Why Does Therapy Actually Work? A New Answer From Neuroscience
If you've ever looked for a therapist, you've probably run into an alphabet soup. CBT. EMDR. IFS. Psychodynamic. Somatic. ACT. Each one comes with its own theory of what's wrong with you and its own explanation of how it fixes it.
Here's the uncomfortable part. When researchers compare these approaches head to head, they mostly perform about the same. The finding is well enough established to have a nickname in the literature: the dodo-bird verdict, after the dodo in Alice in Wonderland who declares that everybody has won and all must have prizes.
Psychotherapy works. That part isn't in question. But why it works has been a genuinely open question for decades. If every approach helps and every approach explains itself differently, then most of those explanations can't be the whole story.
A framework from cognitive neuroscience is starting to offer a better answer.
Your Brain Is Not a Camera
The old picture of the brain went like this. Information comes in through your senses, travels up to your brain, and your brain assembles it into an experience of the world. Bottom up. Passive.
Predictive processing flips that. In this view, your brain is constantly generating predictions about what it's about to encounter, and then checking incoming sensory information against those predictions. What you experience is not raw input. It's your brain's best guess, corrected by whatever mismatch shows up.
Those mismatches are called prediction errors, and your brain can handle them two ways. It can update the prediction to fit reality. Or it can act on the world to make reality fit the prediction.
That second option is the one that matters clinically.
How This Becomes Suffering
Your predictions come from experience. If you grew up in a home where affection was unreliable or anger arrived without warning, your brain built a model that was accurate for that environment. It learned to expect withdrawal, to scan for danger, to assume the other shoe is coming.
That model was not a malfunction. It was good learning in a bad situation.
The problem shows up later. Predictions built in an unpredictable environment get carried into environments that aren't unpredictable anymore. Your circumstances changed and the model didn't. It's one way of describing what I've written about elsewhere as the anxiety that isn't really about now.
Two things keep it stuck.
First, some predictions get held with too much confidence. When your brain treats a belief as highly reliable, contradicting evidence gets dismissed as noise. In depression, for instance, the model predicts that things will go badly, and positive experiences don't land hard enough to revise it. Worse, prediction can be self-fulfilling. Expecting rejection changes how you show up, which shapes how people respond, which confirms the original expectation.
Second, the behaviors that come with distress shrink the input. Withdrawal, avoidance, and reduced activity all limit exactly the kind of new experience that could update the model in the first place. This is part of why the brain seems to prefer familiar suffering to unfamiliar healing.
So you end up running old software on new hardware, with the update permanently postponed.
What Therapy Is Doing About It
A 2024 paper by Daniel Villiger in the Journal of Contemporary Psychotherapy, "An Integrative Model of Psychotherapeutic Interventions Based on a Predictive Processing Framework", takes this framework and applies it across cognitive behavioural therapy, client-centred therapy, and psychoanalysis. He calls the result the 3PI model. His conclusion is that therapy has two jobs, not one.
Job one is providing a safe and positive environment the patient can adapt to. A therapy relationship is unusually safe, unusually predictable, and unusually non-punishing. You disclose something difficult and the relationship doesn't rupture. You disagree and you don't get overruled. Week after week, this produces sensory evidence that contradicts the old model. Eventually that evidence accumulates past the threshold where the model has to give.
Job two is updating the model itself. This is the part that looks like technique. Exposure work generates prediction errors on purpose by staying in contact with the feared thing until nothing bad happens. Cognitive work surfaces core beliefs and holds them up to disconfirming evidence. Psychodynamic work uses transference, where you unconsciously apply an old relational model to your therapist, as live material to examine. Mindfulness turns up the volume on present-moment sensory experience so it can outweigh the prediction.
Villiger's argument is that the common factors across all therapies, meaning the alliance, the empathy, the goal consensus, the psychoeducation, primarily promote the first function. The specific techniques mainly support the second. Both are necessary. Neither is the whole thing.
That reframe is what makes this interesting. The old debate treated common factors and specific techniques as competitors. This says they were never doing the same work.
Why This Maps Onto How I Practice
I work psychodynamically, somatically, with Internal Family Systems, with EMDR, and with mindfulness. That combination can look eclectic from the outside. Under this framework it stops looking eclectic.
The somatic and interoceptive work is a good example. Interoception is your sense of your own internal state, and researchers working in this area have argued that bodily cues may be among the few consciously accessible parts of predictions that are otherwise buried. Careful attention to what's happening in your chest and gut, inside a relationship that's safe enough to stay with it, is a route to material that talking alone doesn't reach.
EMDR has been described in these terms too, as a way of forcing contact with the present moment so a traumatic memory's predictions can finally be tested against current evidence rather than avoided.
And the reason the relationship keeps mattering in every approach is no longer a soft claim about rapport. It's the mechanism for job one.
A Caveat Worth Keeping
Some coverage of this area has described predictive processing as a unified theory for psychotherapy. I'd hold that loosely.
Villiger's paper is a theoretical one. It maps existing techniques onto an explanatory framework. It doesn't run an experiment, and it doesn't demonstrate that any particular method outperforms another. There's also real disagreement among researchers about how unified predictive processing actually is.
What this framework offers is a coherent account of mechanism. That's genuinely valuable. It's not the same thing as proof, and anyone selling it that way is getting ahead of the evidence.
What It Means If You're the One Looking for a Therapist
If the model is roughly right, a few practical things follow.
The fit matters enormously, and not because fit is a nice bonus. A relationship that feels safe and non-judgmental to you is doing half the work directly.
Technique still matters, but it isn't the whole picture, and the specific brand name on your therapist's website is probably less predictive of your outcome than whether you can be honest in the room.
Change is slow because updating is slow. Your model took years to build in conditions where it made sense. It doesn't revise on a single piece of contradicting evidence. It revises on accumulation. That's also why insight alone rarely changes anything: one good realization is a single data point against a model built from thousands.
And the thing you're doing in therapy is not being talked out of your beliefs. It's giving your nervous system enough new experience, in enough safety, that it can finally afford to update.
If you're looking for a place to do that, I'd be glad to hear from you. I work in Pasadena and online across California, and you can book a free consultation here.