There is a common thing people say about anxiety: I know there's nothing to worry about, but I can't stop worrying. The person says this with mild embarrassment, as if the feeling were evidence of irrationality — as if they were failing the basic cognitive task of aligning their internal state to external facts.
What neuroscience suggests is that this framing has the whole thing exactly backwards.
The brain is not designed to align your feelings to present facts. It is designed to generate predictions about future facts — and anxiety is what happens when those predictions detect threat, regardless of whether the threat has materialized yet. Anxiety is not irrational. It is temporally ahead of the moment.
Friston's free energy principle
Karl Friston's free energy principle is among the most ambitious theoretical frameworks in contemporary neuroscience. At its core is the claim that the brain is a prediction machine — one that generates generative models of the world, samples sensory data, and continuously works to minimize the discrepancy between prediction and perception.
This discrepancy — "prediction error" — is the signal on which the brain acts. When prediction and perception align, prediction error is low, and the system can coast. When they diverge, prediction error spikes, and the system must either update its model (learning) or act to bring the world into conformity with the prediction (action).
Crucially, this process runs in advance of events, not after them. The brain is constantly anticipating — running simulations of probable futures, weighting them by prior probability, and pre-loading responses. This is why you flinch before you see the punch; why your stomach tightens before you make the phone call; why you feel dread about Tuesday's meeting on a calm Saturday morning.
Anxiety, in this framework, is what happens when the predictive system detects a high-probability adverse future and generates a preparatory emotional state. It is not about the present. It is about a model of what the present is likely to become.
Why "there's nothing to worry about" doesn't work
Understanding this explains something otherwise puzzling: why reassurance is so ineffective for anxious people.
If anxiety were primarily a response to present facts, then presenting accurate facts — the surgery went well, the statistics are in your favor, your boss likes you — should reduce it. Sometimes it does. But often it doesn't, because the anxiety is not running on present facts. It is running on a predictive model, and models can persist in generating predictions long after the base rates have changed.
A person who grew up in an environment of genuine uncertainty — financial instability, emotional unpredictability in caregivers, repeated small betrayals of trust — will have built a predictive model calibrated to that environment. That model generates anxious predictions not because the current environment is dangerous, but because the model hasn't been updated. The prior is stale.
This is Friston's most practically useful insight: anxiety is not about defective reasoning. It is about inaccurate priors. The question is not why are you being irrational? but what is your model calibrated to? — and more importantly, what evidence would it take to update it?
The temporal structure of worry
This also explains the phenomenology of anxious worry in a way that cognitivist accounts tend to miss.
Worry is future-oriented by nature — you can observe this simply by paying attention to the content of anxious thoughts. Virtually all of it involves anticipation: what might happen, what could go wrong, how a situation might unfold. Present-moment anxiety is usually triggered by a cue that the predictive system reads as pointing toward an adverse future, not by present conditions per se.
The body keeps score in a specific way: the somatic markers Damasio describes encode not just past experience but anticipated costs — the anticipatory anxiety is itself a form of resource allocation, a mobilization of the body for an as-yet-unrealized event. This is why chronic anxiety is exhausting: you are spending metabolic resources preparing for events that, statistically, will mostly not occur.
But the system did not evolve in a context where most threats were statistical abstractions. It evolved to handle immediate, embodied threats in low-certainty environments. Sapolsky documents at length how this mismatch — a stress response system calibrated for leopards, now running on job performance reviews — produces the pathologies of contemporary anxious life.
What updating looks like
If anxiety is an unupdated prediction, then the resolution is not reassurance — it is evidence that actually revises the model.
This is why exposure therapy works: not because the person learns to tolerate anxiety (though that happens), but because repeated non-confirmation of the feared outcome eventually shifts the prior. The model updates. The prediction error associated with the feared cue diminishes because the cue no longer reliably predicts the adverse outcome.
But exposure is not the only route. Insight can function similarly — if by insight you mean not just intellectual understanding but the kind of understanding that is felt, that lands in the body, that genuinely revises the predictive model rather than merely annotating it. When someone suddenly understands that their current anxiety is running on a model calibrated to a parent who no longer exists, in conditions that no longer obtain, that understanding can — not always, not automatically, but sometimes — shift the prior.
The practice this points toward is not relaxation but precision: the capacity to ask, when anxious, what is my nervous system predicting? — and to investigate whether that prediction reflects the current situation or an inherited model that has outlasted its usefulness.
Anxiety is not irrational. It is loyal — to an older version of the world.
Further reading: Your Emotions Aren't Lying to You — You Just Haven't Learned Their Language
The Emotional Manual by Pierre Dufraisse is available as an ebook and in print. Get your copy