The ability of several process variables to predict therapy outcome was tested with 30 depressed clients who received cognitive therapy with or without medication. Two types of process variables were studied: 1 variable that is unique to cognitive therapy and 2 variables that this approach is assumed to share with other forms of treatment. The client's improvement was found to be predicted by the 2 common factors measured: the therapeutic alliance and the client's emotional involvement (experiencing). The results also indicated, however, that a unique aspect of cognitive therapy (i.e., therapist's focus on the impact of distorted cognitions on depressive symptoms) correlated negatively with outcome at the end of treatment. Descriptive analyses that were conducted to understand this negative correlation suggest that therapists sometimes increased their adherence to cognitive rationales and techniques to correct problems in the therapeutic alliance. Such increased focus, however, seems to worsen alliance strains, thereby interfering with therapeutic change.