In spite of the success of cognitive behavioral therapy for social phobia in research studies, it remains unclear whether interventions will remain successful in the routine of clinical practice, where patients might differ from those in research samples. This study investigates the possibility that higher effect sizes are achieved under typical research conditions. Thirty studies testing treatment effects for social phobia are re-examined by categorizing them according to the quality and amount of applied sample restriction and laboratory study characteristics. The results indicate that even accumulation of sample restriction, such as excluding patients with comorbid disorders or being outside a certain age range, does not have any predictive value for treatment effect. However, there was a relevant tendency for studies applying a row of laboratory treatment conditions, such as recruiting patients by advertisements, applying treatment in university settings, and using specifically trained therapists, to achieve higher effect sizes.