Therapeutic alliance is associated with treatment outcomes across a diverse range of intervention models and clinical populations. Likewise, the perceived quality of working alliance in adult psychotherapy has been a consistent predictor of engagement, while research with children is limited. This study draws from a sample of 180 children who were involved in a study of the Coping Power preventive intervention for elementary students with aggressive behavior. Approximately 1,080 individual intervention sessions were coded using the Therapeutic Process Observational Coding System (TPOCS-A). Results suggest that bonding, in the early sessions of Coping Power, predicted a reduction in teacher-reported externalizing behaviors. In addition, data analyses revealed that bonding interacted with children's gender in the latter portions of Coping Power. Higher bonding for girls in later sessions predicted lower teacher-rated externalizing behaviors at 1-year follow-up. Finally, initial levels of aggression moderated the relationship between bond and parent-rated externalizing behaviors during the early and late sessions of Coping Power, with stronger effects for moderately aggressive children. Task orientation, a component of therapeutic alliance, in late sessions had counter-intuitive effect on parent-rated externalizing behaviors, possibly due to therapists' efforts to impose more structure with children who were not progressing well.