In psychiatry, generalization is a process which requires careful handling. It must be firmly rooted in everyday experience of the practical empiricism which gets results, and must be free, as far as possible, from the numerous fallacies which have brought “armchair” philosophizing into disrepute. These are formidable requirements, but fortunately one way of meeting them lies ready to hand. There are doubtless many others apparent to more fertile minds, but the one with which this paper deals rests on the fact that a patient who has been successfully treated by a psychiatrist behaves, when well again, in a different manner from that which he displayed when ill. He is cheerful, not depressed, relaxed, instead of being tense; realistic and socially enterprising as opposed to being autistic. If the term “Behaviour” is used in a sufficiently wide sense, it can cover the great majority of psychiatric phenomena without losing its cardinal virtue for the present purpose, namely that behaviour can be observed and its manifestations measured with considerable accuracy. The products of observation can be handled statistically, thereby avoiding armchair theorizing. Admittedly the interpretation of statistically derived findings presents some difficulty, but this is not insuperable. It is usually easier than the process of generalizing from observed data without the benefit of statistical pre-digestion.