An experimental evaluation of a cognitive self-instructional procedure for increasing self-control of eating was carried out in a weight reduction programme for obese women. All subjects received a 'core', 16 week long treatment programme of nutritional advice and encouragement provided by five group meetings and one individual session. Subjects were randomly assigned to receive only the core programme, or the core programme plus seven individual sessions of one of three types of treatment. Thus there were four treatment conditions: (a) Cognitive Rehearsal (Self-instructional training attempting to teach subjects to resist temptation through the self-directing power of inner speech), (b) Insight Control (teaching subjects to become aware of, but not to change, unhelpful cognitions), (c) Individual Contact Control (encouragement and nondirective discussion of dieting) and (d) Group Contact Control (core programme only). Of the 94 subjects entering the programme 74 completed the 16 weeks of treatment and 69 were followed up a year later. There were no significant differences between treatments in obesity reduction either at the end of treatment or at followup. There were significant reductions in anxiety and depression and in initially unrealistically high outcome expectations. Initial weight loss, an apparently important variable, predicted subsequent losses.