We explored the processes of change in cognitive therapy of depression in a sample of 25 adult outpatients. Raters blind to treatment outcome listened to audiotapes of sessions and rated items that pertained to (a) the therapist-offered facilitative conditions (warmth, empathy, etc.); (b) the patient—therapist relationship (the helping alliance); and (c) the therapist's adherence to the methods of cognitive therapy. Adherence separated into two factors in a factor analysis. One factor, representing concrete, symptom-focused methods of cognitive therapy, predicted subsequent symptom reduction when assessed early in treatment, but not later. The other factor, representing less focused, more abstract discussions, did not predict improvement. Neither did facilitative conditions nor the helping alliance predict change that occurred after the rated session. However, later in treatment, the helping alliance was predicted by prior symptom reduction. Methods used to discern the temporal relations of in-session behavior to outcome are highlighted.