The present study examined whether treatment change among offenders under community supervision would predict reductions in recidivism. The intervention program, based on cognitive-behavioral principles, focused on changing antisocial attitudes. Compared to a matched control group, the likelihood of reduced recidivism was 57% for the binary outcome and 70% for the multiple count outcome, after controlling for past program participation, propensity score, and days of opportunity to offend. The within-person pre-/postchange scores showed less promise in predicting recidivism. Only changes with a single antisocial associate measure, which were not central to the treatment program, predicted future recidivism. Caution must be used in the idiographic assessment of dynamic risk and treatment change.