Objective: Little is known about the variability of the alliance-outcome correlation across identifiable client subsets. This question was explored in a sample of 60 clients receiving cognitive therapy for depression, from which an overall correlation of .23 was observed between alliance ratings and subsequent symptom change.
Method: We examined interactions between the observer-rated version of the Working Alliance Inventory-Short Observer-Rated version (WAI-O; Tracey & Kokotovic, 1989) and client demographics, features of depression, personality, and other clinical features in predicting subsequent symptom change.
Results: After correcting for multiple comparisons, interactions between the WAI-O and the number of prior depressive episodes, as well as the severity of baseline anxiety symptoms, were significant predictors of symptom change. When both interactions were controlled for, number of prior depressive episodes emerged as a statistically significant moderator. The alliance predicted outcome in the subgroup of clients with 0-2 prior episodes (r = .52), but not in those with 3 or more prior episodes (r = -.02). These findings were obtained despite similar univariate distributions on the alliance and symptom change in the 2 subgroups.
Discussion: Differences that were observed in the predictive relation of alliance to outcome as a function of number of prior episodes suggest that different therapy processes may account for change in these subgroups. If the pattern observed in the present study is replicated, it would suggest that the alliance-outcome association has been both under- and overestimated.