Much has been said and written about the logical status of diagnoses; their reliability and validity have often been measured and questioned; and several mathematical models of the diagnostic process have been developed for use in computers. But surprisingly little interest has been taken in the practical aspects of diagnosis as a decision-making process. A large and undistinguished literature on ‘psycho-diagnostics' was generated by clinical psychologists in the 1950s, but this was largely concerned with diagnoses derived from cognitive and projective test results and so has little relevance to diagnosis by the traditional medical method of asking the patient about his symptoms. More recently a few people, like Gauron and Dickinson in Iowa and Sandifer in North Carolina, have studied the diagnostic activity of psychiatrists in specially designed experimental situations, and their findings have been both interesting and instructive. But we are still uncertain at what stage in a diagnostic interview a provisional diagnosis is usually arrived at, or how often this initial impression is correct, or is changed later on. Nor do we know the relative importance of the patient's behaviour and the factual information he provides in reply to questions, or what differences exist between psychiatrists in the means by which they reach a diagnosis, or in the speed and confidence with which they do so. The studies described here were designed with the aim of shedding some light on these matters.