The sexual abuse of children is an issue that society must address with urgency and commitment, given the profoundly damaging effects and widespread occurrence of this kind of crime. Providing psychological treatment to identified offenders is an important endeavour of the criminal justice system, with the aim of reducing recidivism and thereby preventing future victims. This dissertation explores a number of areas relevant to the treatment of sexual offenders on a sample of 223 adult males who completed a prison-based programme for child sexual offenders in New Zealand. Specifically, the assessment of treatment outcome and its relationship with recidivism, risk assessment, and the influence of specific offender factors on estimates of treatment outcome and risk were investigated. Study 1 (N = 218) is an independent validation of the validity of the Violence Risk Scale: Sexual Offender Version (VRS:SO; Olver, Wong, Nicholaichuk, & Gordon, 2007), a recently developed risk assessment instrument for sexual offenders that incorporates both static and dynamic risk factors and contains protocols for the assessment of change as a result of treatment. Results indicate support for the inter-rater reliability, concurrent validity, and predictive validity of the VRS:SO with regard to sexual recidivism, with pre-treatment and post-treatment scores showing superior predictive validity relative to a widely used measure of static risk (Static-99; Hanson & Thornton, 1999) and a measure of "Deviance" based on a 4-factor battery of relevant psychometric tests (Allan, Grace, Rutherford, & Hudson, 2007). In Study 2 (N = 218), three separate methods of assessing proximal treatment outcome (representative of three categories of treatment outcome measures that have previously been applied in the literature) are applied and compared in terms of their predictive validity with regard to sexual recidivism, and the relative advantages and disadvantages of their use. These measures are: change on a battery of relevant psychometric tests administered prior to and following treatment; change across treatment on the VRS:SO; and post-treatment ratings of the attainment of treatment goals as measured by a modified version of Hogue’s (1994) Standard Goal Attainment Scaling for Sexual Offenders (SGAS). Results indicate that positive treatment outcomes as measured by all of these methods are associated with reduced sexual recidivism. SGAS scores are identified as being relatively simple and efficient to obtain, however the VRS:SO and the psychometric battery are both able to provide useful pre-treatment clinical information regarding potential treatment targets for a particular offender. Study 3 (N = 223) and Study 4 (N = 216) are explorations of the influence of particular offender characteristics on response to treatment and risk of recidivism. Of particular interest was the personality construct of psychopathy (measured using the Psychopathy Checklist-Revised, PCL-R; Hare, 1991), and both studies are attempted replications and extensions of previously reported interaction effects involving this construct (Heilbrun, 1979; Seto & Barbaree, 1999). The results of Study 3 indicate that there is no interaction effect between PCL-R scores and treatment outcome (as measured by the SGAS) on sexual recidivism, in contrast to an influential study by Seto and Barbaree (1999). Study 4 reports an interaction effect between PCL-R scores and intelligence on recidivism, such that higher than average IQ scores appear to moderate the well-known association between psychopathy and risk. Overall, the findings reported in this dissertation suggest the importance of considering dynamic factors as well as static factors in sex offender risk assessments, and support the premise that dynamic factors are changeable, with change being associated with changes in recidivism. The potential for certain offender characteristics to influence treatment response and risk of recidivism is highlighted, and several areas for further exploration are identified.