One of the major dilemmas facing the development of new social work practice is that any new approach cannot readily demonstrate its effectiveness by the methods normally available to clinical practice. In clinical practice, even allowing for the ethical problems involved, it is routine to establish trials with matched samples and random allocations. However, social work is much more methodologically complex since by definition it is attempting to respond to users, within their environment and the interaction between the user and their environment. This means that the number of variables involved are beyond management in any ‘experimental’ situation. This is never more true than in the field of work with offenders, since in this work the ‘user’ is rarely if ever defined by the social work agency but can be defined by the public, the police, crown prosecutor, and courts and can change dramatically in their status by a decision being reversed by any one of these agencies.