Many behaviour, psychotherapy and healthy life-style programmes require subjects to take responsibility for the control of the old unwanted behaviours or to be responsible for maintaining new desired behaviours after therapy has ended. A scale to measure the locus of control of behaviour would be valuable if it could predict persons likely to relapse following apparently successful therapy. A 17-item Likert-type scale to measure this construct was developed and shown to have satisfactory internal reliability, to be test-retest reliable in the absence of treatment, to be independent of age, sex and social desirability, and to distinguish clinical disorder from normal non-clinical subjects. Furthermore, change towards internality (a reduced LCB score) during therapy was shown to predict maintenance or, alternatively, change towards externality (an increase or no change in the LCB score) was shown to predict relapse 10 months later in treated stutterers. The scale was shown to be related to the personal control factor of the Rotter I–E scale but to be more powerful a predictor of relapse than this personal control subscale or the full Rotter scale.