings are discussed.