The impact of personality disorders (PDs) on exposure in vivo treatment for social phobia was investigated in three groups of social phobics: social phobia without any PD (n = 30), social phobia with a single diagnosis of avoidant PD (n = 18) and social phobia with multiple PDs (n = 13). We hypothesized parallel change for social phobia with and without an avoidant PD with the latter group being more impaired before and after treatment. In order to test this hypothesis, confidence intervals for change were computed. In line with our hypothesis, social phobics in all three groups improved significantly during treatment and no interaction effects were found on the repeated MANOVAs. By using a confidence interval, parallel change was found on most measures. The impact of additional anxiety and mood disorders on treatment outcome was investigated separately. The analyses showed that an additional anxiety or mood disorder also did not predict outcome of exposure treatment.