This study examined therapeutic mechanisms of action at the single-participant level in a behavior therapy (BT) for youth depression. By controlling for non-specific early responses, identifying potential mechanisms of action a priori, taking frequent measures of hypothesized mechanisms and dependent variables, rigorously evaluating internal validity, and using a variety of analytic methods, a unique model for analysis of potential mediators was created. Eleven children (M age = 9.84) meeting criteria on the Children’s Depression Rating Scale–Revised (M = 55.36) and Children’s Depression Inventory (M = 23.45) received non-directive therapy (NDT), followed by BT for those still displaying significant symptoms. Four participants (36%) had a clinically significant response to NDT. For the remaining seven, statistically significant changes in depressive symptoms and family interactions during the BT interval were found at the group level. At the single-participant level, evidence suggesting that outcome was at least partially mediated by changes in treatment targets was obtained for four of seven (57%). As the field further embraces efforts to learn not only whether treatments work but also how they work, the single-participant approach to evaluating mediators provides a useful framework for evaluating theories of therapeutic change.