The Reticular Activating System: The Switch Between Asleep and Awake

Introduction

Last week a storm went through and you slept for eight hours. Thunder, rain against the window, a car alarm somewhere down the street. None of it reached you.

Then someone said your name from another room, quietly, and you were awake before you understood why.

Almost everyone has had some version of that, and almost everyone has been given the same explanation for it. Somewhere in your brainstem there is a filter, a gatekeeper, a switch. It decides what gets through. Set it correctly and it will start showing you the things you care about. That explanation has a name attached to it, the reticular activating system, and it is repeated on wellness blogs, in coaching newsletters, in productivity threads and in a surprising number of places that should know better.

Here is the awkward part. The reticular activating system is real. It is one of the best-studied structures in neuroscience, it has been mapped in cats, rats and human brains, and damage to it is the difference between a person who is awake and a person who is not. But almost every popular claim made about it is either a mid-century summary that the field has since dismantled, or a straightforward misreading of what it does.

The most striking piece of that dismantling is a 2011 experiment in rats. A group at Harvard destroyed the thalamus, the relay drawn at the centre of every diagram of this system, the box that every arrow passes through. The animals stayed awake. Normal waking brain rhythms, normal behaviour, normal activity markers in the cortex. Then they destroyed a different structure that no textbook diagram bothers to include, and the animals stopped responding to the world entirely [1].

That result is fifteen years old at the time of writing and it does not appear anywhere on the first page of Google results for this topic.

This article is about what the reticular activating system actually is, where the story came from, which parts of it survived contact with modern evidence, and which parts did not. It is also, unavoidably, about the difference between a system that decides how awake you are and a system that decides what you pay attention to. Those are not the same thing, and conflating them is where most of the confusion starts.

What the Reticular Activating System Actually Is

Start with the word. Reticular means net-like, from the Latin for a small net, and it describes what nineteenth-century anatomists saw when they cut into the core of the brainstem and found no clean nuclei, no tidy layered structure, just a diffuse tangle of cells and fibres running through the middle of everything.

That tangle is the reticular formation. It runs the length of the brainstem, from the medulla at the top of the spinal cord, up through the pons, into the midbrain, and it is not a small structure. It sits at the centre of the brainstem the way the pith sits at the centre of a stem, and everything that passes between the spinal cord and the rest of the brain runs past it.

The reticular formation does a great many things. It contributes to breathing rhythm, to cardiovascular control, to muscle tone and posture [2], to pain modulation and to reflexes. Most of that has nothing to do with sleep.

The ascending reticular activating system is the specific claim that a subset of that tangle sends signals upward, toward the thalamus and the cortex, and that those upward signals are what keep the cortex in its waking state. The acronym in the literature is ARAS. When somebody says "the RAS" in a general-audience context, this is what they mean, or at least what they think they mean.

This is the first place a reader gets tripped, and it is worth being precise, because the confusion is not a pedantic one. The reticular formation is an anatomical structure that you can point at in a brain. The ascending reticular activating system is a functional hypothesis about what one part of that structure does. The first is a place. The second is a theory. Theories get revised, and this one has been revised heavily.

The core observation behind the theory has never been in doubt. If you damage the paramedian upper brainstem, roughly where the pons meets the midbrain, a person stops being conscious. Not confused, not drowsy. Comatose. That has been known from clinical material since the 1920s and it has been confirmed by every modern method applied to it, including voxel-based lesion mapping in living patients [3] and detailed human post-mortem tract tracing [4].

What has changed is the answer to two questions. Which cells do it, and by what route.

The Thirty-Two Years Before the Famous Experiment

Nearly every account of this subject opens with the same scene. In 1949, Giuseppe Moruzzi and Horace Magoun put an electrode into the brainstem of a sleeping cat, passed a current, and watched the cortex switch from the slow high-voltage rhythm of sleep to the fast low-voltage rhythm of alertness.

It is a good scene. It is also not the beginning.

The beginning is a disease. Between 1916 and roughly 1927 an epidemic of encephalitis lethargica moved through Europe and North America, and it did something that no experiment could ethically do. It damaged specific parts of people's brains and then left them alive to be observed. Constantin von Economo, a Viennese neurologist, spent years collecting those cases and correlating what he saw at the bedside with what he found at autopsy.

The pattern he found was almost too neat. Patients whose damage sat at the junction of the midbrain and the diencephalon, high in the brainstem, became profoundly and lastingly sleepy. Some slept for most of the day. Patients whose damage sat further forward, in the anterior hypothalamus and preoptic region, had the opposite problem. They could not sleep at all.

From that, in 1930, von Economo drew a conclusion that took the rest of the century to catch up with. There must be a waking region toward the back and a sleep-promoting region toward the front, and sleep is not the absence of waking but an active process with its own machinery [5]. He had no electrodes and no imaging. He had a clinical series from an epidemic, which is not a controlled experiment and which nobody should pretend gave a clean sample size, and he read it correctly.

Five years later, Frédéric Bremer went at the same question with a scalpel. Working in cats, he made two different cuts. One separated the forebrain from the brainstem at the level of the midbrain, a preparation he called cerveau isolé, the isolated brain. The other cut lower, at the junction of the brainstem and spinal cord, which he called encéphale isolé, the isolated encephalon. The results diverged sharply. The high cut left the cortex in a permanent sleep-like rhythm. The low cut left sleep and waking cycling normally. Whatever kept the cortex awake was entering it from somewhere between those two cuts. Bremer published this in 1935, in Comptes Rendus des Séances de la Société de Biologie et de ses Filiales, volume 118, pages 1235 to 1241, a pre-DOI publication that has no digital identifier to link to.

Then in 1946 Walle Nauta went at von Economo's inference directly, in rats. Lesions of the anterior hypothalamus produced sustained insomnia. Lesions further back produced somnolence [6]. The two-centre picture that von Economo had inferred from autopsy material now had experimental support, three years before anyone stimulated a cat.

None of this appears in the popular version. The popular version starts in 1949 and moves straight to the switch metaphor.

The shape of that list matters. It is not a story of one discovery. It is a story of a hypothesis that got progressively more specific, and then got progressively smaller.

1949: The Cat That Woke Up

What Moruzzi and Magoun actually did was more careful than the anecdote suggests.

They worked in cats, in a series of animals, and the paper does not report a single tidy sample size, so nobody should quote one. They used high-frequency electrical stimulation of the brainstem reticular formation and recorded the electroencephalogram from the cortex. In a drowsy or lightly sleeping animal the cortical EEG shows large slow synchronised waves. When they stimulated, that pattern was replaced within moments by the low-voltage fast activity that characterises an alert animal [7].

The technical term for the change is desynchronisation, and the name is worth understanding because it explains the whole logic. A synchronised EEG means enormous numbers of cortical neurons are rising and falling together, which produces big slow waves and, crucially, means the cortex is not doing much information processing. A desynchronised EEG means those neurons have broken out of lockstep and are firing more independently. The waves get small and fast. That is what a working cortex looks like.

So the finding was not "stimulation woke the cat up" in a loose behavioural sense. It was that a specific brainstem region could impose the electrical signature of wakefulness on the cortex, on demand. And the converse held: lesions in the same region produced a state from which the animal could not be roused.

Magoun spent the next year working out the boundaries of the effect [8], and in 1951 Starzl, Taylor and Magoun traced the ascending conduction upward into the diencephalon [9]. That last paper matters more than its citation count suggests, because it is where the thalamus entered the picture. The route became reticular formation to thalamus to cortex, and the arrow through the thalamus has been drawn in every diagram since.

Hold on to that. It is the arrow that later gets removed.

By the mid-1950s Alf Brodal had mapped the ascending fibres anatomically in the cat [10], and by the 1960s the first chemistry arrived. Bradley and Key showed that drugs could selectively alter the arousal response to reticular stimulation [11], and Kanai and Szerb showed that stimulating the midbrain reticular formation increased acetylcholine output from the cortex [12]. Something chemical was being delivered upward, not just an electrical signal.

The picture had become a system: one region, one ascending route, one relay, one effect. That is the reticular activating system as most people have met it, and it is the version that made it into the textbooks and out into the wider culture.

Doubt started early. In 1968 Aryeh Routtenberg published an argument that what everyone was calling arousal was at least two different things being measured with one instrument [13]. That did not stop the single-system reading from spreading. Simple pictures travel further than careful ones.

What the 1949 Picture Got Right

Before taking it apart, it is worth being clear about what has held up, because the overcorrection is as wrong as the myth and there is a version of this article circulating on the internet that claims the reticular activating system is a fiction. It is not.

These things are settled.

Damage to the paramedian upper brainstem causes coma. Von Economo's clinical series, Bremer's transections, Moruzzi and Magoun's lesions, modern lesion-symptom mapping and human connectivity work all converge on it, using completely different methods across nearly a century [14].

Waking cortex is electrically different from sleeping cortex in a specific and measurable way, and brainstem activity drives the difference. That has not moved since 1949.

There is an ascending pathway. Modern human anatomy has traced it in some detail. Post-mortem high-resolution diffusion imaging combined with immunohistochemistry has produced a detailed map of the human ascending arousal system [4], and diffusion tractography in living people has followed the route from the pontine reticular formation up toward the thalamus [15]. A second branch runs around the thalamus entirely, through the hypothalamus and basal forebrain [16].

Note the word "second". That is the first crack.

The Chemistry Nobody Drew

The 1949 experiment used an electrode, which is a blunt instrument. It stimulates whatever is near it. When later work went looking for which cells were actually responsible, it did not find a homogeneous net of reticular neurons. It found distinct chemical populations, each with its own cell group, its own transmitter, its own targets and its own behaviour across the sleep-wake cycle.

This is where the single-system picture starts to lose its shape, because these populations are not one thing that could be switched on or off. They are a set of partly overlapping systems that can dissociate from each other.

Each row of that table rests on specific work. The cholinergic projections from the pedunculopontine and laterodorsal tegmental nuclei were traced in detail in the late 1980s and 1990s [17] [18] and their role in state control is still being refined [19]. The pontomesencephalic tegmentum turned out to contain GABAergic neurons intermixed with the cholinergic ones, which complicated every simple story about that region [20], and the balance of glutamate and GABA in the pontine reticular formation tracks sleep duration directly [21].

The histaminergic tuberomammillary nucleus is the one most readers have already met without knowing it. It is the reason a first-generation antihistamine makes you sleepy: the drug crosses into the brain and blocks a wake-promoting transmitter [22] [23]. The dopaminergic contribution was the last to be properly characterised and is still being worked out [24]. If you want the background on how any of these signals actually pass between cells, that is covered separately in how neurons communicate, and dopamine's other job in learning is covered in dopamine and learning.

The modern synthesis of all this, published in Nature Neuroscience in 2022, does not describe a system at all. It describes an orchestration: multiple cell groups with different transmitters, different targets and different timing, whose combined output produces the states we call waking and sleeping [25].

An orchestra is not a switch. Nobody flips an orchestra.

The Experiment That Broke the Diagram

Now the thalamus.

Every drawing of the ascending reticular activating system routes the signal through the thalamus. It is anatomically reasonable. The thalamus is the relay through which nearly all sensory information reaches the cortex, and Starzl and Magoun put it into the diagram in 1951 for good reasons. The intralaminar thalamic nuclei project diffusely to cortex, which is exactly what a general arousal signal would want.

In 2011, Patrick Fuller, David Sherman, Nigel Pedersen, Clifford Saper and Jun Lu tested it in rats, and the result is the reason this article exists.

They started from a puzzle. Acute lesions that cut across the paramedian midbrain produce deep coma. But cell-specific lesions of the monoaminergic and cholinergic cell groups in that region, the ones everyone assumed were doing the work, had never reproduced that coma. Kill the cells and the animal stays awake. Cut the axons passing through and it does not. Something else was running through there.

So they asked which cortical input mattered more, the one from the thalamus or the one from the basal forebrain, and they made large cell-body-specific lesions of each.

Extensive thalamic lesions had little effect. Not a subtle effect. Little effect on the waking EEG, little effect on behavioural measures of wakefulness, little effect on the molecular activity markers that show cortical neurons have been active during waking.

Large basal forebrain lesions did something entirely different. Those animals were behaviourally unresponsive, their cortical EEG collapsed into a monotonous rhythm below one cycle per second, and their cortical activity markers stayed low even under continuous gentle handling.

Then they traced backward from the basal forebrain to find what feeds it from the brainstem, and found a substantial input from glutamatergic neurons in the parabrachial nucleus and the adjacent precoeruleus area. Cell-specific lesions of that parabrachial and precoeruleus complex reproduced the whole picture: unresponsive behaviour, a monotonous sub-one-hertz cortex, and no cortical activity markers [1].

Their own summary is that in rats the reticulo-thalamo-cortical pathway may play a very limited role in behavioural or electrocortical arousal, and that the route from parabrachial nucleus and precoeruleus, relayed through the basal forebrain, may be the critical one.

Read that against the diagram in your memory. The box in the middle of the diagram turned out to be, for this purpose, close to optional. The structure that actually carried the signal was not in the diagram at all.

This was not a bolt from nowhere. Fifteen years earlier Mircea Steriade, who spent a career on thalamocortical physiology and knew that circuit better than almost anyone, published a short piece in Science under the title "Arousal: revisiting the reticular activating system", arguing that the term had outlived its usefulness and was obscuring more than it explained [26]. In 2019 a clinical reassessment made a similar point from the bedside end, noting how much diagnostic weight the phrase carries relative to how loosely it is defined [27].

The best single account of the whole arc is a 2022 paper with an unusually honest title, "A Century Searching for the Neurons Necessary for Wakefulness" [28]. A century of searching implies the search is not over, which is exactly right.

So Where Is the Switch?

Something in your brain does behave like a switch. Falling asleep is not a slow dimmer. It is quick, it is discrete, and the boundary is sharp enough that people routinely fail to notice they crossed it.

That switch exists. It is just not the reticular activating system, and it is not in the brainstem.

In 1996 Sherin, Shiromani, McCarley and Saper went looking for cells that are active during sleep rather than during waking. Using an immediate-early gene product as a marker of recent activity in rats, they found a compact group of neurons in the ventrolateral preoptic nucleus of the hypothalamus, usually shortened to VLPO, that switch on specifically when the animal sleeps. Then they used a retrograde tracer to see where those cells project, and found they innervate the tuberomammillary nucleus, the histaminergic arousal group from the table above [29].

That is a sleep-active cell group whose job is to inhibit a wake-active cell group. And the arousal cell groups inhibit the VLPO right back.

Two populations that inhibit each other produce a specific kind of behaviour, and it is the behaviour of an electronic circuit called a flip-flop. If the arousal side is slightly stronger, it suppresses the sleep side, which removes the inhibition on itself, which makes it stronger still. The system runs away to a fully awake state and stays there. If the sleep side gets slightly stronger, the same runaway happens in the opposite direction. The intermediate states are unstable. The system cannot sit in the middle, so it does not.

Saper, Chou and Scammell proposed this explicitly in 2001 [30] and developed it into a full account of state switching in 2010 [31]. A twenty-five-year review of what has held up was published in 2022, and it is worth reading for its caveats rather than its headline, because the VLPO has turned out to be more heterogeneous than the original model assumed [32].

The model explains several things at once. It explains why transitions are fast. It explains why the state you are in tends to persist. And it explains why a weakened switch produces something stranger than grogginess. A degraded flip-flop does not settle in the middle. It flips at the wrong times, which is a description of a specific human condition.

Orexin: The Finger That Holds the Switch

In 1998, two laboratories working on completely different problems described the same thing.

One group, studying genes expressed selectively in the hypothalamus, found a pair of peptides they named hypocretins, for hypothalamic secretin-like [33]. The other, screening for ligands of orphan G-protein-coupled receptors and interested in feeding behaviour, found the same molecules and named them orexins, from the Greek for appetite [34].

Same year, same peptides, two names. That is why the literature is inconsistent to this day, and why a reader encountering both words is not looking at two systems.

Neither group was studying sleep. Sleep found them.

In 1999, work on a colony of Doberman pinschers with inherited narcolepsy traced the condition to a mutation in the hypocretin receptor 2 gene [35]. A single receptor, broken, producing a dog that collapses into sleep during excitement.

Human narcolepsy, though, is not usually inherited. It is discordant in identical twins and rarely runs in families, so the canine result did not obviously transfer. Then in 2000 a team examined post-mortem brains from people who had narcolepsy. They counted hypocretin neurons and found a reduction of roughly eighty-five to ninety-five per cent, with the melanin-concentrating hormone neurons that sit intermingled among them left intact, and signs of gliosis suggesting the cells had degenerated rather than never formed [36].

That study examined four narcoleptic brains. Four. It is worth saying that number out loud, because a finding this consequential resting on four post-mortem specimens is the kind of thing that should be flagged rather than smoothed over. It has since been supported by a large amount of independent work, including cerebrospinal fluid measurements in living patients [37], but the original observation was n equals four.

What orexin does in the switch is the interesting part, and it is not what the word "switch" suggests. Orexin neurons project to essentially every arousal cell group in the table above and excite them. They are not one more arousal system running in parallel. They are the input that reinforces all the others.

The cleanest demonstration of that is a 2001 experiment showing that orexin A's arousal effect depends on an intact histaminergic system [38]. Block histamine and orexin's effect on wakefulness largely goes with it. Orexin works through the arousal systems, not instead of them. Later work mapped its actions onto the dorsal raphe and laterodorsal tegmentum directly [39] [40], showed the reverse traffic from the monoamines back onto orexin neurons [41], traced orexin projections onto the sleep-promoting VLPO itself [42], and eventually delivered the causal test: optogenetic stimulation of hypocretin neurons drives transitions from sleep to waking [43]. Selective ablation of the same neurons produces the narcoleptic phenotype in mice [44].

So the picture is a flip-flop with a stabiliser. The VLPO and the arousal systems form the switch. Orexin holds it wherever it currently sits. Take orexin away and the switch still works, which is why people with narcolepsy sleep and wake normally in the basic sense, but it becomes twitchy, which is why they fall asleep at the wrong moment and wake repeatedly at night [45] [46].

The pharmacology confirms the logic from the other direction. Blocking orexin receptors promotes sleep, first shown across rats, dogs and humans with an experimental antagonist [47], with dual receptor blockade proving more effective than blocking either receptor alone [48]. That mechanism is now a recognised drug class [49], and the measured effect on sleep EEG spectra is not identical to natural sleep [50], which is itself informative about how much of sleep is more than the switch position. None of this is advice about anyone's sleep. It is mechanism, described because the mechanism is the point.

What Decides Which Way It Flips

A bistable switch needs something to push it. Two things do.

The first is a pressure that builds the longer you stay awake and dissipates while you sleep. Alexander Borbély formalised this in 1982 as the two-process model, separating a homeostatic sleep pressure that accumulates with waking from a circadian process that runs on its own roughly twenty-four-hour cycle regardless of whether you slept. He wrote his own retrospective on the model forty years later, and it is a good account of what the model got right and where it has been stretched [51].

The chemistry of the homeostatic side is best understood for adenosine, which accumulates in the brain during prolonged waking and acts on arousal-promoting cells to reduce their output [52]. Caffeine's mechanism is blocking adenosine receptors, which is to say caffeine does not add alertness, it removes an accumulating signal for sleep.

The circadian side is the clock in the suprachiasmatic nucleus and the pathways by which it biases arousal across the day [53]. That is a subject in its own right, and it has direct consequences for when learning actually works, covered in circadian rhythms and the best time to study.

The two processes push on the same switch from different directions, which is why you can be exhausted and unable to sleep, or rested and unable to stay awake at the wrong hour. What happens once the switch has flipped, particularly to memory, is a separate machinery covered in how sleep consolidates spaced learning.

The Filtering Claim, Taken Apart

Back to the storm and the whispered name.

The popular claim goes roughly like this: the reticular activating system filters the flood of incoming sensation, admits what matters and blocks what does not, and because it decides what matters based on what you care about, you can influence it by deciding what to care about. Set a goal, and the system will start surfacing opportunities related to that goal.

There are two claims tangled together there, and they have completely different evidential status.

The first claim is that arousal state changes what sensory information reaches cortex. That is true and it is well measured.

Start with the observation everyone recognises. In 1999, Perrin and colleagues recorded auditory evoked potentials in ten adults while they heard their own first name and seven other first names, in random order, both awake and asleep. Awake, the subject's own name produced an enhanced positive response around five hundred milliseconds, even when they were not attending to the sounds. In stage two sleep, the names evoked K-complexes, and while the late portion of the K-complex was the same for all names, the early portion was selectively enhanced for the subject's own name [54]. Ten people is a small study, and the effect has held up, but the number belongs in the sentence.

So something in a sleeping brain distinguishes your name from other names. That much of the popular story is real.

What was missing until recently is where in the processing chain the gating happens, and a 2025 study answers it with unusual clarity. Fourteen adults took a two and a half hour nap while magnetoencephalography recorded frequency-following responses, an evoked response that tracks how faithfully neurons encode the periodicity of a sound. Because the response can be source-localised, the researchers could measure encoding separately in brainstem, thalamus and auditory cortex across sleep stages.

The result splits the auditory system in two. Subcortical encoding of sound was maintained across non-REM sleep stages. It did not degrade with depth. Cortical encoding did degrade, and the deeper the sleep the weaker it got. The factor that tracked the decline was not slow-wave activity and not sleep spindles, the two things everyone assumed were doing the blocking. It was reduced communication between thalamus and cortex [55].

That is a real filter, in a real place, doing a real job: keep the sensory front end listening, disconnect the part that would wake you up about it. Spindles do contribute to gating in other ways and are worth understanding on their own terms [56], and the transition into sleep has its own measurable dynamics [57]. There is also a curious asymmetry across senses: smell is unusually poor at waking a sleeper, which is one reason smoke alarms are loud rather than scented [58].

Notice, though, what the gate is doing. It is set by state, not by content. It goes down when you are deeply asleep and up when you are not. Nothing in that mechanism knows what your goals are.

What Actually Filters for Your Goals

The second claim, the one about setting a goal and having the system surface relevant things, is not describing brainstem arousal at all. It is describing attention, and attention has its own machinery further up.

Two lines of work cover the territory the popular claim is reaching for.

The first is the salience network. In 2007, Seeley and colleagues used functional imaging to separate two intrinsic brain networks that had been running together in earlier analyses: an executive control network anchored in dorsolateral prefrontal and parietal cortex, and a distinct network anchored in the anterior insula and dorsal anterior cingulate cortex whose activity tracked how personally significant a stimulus was [59]. That second network is cortical. It is not in the brainstem, it is not the reticular formation, and it is the closest thing in the brain to the "flags what matters to you" function the popular story attributes to the RAS.

The second line is the honest bridge between the two, and it goes through the locus coeruleus. Aston-Jones and Cohen's adaptive gain theory describes how the noradrenergic system, which is an arousal system, modulates the gain on cortical processing rather than selecting content. In its phasic mode it briefly amplifies processing of a currently relevant target; in its tonic mode it broadens responsiveness at the cost of focus [60].

That is genuinely a case of an arousal nucleus changing what gets through. But look at what it changes. It sets how strongly the cortex responds, not what the cortex responds to. The selection of what matters is made upstream, in cortex, by systems that have access to your goals. Arousal sets the volume. It does not choose the station.

This distinction is the whole disagreement in one sentence, and it is why the popular version is not simply exaggerated but misattributed. The mechanisms that determine what you notice are covered in more depth in attention and memory.

One more thing is worth saying because it is the part of the popular claim that survives, in modified form. Arousal level does constrain attention, and the relationship is not linear. A 2024 study found that strategically stabilising arousal, keeping it in a narrow band rather than letting it drift, improved sustained attention performance [61]. So there is a real and useful relationship between how awake you are and how well you can hold focus. It is not "your brainstem will find your goals for you". It is closer to "the state you are in sets a ceiling on the attention you can deploy", which is the same territory covered in deep focus and sustained attention.

Awake Is Not the Same as Aware

The clinical material makes a distinction that ordinary language does not, and once you have it, several confusing conditions become clear at once.

Arousal is the level of activation. Awareness is the content of experience. They usually travel together, which is why English uses "conscious" for both, but they are produced by different machinery and they can come apart in either direction.

The vegetative state, now more often called unresponsive wakefulness syndrome, is the case that proves arousal and awareness are separable. These patients open their eyes. They have sleep-wake cycles. The brainstem arousal machinery is doing its job. What is missing is any evidence of experience [62].

Locked-in syndrome is the mirror image and it is the more disturbing of the two. A lesion in the ventral pons interrupts the descending motor pathways while leaving the ascending arousal system and the cortex intact. The patient is fully awake, fully aware, and almost completely unable to move or speak. Vertical eye movement is often all that remains [63]. The reason this matters here is that the lesion sits right next to the arousal system without damaging it, which is about as clean a demonstration as clinical neurology offers that these are separate systems in adjacent tissue.

When the Switch Breaks

Coma is what happens when the arousal side fails.

The precision of modern lesion work here is worth appreciating. In 2016, Fischer and colleagues compared twelve brainstem lesions that had caused coma against twenty-four brainstem lesions that had not, using voxel-based lesion-symptom mapping. A small region in the rostral dorsolateral pontine tegmentum came out significantly associated with coma. They then took that brainstem site and asked what it is functionally connected to in healthy brains, and found a network reaching the ventral anterior insula and pregenual anterior cingulate cortex, regions whose connectivity was also altered in patients with disorders of consciousness [3]. Twelve versus twenty-four is a small study and the authors present it as such, but the convergence with a century of earlier lesion material is what makes it persuasive.

Traumatic injury produces a related picture, with disruption of the ascending arousal network detectable in acute traumatic disorders of consciousness [64], and there is active work on which brainstem circuits can be targeted for recovery [65] [66].

Failures do not have to be catastrophic to be informative. Chronic insomnia is associated with measurable differences in the ascending arousal network [67], which reframes it as a disorder of arousal regulation rather than a simple failure to relax. In Alzheimer's disease, tau pathology accumulates early in exactly the sleep-regulating and wake-regulating cell groups described above, which helps explain why sleep fragmentation appears years before memory symptoms [68]. Delirium in hospital, which is common and serious, shows neurophysiological patterns consistent with arousal-system vulnerability [69]. And there is at least one published case of a tumour damaging this region and producing severe intractable insomnia rather than sleepiness, which is a single patient and should be read as an illustration rather than evidence of frequency [70].

The general lesson across all of these is that this system fails in both directions. Popular accounts imagine failure as sleepiness. Damage in the right place produces the opposite, exactly as von Economo saw in 1930.

Anaesthesia: The Switch Forced

General anaesthesia is often described as deep sleep. It is not, and the difference illuminates the whole system.

The reference statement on this separates anaesthesia, sleep and coma as distinct states that share circuitry [71]. Anaesthetic drugs act on many of the same targets as the natural arousal systems, but they impose their states rather than allowing the switch to move on its own, and the resulting brain states are dose-dependent and progress through patterns like burst suppression that natural sleep never produces. The circuit-level story is still being assembled [72].

The most revealing finding in this area is that emergence from anaesthesia is not induction played backwards. The concentration of drug at which a patient loses consciousness differs from the concentration at which they regain it, and the state shows a resistance to change that has been named neural inertia [73] [74]. Induction and emergence engage measurably different mechanisms [75], and there is now serious work on pharmacologically driving emergence rather than waiting for it [76].

Hysteresis is exactly what a bistable flip-flop predicts. A switch that resists leaving whichever state it is in is a switch that will not flip at the same threshold in both directions. The clinical observation and the circuit model agree, which is the sort of agreement that makes a model worth keeping. The interaction between these drugs and memory formation specifically is covered in anaesthesia and memory.

What Is Still Unsettled

An article that corrects a popular myth has an obligation not to replace it with a tidier myth. The honest position on several points here is that the field disagrees.

The largest open question is the one this article has been circling: which relay actually carries arousal to the cortex.

On one side, the thalamus is essential. That position runs from Starzl and Magoun's 1951 tracing through arguments that thalamic circuits are central to attention and consciousness [77], and it is very much alive. A 2024 review in Neuron argues for thalamic contributions to both the state and the contents of consciousness [78], and a 2025 systematic review maps which thalamic nuclei are implicated in consciousness across the clinical literature [79].

On the other side, the thalamic relay is largely dispensable for arousal specifically. That is the Fuller result, supported by Steriade's earlier scepticism and by the historiography in Grady and Boes.

There is a plausible reconciliation, and it should be offered as a possibility rather than a verdict. The two camps may be answering different questions. Arousal level, meaning how activated the cortex is, may depend mainly on the parabrachial and basal forebrain route. The contents of consciousness, meaning what is actually represented, may depend heavily on thalamocortical loops. The title of the 2024 Neuron paper separates state from contents, which suggests the field is already thinking along these lines. But nobody has demonstrated it, and treating a plausible reconciliation as an established one is how the last version of this story went wrong.

A second and smaller disagreement is whether the phrase "reticular activating system" should still be used. Steriade argued against it in 1996 and the clinical reassessment in 2019 raised similar concerns, yet the term remains standard in clinical writing, including in careful reviews of its relevance to brain death determination [80]. Terms outlive the models that produced them. That is not unique to this one.

What is not in dispute, and should not be dressed up as if it were, is that ascending arousal projections from the brainstem exist and are necessary for consciousness. There is a version of this correction circulating that overshoots into "the RAS is a myth". It is not a myth. It is a system that turned out to be plural where it was described as singular, and routed differently from the way it was drawn.

Some newer work is trying to build a fuller subcortical account of where consciousness arises at all [81], and the broader neurobiology of sleep continues to be revised at a pace that makes any confident summary a hostage to fortune [82] [83].

What to Take From This

The reticular activating system is a real anatomical claim that has been narrowed, corrected and partly rerouted over a century of work.

What survives is the core: something ascending from the brainstem keeps the cortex in its waking state, and damage there causes coma. What has changed is nearly everything about the details. It is not one system but several chemically distinct ones. Its critical relay to cortex may not be the thalamus at all. The actual bistable switch between sleeping and waking sits in the hypothalamus, not the brainstem, and it is stabilised by a peptide nobody knew existed until 1998.

And the filtering story, which is what brings most people to this topic, splits cleanly in two once you look at it. Arousal state really does gate what reaches your cortex, with the gate sitting at the thalamocortical connection rather than in the brainstem, which is why you slept through the storm. But nothing in that mechanism knows your goals. What flags personal significance is cortical, and what selects targets is attention. The brainstem sets how loudly the world arrives. It does not decide what the world is about.

That is a less satisfying story than a switch you can program. It is also considerably more interesting, and it has the advantage of being what the evidence says.