The Digital Overbreathing Hypothesis

Respiratory Dysregulation as a Proposed Pathway from Arousal-Weighted Digital Engagement to Sleep and Emotional Dysfunction

Locke

Revised October 18, 2026

Abstract

This paper proposes a mechanistic framework connecting sustained engagement with modern digital environments to a population-scale pattern of respiratory dysregulation, with downstream consequences for sleep architecture and sleep-mediated emotional regulation. The central claim is that the sustained cognitive and emotional arousal characteristic of contemporary technology use drives ventilation in excess of metabolic demand, producing repeated mild hyperventilation and hypocapnia in the waking state. Repeated exposure is proposed to alter chemoreflex responsiveness, with more sustained habitual overbreathing additionally recruiting renal and cerebrospinal-fluid bicarbonate adaptation. The resulting respiratory-control instability fragments the stable non-REM periods that support progression to deep sleep and access to REM, reducing restorative throughput and the opportunity for overnight emotional processing. The operative exposure is proposed to be arousal load rather than undifferentiated screen time. Its acute evening effect is predicted to depend on baseline respiratory stability: the same perturbation may have a large effect near the stability boundary and little effect when the operating margin is wide. The component links in this chain are supported, to varying degrees, by existing peer-reviewed literature drawn from psychophysiology, respiratory physiology, and sleep medicine. The integration of those links into a single chain, and the attribution of its upstream cause to digital exposure at population scale, is the novel and unproven claim. The paper sets out the predictions and the two-arm measurement program that would test it.

Introduction

The health effects of modern technology use are an active area of concern, but the dominant framing is narrow. Where technology is thought to affect sleep, the mechanism is almost always attributed to blue light and circadian displacement; where it is thought to affect wellbeing, to psychological causes, social comparison, distraction, and attentional fragmentation. These pathways are real, but taken alone they are incomplete, and they may not be the most physiologically consequential route by which sustained digital engagement reaches the body.

This paper proposes a different and largely unexamined mechanism: that sustained technology use drives a measurable change in respiratory physiology, and that this change is the upstream cause of a cascade affecting sleep and, through sleep, emotional regulation. It does not ask the established mechanisms to step aside; it adds a respiratory pathway that has been almost entirely absent from the discussion. Because the mechanism has never been isolated and measured, the task the paper sets itself is to specify it precisely enough to be tested. Each link in the chain that follows is drawn from published work; the assembly of those links into a single pathway, and the attribution of its origin to digital exposure, is the hypothesis.

The present paper follows the respiratory-sleep pathway and its secondary emotional consequences. Direct amygdalar acid-sensing mechanisms are addressed separately in the companion amygdala paper and are not required for the pathway developed here.

The phenomenon this describes is best treated not as a discrete clinical illness affecting an unlucky minority but as a graded condition distributed across the population, closer in structure to obesity than to a rare disorder: graded rather than binary, driven by a changed environment acting on an older physiology, and shifted across a whole population by a shared input rather than confined to a predisposed few. What has been missing is not the observation that modern life degrades sleep and emotional health, which is widespread, but a specific and measurable cause for it. Stress, screens, and the pace of modern life name nothing that can be measured, dosed, or changed. Attributing the effect to arousal-weighted digital exposure acting through a defined respiratory mechanism names something that can: it specifies what to measure, what to change, and how one would know whether the change worked. That is the difference between a complaint about modern life and a research program with an endpoint.

The Foundational Premise: Respiratory Control as a Modifiable Trait

section “Introduction”'s population-scale claim rests on a premise this section establishes: that respiratory-control efficiency, and carbon dioxide tolerance with it, is not a fixed biological constant but a graded, continuously distributed trait that environment can degrade, so that a shared environmental input shifts the whole population's distribution rather than afflicting a predisposed few. Two things should be kept apart. That the individual respiratory-control variables are continuously distributed and modifiable is established in the clinical and extreme-physiology literature; that they compose a single graded trait whose degradation predicts poor sleep across the population is a proposed construct, not yet validated at that scale.

The Binary Fallacy

Sleep medicine operates on a diagnostic threshold. An individual with an apnea-hypopnea index below five events per hour is classified as normal; an individual at five or above is classified as having obstructive sleep apnea. That threshold determines who is diagnosed, who is treated, and who is told their sleep is fine. It is a scoring convention rather than a physiological discontinuity. Minor changes to hypopnea scoring, such as requiring a three percent rather than a four percent oxygen desaturation, or counting associated arousals, reclassify the same person from healthy to disordered, and the change is consequential enough to raise the measured population prevalence of the disorder and to disproportionately reclassify women, who more often present with the milder desaturations and arousal-based events that a four-percent rule misses¹. A boundary that shifts that far with a change in convention is not marking a natural kind.

More consequentially, the respiratory-control variables that produce frank apnea at one end of the distribution do not vanish below the diagnostic threshold. They persist, on a continuum, across the whole population. An individual with an index of three may fragment their sleep, feel more fatigued, and perform worse cognitively than an individual with an index of ten, depending on the balance of flow limitation, respiratory-effort-related arousals, and sub-threshold ventilatory instability that the index was never designed to detect. The threshold sorts people into treated and untreated; it does not sort the underlying physiology into present and absent.

The evidence that sub-threshold physiology carries real consequences is direct. Upper-airway resistance syndrome describes patients whose index sits well below five but who accumulate frequent respiratory-effort-related arousals, and controlled comparison finds them to have worse sleep quality and daytime function than patients with formally diagnosed mild apnea². Inspiratory flow limitation, a sub-threshold marker of increased upper-airway resistance, is distributed across the continuum rather than switching on at the diagnostic cut. And the pediatric literature is the most pointed: chronic snoring in children, entirely below any apnea threshold, is independently associated with behavioral and neurocognitive problems, including hyperactivity, inattention, and poorer school performance³. Consequence does not begin at the threshold; the threshold is simply where treatment begins.

Continuous Distribution and Bidirectional Modifiability

Several measurable variables govern where an individual sits on this distribution. The ones that matter most for the present argument are loop gain, the width of the carbon dioxide reserve, and chemoreflex sensitivity, with cerebrovascular reactivity, upper-airway mechanical stability, and autonomic balance contributing. That these are continuous is empirical. Loop gain, the tendency of the ventilatory control system to overrespond and oscillate, has been measured on a continuous dimensionless scale, near 0.16 in healthy controls and near 0.28 in obstructive-apnea patients in one controlled comparison⁴, with high loop gain conventionally marked above roughly 0.7 rather than by any natural break in the distribution. The variable does not switch on at a diagnostic boundary; it is distributed, and people at the higher end of normal carry more sub-threshold instability than those at the lower end.

That these buffers causally gate stability, rather than merely correlating with it, has been shown experimentally. Pharmacologically reducing the cerebrovascular response to carbon dioxide by roughly a quarter, using indomethacin, narrowed the carbon dioxide reserve, the margin between the eupneic level and the apneic threshold, from 6.3 to 4.4 millimeters of mercury and increased the susceptibility to central apnea during non-REM sleep in otherwise healthy subjects⁵. When the buffer is degraded, the chemoreceptors see every systemic fluctuation unfiltered, and stability fails. The variables are not epiphenomena; they set the operating margin.

The premise requires more than continuity, however. It requires modifiability in both directions, because the framework needs the trait to be degradable by chronic environmental input and restorable by training. Both directions have support. On the upward side, elite breath-hold divers exhibit markedly blunted ventilatory responses to carbon dioxide, evidence that controller gain is not a fixed constant⁶, and in clinical and general populations slow-breathing retraining has produced measurable polysomnographic change, including fewer awakenings and fewer sleep-stage transitions, alongside enhanced parasympathetic cardiac vagal activity⁷. A control system often treated as fixed is in fact responsive to deliberate conditioning.

Breathing re-education as a way to modify respiratory control in sleep apnea has relevant prior art. McKeown, O'Connor-Reina, and Plaza explicitly discussed breath-hold measures, carbon dioxide chemosensitivity, and loop gain, and proposed breathing re-education to reduce chemosensitivity and improve ventilatory stability⁸. The present framework builds on that literature; its proposed contribution is the connection to arousal-weighted digital exposure, daytime-to-nighttime carryover, and a population continuum of sub-threshold sleep instability.

The downward direction is established by the same physiology read in reverse, and it is the direction this framework cares about most. Chronic stress, habitual overbreathing, and sustained sympathetic activation narrow the carbon dioxide reserve and raise controller gain, moving an individual from a stable region of the distribution toward an unstable one; passive hyperventilation produces exactly that pairing in the laboratory, a narrowed reserve and a steeper ventilatory response⁹. This is the ordinary consequence of ventilating chronically in excess of metabolic need. Plasticity has no preferred direction. The same responsiveness that lets training widen the reserve lets chronic overbreathing narrow it, and it is precisely that symmetry that makes an environmental degradation claim physiologically plausible.

The practical upshot of a continuous, sub-architectural respiratory axis is that two people can sleep the same hours with the same staging and recover very differently. Consider two sleepers with identical total sleep time and identical stage percentages on a standard study. The first has low loop gain, a wide reserve, robust cerebrovascular buffering, and parasympathetic dominance; their breathing is smooth, cerebral blood flow holds, and each hour delivers close to its full restorative value. The second has higher but still normal loop gain, a narrower reserve, blunted buffering, and residual sympathetic tone; their study looks the same, but their nights contain sub-threshold oscillations, flow limitation, and micro-arousals too brief to score, so on the account this framework develops, each hour delivers less. Standard polysomnography cannot separate them, because the difference lives beneath the resolution of staging, in the respiratory-control variables that determine how much restorative work each minute of sleep accomplishes.

A Population-Scale Continuum

Assembled, these facts support treating respiratory control as a graded, environmentally modifiable trait, in the same structural sense that cardiovascular capacity is one. Cardiovascular capacity sits on a continuum that a sedentary, calorie-dense environment degrades across a whole population, not only in those with diagnosed heart disease: the deconditioning is a population-wide drift, not a discrete illness. The proposal here is that respiratory-control efficiency, comprising chemoreflex sensitivity, carbon dioxide tolerance, reserve width, and the mechanical and autonomic contributors, is degradable in the same way and by an analogous environmental input, and that where a person sits on that continuum governs how efficiently they sleep. A degraded respiratory control system yields unstable, inefficient sleep; an intact one yields stable, efficient sleep; and a person can move along the continuum in either direction over months, the downward direction being the one this framework concerns.

Framed this way, the premise also names a stake that section “Introduction” anticipated. The population stranded between diagnosable disease and intact function is enormous. A large fraction of adults carry measurable respiratory disruption that never meets diagnostic criteria, and below even that sit the far larger numbers of poor sleepers who are told their sleep is normal because no threshold flags. The economic residue alone is substantial, with inadequate sleep costed in the hundreds of billions of dollars annually in a single national economy, driven largely by the cognitive underperformance of people who are present but unrefreshed¹⁰. A framework that treats these people as carrying a measurably degraded but modifiable respiratory control system offers them an explanation and a target where the categorical model offers neither.

The honesty calibration stated at the outset applies here in full. The components of this construct rest on published evidence: the variables are real, measurable, continuously distributed, and modifiable, and shifting them changes sleep quality in the populations studied. The integration of these components into a single graded trait, and the claim that it predicts and modifies sleep quality across the entire population rather than only in clinical or extreme groups, is the proposed and as yet unvalidated contribution. The evidence for the pieces is strong; the evidence for the whole, as a population-scale construct, awaits the controlled trials that would confirm it. That distinction is the same one the framework draws everywhere between established links and the novel synthesis that connects them, and it is maintained deliberately.

With the premise in place, the central question of the paper becomes tractable. If respiratory control is a modifiable trait on a continuous distribution, then it can be asked what environmental input is moving the population along that distribution, in which direction, and through what mechanism. The answer this paper proposes is sustained, arousal-driven overbreathing in digital environments.

The Upstream Cause: Digital Environments and Ventilation-Metabolism Mismatch

The novel element of this framework is its upstream cause. sections “Introduction” and “The Foundational Premise: Respiratory Control as a Modifiable Trait” established that respiratory control is a graded, modifiable trait and that a population-scale environmental input could shift its distribution; this section names the input. The proposal is that sustained engagement with modern digital environments drives ventilation above metabolic requirement, producing chronic mild hyperventilation and consequent hypocapnia, and that this is the entry point of the cascade the rest of the paper traces. The claim is specific and, in one respect, unusual: it locates a respiratory cause where the field has looked almost exclusively for photic and psychological ones. Nothing about the pathway is biologically exotic. Each of its steps is drawn from established physiology. What is new is the assembly of those steps into a single upstream driver, and the attribution of that driver to a now-ubiquitous feature of daily life.

The Proposed Mechanism

The pathway runs as follows. Modern digital environments are characterized by continuous informational load, notification-driven interruption, rapid task-switching, novelty, and social and evaluative stimulation. Taken together these produce a state of sustained cognitive and often emotional arousal, and that arousal is associated with elevated activity in the orexinergic and sympathetic systems. Orexin, the hypothalamic neuropeptide that stabilizes wakefulness and marshals the arousal response, is released not only by physical demand but by cognitive and emotional engagement, and chronic engagement sustains elevated orexinergic tone^11,12.

The first premise, that these environments produce sustained arousal, deserves specification rather than assertion. The features that characterize heavy digital engagement are, individually, arousal-inducing, and they operate more or less continuously. Feeds and messaging systems deliver rewards on intermittent, unpredictable schedules, the reinforcement pattern most effective at sustaining engagement and vigilance. Notifications interrupt, and each interruption imposes a small orienting response and a task-switch, with its attendant cognitive cost and momentary sympathetic bump. The content is frequently social and evaluative, carrying the particular arousal that comparison, judgment, and self-presentation reliably produce. And the pace is fast and self-directed, so the user rarely settles into the low-arousal, absorbed state that sustained attention to a single calm object would allow. None of these features is exotic, and none is inherently pathological. The claim is only that in combination, sustained across hours, they hold the arousal systems at a level of activation that an earlier environment would not have, and that this activation is the input the rest of the mechanism operates on.

The step that turns arousal into a respiratory problem is the coupling between orexin and the respiratory chemoreceptors. Orexin projects to and binds receptors on the central chemoreceptor neurons of the retrotrapezoid nucleus, and this binding increases their sensitivity to carbon dioxide and hydrogen ion¹³. A chemoreceptor driven harder by orexin treats a normal arterial carbon dioxide level as insufficient and calls for more ventilation than the blood gas warrants. The effect is measurable in the opposite direction: blocking orexin receptors attenuates the ventilatory response to carbon dioxide during wakefulness, by roughly a quarter in one controlled study¹⁴. Orexin also sensitizes the peripheral chemoreceptors of the carotid body under an intermittent, pulsatile pattern of release, though the controlled demonstration is of enhanced hypoxic rather than hypercapnic sensitivity¹⁵, so this strand is suggestive adjacent evidence rather than a direct carbon dioxide mechanism. The consequence is that under sustained arousal a person may overbreathe not by conscious decision but because the sensors governing respiration have been driven, by their own neuromodulator, to demand more air than metabolism requires. Orexin is offered here as one candidate mediator of the arousal-to-ventilation coupling, and much of its chemoreceptor evidence is preclinical; the upstream claim does not rest on it alone, because the direct human finding that cognitive load lowers end-tidal carbon dioxide holds regardless of which mediator carries it.

The crux of the mechanism, and the reason it is specific to the modern condition rather than a general feature of arousal, is a mismatch between ventilation and movement. The coupling between arousal and increased ventilation is ancient and, in its original context, self-correcting. For the entire evolutionary history in which that coupling was shaped, a state of heightened arousal accompanied or immediately preceded physical action: running, fighting, climbing, fleeing. The arousal drove ventilation up, and the physical action drove metabolic carbon dioxide production up to match it. Ventilation rose because production rose; the chemoreceptors were responding to a real metabolic signal, and blood gases stayed in balance. The increased breathing was paying for work the muscles were actually doing.

In the sedentary digital context the first half of that sequence is intact and the second half is absent. The stressor arrives, the notification fires, the charged exchange unfolds, and orexin and sympathetic drive rise exactly as they always have. But the body does not run, fight, or climb. It sits. The arousal and its ventilatory drive are present, but the physical action, and therefore the metabolic carbon dioxide production that would justify the increased ventilation, never comes. Ventilation outruns production. The breathing clears carbon dioxide that the muscles are not generating. The result is ventilation in excess of metabolic demand, which is to say overbreathing relative to carbon dioxide production, which is to say hypocapnia. The mismatch is the mechanism, and it is a mismatch that a sedentary, arousing, screen-mediated environment produces continuously and that an active one does not.

Arousal Load, Not Screen Time

A single distinction does most of the work in this framework, and getting it wrong is much of the reason the respiratory pathway has been invisible. The operative variable is not time spent looking at a screen. It is arousal load. Screen time is a biologically underspecified metric, because it aggregates activities whose physiological signatures point in opposite directions and reports their sum as though it were a single exposure. Duration, timing, and light content are not meaningless, and the rival accounts rightly rely on them; the claim is the narrower one that screen time alone does not capture the arousal dimension this pathway turns on.

Consider two hours of screen use. In the first, a person moves through a high-arousal stream: a social feed with its intermittent, unpredictable rewards, a stack of notifications each demanding a small decision, an emotionally charged argument in a comment thread, a half-dozen tasks switched between faster than any of them can be completed. In the second, the same person reads a long, calm passage of continuous text at an unhurried pace, absorbed but not agitated, with no interruptions and nothing to react to. Both register as one hour of screen time. On the mechanism proposed here they are not remotely equivalent. The first drives the arousal, the orexinergic tone, and the sympathetic activation that sensitize the chemoreflex and pull ventilation above metabolic need. The second does not, or does so far less. The physiologically meaningful quantity is not how long the screen was on but how much arousal the engagement carried, and screen time is blind to precisely that. This is why the framework predicts that arousal-weighted exposure, not undifferentiated screen time, will track respiratory dysregulation. High-arousal engagement should drive the pathway; low-arousal engagement matched for duration and posture should not. The prediction is sharp and falsifiable, and it is also, at present, untested. No study has decomposed technology use in this way. The existing evidence links cognitive load in general to overbreathing, which is the necessary groundwork, but no one has isolated the arousal dimension and shown that it, rather than exposure per se, is what moves respiration. That decomposition is the single most important empirical step the framework calls for, and it anchors the measurement program. It also offers an explanation for why the epidemiology of screens and sleep is as noisy and contested as it is: a literature that measures undifferentiated screen time is averaging over the very variable that carries the effect, so a real signal in arousal-weighted exposure would be diluted, sometimes to invisibility, when exposure is counted by the hour.

The Evidence for the Component Links

Several links in this upstream mechanism are directly supported by existing psychophysiology, and stating what is established separates the grounded parts of the claim from the novel synthesis that connects them.

That cognitive load drives overbreathing and lowers end-tidal carbon dioxide is established with instrumented measurement under matched control conditions. Studies of computer-based cognitive work and mental workload have recorded significant reductions in end-tidal carbon dioxide during active engagement relative to matched rest¹⁶, and a systematic review of respiration under cognitive load confirms the general pattern: mentally demanding episodes are marked by faster breathing and higher minute ventilation while respiratory amplitude stays roughly stable, and cognitive load can produce overbreathing as indexed by decreased end-tidal carbon dioxide¹⁷. That the increase in ventilation is carried by rate rather than by tidal volume is itself informative, and consistent with the mechanism: metabolic demand, when it rises, tends to recruit tidal volume, whereas arousal recruits rate, so a rapid, shallow pattern accompanying falling carbon dioxide is the fingerprint of ventilation driven by arousal rather than by metabolism. This is the empirical spine of the upstream claim: the acute hypocapnia of cognitive and computer work is not conjectured, it is measured.

That orexin is released in response to cognitive and emotional arousal, and not solely physical activity, is supported by multiple lines of evidence. Orexinergic neurons participate in attentional processing, arousal, and the response to psychological stress, and their release accompanies waking cognitive engagement independent of physical exertion^11,12. That orexin then modulates chemoreceptor gain is the link established above, through the direct action on retrotrapezoid neurons and the sensitization of the carotid body^13,14,15. The two established facts, that arousal releases orexin and that orexin sensitizes the chemoreflex, compose the bridge from a cognitive state to a respiratory one.

Finally, and most directly, the claim at the center of the proposal, that the ventilatory response to a cognitive stressor exceeds metabolic requirement, has been demonstrated. Using graded exercise to establish each subject's metabolic reference, one controlled study found that during mental arithmetic minute ventilation and respiratory rate ran 7 to 32 percent above the level predicted from oxygen consumption, that is, above genuine metabolic demand¹⁸. The ventilation-metabolism mismatch at the heart of the framework is therefore not a conjecture but an established feature of cognitive stress in the laboratory. What has not been done is to establish it for digital engagement as the exposure of interest, decomposed by arousal.

What Remains Unestablished

Two specific elements of the upstream mechanism are not established, and they are the primary empirical gaps in the entire framework. Naming them precisely is more useful than gesturing at them, because each corresponds to a study that could be run.

The first gap is arousal decomposition. The existing evidence links cognitive load in general to overbreathing, but no study has tested the specific prediction that high-arousal digital engagement, such as social media, notification-heavy work, and emotionally charged interaction, produces greater respiratory dysregulation than low-arousal engagement, such as calm reading, matched for screen time and posture. This is the load-bearing prediction of the framework, and it is the one that most directly distinguishes it from the existing accounts, because it locates the effect in arousal rather than in exposure. Until it is tested, arousal-weighted exposure remains a proposed operative variable rather than a demonstrated one.

The second gap is narrower, because the mismatch itself is established for cognitive stress, as just shown, but it has not been demonstrated for digital engagement specifically, with simultaneous measurement of ventilation and metabolic rate during real, arousal-decomposed screen use. The distinction matters because cognitive load does modestly raise metabolic rate¹⁷, so for any given task a portion of the ventilatory increase may be appropriate rather than excessive; the definitive test isolates that excess for the digital exposure of interest, and that measurement remains to be made.

The Respiratory Core: From Acute Hypocapnia to Durable Recalibration

The upstream cause produces acute hypocapnia, a fall in carbon dioxide during each episode of high-arousal engagement. That much, section “The Upstream Cause: Digital Environments and Ventilation-Metabolism Mismatch” argued, rests on established measurement. The core claim is stronger: that repeated exposure produces a change in respiratory control that outlasts the episode and reaches sleep. The proposed carryover has two tiers: altered neural chemoreflex responsiveness, and, with more sustained habitual overbreathing, additional bicarbonate adaptation that helps maintain a changed operating state. These tiers need not appear at the same exposure dose or have the same measurable signature. Whether either form of carryover occurs, rather than the system fully resetting between episodes, is the chronicity question this section addresses.

The Chronicity Question

A daytime cause can only produce a nighttime effect if something carries across the boundary between them. Transient daytime hypocapnia that fully normalized each evening would leave the sleeping respiratory control system exactly as it found it, and no amount of daytime overbreathing would then reach sleep. For the framework to work, the daytime disturbance must leave a trace in controller responsiveness, acid-base regulation, or both that persists into the relevant sleep period.

The objection this raises is the most serious the framework faces. Human physiology is homeostatic and equipped with fast compensations. Daytime hypocapnia is punctuated by breaks, by meals, by movement, and above all by a night of sleep during which the arousing stimulus is absent. The natural expectation is that the system would normalize in those windows, that each episode of overbreathing would be a transient dip erased before the next, and that no lasting setpoint change would accumulate. If that expectation held, the cascade would be broken at its most important joint.

A precedent for the biochemical tier comes from the clinical literature on chronic behaviorally-driven hyperventilation. Individuals with chronic hyperventilation, including patients with chronic hyperventilation syndrome and with panic disorder, do not merely hyperventilate during acute episodes and normalize between them. In chronic hyperventilation, a persistently low carbon dioxide level is accompanied by a renally lowered serum bicarbonate and a near-normal pH, a compensated respiratory alkalosis measurable at rest¹⁹. In normal human subjects, sustained hyperventilation has been shown to lower plasma bicarbonate durably, primarily by reducing renal net acid excretion, with blood pH remaining elevated throughout²⁰. Panic-disorder patients, whose pathophysiology involves chronic hyperventilation, have been reported at baseline, between attacks, to carry lower resting carbon dioxide and lower bicarbonate than healthy controls. That resting measurement is the crucial one, because it demonstrates a durably reset operating point rather than a series of transient drops that fully recover. The compensation that would be expected to restore balance is precisely what maintains the new state: once the kidneys have excreted bicarbonate in response to sustained or repeated hypocapnia, the lowered buffer and the lowered setpoint persist, held in place by the same renal machinery that established them. This is the finding that makes it physiologically plausible for a daytime cause to carry into nighttime respiratory instability and its sleep-mediated consequences, and it is the reason the rest of the chain is coherent.

Two things must be kept distinct here. That behaviorally-driven hyperventilation can durably recalibrate the control system is established in the clinical and panic literature. That the intermittent, daytime-confined overbreathing of digital exposure is a sufficient and population-scale instance of the same process is the novel claim, and it inherits the gaps of section “The Upstream Cause: Digital Environments and Ventilation-Metabolism Mismatch”: the required repeated measurements of chemoreflex responsiveness, resting carbon dioxide, and bicarbonate as a function of arousal-weighted digital exposure remain to be made.

The digital case differs from the clinical one in a way that matters. The clinical evidence for durable recalibration comes from patients whose hyperventilation is not confined to waking hours, whereas the digital case is one of daytime overbreathing set against a night of ordinary, unstimulated breathing. Renal compensation for a respiratory alkalosis develops over roughly two to three days, so sustained daily overbreathing could in principle hold the buffer and the setpoint down despite nightly pauses, the accumulated adjustment outlasting any single night. But the opposite is also possible: overbreathing confined to the day, with normal nocturnal breathing permitting partial renal re-accumulation each night, might compensate less completely and recalibrate less durably than the round-the-clock hyperventilation of the clinical syndromes. Which obtains is not established. Prediction 3 therefore measures controller responsiveness as well as resting carbon dioxide and bicarbonate: absent morning bicarbonate reduction would limit the biochemical tier at that exposure dose, but would not by itself exclude neural carryover into sleep.

The Mechanism of Recalibration

At sustained exposure, bicarbonate adaptation supplies the biochemical tier of the proposed durable shift. When hypocapnia is sustained over days to weeks, the body defends blood pH against the respiratory alkalosis that low carbon dioxide would otherwise produce. The kidneys excrete bicarbonate, and the choroid plexus simultaneously reduces the bicarbonate concentration of the cerebrospinal fluid. Blood pH is thereby returned toward normal, which is the point of the compensation, but the correction is achieved by lowering the buffer rather than by restoring the carbon dioxide, and the lowered buffer is what matters for what follows.

The consequence is felt most sharply at the central chemoreceptors, because the cerebrospinal fluid that bathes them is essentially devoid of the robust protein buffers found in systemic blood. Bicarbonate is the principal chemical shock absorber available in that compartment, and stripping it removes the cushion that normally smooths carbon dioxide fluctuations into gentle, absorbable pH changes. In a normally buffered system, a small rise or fall in carbon dioxide produces a small, damped change in the pH surrounding the sensors, and the ventilatory response is correspondingly modest. In a bicarbonate-depleted system, the same small carbon dioxide fluctuation produces a large, unbuffered pH swing, and the sensors fire hard in response. The chemoreceptors then recalibrate to this depleted environment as their new baseline, treating the artificially low carbon dioxide level as the reference point and any rise back toward a physiologically normal value as a deviation to be corrected. Four quantities must be kept separate here, because the framework's confidence differs across them. That sustained hypocapnia lowers serum and cerebrospinal-fluid bicarbonate is established acid-base physiology. That a thinner bicarbonate buffer yields a larger pH excursion for a given carbon dioxide change follows directly from the buffering chemistry. That this amounts to a durably lowered carbon dioxide setpoint and a raised loop gain is the inference the framework draws on top of those facts, and it is an inference rather than a measurement: bicarbonate compensation is in the first instance a response to sustained hypocapnia, and it does not by itself prove that the controller has adopted a new preferred operating point or that whole-loop gain has risen, both of which depend on further factors and must be measured rather than assumed. None of this requires the carbon dioxide itself to remain low once compensation is complete; the durable change proposed is in the buffer and the setpoint, not in a persistently abnormal blood gas, which is why the prediction is that it can be measured at rest.

Two properties of this biochemical component of recalibration bear on how it should be understood. First, it is a process with a timescale: the renal and choroid-plexus adjustments that establish the depleted-buffer state develop over days to weeks of sustained or repeated hypocapnia, and once established they are stable, which is why the condition is neither instantaneous in onset nor quick to resolve. Second, and importantly, it is bidirectional. The same plasticity that allows chronic overbreathing to shift the setpoint down allows deliberate carbon dioxide conditioning to shift it back, as the training evidence in section “The Foundational Premise: Respiratory Control as a Modifiable Trait” shows. That reversibility is not incidental; it is corroboration. A state that can be trained back is an acquired recalibration of a setpoint, not fixed structural damage, which is the interpretation the framework requires and the property that makes the recalibrated state a reversible target rather than a fixed lesion.

Chemoreflex Sensitization and the Narrowed Reserve

A lowered operating point with a depleted buffer is, functionally, a sensitized and less stable chemoreflex. Two features combine to make it so. The buffer is thinner, so any given carbon dioxide fluctuation produces a larger pH excursion and a larger ventilatory response, which is heightened controller gain. And the operating point now sits closer to the threshold at which ventilatory changes are triggered, so less perturbation is required to cross it. A system with more gain and less margin is a system that overreacts, and the parameter that captures its tendency to overreact and oscillate is loop gain.

This account has direct human support, which also sharpens it. Held in passive hyperventilation for six hours, healthy subjects showed a ventilatory response to carbon dioxide that both shifted leftward and grew steeper, and the two changes dissociated cleanly: the leftward shift, which carries the response curve toward the operating point, came from the accompanying hypocapnia and alkalosis, while the increase in slope, the heightened gain itself, came from the act of hyperventilation and not from the change in pH⁹. That locates the gain increase in the hyperventilation rather than solely in the thinned buffer, making the buffer-driven amplification described above one contributor among more than one. The recalibrated state proposed here is the chronic, self-sustaining counterpart of that six-hour result: a leftward-shifted and steeper ventilatory response held in place by ongoing daily overbreathing rather than by a single exposure.

This dissociation motivates a two-tier dose-response prediction. At moderate or intermittent exposure, the earliest carryover may be altered neural controller responsiveness, detectable in chemoreflex testing and early-night respiratory stability even when morning serum bicarbonate remains within its usual range. At more sustained exposure, especially if habitual overbreathing persists between digital sessions, renal and cerebrospinal-fluid bicarbonate adaptation may provide an additional, longer-lived component. The six-hour experiment supports separating gain change from alkalosis; extending that result to ordinary repeated digital exposure is the hypothesis to be tested. The study design should therefore resolve neural and biochemical changes separately rather than requiring both to be present at every dose.

The mechanism of that hyperventilation-driven gain increase is not established, and one candidate stands out as a specific molecular prediction the framework does not depend on. Roughly half of the chemoreceptor neurons of the retrotrapezoid nucleus co-release the inhibitory neuropeptide galanin, whose microinjection into the respiratory column blunts the ventilatory response to carbon dioxide^21,22, and chronic hypercapnia, the conditioning direction, upregulates preprogalanin in these neurons, apparently as a homeostatic brake that downscales an over-stimulated chemoreflex²³. The symmetric possibility, that chronic hypocapnia withdraws this brake and thereby disinhibits and sensitizes the chemoreflex, would supply a molecular substrate for the deconditioning proposed here. It is at present untested: galaninergic tone in these neurons has not been measured after a period of chronic or repeated hyperventilation, and the direction cannot be assumed symmetric, since the system is organized to defend against hypoventilation and may carry no equivalent program for the opposite extreme. The point is therefore offered as a target for measurement, not as a link the argument depends on; that argument rests on the buffer and hyperventilation effects already established.

The margin in question has a name and a size. The carbon dioxide reserve is the difference between the resting eupneic carbon dioxide level and the apneic threshold, the level below which the brainstem withdraws the drive to breathe. In a healthy person that reserve is on the order of a few millimeters of mercury, typically reported in the low single digits, a buffer that absorbs the ordinary ventilatory fluctuations of sleep without consequence²⁴; a chronically lowered resting carbon dioxide level does not by itself guarantee a narrower reserve, since the apneic threshold may shift downward with it, and the framework's claim is the more specific one that heightened chemoreflex gain and reduced buffering move the eupneic level and the threshold closer together. That narrowing is therefore a measurable prediction about the sleeping reserve, to be confirmed by direct measurement, rather than an automatic consequence of a lower daytime carbon dioxide. The reserve matters far more during sleep than during waking, and this asymmetry is central to why the disturbance manifests where it does. During wakefulness a separate, behavioral drive to breathe overrides the chemical threshold and masks how close the system sits to it. During sleep that behavioral drive is withdrawn, respiration falls entirely under chemical control, and the apneic threshold becomes a live, operative boundary. A reserve that is invisible and inconsequential by day becomes the governing constraint by night.

This narrowed reserve and heightened sensitivity connect the respiratory core to sleep instability. They also imply a moderation prediction: the effect of an acute pre-bed arousal perturbation depends on the baseline operating margin. Orexin-linked chemoreflex sensitization of the kind discussed in section “The Upstream Cause: Digital Environments and Ventilation-Metabolism Mismatch” may substantially destabilize an already marginal system, yet produce little observable sleep disruption in a system with a wider reserve and lower baseline gain. The same evening exposure can therefore have different effects at different points on the respiratory-control continuum, including within the same person as their baseline changes. A reduction in screen sensitivity after respiratory improvement is consistent with this account even if the exposure continues to generate arousal.

The prediction concerns an interaction between measured baseline stability and an acute perturbation, not a fixed numerical threshold or a conversion of breath-hold seconds into reserve width or loop gain. Standardized breath-hold measures can be retained as candidate indicators, while the proposed moderator is tested with chemoreflex responsiveness, sleeping carbon dioxide reserve, and loop gain. This preserves the population-continuum model: the size of the evening effect varies along a modifiable distribution rather than defining a separate diagnostic category.

A Clinical Precedent: Behavioral Hyperventilation-Induced Central Sleep Apnea

The chronicity hinge is not purely theoretical; there is a documented clinical entity in which behavioral hyperventilation drives central respiratory instability, and it provides both a precedent for the mechanism and a clean illustration of its logic. In a published case, pandemic-related psychological stress drove a patient's carbon dioxide to a nocturnal nadir near 31.7 millimeters of mercury through stress-induced hyperventilation during sleep, and in doing so converted a moderate obstructive sleep apnea into a severe central sleep apnea, with the great majority of events central rather than obstructive; when the psychological stressor resolved, the central apnea reversed^25,26. This is behavioral hyperventilation-induced central sleep apnea, and it demonstrates in a single clinical observation that a behavioral, psychologically driven change in breathing can destabilize the sleeping respiratory control system through a carbon dioxide mechanism, and can do so severely.

What the framework proposes is a chronic extension of this documented acute mechanism, and the borrowed part and the novel part should not be confused. The acute entity is documented: an acute stressor drives hypocapnic central instability, and when the stressor lifts the instability resolves. The framework's claim is that when the driver is not a discrete acute stressor but the sustained, low-grade arousal of chronic digital engagement, and when it operates long enough for neural recalibration and, at more sustained exposure, bicarbonate adaptation to take hold, the instability ceases to be self-resolving and becomes locked in place by the recalibrated chemoreflex itself. The acute mechanism is borrowed from the published literature; the chronic, self-sustaining, recalibration-maintained version is the novel contribution. One further difference follows from the chronic framing. Where the documented acute entity produced frank, severe, scoreable central apnea, the chronic version driven by mild sustained overbreathing may manifest instead as sub-threshold instability, a fragmenting of sleep by ventilatory oscillations and micro-arousals that never reach the duration or desaturation required for formal event scoring. That distinction is developed in the following section, where it explains why a controlled apnea-hypopnea index can coexist with badly fragmented sleep. One caveat on the precedent should be explicit. The documented case involves hyperventilation occurring during sleep itself and directly destabilizing it, whereas the framework's mechanism is daytime overbreathing that must recalibrate the system for the effect to appear at night. The precedent therefore establishes that behavioral hyperventilation can destabilize the sleeping respiratory control system through a carbon dioxide mechanism; it does not by itself establish that a daytime-confined pattern recalibrates the system in the way section “The Chronicity Question” requires, which is the separate and still-open question named there.

With the respiratory core in place, the remaining chain follows from its effect on sleep. A sensitized, narrow-reserve control system is predicted to destabilize non-REM breathing, fragment stable sleep, and impair progression to deep sleep and access to REM. The emotional consequence developed below follows through this sleep disruption; no separate limbic acid sensor is required.

Sleep Disruption and Its Downstream Consequences

The immediate downstream consequence of the recalibrated chemoreflex is degraded sleep, produced by respiratory-control instability during the night. The logic of the pathway is that the sensitized, buffer-depleted, narrow-reserve state established in section “The Respiratory Core: From Acute Hypocapnia to Durable Recalibration” is precisely the condition under which the sleeping respiratory control system becomes unstable, and that this instability fragments sleep in a specific and largely invisible way. Two features make the sleep mechanism more subtle than it first appears, and both require care. The instability is often sub-threshold, so it fragments sleep without producing the scoreable events that standard testing counts, which is why a person can have a controlled apnea-hypopnea index and badly unrefreshing sleep at the same time. And the mechanism by which the instability degrades deep sleep is one of temporal stability rather than gas exchange, a distinction easy to misread. This section develops the instability, its sub-threshold signature, the deep-sleep mechanism, and the consequence of all of it for restorative throughput and sleep-mediated emotional processing.

Loop Gain and the Instability of NREM Sleep

The parameter that governs the stability of breathing during sleep is loop gain, the ratio of the ventilatory system's corrective response to the size of the disturbance that provoked it. A loop gain below one means each correction is smaller than the perturbation that triggered it, so disturbances are damped and breathing converges on stability. A loop gain approaching or exceeding one means each correction is as large as or larger than the perturbation, so disturbances are amplified and breathing oscillates. The recalibrated chemoreflex of section “The Respiratory Core: From Acute Hypocapnia to Durable Recalibration”, with its heightened controller gain and its narrowed reserve, is on this account predicted to raise loop gain, which integrates controller gain with plant gain and circulatory delay: such a system responds too strongly to carbon dioxide fluctuations and sits too close to the apneic threshold to absorb the resulting swings. During wakefulness this instability is masked, because a behavioral drive to breathe supplements the chemical one and holds ventilation steady regardless. During sleep that behavioral drive is withdrawn and respiration falls entirely under chemical control, and high loop gain becomes the dominant characteristic of the system. That waking chemoreflex sensitivity predicts sleep respiratory instability, and that this instability is captured by loop gain, is established in the sleep-medicine literature^27,4. Loop gain is not uniform across sleep; it is higher in non-REM than in REM, which is part of why non-REM sleep is the more vulnerable state.

Two influences act on that loop gain and pull opposite ways. The heightened controller gain and the narrowed reserve are the destabilizing ones, the same pairing produced when central carbon dioxide buffering is reduced experimentally⁵, and they act at the steep, threshold-bound part of the response, where an ordinary carbon dioxide fluctuation, arriving with the eupneic level already close to the apneic threshold, produces disproportionately large instability. The lower operating point of the recalibrated state pulls the other way: it flattens the relationship between ventilation and carbon dioxide and so reduces plant gain, which in isolation is stabilizing. That offset is gradual, though, where the destabilizing influences are steep, and across the range that matters for sleep it does not overcome them. The prediction is therefore a net rise in loop gain, with reduced plant gain a partial offset rather than a countervailing force of equal weight.

High loop gain destabilizes sleep through two distinct pathways, and distinguishing them matters because they operate in different stages and produce different events. The first is passive hypocapnic drift, which operates in the lighter stages and at sleep-stage transitions. When a person passes from wakefulness into light non-REM, muscular carbon dioxide production falls with the drop in motor activity, settling toward a lower resting level rather than ceasing, but ventilation does not immediately fall even to match that reduced production: a neural short-term potentiation keeps respiratory output elevated for seconds to minutes after the transition, like a flywheel still turning, and circulatory delay means the chemoreceptors are reading a blood gas from moments earlier. The lungs go on clearing carbon dioxide faster than the quiet tissues are producing it, and with a narrow reserve the carbon dioxide simply drifts below the apneic threshold with no overcorrection required. Breathing stops passively, the drift-induced pause triggers a micro-arousal, the arousal reinstates ventilation and clears carbon dioxide further, and when sleep resumes the cycle repeats. The frequency of these events is set by the frequency of state transitions, which is why unstable sleep is self-perpetuating: each arousal creates the transition that seeds the next drift.

The second pathway is oscillatory overcorrection, which operates in deeper non-REM. As the brain descends and carbon dioxide begins its normal rise, a hypersensitive chemoreflex reads that ordinary, physiological rise as a threat and fires the emergency response. Either it drives an arousal directly, waking the sleeper from a carbon dioxide rise that a normally calibrated system would not have noticed, or it overcorrects, driving ventilation up and, because the body is still and there is no metabolic demand to absorb the extra breathing, blowing carbon dioxide back below the apneic threshold. Breathing then ceases centrally, carbon dioxide accumulates unopposed and overshoots on the far side, and the sleeper wakes to a violent carbon dioxide rebound. The overshoot is often larger than the original perturbation, so the arousal is more forceful and the recovery longer. These two pathways, passive drift in the light stages and oscillatory overcorrection in the deep ones, between them cover the whole non-REM night, while REM is largely spared.

The Sub-Threshold Signature: Why a Controlled Index Can Coexist with Fragmented Sleep

A central feature of this sleep mechanism, and the reason it is easily missed, is that most of the instability it produces is sub-threshold. Not every ventilatory oscillation crosses the apneic threshold, and not every pause lasts the ten seconds or produces the oxygen desaturation that standard event scoring requires. A hypersensitive chemoreflex can generate a micro-arousal, a stage shift, or a ventilatory overcorrection without ever meeting the criteria for a scoreable apnea or hypopnea. These sub-threshold events fragment sleep architecture while leaving the apnea-hypopnea index and the oxygen saturation deceptively normal. This is the mechanism by which a person can be well treated for obstruction, show a controlled index on their device, and still sleep badly: the events that are degrading their sleep are below the resolution of the instrument that is declaring them absent.

The instability can be organized into three classes that share a common cause but differ in severity, oxygen involvement, and visibility, and that are best understood as escalation states rather than fixed categories. The first class is sub-threshold micro-instability without apnea or desaturation: small ventilatory oscillations and ordinary carbon dioxide movements that trip the hypersensitive chemoreflex and produce micro-arousals or stage shifts before breathing ceases or oxygen falls. This is the quiet, polluting layer, and it is the one that most likely dominates the fragmentation burden in a person whose obstruction is controlled. It leaves a low index, a normal oxygen trace, a fragmented architecture, poor deep-sleep and REM continuity, and the subjective experience of unrefreshing sleep, and it may not be remembered at all, presenting only as light sleep and a sense of never having rested. The second class is an apneic-threshold crossing: carbon dioxide drops below the threshold through drift or overcorrection, central drive pauses, and the pause resolves through carbon dioxide reaccumulation or a brief arousal before meaningful oxygen loss develops. Oxygen may stay stable throughout, and the arousal, when it comes, is driven by the carbon dioxide rebound rather than by any fall in oxygen. The third class is the oxygen-involved event: a pause or oscillation that lasts long enough for oxygen to fall meaningfully, at which point the carotid bodies detect the desaturation and fire a fast sympathetic rescue on top of the carbon dioxide signal, producing the violent, memorable, adrenaline-flooded awakening. The critical point of the taxonomy is that central respiratory instability is defined by the loss or oscillation of central drive, not by oxygen desaturation; oxygen involvement is contingent on how long an event runs and in what gas-exchange context, not intrinsic to the mechanism. In a person with controlled obstruction and relatively normal gas exchange, the first two classes account for the great majority of the fragmentation, and the third class, the most severe, is the least frequent even though it is the most conspicuous.

This pattern is consistent with a recognized clinical entity. Treatment-emergent or complex central sleep apnea describes patients in whom positive airway pressure resolves the obstructive events but unmasks a residual instability, consistent with a high-loop-gain contribution that airway treatment does not address²⁸. Endotyping work points the same way: a substantial fraction of apnea patients have high loop gain as a primary contributor, and among patients with only mild anatomical compromise the majority are loop-gain-dominant²⁸. These are people whose problem was never mainly the airway, and for whom controlling the airway leaves the generator of their fragmentation untouched. The measurement implication follows directly. An index built to count discrete obstructive events is the wrong instrument for a disturbance that operates below its threshold, and a measure that captures the autonomic and cardiopulmonary signature of respiratory instability regardless of whether events are scoreable, such as cardiopulmonary coupling analysis, is a more appropriate one, best used alongside direct measures of loop gain, carbon dioxide reserve, and arousal rather than as a stand-alone proxy for the mechanism. This sub-threshold sleep mechanism is not visible to the tool the field relies on, which is part of why it has not been seen.

The Deep-Sleep Mechanism: Stability, Not Gas Exchange

The mechanism by which respiratory instability degrades deep sleep requires precise statement, because it is easy to misattribute to gas exchange. The improvement that follows from more stable breathing is not an oxygen-delivery effect, of the kind a shift in the oxyhemoglobin dissociation curve would produce, and it requires no change in respiration rate. It concerns the temporal stability of the sleep progression rather than gas exchange, and that distinction governs what follows.

The correction rests on a distinction that the cardiopulmonary coupling literature draws sharply: the difference between stable and unstable non-REM sleep does not map onto the boundary between the lighter N2 stage and the deeper N3 stage. Measured by the autonomic and cardiopulmonary signature of stability, most of healthy N2 is stable sleep, as most of N3 is, and this stability dimension correlates only weakly with the conventional N2-to-N3 boundary, which itself rests on an arbitrary slow-wave threshold²⁹. Stable non-REM, whether it is scored N2 or N3, is the sleep that carries temporally stable breathing, stable arousal thresholds, and increased slow-wave power³⁰. In a respiratory-stable person, N2 is physiologically much closer to N3 than the staging hierarchy implies, and the difference between stable and unstable N2 is larger than the difference between stable N2 and N3. Autonomic and respiratory stability is a restorative axis in its own right, one that cuts across the electroencephalographic staging rather than running along it.

With that distinction in place, the deep-sleep mechanism is straightforward. Progression from N2 to N3 depends on a sustained period of uninterrupted, stable N2 preceding the transition. Respiratory-control instability produces micro-arousals that repeatedly interrupt that stable period before the transition to N3 can consolidate, resetting the progression each time and preventing efficient entry into deep sleep. When the instability is reduced, the stable N2 period is preserved, the transition to N3 occurs reliably, and deep sleep is reached and held. The improvement is in the temporal stability of the progression, not in gas exchange. The supporting evidence is that the stable-sleep cardiopulmonary signature is associated with, and tends to precede, the emergence of deep-sleep slow-wave power³⁰, which is most naturally read not as a deterministic threshold but probabilistically: a sustained stretch of autonomic stability raises the probability that deep-sleep slow waves emerge, and interrupting that stretch lowers it. Consistent with the same picture, respiratory instability is associated with more frequent transitions backward from consolidated non-REM toward lighter sleep and wakefulness, which is the aborted-progression signature the mechanism predicts. The account is therefore that instability degrades deep sleep by repeatedly resetting the stable non-REM runway that deep sleep requires, and that stabilizing breathing restores deep sleep by protecting that runway. No oxygen-delivery mechanism is required.

Stable Throughput and Restorative Efficiency

The corrected mechanism predicts improved restorative throughput when respiratory instability is reduced. If instability consumes non-REM time in ventilatory cycling and aborted transitions, stabilizing breathing should increase the proportion of sleep spent in sustained, stable non-REM and permit more reliable slow-wave expression. This is the removal of an acquired inefficiency, not an acceleration of the underlying restorative processes. Independent evidence supports separating sleep quality from duration: selectively suppressing slow-wave activity without reducing total sleep time produces metabolic impairment proportional to the slow-wave activity lost³¹. The primary claim here is therefore that respiratory stabilization improves the physiological quality and functional yield of sleep at an adequate sleep opportunity.

The relevant endpoints are stable non-REM continuity, cardiopulmonary-coupling high-frequency-coupling minutes and density, slow-wave activity, access to REM, and next-day function. These measures should be assessed together: an increase in stable-sleep minutes or delta power is informative, while preserved or improved daytime performance establishes whether the physiological change translates into a functional benefit. Total sleep time and sleep opportunity should be recorded alongside them so that improved continuity can be distinguished from a change in duration alone.

A shorter preferred sleep duration after recovery may be recorded as an ancillary observation, but it is neither the target nor a required outcome of this framework. No general reduction below a specified number of hours is predicted, and shorter sleep alone would not establish restoration. The respiratory-sleep mechanism stands independently of whether any individual's preferred duration changes; its primary prediction is greater restorative efficiency with adequate opportunity for both non-REM and REM sleep.

The Route to Emotional Consequence Through REM

The sleep pathway also provides a route to emotional dysregulation. It is a natural assumption that a respiratory-control problem would disrupt REM sleep directly, but the opposite is true: REM is one of the most centrally stable respiratory states, with loop gain falling substantially and chemosensitivity dropping to a fraction of its waking value, and central respiratory events cluster in non-REM rather than in REM. High loop gain therefore does not attack REM directly. It attacks REM by preventing the antecedent non-REM continuity that REM requires. Reaching substantial REM depends on proceeding through consolidated early non-REM cycles. Each fragmentation-driven arousal interrupts that progression and reactivates the arousal system, so the brain repeatedly fails to assemble the sustained non-REM continuity from which the transition into REM is made. Fragmented non-REM thus starves REM indirectly, by never assembling the stable runway from which REM is entered.

The consequence of chronic REM loss is a specific impairment of emotional processing. REM is the state in which emotionally charged material is consolidated while the noradrenergic tone that accompanies waking distress is at its lowest, which allows the content of an emotional memory to be integrated without its affective charge. Deprived of adequate REM, that overnight processing does not complete, and emotional material returns the next day with its charge intact, as heightened reactivity, stickier rumination, and weaker extinction of prior distress^32,33. This is the emotional route retained in the present framework: respiratory instability fragments non-REM, impairs access to REM, and thereby reduces the opportunity for the emotional processing that REM supports.

The sleep-mediated route can also close a self-reinforcing loop. Poor overnight emotional processing may increase next-day threat sensitivity, rumination, and anxiety; that additional arousal can in turn sustain the respiratory drive described in section “The Upstream Cause: Digital Environments and Ventilation-Metabolism Mismatch”. On this account, respiratory instability and emotional distress reinforce one another through sleep and waking arousal without requiring direct amygdalar proton gating.

Predictions and the Measurement Program

The value of a framework of this kind rests on its testability, and the purpose of this section is to state what the framework predicts precisely enough that the predictions could fail, and to specify the measurement that would confirm or refute them. Each individual link in the chain, as the preceding sections established, is already supported by published evidence; what has not been tested is the integrated chain and, above all, the attribution of its upstream cause to arousal-weighted digital exposure at population scale. The predictions below are the specific, falsifiable commitments that attribution entails, and the measurement program is the concrete study design capable of adjudicating them. The section closes by stating what results would refute the framework.

Core Predictions

The framework generates the following specific and falsifiable predictions, ordered from the upstream cause through sleep and its emotional consequences to reversibility.

• Prediction 1 (upstream, arousal-differentiated). High-arousal digital engagement will produce a greater reduction in end-tidal carbon dioxide, relative to a matched sedentary control, than low-arousal digital engagement matched for screen time and posture. Arousal load, not screen time, will be the operative predictor.

• Prediction 2 (ventilation-metabolism mismatch). During high-arousal digital engagement, ventilation will exceed metabolic requirement, demonstrable by simultaneous measurement of ventilation and metabolic rate, producing hypocapnia not fully accounted for by any accompanying rise in metabolic rate.

• Prediction 3 (two-tier chronicity). Repeated high-arousal digital exposure will produce a persistent change in chemoreflex responsiveness and nocturnal respiratory stability beyond the acute exposure period. At moderate or intermittent doses this may occur without a detectable reduction in morning serum bicarbonate. More sustained habitual overbreathing, including between sessions, will additionally predict lower resting carbon dioxide and bicarbonate. Repeated neural, respiratory, and biochemical measurements will distinguish these tiers and establish whether each change persists into sleep.

• Prediction 4 (sleep and restorative throughput). Chronic high-arousal digital exposure will predict elevated sleep loop gain, greater non-REM instability, less efficient progression to deep sleep and access to REM, and reduced restorative efficiency, with these effects statistically separable from blue-light and behavioral-displacement confounds. Improvement will be assessed by stable non-REM continuity, slow-wave activity, and daytime function, rather than by shorter sleep duration.

• Prediction 5 (sleep-mediated emotional consequences). Chronic high-arousal digital exposure will predict altered emotional regulation in part through the measured sequence of respiratory dysregulation, fragmented non-REM, and reduced access to REM described in section “The Route to Emotional Consequence Through REM”. Testing will estimate this sleep-mediated pathway while accounting for baseline emotional state and psychological effects of the exposure. A residual emotional effect independent of sleep is not required by this paper.

• Prediction 6 (baseline stability by acute evening arousal). Under matched light, duration, posture, and sleep opportunity, high-arousal pre-bed digital engagement will disrupt non-REM stability and subsequent access to REM more strongly when baseline respiratory stability is low than when it is high. The same person's sensitivity to the evening exposure should diminish as their measured respiratory operating margin improves, even if the exposure continues to produce an arousal response. The interaction will be tested continuously using respiratory measurements rather than a presumed breath-hold-to-loop-gain conversion.

• Prediction 7 (reversibility). Behavioral carbon dioxide tolerance conditioning that recalibrates the measured respiratory state will improve sleep stability and restorative throughput. To the extent that emotional impairment is mediated by fragmented sleep and restricted access to REM, emotional outcomes should improve through that sleep-mediated sequence. The study will measure the time course of respiratory, sleep, and emotional changes rather than requiring a direct limbic effect.

These predictions permit the chain to be tested link by link, with the acute evening interaction testing a distinct moderation claim. The discriminating features are the arousal differentiation of Prediction 1, the neural and biochemical carryover distinguished in Prediction 3, the measured sleep-mediated sequence in Prediction 5, and the baseline-stability interaction in Prediction 6. Their joint test is more informative than a simple association between screen time and poor sleep.

The Measurement Program

The program has six measurement requirements. First and most important, arousal-differentiated exposure: device-level behavioral data, notification density, application category, task-switching, and session pattern, combined into an arousal-weighted metric that separates the high-arousal stream from calm, absorbed use, which is the measurement the existing epidemiology most conspicuously lacks. Second, carbon dioxide status in two forms, acute end-tidal carbon dioxide during engagement and resting carbon dioxide with serum bicarbonate as biochemical measures between episodes, interpreted alongside chemoreflex testing rather than as a sufficient account of the chronic state. Third, ventilation measured simultaneously against metabolic rate during engagement, the only way to show the hypocapnia reflects overbreathing beyond metabolic need rather than ventilation matched to raised demand. Fourth, chronicity, through repeated assessment of the resting respiratory state and chemoreflex sensitivity, to distinguish a durable shift from a transient one. Fifth, sleep measured for its respiratory-instability signature, loop gain, arousal burden, non-REM stability, and deep-sleep progression, REM access, and next-day function, with blue light and behavioral displacement controlled so the respiratory pathway's independent contribution is isolated. Sixth, emotional and autonomic state, assessed longitudinally to estimate the pathway from exposure through respiratory change and sleep disruption to emotional outcomes. Sleep is a proposed mediator in that analysis, rather than a variable to remove in order to demand a direct limbic effect.

The program divides into two arms of very different feasibility. The acute arm measures carbon dioxide, ventilation, and metabolic rate during controlled, arousal-differentiated engagement in the laboratory; it is within reach of established psychophysiology methods, could be run now, and is the most direct test of the upstream cause without requiring the population-scale claim to be granted first. The chronic arm correlates arousal-differentiated real-world exposure with resting respiratory, sleep, and emotional outcomes at scale, and is the more ambitious. The framework requires both. The instrumentation each needs already exists, in capnography and metabolic measurement for the acute arm and in cardiopulmonary-coupling analysis, chemoreflex testing, and daytime-performance measurement for the chronic; what is missing is the assembly, not the tools.

The acute arm admits a design capable of a decisive result. Within-person, each participant completes four conditions matched for time of day, posture, ambient light, prior sleep, caffeine, and activity: high-arousal digital engagement, low-arousal digital engagement such as calm reading on the same device, high-arousal non-digital engagement such as a sedentary social-evaluative stressor without a screen, and a low-arousal non-digital control. The high-arousal non-digital condition is essential, because it separates the digital-specific claim from the broader claim that sedentary arousal of any kind produces the mismatch; if it reproduces the respiratory signature, the attribution narrows from digital exposure to sedentary arousal, of which digital engagement is a major modern instance. The high-arousal condition must first be verified to raise independent arousal markers, such as heart-rate variability, electrodermal activity, and salivary cortisol, so a null downstream result cannot be dismissed as a failure to manipulate arousal. Ventilation, oxygen consumption, carbon dioxide production, and end-tidal carbon dioxide are measured simultaneously, with speech and breath-holding controlled, so hypocapnia is tested against metabolic rate rather than assumed; and across weeks the same participants are assessed for resting carbon dioxide and bicarbonate, chemoreflex responsiveness, overnight transcutaneous carbon dioxide, and sleep loop gain and non-REM stability. The design yields a clean negative if high-arousal engagement produces no ventilation-metabolism mismatch, or, more decisively, if it produces verified acute hypocapnia with no downstream shift in the resting or nocturnal measures after repeated exposure, a result that would place the effect as real but self-resetting and stop the chain before it reaches sleep or emotion.

A pre-bed crossover extension tests Prediction 6 directly. Participants spanning the baseline respiratory-stability distribution undergo matched high- and low-arousal evening conditions, followed by measurement of overnight carbon dioxide, respiratory events and arousals, cardiopulmonary coupling, slow-wave activity, and REM access. The analysis pre-specifies the exposure-by-baseline-stability interaction. Repeating the same exposure challenge after an intervention that measurably changes respiratory stability tests the within-person prediction while documenting whether the acute arousal response itself remains present. Light, posture, exposure duration, and sleep opportunity remain matched; breath-hold duration is recorded under a standardized protocol alongside, rather than substituted for, the respiratory measurements.

The chronic arm carries the harder identification problem and should be built around it. Its central threats are reverse causation, since anxious or insomniac individuals may seek high-arousal media, and confounding, since exposure, respiratory state, sleep, and emotion are entangled in ordinary life with fitness, medication, obesity, airway anatomy, and hormonal status. Adjustment alone cannot resolve this, so the framework commits to an explicit causal formulation: a directed acyclic graph, pre-specified adjustment sets and estimands, within-person repeated measures, and, decisively, a mediator-intervention arm in which carbon dioxide tolerance is shifted without altering digital exposure, since an intervention that moves the proposed mediator and moves the outcomes supplies the identification that observational mediation cannot. Sequencing follows: the acute arm first, as the cheapest and most decisive test, since a clear negative there halts the program before the expensive population work; the reversibility prediction in parallel, by within-subject designs in which each participant is their own control; and the population arm last, once the mechanism is established, when the question becomes not whether the pathway exists but how much of the burden it carries.

What Would Falsify the Framework

A framework that could accommodate any result would not be worth testing, so these are the results that would count against this one. The most direct refutation would come from the acute arm: if high-arousal digital engagement did not reduce end-tidal carbon dioxide more than low-arousal engagement matched for screen time and posture, the arousal-differentiation claim at the center of the framework would be refuted, and screen time rather than arousal load would stand as the relevant variable after all. If simultaneous measurement showed the ventilatory increase during engagement to be fully accounted for by a rise in metabolic rate, with no true hypocapnia beyond metabolic need, the ventilation-metabolism mismatch would be disconfirmed and the upstream mechanism would lose its footing. At the chronic joint, verified repeated acute mismatch followed by no persistent change in controller responsiveness, acid-base measures, or nocturnal respiratory stability would count against the proposed carryover. Normal morning bicarbonate alone would not refute the neural tier; each tier must be assessed using its corresponding measurements and exposure timescale. At the downstream joints, if arousal-weighted exposure predicted sleep and emotional outcomes no better than the blue-light and psychological confounds already do, once those confounds were properly controlled, the respiratory pathway would have failed to demonstrate the independent contribution the framework claims for it. The moderation prediction has a separate failure condition: if a well-controlled evening challenge spanning a measured range of baseline respiratory stability showed no exposure-by-stability interaction, or if a verified within-person change in that stability left exposure sensitivity unchanged, Prediction 6 would be weakened. Failure of the sleep-mediated emotional sequence would challenge that downstream extension without requiring failure of the respiratory-sleep mechanism itself. Each of these is a result the framework forbids, and each corresponds to a measurement the program specifies. The framework is offered on the understanding that it stands or falls by them.

One methodological safeguard belongs with these conditions, because without it they could be evaded. The arousal-weighted exposure metric must be specified and validated against independent arousal markers before outcomes are examined, not tuned afterward to fit them. A framework that could rescue a failed prediction by redefining arousal load after seeing the data would forfeit exactly the falsifiability claimed here; the metric is therefore fixed in advance, and if the pre-specified metric does not predict the respiratory outcomes, the upstream claim does not stand.

Relationship to Established Explanations, and Limitations

This section situates the framework against the explanations the field already relies on, and then states its limitations plainly. The two tasks belong together, because the framework's relationship to the established accounts is precisely what determines the standard it must meet, and the limitations are largely a matter of how far short of that standard the current evidence falls.

What the Framework Adds, and What It Does Not Displace

The framework does not ask the established explanations to step aside; they are well-supported and largely correct as far as they go. For sleep, the dominant account attributes technology's effects to blue light and circadian displacement, and both are real: evening light exposure does shift circadian timing and delay sleep onset, and the mechanism is well characterized. For wellbeing, the dominant accounts are psychological, invoking social comparison, attentional fragmentation, distraction, and the dynamics of online interaction, and these too are real and supported. The claim of this framework is not that these pathways are wrong but that a third, respiratory pathway operates alongside them, that it has not been identified or measured, and that its absence from the discussion is not evidence of its unimportance but only of its having gone unlooked-for. The established effects, real as they are, are often modest in magnitude relative to the burden of unrefreshing sleep and background reactivity that people actually report, which leaves room, though it does not require, for a further contributor to occupy.

There is a further, quieter relationship to the established sleep-medicine model itself, developed in sections “The Foundational Premise: Respiratory Control as a Modifiable Trait” and “Sleep Disruption and Its Downstream Consequences”. The prevailing model sorts sleep into disordered and normal by a diagnostic threshold and treats what falls below that threshold as fine. The framework proposes that a graded, sub-threshold respiratory instability degrades sleep across the whole population, including in people the threshold declares healthy, which is less a competing explanation than a reframing of what counts as a thing to be explained in the first place. On this view the field has not so much explained sub-optimal sleep wrongly as declined to treat it as a phenomenon requiring explanation, and the framework's contribution is partly to insist that it is one.

The reason the respiratory pathway might matter more than its current invisibility suggests is that it offers to explain features the established accounts leave unaddressed. The rival accounts can each reach some of these features: circadian and insomnia models can produce unrefreshing sleep, stress and social-evaluative models are somatic, and the bidirectional link between sleep and mood already predicts that the two problem domains travel together. What the respiratory pathway adds is not the only possible explanation of the somatic quality or the co-occurrence, but a single mechanism that yields both together and carries distinctive, measurable gas-exchange predictions the other accounts do not make. That is not proof that the respiratory pathway dominates, and it may well prove to be the smaller contributor; but a mechanism with its own testable signature has earned the right to be measured rather than assumed away.

The Demanding Standard: Statistical Separation

The framework's relationship to the established explanations sets the bar it must clear, and the bar is high. It is not enough to show that technology exposure correlates with sleep disruption or with emotional dysregulation, because the established explanations already predict exactly that, and a correlation between screens and poor sleep is equally consistent with blue light, with behavioral displacement, with psychological distress, and with the respiratory pathway proposed here. Demonstrating this framework therefore requires more than demonstrating the correlation it shares with its rivals. It requires showing that the respiratory pathway contributes independently: that arousal-differentiated exposure predicts respiratory dysregulation, and that respiratory dysregulation predicts the sleep and emotional outcomes, over and above what blue light, behavioral displacement, and psychological factors account for once those are properly measured and controlled. This is a demanding standard, and it is the correct one.

The difficulty is that the candidate predictors are themselves correlated. A person who spends their evenings in a high-arousal digital stream is also, by that fact, exposing themselves to evening light and often to the psychological content the rival accounts implicate; arousal-weighted exposure, photic exposure, and psychological distress travel together in ordinary life, and disentangling their separate contributions is not a matter of adding covariates to a regression and hoping. It requires designs that break the natural correlation, matching high-arousal and low-arousal engagement for light and duration in the laboratory, and measuring the respiratory mediators directly rather than inferring them, so that the respiratory pathway is given its own measured predictor and its own measured intermediary rather than being asked to appear as unexplained variance. This is why the acute, controlled arm of the measurement program matters so much: it is the setting in which the confound can be held fixed by design rather than modeled away, and it is where the framework can be tested most decisively, the setting in which a clean negative result would be unambiguous.

Limitations

The limitations of the framework follow from the honesty calibration maintained throughout, and stating them in one place is more useful than distributing them. The most fundamental is that the integrated chain has not been demonstrated as a whole. Its individual links are each supported by published evidence, and it is that support the paper has drawn on, but the assembly of the links into a single chain, and the attribution of the chain's upstream cause to arousal-weighted digital exposure at population scale, is the novel claim, and it is unproven. No study has yet measured the full pathway in a single population, and until the two acute-arm gaps identified in section “The Upstream Cause: Digital Environments and Ventilation-Metabolism Mismatch” are closed, the arousal-decomposition of exposure and the simultaneous measurement of ventilation against metabolic rate, even the upstream cause remains a well-motivated hypothesis rather than an established finding.

Several more specific limitations qualify particular parts of the argument. The sleep-mediated emotional extension requires longitudinal testing of the sequence from respiratory change through sleep disruption to emotional outcomes. The restorative-throughput prediction requires physiological and functional endpoints, not a reduction in sleep duration. The dose and persistence of the neural and biochemical tiers, and the magnitude of the acute-evening moderation effect, remain specific targets of the measurement program. The framework does not claim that digital technology is the sole cause of the respiratory dysregulation it describes; chronic psychological stress, habitual posture, and dysfunctional breathing patterns plausibly contribute, and the claim is only that digital exposure is a specific, measurable, and previously untested driver among them. And the alternative explanations for the phenomena in question, the photic and psychological accounts of sleep and mood, and further candidates such as hypothalamic-pituitary-adrenal dysregulation and inflammatory pathways, are not excluded by anything argued here; they are the very confounds the demanding standard of the previous subsection requires be controlled.

A final limitation concerns the evidentiary status of the specifics. This paper synthesizes across several literatures. Its citations have been checked against primary sources, and where a specific figure could not be traced to one it has been removed in favor of the qualitative claim it illustrated rather than retained on the authority of a secondary synthesis. A small number of well-established claims carry references still to be formatted. The qualitative structure of the framework, the chain and its architecture, does not hinge on any single figure, and no quantitative statement here should be treated as more than its source supports. Generalizability is likewise uncertain: the mechanisms are drawn largely from clinical and physiological work whose applicability across the full range of ages, sexes, and clinical subgroups has not been established. Developmental timing is a particular unknown, since adolescents are among the heaviest users of high-arousal digital media and are also still consolidating respiratory and sleep-control maturation, so the lack of evidence on developmental windows and on the magnitude of the proposed interaction with baseline respiratory-control traits is a real gap rather than a detail. The population-scale claim in particular awaits the population-scale measurement that would test it. Each of these is a place where the framework must meet data, which is what it means to offer it as a hypothesis.

Conclusion

This framework proposes that the sustained cognitive and emotional arousal of modern digital environments drives a chronic mismatch between ventilation and metabolic demand, producing repeated hypocapnia that may recalibrate respiratory control. Neural changes in controller responsiveness, with additional bicarbonate adaptation at more sustained exposure, are proposed to carry that disturbance into sleep. The resulting instability fragments stable non-REM, reduces restorative throughput, and impairs access to REM, providing a sleep-mediated route to emotional dysregulation. The operative exposure is arousal load rather than screen time. Its acute evening effect is predicted to interact with the baseline respiratory operating margin, so the same exposure can have markedly different consequences as that modifiable baseline changes. The distinctive tests measure the respiratory mismatch, its persistence into sleep, and this exposure-by-stability interaction.

Each individual link rests on existing peer-reviewed literature. The integration of the links into a single chain, and the attribution of the chain's upstream cause to digital technology at population scale, is novel and offered as a testable hypothesis, with the open questions named and the falsifying results specified in section “Predictions and the Measurement Program”. Its significance, if it survives that testing, is that it converts an untreatable generality into a tractable target: a complaint about the pace of modern life supports no intervention and no measurement, whereas a specific respiratory mechanism driven by a specific, quantifiable exposure supports both.

That target is unusually robust to the uncertainty surrounding the rest of the argument. Whatever the verdict on the digital attribution, the recalibrated respiratory state is a single, defined, and modifiable end-state, and it is the end-state, not its cause, that any intervention would address, so the practical arm of the framework does not stand or fall with the technological hypothesis.

The framework is offered, then, not as a finished claim but as a coherent, evidence-grounded hypothesis with a program for its own testing. Its components rest on established science; its integrated form and its population-scale attribution await the measurement that would confirm or refute them; and its contribution is to assemble the components into a chain precise enough to be tested and to name the measurable cause that makes it tractable.

References

Original bibliography wording is preserved. Linked source records identify verified bibliographic corrections; searches are labeled as searches.

1. Chou CA, Khan A, Singh A. The impact of 4% versus 3% hypopnea scoring criteria on obstructive sleep apnea diagnosis in women: a scoping review and commentary. Sleep Sci Pract. 2025;9(1):29.
   Read sourceSource record & notes
2. Guilleminault C, Stoohs R, Clerk A, Cetel M, Maistros P. A cause of excessive daytime sleepiness: the upper airway resistance syndrome. Chest. 1993;104(3):781-787.
   Read sourceAbstractSource record & notes
3. Guilleminault C, Winkle R, Korobkin R, Simmons B. Children and nocturnal snoring: evaluation of the effects of sleep related respiratory resistive load and daytime functioning. Eur J Pediatr. 1982;139(3):165-171.
   Read sourceAbstractSource record & notes
4. Deacon-Diaz NL, Sands SA, McEvoy RD, Catcheside PG. Daytime loop gain is elevated in obstructive sleep apnea but not reduced by CPAP treatment. J Appl Physiol. 2018;125(5):1490-1497.
   Read sourceAbstractOpen full textSource record & notes
5. Xie A, Skatrud JB, Barczi SR, Reichmuth K, Morgan BJ, Mont S, Dempsey JA. Influence of cerebral blood flow on breathing stability. J Appl Physiol. 2009;106(3):850-856.
   Read sourceAbstractOpen full textSource record & notes
6. Grassi B, Ferretti G, Costa M, Ferrigno M, Panzacchi A, Lundgren CEG, Marconi C, Cerretelli P. Ventilatory responses to hypercapnia and hypoxia in elite breath-hold divers. Respir Physiol. 1994;97(3):323-332.
   Read sourceAbstractSource record & notes
7. Tsai HJ, Kuo TBJ, Lee GS, Yang CCH. Efficacy of paced breathing for insomnia: enhances vagal activity and improves sleep quality. Psychophysiology. 2015;52(3):388-396.
   Read sourceAbstractSource record & notes
8. McKeown P, O'Connor-Reina C, Plaza G. Breathing Re-Education and Phenotypes of Sleep Apnea: A Review. J Clin Med. 2021;10(3):471. doi:10.3390/jcm10030471.
   Read sourceAbstractOpen full textSource record & notes
9. Ren X, Robbins PA. Ventilatory responses to hypercapnia and hypoxia after 6 h passive hyperventilation in humans. J Physiol. 1999;514(Pt 3):885-894.
   Read sourceAbstractOpen full textSource record & notes
10. Hafner M, Stepanek M, Taylor J, Troxel WM, van Stolk C. Why sleep matters: the economic costs of insufficient sleep. Rand Health Q. 2017;6(4):11.
    Read sourceAbstractOpen full textSource record & notes
11. Sakurai T. The neural circuit of orexin (hypocretin): maintaining sleep and wakefulness. Nat Rev Neurosci. 2007;8(3):171-181.
    Read sourceAbstractSource record & notes
12. Mahler SV, Moorman DE, Smith RJ, James MH, Aston-Jones G. Motivational activation: a unifying hypothesis of orexin/hypocretin function. Nat Neurosci. 2014;17(10):1298-1303.
    Read sourceAbstractOpen full textSource record & notes
13. Lazarenko RM, Stornetta RL, Bayliss DA, Guyenet PG. Orexin A activates retrotrapezoid neurons in mice. Respir Physiol Neurobiol. 2011;175(2):283-287.
    Read sourceAbstractOpen full textSource record & notes
14. Li A, Nattie E. Antagonism of rat orexin receptors by almorexant attenuates central chemoreception in wakefulness in the active period of the diurnal cycle. J Physiol. 2010;588(Pt 15):2935-2944.
    Read sourceAbstractOpen full textSource record & notes
15. Kim SJ, Pilowsky PM, Farnham MM. Intrathecal intermittent orexin-A causes sympathetic long-term facilitation and sensitizes the peripheral chemoreceptor response to hypoxia in rats. J Pharmacol Exp Ther. 2016;358(3):492-501.
    Read sourceAbstractOpen full textSource record & notes
16. Schleifer LM, Spalding TW, Hatfield BD, Kerick SE, Ley R. Effects of mental workload stress on end-tidal CO2 in computer work. Biol Psychol. 2003;62:234-235. (Related: Schleifer LM, Ley R. End-tidal PCO2 as an index of psychophysiological activity during VDT data-entry work and relaxation. Ergonomics. 1994;37(2):245-254; Schleifer LM, Ley R, Spalding TW. A hyperventilation theory of job stress and musculoskeletal disorders. Am J Ind Med. 2002;41(5):420-432.)
    Search PubMed by titleSource record & notes
    Bibliographic note: Original citation not exactly verified. Related verified sources are linked separately; the main link remains a title search.
17. Grassmann M, Vlemincx E, von Leupoldt A, Mittelstadt JM, Van den Bergh O. Respiratory changes in response to cognitive load: a systematic review. Neural Plast. 2016;2016:8146809.
    Read sourceAbstractOpen full textSource record & notes
18. Delistraty DA, Greene WA, Carlberg KA, Raver KK. Use of graded exercise to evaluate physiological hyperreactivity to mental stress. Med Sci Sports Exerc. 1991;23(4):476-481.
    Read sourceAbstractSource record & notes
19. Gardner WN. The pathophysiology of hyperventilation disorders. Chest. 1996;109(2):516-534.
    Read sourceAbstractSource record & notes
20. Krapf R, Beeler I, Hertner D, Hulter HN. Chronic respiratory alkalosis: the effect of sustained hyperventilation on renal regulation of acid-base equilibrium. N Engl J Med. 1991;324(20):1394-1401.
    Read sourceAbstractSource record & notes
21. Stornetta RL, Spirovski D, Moreira TS, Takakura AC, West GH, Gwilt JM, Pilowsky PM, Guyenet PG. Galanin is a selective marker of the retrotrapezoid nucleus in rats. J Comp Neurol. 2009;512(3):373-383.
    Read sourceAbstractOpen full textSource record & notes
    Bibliographic note: The source is J Comp Neurol 512(3):373–383 (2009; online 2008). Amygdala bibliography uses 2008.
22. Abbott SB, Burke PG, Pilowsky PM. Galanin microinjection into the pre-Botzinger or the Botzinger complex terminates central inspiratory activity and reduces responses to hypoxia and hypercapnia in rat. Respir Physiol Neurobiol. 2009;167(3):299-306.
    Read sourceAbstractSource record & notes
23. Dereli AS, Yaseen Z, Carrive P, Kumar NN. Adaptation of respiratory-related brain regions to long-term hypercapnia: focus on neuropeptides in the RTN. Front Neurosci. 2019;13:1343.
    Read sourceAbstractOpen full textSource record & notes
24. Dempsey JA, Veasey SC, Morgan BJ, O'Donnell CP. Pathophysiology of sleep apnea. Physiol Rev. 2010;90(1):47-112.
    Read sourceAbstractOpen full textSource record & notes
25. Shiroshita N, Shiota S, Kawana F, Uemura R, Sugiyama A, Kato T, Nagaoka K, Inoshita A, Kasai T, Takahashi K. Reversible conversion of moderate obstructive sleep apnea to severe behavioral hyperventilation-induced central sleep apnea under pandemic-related psychological stress: a case report. Sleep Sci Pract. 2026;10:5.
    Read sourceSource record & notes
26. Pevernagie D, Mariman A, Vandenbussche N, Tobback E, Overeem S, Delesie L, Janssen H, Vogelaers D. Behavioural hyperventilation as a novel clinical condition associated with central sleep apnoea: a report of three cases. Sleep Med. 2012;13(10):1317-1320.
    Read sourceAbstractSource record & notes
27. Messineo L, Taranto-Montemurro L, Azarbarzin A, Oliveira Marques MD, Calianese N, White DP, Wellman A, Sands SA. Breath-holding as a means to estimate the loop gain contribution to obstructive sleep apnoea. J Physiol. 2018;596(17):4043-4056.
    Read sourceAbstractOpen full textSource record & notes
    Bibliographic note: Pharmacological recalibration reference gives a different title and incorrect volume/pages; verified article is 596(17):4043–4056. Maximal breath-hold testing was evaluated as a marker of loop gain, not as a therapeutic training intervention.
28. Eckert DJ, White DP, Jordan AS, Malhotra A, Wellman A. Defining phenotypic causes of obstructive sleep apnea: identification of novel therapeutic targets. Am J Respir Crit Care Med. 2013;188(8):996-1004.
    Read sourceAbstractOpen full textSource record & notes
29. Thomas RJ, Mietus JE, Peng CK, Goldberger AL. An electrocardiogram-based technique to assess cardiopulmonary coupling during sleep. Sleep. 2005;28(9):1151-1161.
    Read sourceAbstractSource record & notes
30. Thomas RJ, Mietus JE, Peng CK, Guo D, Gozal D, Montgomery-Downs H, Bandla H. Relationship between delta power and the electrocardiogram-derived cardiopulmonary spectrogram: possible implications for assessing the effectiveness of sleep. Sleep Med. 2014;15(1):125-131.
    Read sourceAbstractOpen full textSource record & notes
31. Tasali E, Leproult R, Ehrmann DA, Van Cauter E. Slow-wave sleep and the risk of type 2 diabetes in humans. Proc Natl Acad Sci U S A. 2008;105(3):1044-1049.
    Read sourceAbstractOpen full textSource record & notes
32. Walker MP, van der Helm E. Overnight therapy? The role of sleep in emotional brain processing. Psychol Bull. 2009;135(5):731-748.
    Read sourceAbstractOpen full textSource record & notes
33. Goldstein AN, Walker MP. The role of sleep in emotional brain function. Annu Rev Clin Psychol. 2014;10:679-708.
    Read sourceAbstractOpen full textSource record & notes