The purpose of this study was to ascertain whether and how alliance and specific cognitive process may interact to influence outcome. Eighty social phobic patients were randomized to 10-week residential cognitive or interpersonal therapy, conducted mostly in groups. They completed process and outcome measures weekly. The ratings were analyzed with mixed models. It was found that initial patient-rated alliance predicted the course of social anxiety throughout therapy and that this effect was indirect through the cognitive process. However, this indirect effect did not interact with treatment. There was a trend toward an indirect effect of weekly variations in alliance rated by the individual therapist through weekly variations in subsequent cognitive process on weekly variations in subsequent social anxiety. Thus, the results support a facilitative rather than an active ingredient perspective on the role of alliance.