Early sessions from three variants of cognitive-behavioral therapy (CBT) were examined to replicate work done in psychodynamic-interpersonal treatments linking interpersonal process to outcome (W. P. Henry, T. E. Schacht, & H. H. Strupp, 1986, 1990). Cases were available from a component study of CBT for generalized anxiety disorder (T. D. Borkovec, M. G. Newman, A. L. Pincus, & R. Lytle, 2002) and were selected to form good and poor outcome groups maintained through a 1-year follow-up. A third group was also examined that had initial positive outcomes and marked decline by follow-up (n = 8 for each). Structural analysis of social behavior (SASB) was used to identify interpersonal behaviors. Contrary to the authors' expectation, SASB variables were not strong predictors of outcome, and lower levels of interpersonal hostility were found than was the case in previous work. Findings are discussed in light of differences observed between treatment variants and the role that manuals may have in standardizing some aspects of the therapeutic relationship.