In 1963–4 the Oxford University Penal Research Unit managed to collect information about 90 per cent of the offender-patients who were admitted to N.H.S. and Special Hospitals under hospital orders made by criminal courts: this cohort has been described in Crime and Insanity, Vol. II, by Walker and McCabe (1973). One of the by-products of the follow-up of these offender-patients was a rough and ready scoring system for predicting reconvictions (within a two-year follow-up) of 456 offender-patients who were allowed to leave hospital within a year of admission. A prediction system can be used to assist human decision-making (though it should not be a substitute for it)∗ and can also be used to assess the efficacy of measures—such as after-care—by seeing whether individuals with equal predicted reconviction-rates do better with than without the measure (Mannheim and Wilkins, 1955).