In breathwork and functional breathing, we are very good at talking about mechanics.

We talk about breathing rate. Tidal volume. Nose or mouth. Chest or belly. We talk about CO2 tolerance, about chemosensitivity, about people who breathe harder and faster than they need to. All of that matters. It is real. It is measurable. It is often where useful change begins.

But there is a whole dimension we rarely talk about. Not how someone breathes. How they relate to their breathing. How they perceive it. What the sensation of air hunger, or tightness, or effort actually means to them in the moment it arrives.

Two people can have the same breathing pattern. The same blood gases. The same signal. And a completely different experience of it. For one person, the rising urge to breathe is information. For another, it is a threat. That difference does not live in the lungs. It lives in how the brain reads the body.

This is respiratory interoception. And it is one of the most useful frontiers in the science for any practitioner who wants to understand why their clients respond the way they do.

What interoception actually is

Interoception is the perception of the internal state of the body. The sense of what is happening inside you. Your heartbeat. The fullness of your stomach. The temperature of your skin. And, for our work, the sensations of breathing.

Here is the first thing worth understanding. Interoception is not a simple readout of what the body is doing. It is not a gauge that faithfully reports a number. The brain does not sit there passively receiving body signals and displaying them. The brain predicts them.

This is the core idea of predictive processing. The brain is constantly generating expectations about what your body is about to feel. Then it compares those expectations against the signals that actually arrive. What you consciously perceive is not the raw signal. It is the brain’s best model. A blend of prediction and evidence, weighted by how much the brain trusts each one.

That weighting is the whole thing. Say the brain holds a very strong expectation. “This is going to be bad.” And say it trusts that expectation more than the actual signal. Now the experience is shaped by the prediction, not the physiology. The body can be doing something quite ordinary. And it can still be experienced as alarming. Because the prediction carried the alarm.

Hold on to that idea. It runs through everything below.

Interoception is not one thing

The second thing that changes how you think about this. Respiratory interoception is not a single ability someone has or lacks. It is a hierarchy. It has at least three levels. And they can come apart.

There is sensitivity, sometimes called accuracy. Can you actually detect a change in your breathing. Add a small resistance to the breath. Can you feel it. This is raw signal detection.

There is metacognition, sometimes called insight. How well does your confidence track your accuracy. Do you know when you know, and know when you are guessing. This is your awareness of your own perception.

There is interpretation, sometimes called belief. What does the sensation mean to you. Neutral information, or threat.

Here is why this matters. When breathing distress shows up, our instinct is to assume the person is feeling too much. That their detection is turned up too high. That they are hyper-aware of their body. The science suggests the opposite is often true. The problem is usually not at the sensitivity layer at all. It is higher up. In metacognition, and in interpretation.

The Harrison work: it is not the signal, it is the confidence and the meaning

This is where the research of Olivia Harrison, who published earlier work under the name Faull, becomes so clarifying. Harrison and colleagues have spent years building rigorous, quantifiable ways to probe respiratory interoception. Measuring not just whether people can detect breathing changes, but how confident they are, and how anxiety relates to each level.

Their landmark study, published in Neuron in 2021, looked at anxiety and breathing-related interoception across the whole hierarchy at once. From low-level perceptual sensitivity, up through metacognition, up to brain activity. In healthy adults with low versus moderate trait anxiety.

The findings are worth stating carefully. They are easy to get slightly wrong, and the precise version is more interesting than the loose one.

People with moderate anxiety were less sensitive at the raw perceptual level. They had a higher detection threshold. It took a bigger change before they reliably noticed it. So they were not picking up more signal. If anything, less. And their metacognitive confidence was lower. They were less sure of their own perceptions. A more recent line of this work, from 2025, sharpens the headline further. Interoceptive accuracy can stay essentially stable, while anxiety significantly modulates the metacognitive confidence layered on top of it.

Sit with what that means. The anxious breather is not a person with superior bodily radar. The distress does not come from detecting the signal more accurately. It comes from how the signal is weighted, trusted, and interpreted. The confidence placed in it. The meaning assigned to it. Anxiety lives in the interpretive and predictive layer. Not the sensory one.

And Harrison’s imaging work locates this physically. Activity in the anterior insula, a key hub for building the body’s internal model, reflects breathing-related predictions and prediction errors. Not just raw sensation. In people with higher anxiety, the insula responds differently to predicted changes in breathing. The difference is not in the incoming signal. It is in the machinery of prediction.

This is the predictive-processing story made concrete. The anxious experience of breathing is a story the brain is telling about the breath. More than a report of it.

The athlete puzzle: fit lungs, frightened breath

Here is where it gets genuinely strange. And where my own work picks up the thread.

Say breathlessness anxiety were fundamentally about physiology. About CO2 sensitivity. About how close you are to your limits. About a hair-trigger chemoreflex. Then the fittest people, with the most robust respiratory systems, should be the least anxious about their breathing. The opposite turns out to be true. In a way a purely physiological model cannot explain.

Faull, now Harrison, and colleagues found that endurance athletes reported significantly greater breathlessness anxiety at maximal exercise than sedentary people. Roughly twice as much. Despite rating the breathlessness itself as no more intense. Same perceived intensity. Much more anxiety. The affective dimension and the sensory dimension came apart.

Then the crucial control. When Harrison and colleagues re-analysed these athletes’ CO2 chemosensitivity, how strongly their breathing responds to rising carbon dioxide, there was no difference between the athletes and the sedentary controls. On any measure. The chemical alarm system was not more sensitive in the anxious-breathing athletes. Carbon dioxide sensitivity, the thing our field talks about most, simply was not the explanatory variable.

So what is. The signal is equivalent. The chemistry is equivalent. The anxiety is not. So the difference has to live somewhere else. In how the respiratory sensation is predicted, weighted, and interpreted. Exactly where Harrison’s anxiety work pointed.

A predictive-interoceptive model of breathlessness anxiety

This is the frontier I have been building in my own theoretical work. It is an attempt to give that “somewhere else” a mechanism.

The proposal, in plain terms. Breathlessness anxiety in athletes emerges not from a more sensitive body, but from hyperprecise prior expectations around air hunger. Through repeated exposure to severe respiratory distress under competitive threat, lost control, or failed prediction, the brain can build an unusually strong, over-confident prior. This sensation is dangerous. “Think identity rather than dangerous to the body” Once that prior is set, and held too rigidly to be updated by contradicting evidence, it generates a large affective response to a respiratory signal that is not objectively more threatening than anyone else’s. The alarm is in the prediction. Not the lungs.

A few threads from the science tie this together.

Air hunger is a central, predictive phenomenon. Not a peripheral one. The most unpleasant, most anxiety-linked quality of breathlessness, air hunger, does not arise from the breathing muscles reporting up. It arises from a copy of the brain’s own respiratory command, compared against what the body actually achieved. It is a mismatch signal. Computed centrally. Which is exactly why it can be changed centrally. Studies show that hypnosis, and even deep brain stimulation, can substantially reduce air hunger without changing a single ventilatory variable. The stimulus stays fixed. The percept changes. That is the interpretive layer, laid bare.

There are two pathways. Anxiety rides the affective one. Breathlessness is processed along a discriminative pathway, the intensity, the how much. And a separate affective pathway through limbic regions like the amygdala and insula, the how bad, the how threatening. Athletes with breathlessness anxiety appear to run hot on the affective pathway, while the discriminative signal stays unremarkable. They are not more accurate. They are more alarmed. And high symptom-reporters often show reduced discriminative accuracy. So the anxiety is a marker of miscalibrated threat appraisal. Not superior awareness. Same shape as Harrison’s finding. Arriving from the athlete direction.

The brain builds this predictively. Imaging of these same athletes showed that anticipatory brain activity, activity before the sensation even arrived, tracked their breathlessness ratings. In a pattern reversed in non-athletes. The trained brain runs a feedforward model of incoming respiratory sensation. In someone with well-calibrated priors, that is adaptive anticipation. In someone with hyperprecise threat priors, it is pre-loaded alarm.

Why this reframes the intervention

Here is why this is not just interesting neuroscience. It changes what you are actually working on as a practitioner.

Say breathlessness anxiety were a chemosensitivity problem. Then the answer would be CO2 tolerance work. Full stop. Say it were a mechanics problem. Then the answer would be pattern retraining. Those things have value. But say the real driver is often a hyperprecise, over-trusted prior. A prediction that a normal sensation is dangerous. Now the target of the work is different. You are not primarily trying to change the signal. You are trying to change the brain’s relationship to the signal.

And the science suggests how that recalibration happens. It is not erasing the sensation. It is not grinding through exposure until it habituates. The mechanism appears to be the formation of competing, low-threat predictions. New experiences that violate the old expectation. “I felt air hunger. I stayed regulated. I was fine.” These gradually inhibit the original threat prior.

And it only works under perceived control. Clinical work on air hunger exposure shows the benefit appears when exposure happens without high concurrent anxiety. And it vanishes when anxiety during exposure is high. Exposure frequency alone does not do it. Plenty of athletes get enormous respiratory exposure and stay anxious. Because the exposure kept happening under threat. Not control.

This is what I have called air hunger tolerance. Not CO2 tolerance because it’s nothing to do with chemosensitivity. It is however, a trainable capacity to stay regulated in the presence of strong respiratory signals. Built through structured exposure under perceived control, resolved threat, and accurate prediction. It is the difference between a body that has learned the sensation is safe, and one that has only ever learned, over and over, that it is not.

The through-line

Everything here converges on a single idea. It is worth carrying into your practice. The breath you work with is not only a mechanical event. It is a predicted, interpreted, weighted experience.

The mechanics matter. CO2 matters. But two people with identical physiology can live in completely different breathing worlds. Because the brain does not simply read the body. It forecasts it. Then it largely believes its own forecast. When someone is caught in breathing distress, the most powerful lever is often not further down in the mechanics. It is up in the interpretation. In the prior. The confidence. The meaning. The sense of control.

Which is why, in this work, the most important question is often not how are you breathing. It is what does this sensation mean to you. And can we teach your brain, under the right conditions, to predict it differently.

That is the layer our field talks about least. It may be the one that matters most.

This piece draws on the published research of Olivia Harrison and colleagues, and on my own theoretical work, “Beyond Chemosensitivity: A Predictive-Interoceptive Model of Breathlessness Anxiety in Athletes.” Understanding this predictive, interpretive layer, how to assess it and how to work with it under perceived control, is a core part of how we train practitioners in the Breath Science Practitioner Certification.