Everybody Has Won

In 1936 Saul Rosenzweig published a short paper suggesting that the competing schools of psychotherapy, which disagreed violently about almost everything, seemed to produce roughly similar results. He borrowed a line from Alice in Wonderland, where the Dodo adjudicates a chaotic race: everybody has won, and all must have prizes.

Nearly ninety years of meta-analysis later, the dodo bird verdict is contested, heavily attacked, and still standing in roughly the form he proposed. When bona fide therapies (coherent, competently delivered, intended by their practitioners to be therapeutic) are compared directly, the differences between them are consistently far smaller than their theoretical differences would predict.

This is a strange result. Psychoanalysis and cognitive behavioural therapy disagree about what causes distress, what maintains it, what the therapist should do, what should be discussed, and what counts as improvement. That two such different procedures produce similar outcomes demands an explanation.

The explanation offered is that what they share is doing most of the work.

The common factors are these. An alliance: agreement between therapist and patient about the goals and the tasks, plus a genuine bond. An expectation that this will help. A credible rationale: a plausible story about what is wrong and why this procedure addresses it. A therapist who believes in what they are doing. And a structured set of activities that follow from the rationale.

The therapeutic alliance is among the most consistently replicated predictors of outcome in the whole literature, and it predicts better than modality does. It is also measured early and predicts later change, which helps with the obvious worry that a good alliance is simply what improvement feels like.

Then there is a finding that is more awkward for the profession than the dodo verdict itself.

Differences between individual therapists working within the same modality, in the same clinic, with comparable caseloads, are typically larger than differences between modalities. Some therapists reliably get better results than others, and the gap is substantial.

What predicts being one of them is not what you would hope. Not years of experience: the correlation between experience and outcome is close to zero, which is remarkable. Not theoretical orientation. Not qualifications. The best predictor identified so far is interpersonal skill measured in difficult moments: how a therapist responds when a patient is hostile, withdrawn or challenging, assessed from recordings of exactly those moments.

There is also evidence that deliberate practice (reviewing recordings of one's own difficult sessions with feedback, as musicians and surgeons do) improves outcomes, and that ordinary accumulated experience does not.

Two important qualifications, because "all therapies are the same" is a bad summary of the dodo verdict and the exceptions are where the mechanisms are.

First, the verdict applies to bona fide therapies compared with each other, not to therapy versus nothing. Psychotherapy substantially outperforms no treatment and waiting lists for most conditions studied.

Second, there are genuine exceptions, and they cluster where a specific mechanism is well understood. Exposure and response prevention for OCD, and exposure-based treatments for specific phobias and PTSD, outperform credible control conditions that share the alliance, the rationale and the expectation. The specific procedure is doing something the common factors do not supply.

That pattern is roughly what you would expect. Where we understand the maintaining mechanism (avoidance preventing disconfirmation) a technique targeting it directly should beat a technique that does not. Where we do not understand the mechanism, the common factors are what is left. So the dodo verdict is best held as a default that specific evidence can overturn, rather than as a law.

One more thing that follows from therapy being a treatment with real effects: it can harm.

Deterioration rates of roughly five to ten per cent are typical in trials, and some interventions have produced reliable harm. Critical incident stress debriefing, having people recount a trauma in detail shortly after it happens, performed worse than no intervention in several controlled trials, apparently by interfering with natural recovery, and is no longer recommended. Scared Straight programmes increased subsequent offending. Some grief interventions do worse than nothing for people who were coping.

This is why routine outcome monitoring matters, and why its absence in most practice is a genuine problem. Sessional measurement identifies patients who are deteriorating while there is still time to change course, and trials of feedback-informed treatment find it reduces deterioration meaningfully. Without it, a therapist's impression is the only signal, and impressions are formed by the same machinery this book has spent several chapters describing.

Which leaves a practical conclusion for anybody choosing. Given small average differences between established therapies and large differences between therapists, the fit matters more than the brand: whether the explanation makes sense to you, and whether you can work with this person. And since the alliance forms early and predicts outcome, changing therapist after a few genuinely unproductive sessions is a reasonable response to evidence rather than a failure of commitment.