Over recent years there has been considerable publicity and research on the issue of child sexual abuse. Although figures vary, there is apparently consensus among the general public and health professionals that in terms of both the extent and the effect, the sexual abuse of children is a grave problem. Studies on the efficacy of the treatment of child molesters have indicated differential results both according to treatment modality and the classification of this heterogeneous population. Relapse Prevention (RP), a multi-modal, cognitive-behavioural approach, for the treatment of child molesters, is a relatively new model which emphasises self-management and personal responsibility for avoiding and coping with situations that threaten self­-control, rather than cure. Early outcome and interim assessment data suggest that RP reduces recidivism and encourages sex offenders to accept responsibility for their deviant sexual behaviour. The present study examines interim causal change in male prisoners who had sexually abused children, by using the 4 Attribution Dimension Scale (4-ADS). The 4- ADS measures perceptions of the cause of an event according to the attributional dimensions of controllability, locus, stability, and globality. Fifteen inmates in a medium security sex offender treatment unit were assessed at four points over the course of the programme to evaluate their response to the entire treatment programme as well as specific components. At each assessment point their causal perceptions at four specific stages of their relapse processes were measured also; entering the high risk situation (HRS), HRS, lapse, and relapse. The subjects were classified as fixated or regressed, and as incestuous or non-familial. The results indicate that the whole group made significantly positive change on all causal dimensions, over the course of treatment. There were differences between the causal dimensions in terms of the effect of the specific components. Subjects tended to see the cause of their offending to differ in stability across the relapse scenario but there were no significant changes on the other dimensions. The results also indicated that fixated subjects saw the cause of their sex offending as less controllable, more stable at the time of the offence, and more global, than regressed subjects. Compared with non-familial subjects, incestuous subjects evaluated the cause of their offending as less stable at the time of the offence, and less global across all the assessment points. These results are discussed in terms of the utility of using the 4-ADS as a measure of interim progress in a Relapse Prevention treatment programme for sex offenders against children. Implications for future research are also explored.