A meta-analysis of the efficacy of five selective serotonin re-uptake inhibitors (SSRIs) against non-selective and noradrenergic re-uptake inhibitors (mainly tricyclic antidepressants, TCAs) is presented. Fifty five double- blind studies were identified after excluding those multiply reported or with methodological problems likely to bias the outcome in favour of SSRIs. Standardised effect sizes and 95% confidence intervals were calculated based on the difference in the reduction in mean Hamilton depression rating scale (HDRS) scores for the two antidepressants. For studies not reporting standard deviations, the pooled variance from complete studies was used and a variance-weighted mean effect size calculated. There were no differences in efficacy between SSRIs and comparator antidepressants for SSRIs taken together or individually. If studies were classified into high and low depression scores based on a median split of initial HDRS scores, there was a slight advantage to TCAs in the high HDRS group. In addition, SSRIs were slightly less effective than TCAs in in-patients and against combined serotonin and noradrenaline re-uptake inhibitors (clomipramine and amitriptyline). These findings were accounted for by a clinically significant lower efficacy of paroxetine in these subgroups. In contrast, SSRIs as a group were marginally more effective than noradrenergic antidepressants, a finding accounted for by two studies with sertraline. Fluvoxamine was the only SSRI to have been tested adequately in in-patients, where it displayed equal efficacy to TCAs. This meta-analysis confirms that SSRIs and TCAs are in general equally effective, but suggests that paroxetine's efficacy in in-patients and against clomipramine and amitriptyline is not proven.