In this study, we measured emotional processing and the alliance across 3 phases of therapy (beginning, working, and termination) for 74 clients who each received brief experiential psychotherapy for depression. Using path analysis, we proposed and tested a model of relationships between these 2 processes across phases of therapy and how these processes relate to predict improvement in the domains of depressive and general symptoms, self-esteem, and interpersonal problems after experiential treatment. Both therapy processes significantly increased across phases of therapy. Controlling for both client processes at the beginning of therapy, working phase emotional processing was found to directly and best predict reductions in depressive and general symptoms, and it could directly predict gains in self-esteem. Within working and termination phases of therapy, the alliance significantly contributed to emotional processing and indirectly contributed to outcome. Surprisingly, beginning therapy alliance (measured after Session 1) also directly predicted all outcomes. Furthermore, only clients' beginning therapy process predicted reductions in interpersonal problems. Therefore, although the proposed theory of change was supported, clients' beginning therapy processes may constrain clients' success in experiential treatment and in particular their outcomes in some problem domains related to depression. (Contains 3 figures and 4 tables.)