Beyond Talking: How Therapy Rewires the Brain

If you've ever wondered why some therapeutic approaches create change faster or more deeply than talking through a problem alone, there's a compelling possible answer in one of the leading theories of how the brain works: predictive processing. Predictive processing is a theory, not a settled fact, however it's grounded in neuroscience and increasingly influential across the psychology field, but like any model of something as complex as the brain, it's a current best explanation - not the final word. With that caveat in mind, it's a genuinely useful lens for understanding why certain therapies work the way they do.

Your Brain Isn't a Camera - It's a Prediction Machine

According to this theory, we're wrong to assume our eyes and ears simply show us the world as it is. Instead, the brain has no direct access to what's "out there" - only ambiguous streams of nerve signals from our eyes, ears, skin, and other senses, including our internal organs such as our heart or our gut. Relying on those signals alone, the theory suggests, would be far too slow and uncertain to keep us alive. So instead, the brain builds an internal model of the world from memory and uses that model to predict, moment to moment, what's about to happen. Crucially, we don't experience these predictions as guesses - we experience them as reality itself.

Imagine hearing a loud bang outside your window. Before you've even had time to think, your brain has already generated a prediction about what caused it - built from past experience, context, and whatever memory feels most relevant in that instant. Maybe it lands on a door slamming shut. Maybe, if you live near a busy street, it lands on a car backfiring. Maybe, for someone with a very different set of memories, it lands on something far more alarming, like an explosion. Once you actually look outside or gather more information, that fresh sensory evidence is compared against the prediction. If they match, the prediction becomes what you perceive. If they don't quite match, the brain has to decide which to trust: its existing model, or the new evidence. Most of the time, it updates appropriately, and the model gets a little more accurate.

When the Model Gets Stuck

Now imagine a war veteran, home from deployment, startled awake by the sound of footsteps outside - perhaps just neighbors walking past on the street below. Because the brain weighs survival-relevant memories more heavily than ordinary ones, it may default to predicting danger rather than recognising something as mundane as people out for a late walk. Even after they look outside and the mismatch becomes obvious, the pattern can repeat night after night - the brain choosing to trust its threat prediction over the calmer reality in front of them.

This is a vivid example of something all of us experience to some degree. Many of our deepest predictive patterns - especially about relationships, safety, and our own worth - were formed in early childhood, at a time when connection to caregivers was literally a matter of survival. We carry these patterns into adulthood, often reacting to present-day people and situations as if we were still that small, dependent child. And under stress, this tendency intensifies: the more threatened we feel, the more we default to our oldest, most emotionally loaded predictions rather than what's actually happening now. This is why insight alone - "I know logically I'm safe" - often isn't enough. The model has to actually be updated, and updating it usually requires the right kind of new experience, not just new information.

Why CBT, EMDR, and Imagery Rescripting Work - Through This Lens

If we take the predictive processing model as a useful working framework, each of these approaches can be understood as generating corrective experiences strong and safe enough to genuinely revise the model, rather than being dismissed by it, the way the veteran's brain dismissed their own eyes.

CBT (Cognitive Behavioural Therapy): CBT surfaces the specific predictions driving distress ( for example "If I speak up, I'll be humiliated") and tests them directly against reality through structured ‘behavioural’ experiments. Each time the predicted outcome doesn't occur, the brain receives a real piece of dis-confirming evidence. Enough of these, over time, and the model shifts.

EMDR (Eye Movement Desensitization and Reprocessing): Traumatic memories often stay locked in a raw, fragmented form that keeps firing threat predictions in the present, much like the veteran's middle of the night alarm. EMDR appears to support the brain's natural reprocessing mechanisms, helping these memories integrate into the broader, more accurate model - so the event is remembered, but no longer predicted as an ongoing danger.

Imagery Rescripting: Because emotional learning doesn't fully distinguish between vivid imagination and lived experience, imagery rescripting works directly on the model itself. By deliberately changing the imagined outcome of a distressing memory or feared scenario, it offers the brain new experiential "evidence" - updating emotional predictions in a way that talking alone often can't reach.

The Common Thread

What unites these approaches isn't the technique itself, but the fact that each creates the right kind of experience for the brain to revise its predictions rather than stick with them. This is also why therapeutic change can feel less like "figuring something out" and more like something genuinely shifting - because it is. The internal model is being updated, one well-placed corrective experience at a time.

If you're curious how any of these approaches fits your own goals in therapy, it's always worth asking your therapist directly - understanding the "why" behind the method can make the work itself feel more purposeful. You can read more about CBT here and EMDR here.

Predictive processing is an active and evolving area of research, and scientists are still working out exactly how well it maps onto the full complexity of the brain. But as a working theory, it offers one of the clearer explanations we currently have for why simply "knowing better" often isn't enough, and why certain therapeutic experiences can be more helpful where insight alone falls short.

If you’d like to read more about predictive processing, I highly recommend the books ‘The Experience Machine: How Our Minds Predict and Shape Reality’ by Andy Clark and ‘How Emotions Are Made: The Secret Life of the Brain’ by Lisa Feldman Barrett.

References Kotler S, Mannino M, Fox G and Friston K (2026) The body does not keep the score: trauma, predictive coding, and the restoration of metastability. Front. Syst. Neurosci. 20:1812957. doi: 10.3389/fnsys.2026.1812957