This study examined whether self-efficacy differentiated treatment responders from non-responders in a trial of cognitive rehabilitation (CR) for postconcussive symptoms. 126 service members with mild TBI seen on average 9.5 months since injury completed one of four cognitive rehabilitation treatments for 6 weeks. The four treatment arms were: (1) Psychoeducation control, (2) Self-administered computerized CR, (3) Interdisciplinary CR, and (4) Interdisciplinary CR integrated with CBT. Outcome was assessed across time (baseline, and 6, 12, and 18 weeks post-treatment) for three domains: psychological (Symptom Checklist–90-Revised; SCL-90-R), cognitive (Paced Auditory Serial Addition Test; PASAT), and functional/behavioural (Key Behaviors Change Inventory; KBCI). Mixed model ANOVAs tested for self-efficacy differences across time in treatment responders versus non-responders, as defined by reliable change indices. A significant interaction was found on the SCL-90 such that responders had increasing self-efficacy with respect to psychological symptoms across four time points, whereas non-responders’ self-efficacy did not change. Perceived self-efficacy at the beginning of treatment was associated with treatment engagement within the psychological domain for responders only, suggesting a mediating role in treatment outcome. Overall, results suggest that increasing patients’ level of self-efficacy may be important for successful treatment of psychological distress in those with remote concussion.