Does therapy actually work? What the research says about its effectiveness

“It’s only a conversation. How is a conversation supposed to change anything?” If this question is holding you back from therapy, you’re in good company — scientists have asked it too, and very seriously. Therapy costs something: time, money, courage. It’s fair to ask whether it’s worth it.

The good news: we don’t have to answer this question with feelings, beliefs or advertising stories. Psychotherapy is one of the most researched forms of help in the history of medicine and psychology — several thousand controlled studies and hundreds of meta-analyses are available. In this piece I’ll show you what they find: how large the effect of therapy is in comprehensible numbers, how it compares with medical treatments, whether the effects last — and, because I owe you honesty, also when therapy doesn’t work and what you can do to improve your own chances.

1977: the moment the doubt got an answer

The story actually begins with an attack. In 1952 the psychologist Hans Eysenck published a much-discussed paper claiming that psychotherapy works no better than simply waiting: people were said to recover by themselves in equal proportions, with therapy merely claiming the credit. The claim stung the profession — and, fortunately, forced it into measuring.

The answer came in 1977, when Mary Lee Smith and Gene Glass did something previously unseen: they statistically combined the results of 375 psychotherapy studies into a single overall estimate. In doing so they effectively helped invent the meta-analysis — the method that is today the gold standard for judging treatments across all of medicine. Their finding: the average person who received therapy ended up doing better than roughly 75 per cent of people who had no therapy. Eysenck’s thesis fell — not to an argument, but to data.

Over the following decades this finding was tested again and again, on larger samples, with stricter methods, across different difficulties. The picture has remained remarkably stable.

How is something like this even measured?

Before we get to the numbers, a quick look into the kitchen — because how much you should believe the numbers depends on it. The gold standard for researching therapies is the randomised controlled trial: people with a comparable difficulty are allocated by lot into a group that receives the therapy and a group that (as yet) doesn’t — a waiting list, say. Because chance decides the allocation, the groups start out identical in everything, including things we can’t measure; so the difference at the end belongs to the therapy, not to chance or to more motivated people ending up in one group.

A single such study can be misleading — a small sample, a particular therapist, lucky circumstances. Hence the meta-analysis: it statistically adds together dozens or hundreds of studies and asks what remains once the individual quirks average out. It expresses the result as an effect size — a measure that tells you how much difference a treatment makes, not merely whether it makes any. That is precisely why meta-analyses of psychotherapy are so persuasive: they don’t measure the profession’s opinion of itself, but the summed experience of tens of thousands of people under controlled conditions.

Of course this method isn’t without weaknesses either — studies with nicer results get published more readily than those showing no effect, which can inflate the overall estimate somewhat. Modern meta-analyses measure and correct for this bias; after correction the effect of psychotherapy remains clear and large, just somewhat more modest. That belongs to an honest picture too.

What “an effect of 0.8” means in human language

Science expresses effect size with a number it calls Cohen’s d; for psychotherapy it consistently comes out at around 0.8 in meta-analyses. That figure says nothing by itself — but translated it says a great deal.

An effect of 0.8 counts in statistics as a large effect. To give you a feel for it: that’s roughly the difference in height between an average 14-year-old girl and an average 18-year-old — a difference you notice with the naked eye, not only with a tape measure. Translated into percentages it means: the average person after therapy is better off, in whatever the therapy addresses, than around 79 per cent of people without it. One more translation I’m fond of: if I picked one person at random from a treated group and one from an untreated group, the person who had therapy would have the better outcome in almost two cases out of three.

None of these translations says therapy helps everyone equally — an average conceals a range. But it does say something important: the effect isn’t at the edge of detectability. It is robust, large and measured over and over again.

For comparison with everyday benchmarks: the effect sizes of many widely accepted medicines for common physical complaints run at around 0.3 to 0.5 — appreciably below what psychotherapy consistently measures. Next time you hear that therapy is “only a conversation”, remember: by measured effect, this conversation outperforms plenty of pills that people take without a second thought.

A comparison that surprises: therapy alongside medicine

Perhaps the most telling context is a comparison with treatments nobody doubts. Bruce Wampold, one of the leading psychotherapy researchers, points out in his reviews that the effect of psychotherapy is comparable to or greater than the effects of many established medical interventions — from the impact of certain cardiovascular drugs on mortality to that of many surgical procedures on quality of life. In a large review, Lipsey and Wilson (1993) showed that psychological and behavioural interventions rank alongside established medical ones in effect size — and outperform quite a few of them.

Why does this sound incredible? Because we don’t see a conversation as an “intervention”. But the brain is an organ and relationships are an environment that physically shapes it — I write about how therapy literally changes neural connections in my piece on neuroplasticity. The right kind of conversation, in the right relationship, for long enough, is a biological event.

And here is a finding I’m especially fond of: the effects of therapy as a rule last, and often go on consolidating after the treatment ends. Follow-up meta-analyses show that with psychotherapy the gains are maintained or even grow after it finishes — whereas with medication the effect often fades once you stop taking it. The logic is simple: therapy doesn’t “repair” you, it teaches you a different relationship with yourself, your emotions and other people. You take that knowledge with you and it keeps working — including between sessions and after them.

What therapy works for

“Therapy works” is of course a generalisation — so let’s unpack it. The strongest evidence exists for depression and anxiety disorders, where by far the most research has been done; the effects are also well supported for post-traumatic stress disorder, eating disorders, chronic insomnia (where cognitive behavioural therapy is even the first-line choice ahead of medication — I’ve written about this in my piece on CBT-I), for couple and family distress, and for physical illnesses accompanied by mental distress.

But something else matters too: therapy isn’t confined to diagnoses. A large proportion of people come without any “disorder” at all — with burnout, grief, decisions, repeating patterns in relationships. The research shows meaningful effects there as well, even though such themes are harder to squeeze into research protocols. If you’re wondering whether your difficulty is “serious enough” for therapy, I’ve devoted a piece to therapy without a diagnosis — the short answer: a diagnosis isn’t a ticket.

The dodo bird and a secret that isn’t a secret

When researchers compared different therapeutic modalities with one another, a surprise was waiting: the established modalities mostly differ far less in effectiveness than you would expect. This finding acquired a jokey name in the profession — the “dodo bird verdict” — after the scene in Alice in Wonderland where the dodo announces after the race: “Everybody has won, and all must have prizes.”

What does this mean? Not that method is unimportant — it means that what drives therapy isn’t primarily the brand name of the technique but the common factors alive in every good therapy: a trusting relationship in which you feel seen; a convincing account of your distress and a meaningful way out of it; your hope and your engagement; the therapist’s personality and how well you fit with them. The loudest of these is the therapeutic relationship — its quality is one of the most reliable predictors of outcome, which I’ve explained in detail in my piece on why the therapeutic relationship heals, so I won’t repeat it here.

The practical consequence for you: when choosing a therapist, the question “do I click with this person and can I trust them” is at least as important as the question “which modality do they practise”. I’ve gathered how to choose in my piece on choosing a therapist.

Honestly: when therapy doesn’t work

If I told you only the above, I would be a poor informant. Therapy doesn’t help everyone and doesn’t help always — and the reasons are known.

• A poor fit. Sometimes you and the therapist simply don’t click — and since the relationship carries so much of the effect, that’s no small matter. Research also shows that therapists differ from one another in effectiveness. If after a few sessions you feel it isn’t working, that isn’t a reason to despair of therapy — it’s a reason to talk to your therapist, or to change.
• Dropping out too early. A considerable proportion of people break off therapy after one or two sessions — often at exactly the point where it becomes uncomfortable, just before it would start to work. I’ve described why the process can hurt at first in my piece on why therapy hurts at first.
• Lack of motivation. Therapy isn’t a service you receive but a process you take part in. Someone who comes because they “have to” — because a partner or an employer insists — usually takes little away without an investment of their own. Inner resistance is normal too, and can be worked with, but only if you bring it into the room.
• Deterioration does happen. A small proportion of people — estimates in the literature run at around 5 to 10 per cent — feel worse during therapy. Part of this is temporary and part of the process, but part are genuine negative outcomes, often connected with a poor fit or an unsuitable approach. That is exactly why it matters to speak up about feeling worse rather than quietly disappearing.

In short: therapy is effective on average, not in every individual case. That isn’t an argument against it — it’s true of every treatment, from antibiotics to physiotherapy — but it is a reason to enter it informed: with realistic expectations, a willingness to engage, and permission to speak up whenever you feel it isn’t going in the right direction.

How to improve your chances

From all of the above follows a piece of good news: because the outcome of therapy is carried to such a degree by factors on your side — engagement, relationship, persistence — you aren’t the passive recipient of an intervention but its co-author. Which means you can actively shift the likelihood of a good outcome. Concretely:

• Choose a therapist you can trust — and allow yourself to change if, after a fair attempt, the two of you don’t click.
• Say what’s hard to say. Therapy works in proportion to your honesty — including (and especially) when it concerns doubts about therapy itself or about your therapist.
• Persist through the first discomfort. Give the process a fair few sessions before you judge it. I’ve written up how long it usually takes in my piece on the length of therapy.
• Work between sessions. The hour a week is the seed; the other 167 hours are the watering.
• Track your progress together. Ask each other regularly whether you’re going in the right direction — I’ve gathered what to look out for in my piece on the signs that therapy is working.

So: is it worth it?

To sum up half a century of research in three sentences: psychotherapy works, its effect is large and comparable with established medical treatments, and its effects last beyond the end of treatment. It doesn’t work for everyone and it doesn’t work by itself — it works as a collaboration between you and someone you trust. And the single biggest factor you have influence over is you: your decision, your honesty, your persistence.

The doubt you may have started reading with is legitimate — and the research has answered it better than I could. If you’d like to try out what this “only a conversation” looks like in practice, have a look at the forms of work I offer, or write to me. A first conversation isn’t a commitment — it’s only the first piece of information.