Sexual preference has been identified as an important component in the motivational matrix of many sexual offenders against children. Deviant sexual cues or behaviours are understood to elicit maximal arousal in the offender, who then experiences a preference for these cues or for behaviour motivated by the stronger sexual arousal. This preference for sexual interaction with children is understood to be a learned phenomenon, and thus modifiable. In recent years there has been a huge expansion in knowledge of the assessment and treatment of sexual offenders. One aspect of treatment is techniques designed to alter sexual preference. A number of procedures have been developed, but controlled outcome evaluations are sorely needed. The present study employed a combination of two techniques - directed masturbation and verbal satiation – and utilised phallometry to assess the degree of change. Three groups, each of ten child sex offenders, underwent phallometric assessments and treatment in a multiple base-line, across groups design. Results suggest that directed masturbation, as conducted here, was not effective in enhancing the level of arousal of offenders towards adult stimuli. Verbal satiation however appears to have resulted in a significant reduction in arousal to deviant stimuli. The use of repeated phallometric assessments with some participants in this study also permitted an investigation into the reliability of phallometric measurement. Results indicate, on the whole, that phallometry is a somewhat unreliable form of assessment with child sex offenders. Implications of these findings, with respect to the underlying theoretical assumptions of phallometry, are discussed.