This study had two primary goals: 1) to determine if the CSS-M (Criminal Sentiments Scale-Modified) assessment was a valid predictor of technical violations and rearrest in a sample of probationers and parolees; and 2) to determine whether officer training in the EPICS (Effective Practices of Community Supervision) model affected the levels of offender antisocial thinking and attitudes as evidenced by changes in CSS-M scores. The study found that some support exists for using CSS-M as a means for predicting offender outcomes, and that those offenders who were supervised by officers trained in the EPICS model had positive improvements in their antisocial thinking patterns. Data for the study were obtained from a group of offenders (n=238) who were supervised by probation and parole officers that had been randomly assigned to the group of officers receiving EPICS training or the untrained group of officers. Only probationers and parolees who had moderate to high-risk levels of recidivism were included in the sample. Offenders completed the CSS-M assessment during the first contact session with their probation or parole officer and then again during the final contact session to determine whether the officer's EPICS training affected the results of the assessment. The findings suggest that the EPICS model is an effective tool at changing criminal thinking patterns and that these changes can be measured using the CSS-M. Study limitations are discussed. Tables and references