A new meta-analysis published in PLoS One of publication bias in NIMH-funded trials of psychological therapies for depression shows that once unpublished studies are included in the meta-analysis, the effect size for psychotherapy drops from 0.52 to 0.39. You'll note that the confidence interval for the new effect size estimate very nearly crosses zero (0.08~0.70). It's good to see big names in the field authoring a paper like this. It's important to also keep in mind that outcome measures in depression studies are questionnaires. Only a small number of items on the Hamilton Depression Rating Scale are objectively observable by the clinician (things like psychomotor agitation or retardation) who is doing the rating while the remaining items are rated on the basis of what the patient tells you. Some scales like the Beck Depression Inventory are self-report scales. An unknown % of that improvement is therefore likely to be due to the effects of psychological brainwashing on the patients' reporting of subjective symptoms. Personally, I've suspected for a long time that psychotherapy is worthless even for problems conventionally classified as psychiatric.