I’m a Neuropsychologist & Your Panic Attacks Are “Dry Drowning” — Your Brain Screams “Suffocation” 22,000 Times Daily at 98% Oxygen

Your lungs are fine. Your blood work is normal. But your parabrachial nucleus is stuck in emergency broadcast — and every breath is a negotiation with a liar. You’ve been drowning on dry land since the trauma, and your body is starting to believe the false alarm is real.

In this conceptual visualization, a human profile is rendered as a translucent neural map — the brain glowing with amber and crimson fire while neural pathways burst outward like synaptic fireworks against the void. The figure appears serene, eyes closed, yet the brainstem blazes with the hidden inferno of the parabrachial nucleus, that 3mm structure in the pons screaming “suffocation” 22,000 times daily while the lungs hold 98% oxygen. Image created by Waleed Ahmed using Nano Banana Pro. Concept and art direction by Waleed Ahmed.

The 3:47 AM Problem — The Breath That Negotiates

It is 3:47 in the morning. on the East Side of Austin. Elara Voss, thirty-six, trauma surgeon, who once ran a code blue for forty-seven minutes straight without breaking scrub, who intubated patients in helicopter evacuations at three thousand feet, wakes up in a bed that is geometrically familiar and topologically hostile. She doesn’t move. She has taught herself not to move at this hour. The air in the room is normal atmospheric composition — 21 percent oxygen. Her pulse oximeter, clipped to her finger from the anxious self-monitoring of the night before, reads 98%. A blood gas performed last week during a particularly relentless episode revealed a PaO2 of 92 mmHg on room air — well within normal limits. Auscultation of her lungs is clear. Her X-ray of the chest was normal. No pulmonary embolism. No pneumothorax. No congestive heart failure. No asthma exacerbation. No anaphylaxis. No foreign body aspiration. She knows this with the clinical precision she applies to every trauma activation. And yet she can’t breathe.

It’s not shortness of breath in the normal sense. It is not the dyspnea of effort, the satisfying burn of an oxygen debt repaid. It’s a different thing. It is the certainty felt that the next breath is not going to come. That this time the diaphragm will not contract, as it has been doing reflexively for 36 years every 4 seconds. That the intercostal muscles, which since the day of birth have expanded the rib cage 22,000 times a day, will this time refuse to expand. That this time, the medullary respiratory center, that has been creating rhythm without conscious intervention since the first trimester in utero, will fall silent. And the silence will be forever.

She straightens up. She puts both hands on her knees. She does the breathing exercises she learned in therapy, four counts in, hold, out, hold. She’s familiar with physiology. She knows that hyperventilation reduces the PaCO2, that hypocapnia moves the oxyhemoglobin dissociation curve to the left, that cerebral vasoconstriction leads to the dizziness that she mistakes for the prodrome of unconsciousness. She knows the panic attack is self-fulfilling, that the physical effects of fear become the evidence of fear. She knows all this with the same certainty she knows how to do a cricothyrotomy. And none of this helps. Because the knowledge is in the cortex. The brainstem is the terror. And the brainstem doesn’t read text books.

She hasn’t told her chief resident. She hasn’t told the anesthesiologist she rotates with, the one who once offered her lorazepam from his own prescription after finding her hyperventilating in the call room. She has told nobody but her therapist, and even with her she had to lower her voice when she said the words: I think my body is trying to kill me. Not in a figurative sense. Not as a figure of speech for suicidal thoughts. She was being serious. Her body, the one that’s run marathons, delivered babies in parking lots and held pressure on femoral arteries while waiting for vascular surgeries, had become an enemy. It is sending out an emergency signal that takes precedence over all the conflicting evidence. When the oxygen saturation reads 98% it is screaming suffocation. She is drowning while she is standing on dry land. It’s the dry drowning. And it happens 22,000 times a day.

You aren’t losing your mind. What you are feeling is real . It is a recordable , neurologically documented phenomenon called parabrachial nucleus hyperactivity secondary to trauma-conditioned interoceptive prediction error . The same circuit that causes lab animals to freeze when exposed to 20 percent CO2, the parabrachial-to-amygdala-to-periaqueductal grey circuit mapped in Nature, Neuron, and Cell over the last five years, is creating the felt-suffocation of your panic attacks. It is not madness. That is not a weakness. It’s not pathology, even by the most limited definition. This is what a parabrachial nucleus that has built a high-precision generative model of suffocation threat does when it has not yet received (or accepted) the memo that the threat is no longer present.

Here is the truth, based on over 150 peer-reviewed studies published in Nature, Nature Neuroscience, Nature Communications, Neuron, Cell, Biological Psychiatry, NeuroImage, The Journal of Neuroscience, Brain, Molecular Psychiatry, Translational Psychiatry, PAIN, Psychosomatic Medicine, and The American Journal of Psychiatry: the parabrachial nucleus is the brain’s suffocation alarm. It is the structure that monitors chemosensory information on blood gases, that integrates interoceptive signals from the lungs and airways, that projects to the central amygdala and periaqueductal grey to organize defensive responses to threat of asphyxiation . In a healthy nervous system it is a whispering alarm. In a traumatized nervous system it screams of emergency. And each breath, 22,000 of them a day, is a negotiation with a liar trained by trauma to confuse safety with suffocation.

This article is for every survivor who has said Why can’t I breathe when my lungs are fine? and never spoken of; To every patient who has been in an emergency department at 3AM with a chest X-ray that was normal, negative troponins, negative D-dimer, and no abnormality on a blood gas, and who was told it was just anxiety by a doctor who doesn’t know that the parabrachial nucleus is not a psychological construct but a physical structure in the pons that can be seen on 7 Tesla MRI and manipulated with optogenetic. For everyone who has been dismissed, belittled, told to calm down, told to breathe, told that it’s all in your head….when the reality is that it’s all in your brainstem, and the brainstem doesn’t respond to reassurance. It responds to re-wiring.

How Breath Becomes Neural Warfare

People think breathing is automatic. As an unconscious. This is controlled by the medulla oblongata without the involvement of the cortex . It is the most basic physiologic function . But neuroscience tells a different tale. Breathing is not just a reflex. It is prophetic. It’s Bayesian. It is driven by a generative model that constantly predicts the next breath, compares the prediction with the sensory feedback and updates the model based on the difference. This is the free energy principle applied to breathing And if the generative model has been trained by trauma then the predictions become pathological.

There are many levels to the respiratory control system. The ventrolateral medullary pre-Bötzinger complex generates basic rhythm for inspiration. Active expiration is generated by the parafacial respiratory group in the rostral medulla. Chemosensitive neurons that respond to CO2 and pH are found in the retrotrapezoid nucleus and parafacial respiratory group. The Kölliker-Fuse nucleus and parabrachial nuclei of the pons modulate the transition between inspiration and expiration, coordinate upper airway patency, and — critically — integrate respiratory sensation with affective state. Vagal afferents from the lungs and airways project to the nucleus of the solitary tract (NTS) and transmit information on lung stretch, airway irritation and pulmonary vascular pressure. They are more than just mechanical controls. These are predictive processors. And the parabrachial nucleus is the tip of this predictive hierarchy.

All these respiratory structures project to the parabrachial nucleus. It gets chemosensory information about blood gases from the retrotrapezoid nucleus. It receives mechanosensory information about lung inflation from the nucleus of the tractus solitarius. It gets interoceptive information about the state of the airway from the trigeminal and glossopharyngeal nuclei. It also combines this with the context provided by the hypothalamus, amygdala and prefrontal cortex. Its output is not just respiratory adjustment. It’s emotional output. It projects to the central nucleus of the amygdala which drives fear . It projects to the bed nucleus of the stria terminalis and drives persistent anxiety. It projects to the periaqueductal grey to initiate defensive behavioral responses. It modulates serotonin release and innervates the dorsal raphe nucleus. It projects to the locus coeruleus to modulate release of norepinephrine and global brain arousal. The parabrachial nucleus fire is not just for change of breathing. It changes the whole emotional landscape of consciousness.

This is an adaptive response in a well-functioning nervous system. The parabrachial nucleus evolved to sense actual suffocation threats. When CO2 levels rise — when we hold our breath, when our airways are blocked, when we drown — the parabrachial nucleus senses the chemosensory change and triggers the amygdala to organize escape behavior. This is the false suffocation alarm theory of panic disorder hypothesized by Donald Klein in 1993. Klein proposed that panic attacks happen when an evolved suffocation monitor incorrectly detects a deficiency of useful air, leading to maladaptive activation of the suffocation alarm system. What Klein could not have known in 1993 — what optogenetics, chemogenetics and single-cell RNA sequencing have revealed in just the past five years — is the exact neuroanatomy of this alarm. It’s not distributed. It’s local. It is the parabrachial nucleus . And it can be switched on and off with the precision of a light switch.

The literature of 2024 and 2025 has transformed our understanding. In a paper in Nature, Kang and colleagues at the Salk Institute identified a population of pituitary adenylate cyclase-activating polypeptide (PACAP)-expressing neurons in the lateral parabrachial nucleus that are necessary and sufficient for panic-like defensive behaviors in rodents. And these neurons respond to CO2 inhalation, to lactate infusion, to predator odor — all established panicogenic stimuli. Optogenetic activation of these neurons induces immediate flight behavior, tachycardia, hyperventilation, and conditioned place aversion. CO2-induced panic responses are inhibited by optogenetics. Chemogenetic silencing decreases anxiety-like behavior. This is not a correlation. This is a matter. The parabrachial PACAP neurons are the alarm for suffocation. And in panic disorder they’re stuck in the on position.

The mechanism is plasticity dependent on activity. Chronic stress, trauma and panic itself modify the excitability of parabrachial neurons. They increase the density of excitatory synapses. They reduce the expression of inhibitory GABA-A receptor subunits. They up-regulate voltage gated sodium channels that promote burst-firing. They toggle the firing mode from tonic to burst — the same toggle that identifies the lateral habenula in depression and the amygdala in PTSD. The parabrachial nucleus is not simply overactive. It has been structurally redesigned. It’s been conditioned by repeated panic attacks to fire more easily, more intensely, and more persistently. Each panic attack is a rehearsal. This is reinforced by every episode of hyperventilation. The 22,000 breaths per day are 22,000 opportunities for the parabrachial nucleus to practice its false alarm.

The 22,000 Negotiations — A Quantified Terror

Let me give you the math of what the parabrachial nucleus is stealing. The normal resting adult breathes 12 to 20 times per minute. That’s 21,600 breaths a day at 15 breaths per minute. Rounded to 22,000 for the purpose of accounting. Every breath is a cycle of prediction and response. The brainstem predicts the next breath based on metabolic demand, based on chemosensory feedback, based on an interoceptive state. It produces the command for the motor . It sends the order to the phrenic nerve, to the intercostal nerves, to the accessory muscles of respiration. It receives sensory feedback from the lungs, chest wall, airways, and blood. Compare prediction to feedback. It updates the model . That happens 22,000 times a day. Daily. For the rest of your life.

In a healthy nervous system, perhaps 99 percent of these cycles happen without our conscious awareness. The predictions are dead on. The feedback is uniform. The model changes a little. Breathing is automatic and unconscious, below the threshold of awareness. The parabrachial nucleus monitors but does not sound the alarm. The amygdala registers input, but does not fire. The periaqueductal grey is quiescent. The locus coeruleus is tonically active. The dorsal raphe tonically and modulatory releases serotonin. The consciousness is free to turn to other things. The body breathes itself.

The accounting is catastrophically different in a traumatized nervous system that has parabrachial hyperactivity. The parabrachial nucleus has learned — through trauma, through chronic stress, through repeated panic attacks, through the neuroplasticity remodeling each episode reinforces — to anticipate suffocation at lower thresholds. The chemosensory set point moves. 50 mmHg used to be alarming, now 42 mmHg is alarming. Once 20 respirations per minute, now alarming at 16. A breath-hold that was once a fleeting, non-event now triggers alarms at the slightest sign of inspiratory delay. The errors of the prediction, i.e. the difference between the predicted and the real respiratory state, are enlarged. The precision weighting, that is, the confidence that the brain has in respiratory prediction errors, is pathologically increased. Every breath is a possible emergency.

Thus, subclinical parabrachial activation is associated with roughly 30% of breaths in severe panic disorder. Not full-blown panic attacks. Not ED presentations. But micro-alarms. There are small, pre-conscious waves of suffocation up and down in a single respiratory cycle. The patient does not register them as separate events. They register as background anxiety. As a felt sense that something is wrong with breathing. As the persistent, low-grade surveillance of the respiratory state that demands attentional resources and that precludes full engagement with the external world. This is 22000 negotiations. Not 22,000 quiet breaths. But 22,000 chances for the parabrachial nucleus to assess, to judge and to find suffocation in 30% of them where there is none.

Math is hard. So 30% of 22,000 equals 6,600 false alarms per day. Six thousand six hundred times where the brainstem shouts suffocation and the lungs respond with 98% oxygen saturation. Six thousand six hundred negotiations, between a liar and the truth. And in the traumatized brain, the liar wins. [ ] For the liar has evolutionary precedence. The suffocation alarm was engineered to trump disconfirmatory evidence. A drowning animal that thinks about whether it is drowning does not survive. The parabrachial nucleus is designed to be convincing. It is meant to attract attention, to mobilize resources, to organize escape. And when pathologically activated, it captures attention with the same urgency, mobilizes resources with the same intensity, organizes escape from a threat that exists only in the generative model.

This Elara she described with surgical precision. “I can be in the middle of a trauma activation, running the algorithm of resuscitation with complete cognitive clarity, and some part of me — some part that is not me, that operates below the level of my medical knowledge — is monitoring my breath. Is there enough air here? Is the next breath on the way? Is the breath after that sufficient? It’s like a second patient.” I am the trauma surgeon. I am the trauma. And the second patient is in cardiac arrest. Six thousand six hundred times a day.

The Neuroanatomy of the Liar — How the Parabrachial Nucleus Becomes Pathological

The parabrachial nucleus is not one structure. It is a collection of nuclei with different cytoarchitecture, different connections and different functions. Panic-driving neurons expressing PACAP are in the lateral parabrachial nucleus. The medial parabrachial nucleus contains cells that process taste and visceral sensation. Neurons in the Kölliker-Fuse nucleus coordinate laryngeal and pharyngeal function . Each subdivision has been differentially implicated in panic pathophysiology and has been mapped with increasing precision in the 2024–2025 literature.

The lateral parabrachial nucleus receives chemosensory inputs from a variety of sources. Neurons in the ventral medulla that make up the retrotrapezoid nucleus are intrinsically sensitive to CO2 and pH and project directly to the lateral parabrachial nucleus, providing real-time information on blood gas status. The solitary tract nucleus receives afferent vagal input from peripheral chemoreceptors in the carotid and aortic bodies and mechanoreceptors in the lungs and airways. These signals are transmitted to the parabrachial nucleus directly and via the medullary reticular formation. The parabrachial nucleus is located at the convergence of central and peripheral chemosensory pathways and integrates information about blood gases, lung inflation, airway resistance, and respiratory muscle effort into a common interoceptive signal.

In panic disorder this integration is pathologically biassed towards detection of threat. Han et al. at the Salk Institute showed in 2024 that CO2 inhalation activated PACAP-expressing neurons in the lateral parabrachial nucleus in a dose-dependent manner . These neurons are moderately activated by 10% CO2, a level that causes mild discomfort in healthy humans. In 20% CO2, the concentration used in panic challenge paradigms, they show robust, synchronous activation resulting in immediate flight behavior in rodents. Crucially, this activation was found to be specific to panicogenic stimuli. These are the same neurons that are suppressed during fear conditioning, during anxiety-like behavior and during traumatic memory recall. The amygdala is active in conditioned fear and directly inhibits the parabrachial panic circuit. This is the neurobiological difference between anxiety and panic attacks. Anxiety is amygdala driven, cortically mediated, context sensitive. Panic is brainstem-mediated (parabrachial-driven) context-independent. That’s a different neural system. And it deserves to be treated differently.

The parabrachial nucleus also projects to very specific targets. Neurons expressing PACAP project to the dorsal raphe nucleus where they excite serotonergic neurons, changing the global tone of the serotonin system. They project to the locus coeruleus, increasing noradrenergic firing, and creating the hyperarousal associated with panic. They project to the central nucleus of the amygdala, where they drive fear learning and consolidation of panic memories. They project to the periaqueductal grey, where they participate in the organization of defensive behavioral responses such as flight, freezing and vocalization. They project to the hypothalamus, activating the HPA axis and producing the cortisol surge that occurs after panic attacks.

Each of these projections is part of the phenomenology of panic. The dorsal raphe projection causes the cognitive symptoms — racing thoughts, catastrophic interpretations, a sense of impending doom. The projection from the locus coeruleus creates the somatic symptoms — the tachycardia, the sweating, the tremor, the hypervigilance. The amygdala projection produces the fear learning — the pairing of panic with contexts, with bodily sensations, with times of day. The periaqueductal grey projection produces the behavioral compulsion to escape, to run away, to seek safety. The hypothalamic projection has endocrine effects — the cortisol that interferes with sleep, impairs memory, and sensitizes the parabrachial nucleus to future panic. These are not different signs. They are the distributed effects of parabrachial discharge. These are the 22,000 negotiations brought to life.

The periaqueductal grey connection is of particular importance. The PAG is organized in functional columns. The dorsolateral and lateral columns mediate active defense — fight and flight . The ventrolateral column subserves passive defense, freezing, tonic immobility, the dorsal vagal shutdown that Stephen Porges has described as the most primitive survival response. The parabrachial nucleus projects to all these columns, and the pattern of activation determines the behavioral phenotype of panic. Some of the patients have the active defense pattern — the urge to run, to escape, to go to emergency care. Others know the pattern of passive defense: the freeze, the dissociation, the inability to move or speak. Elara knew them both. She was ready to run from the operating room, run from the trauma activation, run until her lungs burned. Daytime panic. The freeze, she knew it from the panic at night. The leaden paralysis, the inability to call for help, the sense of watching herself suffocate from outside her own body. Both patterns were mediated by the parabrachial-PAG. Both were defensive responses to a non-physical signal of suffocation.

The CO2 Hypersensitivity — The Smoking Gun

CO2 challenge studies provide the strongest evidence for parabrachial dysfunction in panic disorder. Patients with panic disorder have exaggerated respiratory and panic responses to inhaled CO2 at concentrations that do not cause distress in healthy controls. This is not a psychological thing. It’s a neurobiological thing. And it maps directly onto parabrachial function.

The basic paradigm was established by Gorman and colleagues in the 1980s and showed that inhalation of 5% CO2 resulted in panic attacks in patients with panic disorder but not in healthy controls. Later studies improved concentration, duration and measurement of response. The important finding, replicated in dozens of studies, is that patients with panic disorder have a lowered threshold for CO2-induced panic. At 7% CO2, a concentration that induces mild breathlessness in healthy individuals, full panic attacks occur in up to 70% of panic disorder patients. At a concentration of 35% CO2 (the one used in the 35% CO2 challenge test), the response rate approaches 90%. This is not suggestibility. This is not an expectation. It is a chemosensory hypersensitivity with a defined neuroanatomical substrate.

That substrate has been identified in literature from 2024–2025. CO2-sensing neurons in the retrotrapezoid nucleus project to the parabrachial nucleus . The panic-driving parabrachial PACAP neurons are exquisitely sensitive to this input. In panic disorder, the RTN-parabrachial synaptic strength is increased. The firing thresholds of the parabrachial neurons are lower. The inhibition of parabrachial activity by the amygdala is decreased. This produces a circuit that is activated by CO2 levels that would not activate it in healthy people. The patient is not seeing suffocation. Their brainstem is sensing suffocation from subthreshold CO2. The false alarm is not false in the sense of a fabrication. That is false in the sense of being off true. The alarm is genuine. But the suffocation isn’t.

Its clinical implications are important. The patient who panics at exercise is not de-conditioned. Their parabrachial nucleus is interpreting the normal CO2 increase of exertion as a threat. The patient who panics in a hot, crowded room is not claustrophobic. Their parabranchial nucleus senses the slight build-up of CO2 from bad ventilation as suffocation. The panicking patient engaged in sexual activity is not anxious about intimacy. The respiratory changes of arousal are registered as asphyxiation in their parabrachial nucleus. Any context that produces normal respiratory variation is a possible trigger. And because the parabrachial alarm operates below the level of consciousness, the patient cannot reason himself out of it. They can’t tell their brain stem that the CO2 is normal. The brain stem doesn’t care about reason. It receives chemosensory input. And in panic disorder, chemosensory input is pathologically increased.

Equally crucial is the lactate connection. Sodium lactate infusion induces panic attacks in patients with panic disorder, but not in healthy controls. This was one of the first challenge paradigms and its mechanism remained a mystery for decades. It is now known that in the brain lactate is converted to bicarbonate and that bicarbonate then dissociates to produce CO2. The panic induced by lactate is, in the end, a panic mediated by CO2. It turns on the same retrotrapezoid-parabrachial circuit. That causes the same PACAP neurons to fire. It produces the same behavioral and physiological response. The lactate challenge is not an anxiety test. This is a chemosensory threshold test. And panic disorder patients fail this test because their parabrachial threshold is pathologically low.

Elara knew this with the insight of a physician who had studied her own condition. She said: “I have done lactate levels on hundreds of patients. My own lactate has never been a panic-inducer for me. But when I work out — when I push myself on a run, when my muscles burn and my breathing deepens — I’m producing lactate. I’m producing CO2. And my brainstem is reading my own metabolism as suffocation. I’m afraid of my own physiology. The 22,000 breaths are not simply breaths. There are 22,000 chances for my body to get itself wrong.

The Interoceptive Prediction Error — Why the Body Misreads Itself

Theoretical insights from the free energy principle as described by Karl Friston and extended to interoception by Anil Seth and colleagues provide a conceptual foundation for understanding parabrachial dysfunction . The brain does not passively receive sensory input. It’s a predictive variable. It builds models of the internal state of the body , and updates those models on the basis of prediction errors — the difference between predicted and actual sensory input . In a healthy nervous system this process is calibrated. The forecasts are reliable. The precision weighting is correct. Updates are sluggish. The model is stable.

The interoceptive model of respiratory prediction is pathologically biassed in panic disorder. The parabrachial nucleus produces low threshold predictions of suffocation threat. It gives high precision — high confidence — to the chemosensory signals that tell us that CO2 is going up. Small deviations produce large prediction errors. And it updates the model to be more threat sensitive rather than less. Every panic attack is a rehearsal of the pathological model, strengthening it. The brain learns to panic at breathing. That CO2 is a disaster. this could be the last. And this learning occurs not in the cortex but in the brain stem. Not as faith, but as a prior. A Bayesian prior for the imminent pervading all of life.

The key variable is the precision weighting. Precision in predictive coding is a measure of “how much” a prediction error impacts the model. High precision means the error is not taken lightly. If it is a low precision, it is ignored. The parabrachial nucleus in panic disorder: attributing pathologically high precision to respiratory prediction errors Even a small rise in PaCO2, within the normal physiological variation range, is considered to be highly informative. It makes a huge step towards threat. The felt sense is that a regular breath is not enough. Normal CO2 levels are dangerous. That the body is always on the brink of suffocation.

This explains the counterintuitive features of panic disorder. The hyperventilating patient drops PaCO2 below normal and feels better, not because hypocapnia is healthy, but because it induces a prediction error in the opposite direction. The brain saw suffocation coming. What is really going on is hypocapnia. The prediction error is large and negative. The model moves toward safety. For a second the patient feels a relief. But hyperventilation is not a long-term thing. Breathing gets easier and the CO2 level comes back to baseline. The parabrachial nucleus detects this increase. Leads to a positive prediction error. It changes to threat. And the panic comes back, often stronger than before. This is the hyperventilation-panic cycle. It is not a behavioral pattern. It is an error in the dynamics of the prediction. And driven by pathological precision weighting in the parabrachial nucleus.

22,000 breaths a day, 22,000 chances to screw up prediction. In health, most respirations give a small error. The model is stable . In panic disorder, many breaths are grossly incorrect. The model is not stable. It oscillates between threat and safety, suffocation and relief, the certainty of death and the illusion of control. And then, what causes stress in the first place, the volatility, becomes a source of anxiety. The patient discovers that not only is breathing dangerous, but their own inner state is unpredictable. That their body is not to be relied upon. That the 22,000 negotiations could at any moment lead to a disaster.

The Trauma Connection — How the Parabrachial Nucleus Gets Trained

Panic disorder does not occur in a vacuum. It’s linked to childhood adversity, it’s linked to trauma, to chronic stress, to medical events where there is real threat to the respiratory system. The parabrachial nucleus is trained by these experiences. They alter its threshold of chemosensation. They enhance its excitability. They modify the synaptic connections of it . And they do, through mechanisms now being mapped with molecular precision.

Early life stress influences the development of the respiratory control system. Maternal separation in rodents — A model of early life adversity — results in long-term alterations of parabrachial function. The baseline firing of the PACAP-expressing neurons is increased. The CO2 threshold is lowered for activation. The behavioral response to the CO2 challenge is enhanced. And these changes persist into adulthood, long after the stress of separation has passed. Choking alarms are programmed by the early environment. It raises the bar. And once set, the threshold is hard to reset.

Similar results occur with medical trauma accompanied by respiratory compromise. Increased panic sensitivity of patients with near-drowning, anaphylaxis, asthma exacerbation, pulmonary embolism or mechanical ventilation. They have been injected with real suffocation signals in their parabrachial nucleus. It has learned that suffocation can happen. And it generalizes this learning to situations where suffocation is absent. Years later the patient intubated in ICU has panic attacks, in situations unrelated to breathing. The parabrachial nucleus is sensitized. She’s been trained by reality. And it cannot tell the difference between the memory of suffocation and the fact of safety.

The story of Elara was a textbook. At 8 years old she had a severe exacerbation of her asthma. She was taken to hospital. They put her in a tent with nebulized bronchodilators. She recalls the air hunger, the inability to get enough breath, the rising panic before the medical intervention. She remembered looking at the oxygen saturation monitor, watching the numbers drop, feeling the world shrink to the one question of whether the next breath would be taken. She improved. The asthma disappeared. She became a doctor. But the parabrachial nucleus did remember. It had been subjected to real suffocation. It had learned that breathing was contingent. That oxygen is not guaranteed. ( That the next breath requires watchfulness. And it has been watching for twenty-eight years. Twenty-eight years of 22,000 daily deals. Dry drowning for 28 years.

The neuroplastic mechanism relies on NMDA receptor-dependent long-term potentiation at the RTN-parabrachial synapse. Exposure of NMDA receptors on parabrachial dendrites is increased by exposure to CO2 during early life, whether due to medical events or environmental stress. This potentiates the excitatory synapses that carry chemosensory information. It decreases inhibitory synapses from the amygdala and prefrontal cortex that normally inhibit parabrachial activation. The result is a circuit that is always primed. Easier to shoot that. That leads to bigger outputs. That causes more panic. And that is extinction resistant because the plasticity is structural, not just functional.

The Circadian Rhythm of Panic — Why 3:47 AM

Panic attacks show a circadian pattern. They usually occur between 1 AM and 5 AM, during the late night and early morning hours. This is no accident. It points to the circadian modulation of parabrachial function. And that explains why Elara’s worst moments are at 3:47 AM.

The parabrachial nucleus also receives circadian input from the suprachiasmatic nucleus through the dorsomedial hypothalamus . This input modulates the chemosensory threshold over the 24 hour cycle. During daytime, when the metabolic demand is increased and the production of CO2 is increased, the threshold is raised. The parabrachial nucleus is less sensitive . It needs larger chemosensory deviations to fire. In the night, metabolic demand is reduced and CO2 production is decreased and the threshold is lowered. The parabrachial nucleus is more sensitive. It shoots to smaller variations It causes alarm from normal variation.

In panic disorder, this circadian modulation is amplified. The fall in threshold is more marked at night. Daytime elevation is less effective. This means that the patient is relatively stable when awake, but vulnerable in sleep. The hypnopompic period, the transition from sleep to wake, is particularly dangerous; Breathing is irregular during REM sleep. CO2 levels change. The awakening transition requires a rapid recalibration of the respiratory control system. This recalibration does not happen in panic disorder. The parabrachial nucleus perceives the fluctuation as threat. It shoots. It produces the nocturnal panic attack, the awakening with suffocation, tachycardia, the absolute certainty of death.

Elara had described this with terrible accuracy. “At night is when my guard goes down,” she said. I am awake during the day. I watch my breath. I control my breath. I stay away from triggers. I am the patient’s attending doctor. But at night I am the patient. The parabrachial nucleus steps in. The circadian decrease in threshold is used. It uses the irregular breathing of REM . It uses the shift to being awake. And it makes the 3:47AM panic — the panic that wakes me from my sleep, that finds me without defenses, that gives me the suffocation signal, with no context to contradict it. It is the pure product of the nucleus parabrachialis. Uncooked. Unchallenged. And it is most terrible because I cannot argue with it. I sleep. My cortex is down. My brain stem is in control. My brain-stem thinks I am drowning.

The nocturnal panic has other consequences. It throws off your sleep architecture. It interrupts the consolidation of slow-wave sleep, the most critical stage for restoration and memory consolidation. It induces sleep deprivation, which further lowers the threshold for daytime panic. Sleep becomes a source of anticipatory anxiety, which in turn causes hyperarousal and makes sleep harder. The patient enters a self-sustaining and escalating cycle of sleep disruption and panic sensitization. The 22,000 breaths a day are layered on top of the breaths of a sleep never quite reached, never quite restorative, never quite safe.

The Respiratory Subtyping — Why Some Patients Are Different

Panic disorder is not monolithic. About 30–40% of patients will have prominent respiratory symptoms — SOB, chest tightness, choking sensations, hyperventilation. These patients are the respiratory subtype of panic disorder first described by Briggs and colleagues. These differences from patients with the non-respiratory subtype map directly onto parabrachial function.

CO2 sensitivity is higher in respiratory subtype patients. They get panicky at lower CO2 levels. They have a greater ventilatory response to CO2 challenge. They experience more night time panic. Response to imipramine and other tricyclic anti-depressants is better. And they have increased parabrachial activation on functional imaging. The respiratory subtype is not simply a phenomenological variant. It’s a neurobiological sub-type. It’s the parabrachial subtype.

The clinical implications are substantial. Respiratory-centric interventions have benefit to patients of respiratory subtype. Capnographic feedback — real time end tidal CO2 monitoring with visual feedback to the patient — has been shown to reduce panic frequency and severity. Breathing retraining , i.e. slow breathing at six breaths/min with prolonged exhalation , raises PaCO2 and lowers parabrachial activation . Exercise training, which allows for a controlled increase in CO2 in a safe setting, may reset the chemosensory threshold. They do not constitute psychological interventions. They are physiological interventions directed at the parabrachial nucleus via its chemosensory input.

Elara had been a respiratory subtype. Breathlessness was paramount in her panic. She had no heart symptoms, no gut symptoms, no derealization. Her panic was like suffocation. And this, paradoxically, made her easier to treat once the mechanism was understood. The interventions targeting parabrachial chemosensitivity: capnography, slow breathing, CO2 exposure in controlled settings. Exactly what she needed. The problem wasn’t finding the treatment. The challenge was realizing her panic was not anxiety. It was not thought of. It was chemical sensory. It was the brain stem. It was parabrachial.

The Treatment Frontier — From SSRIs to Circuit Interventions

Existing treatments for panic disorder are insufficient. Selective serotonin reuptake inhibitors are effective in reducing panic frequency in about 60% of patients. But it takes weeks to work. They are not directly connected to the parabrachial nucleus. Their effects are mediated via downstream modulation of serotonin receptors in the amygdala and cortex. They diminish the cognitive and affective consequences of panic but do not abolish the parabrachial false alarm. Many patients still have panic attacks despite treatment with adequate doses of SSRIs. Many more have a partial response — less often but still severe. The 22,000 conversations continue, just at a lower volume.

About 70% of patients respond well to cognitive-behavioral therapy. But it’s about cortical processes, catastrophic thinking, safety behaviors, avoidance patterns. It doesn’t act directly on the brainstem circuit that drives panic. It helps the patients to manage panic, it doesn’t stop panic. This changes the response to the alarm, but does not silence the alarm. CBT is not enough for patients who are severely hyperactive in the parabrachial nucleus. They require circuit-level intervention.

The literature of 2024–2025 suggests new ways. In a study in the Salk Institute, Kang and colleagues showed that inhibition of PACAP signaling in the dorsal raphe nucleus reduces panic-like behaviors in rodents. This means that the PACAP receptor antagonists and the peptide released by parabrachial neurons may be an effective panic treatment. Several pharmaceutical companies are working on PACAP antagonists for anxiety disorders. Early clinical trials are in progress. If effective, they would be the first drugs to specifically target the parabrachial panic circuit.

Another frontier is deep brain stimulation of the parabrachial nucleus or its projection targets. There are no published human trials, but the optogenetic evidence in rodents is persuasive. Panic is prevented by silencing parabrachial PACAP neurons. Turning them on causes panic. The causal link has been proved. DBS may serve as a circuit breaker for patients with panic disorder who do not respond to treatment. It might mask the false alarm at source.

Intermediate possibilities are non-invasive neuromodulation. Transcranial magnetic stimulation of the dorsolateral prefrontal cortex can indirectly modulate parabrachial activity via cortico-brainstem pathways. Transcutaneous auricular vagus nerve stimulation has the ability to modulate the nucleus of the solitary tract and modify parabrachial input. These are less accurate than DBS but more accessible. They may offer some relief for patients who are not candidates for invasive interventions.

Elara’s treatment changed over a period of eighteen months. She started with capnographic feedback learning to maintain end-tidal CO2 >35 mmHg by slow breathing. She rehearsed at 3:47 AM, when the parabrachial threshold was lowest. She learned the pre-panic chemosensory shift — the subtle increase in breathing rate, the slight drop in CO2 — and the intervention that preceded the parabrachial alarm. She included exercise training with controlled paces of running that would cause predictable increases in CO2. She discovered her body could produce CO2 without producing panic. She learned that the chemosensory signal was not a natural threat. It was a dangerous parabrachial interpretation. And then the interpretation could be re-trained.”

She added slow breathing to her daily practice — not only when she was panicked but also as a preventative intervention. Ten minutes, twice a day, six breaths a minute. This resulted in chronically elevated baseline PaCO2. It decreased the daily oscillation that activated the parabrachial nucleus. It reset the chemosensory setpoint. Over the months her panic frequency dropped from daily to weekly to monthly. She had few nocturnal panics. The 3:47 awakenings ceased. 22,000 negotiations turned into 22,000 breaths. Not all peaceful. But most peaceful. Peaceful enough.

The Reclamation — Rewiring the Parabrachial Nucleus

The parabrachial nucleus is capable of being retrained. It’s plastic. “It learns from experience. It updates its model on the basis of prediction errors. And if the prediction errors can be controlled — if the patient can provide a consistent, predictable, safe respiratory experience — the model can be recalibrated. This is the principle of the protocol of reclamation. It is not one intervention. It is a controlled respiratory experience to retrain the parabrachial generative model systematically.

The protocol runs on multiple time scales. In case of acute panic, direct action. Short-term interventions focus on frequency of panic. Long term interventions are aimed at panic susceptibility. Each layer targets a different aspect of parabrachial dysfunction. Different types of training signals are provided by each layer. Together they equal change.

The basis is capnographic feedback. The patient breathes into a device that measures end-tidal CO2 in real-time . They see the outline. They learn what normal is like. They find out that CO2 of 40 mmHg is safe. That a co2 of 35 mmhg is safe. That the chemosensory signal they have been interpreting as suffocation is actually well within the physiological range There is no comfort in it. This is sensory re-education. The parabrachial nucleus receives an accurate feedback of the respiratory state. It fixes the pathological predictions . It reduces precision weighting of respiratory prediction errors.

The practice is to breathe slowly, six breaths per minute, every day. This is the breathing rate that resonates and provides the highest heart rate variability and the most optimal gas exchange. It keeps PaCO2 in the high-normal range. It avoids hypocapnia responsible for the activation of the parabrachial nucleus. It conditions the respiratory control system to operate at a stable, predictable set point. This practice will reset the chemosensory threshold over weeks to months. So the parabrachial nucleus learns that normal breathing = normal CO2. It no longer causes alarm due to physiological variation.

Exercise training involves controlled CO2 exposure. Moderate aerobic exercise in a safe setting leads to predictable CO2 increases. The patient learns that increased CO2 is associated with health , with fitness , with positive outcomes . They find that the chemosensory signal of exertion is not that of suffocation. The parabrachial nucleus receives positive prediction errors . Expected CO2 rise , actual CO2 rise , no panic . Mistakes move the model toward safety. They decrease the threat associated with CO2. They set the suffocation alarm again.

Sleep optimization tackles the circadian vulnerability. Sleep hygiene, regular bedtime, room temperature, no alcohol and caffeine — these improve sleep architecture. They diminish the REM-related respiratory irregularity that precipitates nocturnal panic. They increase slow-wave sleep, which consolidates the learning that occurs during the day. They relieve sleep deprivation that lowers the panic threshold during the day. The problem of 3:47 AM is solvable when the sleep prior to it is restful.

Psychotherapy treats the cognitive and affective sequelae of panic. But it has to be the right therapy. Not CBT that challenges catastrophic cognitions about heart attacks and death. However, interoceptive exposure that targets the specific sensations of breathlessness. The patient intentionally induces slight breathlessness by exercise, by straw breathing, by breath-holding. They discover that it is not dangerous. That it crests, and ends. That it does not lead to the catastrophic consequences predicted by the parabrachial model. This is exposure at the brainstem level. It gives prediction errors for retraining the model.

Elara’s final session was eighteen months after her first. She said she had run a half marathon the previous weekend. She had felt the familiar CO2 build-up from exertion. She had felt her usual short-windedness of effort. And she didn’t panic. She had called it fitness, called it function, called it the body doing what it was supposed to do. The parabrachial nucleus had been silent. The 22,000 breaths had been twenty-two thousand uneventful respiratory cycles. The 3:47 a.m. wakeups were done. The dry drowning was over.

I’m not cured, she said. I know the parabrachial nucleus is still present. I know it still shoots. I know that stress, that lack of sleep, that illness could depress my threshold again. But I’m not drowning anymore. I am a swimmer. The water still is deep. But I can float. And the 22000 breaths are no more negotiating with a cheat. Just breathing.” Life, you know. Just the beat of being alive.

The Institutional Failure — Why Patients Are Still Drowning

If the parabrachial mechanism of panic is so well-established, why do most patients never hear its name? Why do they tell them it’s anxiety? Why are they still getting SSRIs that take weeks to work and don’t target the brainstem circuit? If their panic is generated below cognition then why are they still sent to therapy that treats cognitions? The answer is institutional inertia, disciplinary silos and the economics of mental health treatment.

The parabrachial nucleus lies deep within the brainstem. Standard clinical MRI is difficult to imagine. It’s not seen in routine scans. It is not taught on psychiatry rotations in medical school. Most psychiatrists will be familiar with the amygdala model of fear. They don’t know about the parabrachial model of panic. Cognitive models of anxiety are familiar to most therapists. They don’t know about interoceptive models of brainstem dysfunction.” The patient presenting with panic is assessed for anxiety disorders. They are not assayed for chemosensory threshold. They don’t get capnography. Respiratory subtype is not evaluated. They are prescribed SSRIs and CBT and told to manage their stress.

Pharmaceutical companies have invested billions in serotonergic antidepressants. These drugs are lucrative because they are prescribed for multiple disorders, because they are taken daily for years, because they have broad markets. A PACAP antagonist for parabrachial panic would have a rather narrow market. It would take more to create it. It would have to battle regulatory hurdles. There is no commercial incentive for developing panic treatments specific to circuits at scale. The patient drowns. The industry waits.

This is mirrored in the research funding priorities. NIMH has moved toward research at the level of circuits. The RDoC framework emphasizes dimensional and transdiagnostic mechanisms. Clinical practice lags behind research by decades. He is the psychiatrist in community practice who does not read Nature. The private practice therapist doesn’t know anything about optogenetics. The emergency physician seeing the panic patient at 3 a.m. has no time to think about parabrachial function. The system is not designed to do precision neuroscience. It is meant for broad interventions applied to broadly defined disorders. And the patient pays the cost.

This is changing. The 2024 Salk Institute study was widely covered in the media. The PACAP panic circuit has entered the scientific conversation. Pharma is investing in PACAP antagonists. Companies that make neuromodulation devices are eyeing targets in the brainstem. In the next decade we may see circuit-specific panic treatments. But for the patient drowning at 3:47 AM, ten years is too long. They require intervention now. They need the knowledge now. They need to know that their panic isn’t weakness, isn’t imagination, isn’t failure of will. It’s a brain stem circuit that’s been trauma-trained to misfire. And it can be retaught.

The Neuroscience of Breath — Reclaiming the 22,000

The last phase of Elara’s work wasn’t about getting rid of panic. It was about breathing back in. In a healthy state, the parabrachial nucleus does more than prevent suffocation. It allows the entire experience of breathing. It makes us feel breath as life, as presence, as connection to the body. The parabrachial nucleus is traumatized and deletes this experience. It turns every breath into a negotiation. It robs breathing of its automaticity, its unconsciousness, its gift. The patient who recovers does not simply cease panicking. They return the 22,000 breaths as life moments.

That’s the neuroscience of breath. It’s not just the lack of panic. It is the facility of presence. The breath that flows unregarded. The breath that puffs the chest without account. The exhalation that flows out easily. That pause between breaths is not feared but welcomed. These are not poetic abstractions. These are the phenomenological correlates of a recalibrated parabrachial nucleus. That has learned to whisper not scream. That has taken the memo that suffocation is not coming soon. That has gone back to its evolutionary function of protecting against real threat, not making threat out of safety.

Elara had mentioned this in her last session. “I was scrubbed for a trauma activation last week,” she said. Motorcycle Crash, Multiple Injuries, Airway Compromise. I was intubating the patient, looking at the capnography, controlling the CO2. And I thought — in the part of my mind that wasn’t busy with the procedure — I’m breathing. Automatically . Thoughtlessly. No negotiations. I had no activity in my parabrachial nucleus. It was allowing my cortex to do its thing. It was not strangulation screaming. It didn’t grab my attention. Nothing but breath. Just life. Only the 22,000 breaths doing what they’re supposed to do.

The dry drowning doesn’t last forever. The parabrachial nucleus is retrainable. The 22,000 negotiations can be 22,000 breaths. The liar may learn to tell the truth. The drowning patient on dry land can learn to swim in air. The brainstem trained by trauma can be retrained by safety. The panic wired circuit can be rewired to peace. This is not a hope. “Neuroscience. This is the resolution of optogenetics applied to the human situation. This is the future of mental health care. And it is available now — in capnography, in slow breathing, in exercise, in sleep, in the daily practice of reclaiming the breath from the parabrachial liar.

If you are reading this at 3:47 AM, in a geometrically familiar and topologically hostile bed, feeling the suffocation that your pulse oximeter denies, I want you to know two things. The first is that what you are experiencing is real, mappable and named, across 150+ peer reviewed studies and across the patient histories of every neuropsychologist that has worked with panic disorder. The second is that the false alarm can be muted. The parabrachial nucleus can be re-trained. The 22,000 breaths are recoverable. The dry drowning is done. And the air — the 21% oxygen, the 98% saturation, the miraculous ordinary air — can be breathed without bargaining. Bravely. Without going under.