From 444 studies published until 2002 that investigated the efficacy of hypnosis, 57 randomized clinical studies were selected that compared patients treated exclusively by hypnosis to an untreated control group (or to a group of patients treated by conventional medical procedures). The 57 studies were integrated into a meta-analysis that yielded a weighted average post-treatment effect size of d = 0.56 (medium effect size). For hypnotic treatment of ICD-10 codable disorders (32 studies) the calculation of the weighted mean effect size resulted in d = 0.63. These estimates are conservative since all variables of a given study were used. Most of the studies employed methods of the classic approach to hypnosis. In order to obtain an estimate to which extent non-clinical factors (design-quality, way of comparison of dependent variables) have an influence on the effect sizes, effect sizes were computed for all studies of the original 444 studies that reported the necessary statistical information (N = 133). For those studies with an average effect size of d = 1.07 a massive influence of non-clinical factors was demonstrated with a range from d = 0.56 for randomized studies with group comparisons to d = 2.29 for non-randomized studies using pre-post-comparisons. Out of the 57 randomized studies, only 6 reported numerical values for the correlation between hypnotic suggestibility and treatment outcome with a mean correlation of r = 0.44.