Background: Psychoanalytic theory and some empirical research suggest the working alliance follows a “rupture and repair” pattern over the course of therapy, but given its emphasis on collaboration, cognitive behavioral therapy may yield a different trajectory. Aims: The current study compares the trajectory of the working alliance during two types of cognitive behavioral therapy for social anxiety disorder – virtual reality exposure therapy (VRE) and exposure group therapy (EGT), one of which (VRE) has been proposed to show lower levels of working alliance due to the physical barriers posed by the technology (e.g. no eye contact with therapist during exposure). Method: Following randomization, participants (N = 63) diagnosed with social anxiety disorder received eight sessions of manualized EGT or individual VRE and completed a standardized self-report measure of working alliance after each session. Results: Hierarchical linear modeling showed overall high levels of working alliance that changed in rates of growth over time; that is, increases in working alliance scores were steeper at the beginning of therapy and slowed towards the end of therapy. There were no differences in working alliance between the two treatment groups. Conclusion: Results neither support a rupture/repair pattern nor the idea that the working alliance is lower for VRE participants. Findings are consistent with the idea that different therapeutic approaches may yield different working alliance trajectories.