Forty-three social phobics were assigned to exposure (EXP), cognitive restructuring without exposure (CR-alone), or to an intervention combining these techniques (COMB), in a wait-list controlled (WLC) trial. Treatment integrity assessment showed compliance with instructions consistent with the treatments. Within-group analyses showed that the COMB and CR-alone groups improved significantly on all variables, whereas the EXP group showed changes on phobia but not attitudinal measures. Between-group analyses indicated COMB to be superior to EXP on two phobia measures. CR-alone was inferior to EXP and COMB on behavioral approach after treatment, but showed continued improvement relative to the exposure groups on this and other variables by follow-up. The relative ability of treatment-induced changes in fear of negative evaluation (FNE), locus of control, and irrational beliefs to predict long-term improvement was assessed. Changes in these variables were predictive of improvement. The change in FNE accounted for the majority of the explained variance.