The Glymphatic System: How the Sleeping Brain Flushes Its Own Waste

Introduction

Your brain is about two percent of your body weight and it burns roughly a fifth of your energy. Everything that burns, makes waste.

Muscle tissue has an answer for that. So does your gut, your skin, and every other organ you own. Lymph vessels thread through them, collecting the fluid that leaks out of blood capillaries along with whatever the cells have thrown away, and carrying it off to lymph nodes to be sorted and dumped. It is unglamorous municipal plumbing and it runs everywhere in your body.

Except one place.

Open a human brain and you will not find lymph vessels running through the tissue. Anatomists have been looking since the seventeenth century. They are not there. The organ with the highest metabolic rate in your body, the one that cannot afford a single day of neglect, appears to have no rubbish collection at all.

That is the puzzle this article is about. It has an answer, or at least a proposed one, and the proposed answer has a name that was coined in 2012 and started an argument that is still going. The name is the glymphatic system.

Here is what you should know before you read another word about it. The pathway is real. Fluid genuinely moves through it. And the single most repeated claim about it, the one on every wellness blog and in every hospital explainer, was directly contradicted in a top journal in 2024 by a lab that measured the same thing and got the opposite answer.

Nobody told you that. This article tells you that, and it tells you what each side actually measured, because that is the only way you can judge any of it.

The Hundred Years Before Anyone Had a Name For It

The glymphatic system is usually introduced as a discovery from 2012, which makes it sound like it fell out of the sky. It did not.

Cerebrospinal fluid has been known about since antiquity and studied seriously since the 1700s. By the mid twentieth century researchers had established that this fluid is produced continuously, mostly by the choroid plexus, circulates through the ventricles and around the brain and spinal cord, and drains away somewhere [1]. The vague word in that sentence is "somewhere", and it was vague for a very long time.

Two camps formed. One argued the fluid simply diffuses through brain tissue, meaning molecules wander at random and net movement is slow and undirected. The other argued for bulk flow, meaning the fluid is actually pushed in a direction, which would be much faster and would carry large molecules that diffusion cannot shift. This argument is older than most of the people having it now, and it was already being fought carefully in print decades before 2012 [2].

Historical reviews of cerebrospinal fluid drainage make the continuity plain: the routes out of the skull, along cranial nerves and toward the cervical lymph nodes, were being described and disputed long before anybody used the word glymphatic [3].

So when you read that this was discovered in 2012, hold that loosely. What happened in 2012 was that somebody put a new technique on an old question and got a picture vivid enough to change the conversation.

2012: A Pathway, and a Word That Started an Argument

The technique was two-photon microscopy in a living mouse. Inject a fluorescent tracer into the cerebrospinal fluid, then watch, in a live animal, where it goes.

Jeffrey Iliff, Maiken Nedergaard and their colleagues published what they saw in 2012 [4]. The tracer did not sit politely on the brain surface. It ran inward along the outside of arteries, through the narrow sleeves of space that wrap every penetrating vessel. From there it moved into the tissue itself, mixed with the fluid between cells, and left again along the outside of veins. Soluble amyloid beta, the protein that forms plaques in Alzheimer's disease, came out with it.

They called it a paravascular pathway. Then Nedergaard's group gave it the name that stuck.

"Glymphatic" is glia plus lymphatic. It is two claims welded into one word: that this is the brain's equivalent of the lymphatic system, and that glial cells, specifically astrocytes, are what make it work. The following year the group asked the question in a title, plainly: is there a cerebral lymphatic system [5].

Naming a thing before the field agrees it exists is a bold move. It is also, as you will see, part of why the argument got sharp. Some researchers object to the word itself, not just the model behind it.

What the Plumbing Actually Looks Like

Strip away the controversy for a moment and here is the proposed layout.

Every artery that dives into your brain carries a sleeve of space around it. These are perivascular spaces, and unlike almost everything else here they are not in dispute at all. You can see enlarged ones on a routine human MRI, and their size and distribution are studied in their own right as a marker of small vessel disease [6].

Cerebrospinal fluid is proposed to flow inward along these periarterial sleeves. To get from the sleeve into the tissue it has to cross a barrier made of astrocyte endfeet, the flattened terminals that wrap almost the entire vascular surface. Those endfeet are studded with a water channel called aquaporin-4, and the model says aquaporin-4 is what lets the water through [7].

Once inside, the incoming fluid mixes with interstitial fluid, the liquid filling the gaps between your brain cells. That gap space is narrower than most people imagine, roughly a fifth of brain volume, and getting anything through it is genuinely difficult [8]. The mixed fluid then collects along veins and drains out.

In, across, through, out. Four steps.

The most complete current account of brain fluid transport, written by three of the people most associated with the model, runs to a full issue of Physiological Reviews and is worth knowing exists if you ever want the unabridged version [9].

Notice what the diagram does not tell you. It does not tell you how fast, how much, or what makes it move. Those are the questions everything else in this article turns on.

The Heart Is the Pump. Or Part of It.

If fluid is being pushed rather than wandering, something has to push it.

The first proposed answer was your pulse. Iliff and colleagues showed in mice that when they dampened arterial pulsation, paravascular exchange dropped, and when they increased it, exchange rose [10]. The mechanism is intuitive once you picture it. Each heartbeat makes the artery wall bulge outward and spring back. In a narrow fluid-filled sleeve, that repeated squeeze acts like a peristaltic pump.

Six years later a different group did something more direct. Instead of inferring flow, they put microspheres into the cerebrospinal fluid of live mice and filmed individual particles moving in the perivascular spaces of the brain surface [11]. The particles moved. They moved in the same direction as blood flow, they moved in time with the cardiac cycle, and in hypertensive mice they moved less.

That study matters more than its citation count suggests, because it converted an inference into an observation. You do not have to believe a model to believe a film of a bead moving.

The same group also showed the whole pathway could be captured non-invasively with contrast MRI rather than only through a surgical window, which is the step that eventually made human work thinkable [12].

2013: The Experiment That Made It Famous

Here is the study that put this topic on your radar, whether or not you have heard of it.

In 2013, Lulu Xie and colleagues reported in Science that in live mice, natural sleep or anaesthesia was accompanied by roughly a sixty percent increase in the volume of the space between brain cells [13]. More space means less resistance. Less resistance means fluid moves through more easily. They measured a large increase in cerebrospinal fluid exchange during sleep and a faster rate of amyloid beta clearance along with it.

Read that back and you can see why it travelled. It gives sleep a job. Not a vague restorative job, a specific mechanical one. The brain widens its own gaps at night so it can be rinsed.

Later work sharpened the timing. Clearance in mice tracks EEG delta power and low heart rate more closely than it tracks sleep as a category, which points at a particular physiological state rather than at unconsciousness in general [14]. Brain lactate concentration, a rough proxy for metabolic waste, also swings with brain state in a way consistent with clearance [15].

If you want the version of this story that connects to memory rather than to plumbing, our article on what the sleeping brain does with the day's memories covers the consolidation side of the same night.

The Number You Have Read, and What It Actually Was

You have almost certainly encountered that sixty percent figure. It shows up in mattress advertising, supplement copy, and a fair number of otherwise careful health articles. It is usually phrased as "your brain cells shrink by sixty percent during sleep".

That is not what was measured, and it is worth being precise about why.

What was measured: the interstitial space, meaning the gaps between cells, expanded by about sixty percent. Cells did not lose sixty percent of their volume. A modest shrinkage of many cells produces a large proportional change in the small space between them, which is exactly the point.

In what: mice. Live ones, imaged through a cranial window.

Shown in humans: no.

That last line is not a criticism of the study. It is a superb study. It is a criticism of every page that hands you a mouse number as if a technician had measured it in your skull last Tuesday. You will see this pattern repeatedly below, so it is worth installing the habit now: whenever you meet a number in this field, ask which species it came from. Most of the time the honest answer is mouse.

Then Somebody Looked in a Human

For years the human evidence was the weak point, and everybody in the field knew it.

Then in 2019 Nina Fultz and colleagues published something genuinely striking in Science [16]. They put people to sleep inside an MRI scanner and used accelerated imaging to capture several things at once: electrical brain activity, blood oxygenation, and the movement of cerebrospinal fluid at the base of the fourth ventricle.

During non-REM sleep, a pattern appeared. A slow wave of neural activity would pass. Blood volume would drop a beat later. And then a large pulse of cerebrospinal fluid would surge into the head.

Neural, then vascular, then fluid. Over and over, all night, in living humans.

Related work using ultra-fast imaging had already shown that human brain pulsations come in several distinct physiological rhythms rather than one, which is the technical groundwork this rests on [17].

Now here is the sentence that most coverage of the Fultz study leaves out, and it is the most important piece of precision in this article.

Showing that fluid moves is not the same as showing that waste leaves.

The Fultz study is beautiful evidence that large CSF pulses happen in the sleeping human brain and are locked to slow waves. It is not evidence that those pulses carried anything out. Flow and clearance are different measurements, and a great deal of confused writing about this topic comes from treating them as one. Keep them apart in your head and half the noise disappears.

Where Does It All Go?

Suppose fluid does wash through. It still has to leave the skull.

For most of the twentieth century the honest answer was that nobody was sure. Then in 2015, two independent groups published the same surprise within months of each other: there are lymphatic vessels in the dura, the tough outer membrane wrapping the brain, running alongside the venous sinuses [18] [19].

Think about how odd that is. The meninges have been dissected, stained and taught for centuries. Two labs found a whole vessel network in them in 2015.

The finding held up. It was extended to human and non-human primate meninges, where the vessels can be seen non-invasively on MRI [20]. Later work argued the main drainage route is at the skull base rather than the top [21]. Blocking these vessels in mice worsened amyloid pathology and cognitive measures in ageing animals [22], and further work mapped how the meningeal and glymphatic routes relate to each other [23] and how the vessels themselves develop [24].

Two things are worth separating here. The meningeal lymphatic discovery is on much firmer ground than the sleep-clearance claim. It is anatomy, it was found twice independently, and it has been confirmed in humans. When you read a sceptical piece about the glymphatic system, the scepticism is almost never aimed at this part.

Read that timeline again and notice its shape. It is not a straight line from ignorance to knowledge. It is a claim, a challenge, a counter-challenge, and a reframe. That is the actual story, and the next four sections walk through the fight.

The First Serious Objection

In 2017, Alex Smith, Alan Verkman and colleagues published a paper in eLife with a title that leaves nothing to interpretation: a test of the glymphatic hypothesis demonstrating diffusive and aquaporin-4-independent solute transport in rodent brain [25].

They went after three pillars at once.

First, they injected fluorescent dextrans of different molecular sizes and looked at how far each travelled. If transport is convective, meaning the fluid carries everything along together, size should not matter much. If it is diffusive, small molecules should outrun large ones in a predictable way. Size mattered, in the way diffusion predicts.

Second, they measured transport in the extracellular space using a photobleaching technique, then killed the animals. If a heartbeat-driven pump is doing the work, transport should collapse the moment the heart stops. It did not change.

Third, and most damaging to the model as stated, they deleted the aquaporin-4 gene. The glymphatic model says aquaporin-4 is the doorway. Remove the doorway and influx should fall. In their hands it did not fall, in mice or in rats.

Modelling work pointed the same way. Separate groups built physical models of solute transport through brain tissue and concluded that bulk flow was not required to explain the observations, and in one case that the geometry does not support the proposed mechanism at all [26] [27].

This was not a blog post. It was a careful experimental paper in a good journal from a lab with deep expertise in aquaporins. The field had a real problem.

Five Labs, One Question, Opposite Answers

What happened next is the best thing in this entire story, and almost nobody outside the field has heard about it.

Rather than trading letters, five independent research groups agreed to re-examine the aquaporin-4 question together. The result appeared in eLife in 2018 with twenty-six authors [28].

They tested cerebrospinal fluid influx in wild-type mice against four different aquaporin-4 knockout lines, plus a fifth line lacking a protein that anchors aquaporin-4 to the endfeet specifically. Across all of them, influx was higher in the animals with working aquaporin-4. A meta-analysis pooling the available studies found a significant reduction in the knockouts.

On its own that is just a rebuttal. The genuinely useful part is what they did next.

They ran a meta-regression to find out why different labs had been getting different answers. Three variables explained most of it: the anaesthetic used, the age of the animals, and how the tracer was delivered. They also found that injecting tracer directly into brain tissue suppresses glymphatic function by itself, which means one of the most common experimental shortcuts damages the thing it is trying to measure.

Sit with that for a second. Two competent labs, same species, same question, opposite results, and the explanation was not that one of them was careless. It was that a mouse under one anaesthetic is a different preparation from a mouse under another, and nobody had pinned that down.

This is what science actually looks like when it works. Not a knockout blow. A boring, patient audit of the variables, published jointly by people who disagreed.

The disagreement did not end there. Physiologists at large reviewed the whole question of fluid movement across brain barriers and asked in a title whether there is a glymphatic system at all [29]. Stephen Hladky and Margery Barrand, who have been the most consistent published sceptics for over a decade, laid out a full assessment of the theory against the evidence [30]. Fluid dynamicists asked whether the physics of these narrow spaces even permits the flow rates being claimed [31].

An even-handed summary of the state of the argument, written by researchers on the pro-glymphatic side but genuinely fair to the objections, is worth reading if you want one document rather than twenty [32].

A Different Map Entirely

There is a further objection that rarely reaches general audiences, and it is not a quibble about rates. It says the map is wrong.

The glymphatic model has fluid entering along arteries and leaving along veins. A group working on the vascular basement membranes proposed something different: fluid enters and leaves along separate periarterial routes, moving out through the basement membranes of the artery wall itself, in the opposite direction to blood flow [33] [34]. This is called intramural periarterial drainage, or IPAD.

The two models are not decorative variants of each other. They predict different anatomy, different failure modes in disease, and different targets if anyone ever wants to treat this. A 2025 review of the field still lists the routing question as unresolved [35], and computational modelling of the system is an active field precisely because the geometry is still argued over [36].

When you next read a confident diagram of arrows going in one side and out the other, remember that the arrows are a hypothesis.

2024: The Paper That Said the Opposite

Now the part nobody has told you.

In May 2024, Miao and colleagues published in Nature Neuroscience under a title that is about as blunt as scientific titles get: brain clearance is reduced during sleep and anesthesia [37].

The senior authors were Nicholas Franks and William Wisden at Imperial College London, researchers whose main subject is how anaesthetics and sleep affect the brain. The first author was Andawei Miao. They measured the movement and clearance of fluorescent molecules in the brains of male mice, in the awake state, in natural sleep, and under anaesthesia.

They reported two findings. Movement of the tracer was independent of brain state entirely, the same awake, asleep or anaesthetised. And clearance, meaning how much of the tracer actually left, was markedly reduced during sleep and anaesthesia rather than increased.

That is not a small adjustment to the model. It is the reverse of the headline claim.

An author correction to the paper followed in June 2024 [38]. It is worth being clear about what that means, because corrections get misread in both directions online. An author correction is routine housekeeping where authors fix an error in a figure, a value or an attribution. It is not a retraction. The paper stands.

Should you now conclude that sleep does not clean your brain? No, and this is where care matters. One mouse study does not overturn a decade of mouse studies any more than the reverse. What it establishes is that the question is open, that the measurement is harder than the popular story implies, and that serious researchers using serious methods currently disagree about the direction of the effect.

Go and look at how the top search results for this topic handle that. Most do not mention it at all.

2025: What If Neither Side Was Asking the Right Question

Then, in a paper published in Cell, the picture shifted in a way that makes both sets of results less baffling [39].

Natalie Hauglund and colleagues went looking for what actually drives clearance during non-REM sleep, rather than assuming sleep itself was the driver. What they found was a rhythm.

Norepinephrine, the arousal chemical released by a small brainstem nucleus called the locus coeruleus, does not simply switch off when you fall asleep. It oscillates, slowly, in waves lasting tens of seconds. And those oscillations turned out to be tightly synchronised with two other things: the volume of blood in the brain's vessels, and the cerebrospinal fluid signal. Of everything the group measured, these coupled oscillations were the strongest predictor of glymphatic clearance during non-REM sleep.

Then they tested it rather than just correlating it. Stimulating the locus coeruleus optogenetically produced the expected anti-correlated swing in vasomotion and fluid signal. Driving arterial oscillations directly increased cerebrospinal fluid inflow, which is the observation that makes the mechanism a pump rather than a coincidence.

If your arteries are the pump, then sleep is not the thing doing the cleaning. Sleep is the condition under which your arousal chemistry produces the slow rhythm that runs the pump. That is a meaningfully different claim, and it fits data from both camps: it explains why clearance tracks delta power and heart rate rather than sleep as a category, and it explains why an anaesthetised animal can behave very differently from a naturally sleeping one.

Supporting this, other work has shown vasomotion driving paravascular clearance in the awake mouse brain, which decouples the pump from sleep entirely [40].

And then there is the finding in that Cell paper that should be far better known than it is. The group tested zolpidem, one of the most widely prescribed sleep medications in the world. In mice, zolpidem suppressed the norepinephrine oscillations, and it suppressed glymphatic flow along with them.

Two things about that, and you need both. It is a mouse result, not a human one, and nothing here has been shown to affect any human outcome. And it is not a reason for anyone to stop taking a medication a doctor prescribed. What it is, is a very good reason for researchers to look harder at whether drug-induced sleep does the same physiological work as the real thing. That question is now being asked directly [41], and the broader field published a large multi-author synthesis in 2025 attempting to reconcile the whole mess [42].

The Clock, Not the Sleep

There is one more wrinkle that undercuts the simple story, and it arrived quietly in 2020.

Lauren Hablitz and colleagues found that in mice, glymphatic influx follows a circadian rhythm [43]. It peaked during the animals' natural rest phase. Crucially, the rhythm persisted under constant conditions, which is the standard test for whether something is driven by an internal clock rather than by the environment or by behaviour.

That means at least part of what looks like a sleep effect may be a time-of-day effect. Your body clock is anticipating the rest phase and adjusting the plumbing whether or not you actually sleep.

If your interest is in how that clock shapes when you think best, our piece on circadian rhythms and the best time to study deals with the cognitive side of the same daily cycle. And because arousal chemistry is doing so much of the work here, the article on how stress hormones affect memory is a useful companion, since the same arousal system that gates this pump also shapes what you retain.

What This Has to Do With Alzheimer's

Now for the part everyone actually wants to know about, handled carefully.

The link is amyloid beta. Amyloid is produced by normal neural activity and cleared continuously, and Alzheimer's disease begins, as far as anyone can tell, when clearance falls behind production and the protein starts accumulating. If sleep affects clearance, sleep affects amyloid. That is the chain, and every link in it has been tested.

Human study one. Ehsan Shokri-Kojori and colleagues used PET imaging to measure amyloid burden in twenty healthy adults, once after a normal night of sleep and once after a night of total sleep deprivation [44]. After the sleepless night, amyloid burden rose significantly in the right hippocampus and thalamus. The increase was not explained by genetic Alzheimer's risk. n equals twenty.

Human study two. Yo-El Ju and colleagues took seventeen healthy adults aged thirty-five to sixty-five and did something more surgical: they let people sleep, but selectively disrupted slow-wave activity using acoustic tones, with a sham condition for comparison [45]. Cerebrospinal fluid amyloid beta rose in proportion to how much slow-wave activity was disrupted. The effect was specific to slow-wave activity rather than to sleep duration or efficiency, and specific to amyloid rather than to tau or the other proteins they measured. Separately, worse sleep quality across the preceding several nights was associated with higher tau. n equals seventeen.

Human study three. Per Kristian Eide and colleagues did the most direct thing available: they injected a contrast agent into the cerebrospinal fluid and tracked how fast it cleared from the brain over forty-eight hours, comparing people who had been kept awake for a night against people who slept [46]. Clearance was impaired across most brain regions in the sleep-deprived group. Related work from the same group examined how the tracer escapes to the parasagittal dura and how sleep deprivation affects that step [47].

That third study is the strongest direct human evidence in this entire field. Here is its sample size.

Seven sleep-deprived participants. Seventeen controls.

Twenty-four people total. That is not a criticism of the researchers, who did something invasive and difficult that almost nobody else has attempted. It is context you are entitled to before you read another headline about sleep and dementia.

Tau tells a similar story. The sleep-wake cycle regulates interstitial tau in mice and cerebrospinal fluid tau in humans, with levels rising during wakefulness [48]. In human post-mortem tissue, the degree to which aquaporin-4 stays properly localised to the perivascular endfeet is associated with Alzheimer's pathology and cognitive status [49]. Imaging indices of glymphatic function correlate with amyloid deposition and cognition in people with mild cognitive impairment and Alzheimer's disease [50]. In mouse models of the disease, clearance of tau is impaired [51].

Nedergaard and Steven Goldman went furthest and proposed glymphatic failure as a final common pathway to dementia, a single mechanism that several different risk factors converge on [52]. Read that as what it is: a hypothesis paper from the model's own architects, published as a perspective, not as a settled conclusion.

The honest summary is short. Poor sleep and higher amyloid go together in humans, repeatedly, in small studies. The direction of causation is not established, and there is good reason to think it runs both ways, because Alzheimer's pathology itself wrecks sleep. Nobody has shown that improving your sleep prevents dementia through this pathway. That study has not been done.

Newer work keeps pushing at the mechanism. Multisensory gamma stimulation increased amyloid clearance in mice [53]. A 2026 paper traced amyloid-driven glymphatic dysfunction in model mice to a calcium-mediated process [54]. Others have implicated remodelling of the choroid plexus, the tissue that makes the fluid in the first place [55]. Broader reviews now treat glymphatic and meningeal lymphatic dysfunction as a joint therapeutic target [56] [57], and the sleep-specific version of the argument has its own 2026 review [58].

None of that is treatment. All of it is mechanism, mostly in mice.

Why This Might Matter More As You Get Older

One finding here has held up consistently and is rarely disputed.

Benjamin Kress and colleagues compared young mice, aged two to three months, with middle-aged mice at ten to twelve months and old mice at eighteen to twenty months [59]. Exchange between cerebrospinal fluid and brain tissue declined steeply with age. Alongside it, aquaporin-4 lost its tidy arrangement on the astrocyte endfeet and became scattered across the cell, which the authors linked to the functional decline.

Reviews of ageing and clearance have made this one of the more stable planks in the field [60]. It also connects to something you can observe on human scans, since perivascular spaces visibly enlarge with age and vascular disease.

Our article on how ageing changes memory covers what happens to recall itself over a lifetime, which is a different question from what happens to the plumbing, though the two are probably related.

Can Anyone Actually Measure This In You?

Short answer: not in any way you can go and request.

There are two research methods, and you should know the difference because almost every clinical claim you read rests on one of them.

The direct method is what the Oslo group around Per Kristian Eide and Geir Ringstad does. Inject a contrast agent into the cerebrospinal fluid through a lumbar puncture, then image the head repeatedly over hours to days and watch where the contrast goes and how fast it leaves [61]. This is as close to watching human glymphatic function as anybody has come. It has also been used to study idiopathic normal pressure hydrocephalus, a condition where fluid dynamics are obviously abnormal [62], and to track tracer escaping into the parasagittal dura [63] [64]. The earliest version of this approach was worked out in animals with clinically relevant intrathecal infusion [65], and modern work now feeds the imaging into biophysical models to estimate transport [66].

It also involves a lumbar puncture and an off-label contrast injection. No healthy person is getting this for curiosity.

The indirect method is DTI-ALPS, introduced by Toshiaki Taoka and colleagues in 2017 [67]. It uses diffusion MRI to measure water movement along the direction of perivascular spaces next to particular veins, and produces a single index number. It needs no injection and no lumbar puncture, which is why it exploded in use.

Now the caveat, and you should hold onto it. DTI-ALPS is an indirect proxy, measured in one small region, and its reproducibility has been an active methodological concern since shortly after it was introduced [68]. Improved acquisition methods have been proposed to stabilise it [69], and the technique's own originators published a revisiting paper laying out what it can and cannot support [70].

This matters because a very large share of the clinical literature you will encounter, papers connecting glymphatic function to stroke, Parkinson's, small vessel disease, epilepsy, migraine and more, rests on this one index. When you read that glymphatic function is impaired in some condition, there is a good chance the actual measurement was a diffusion ratio in a small patch of white matter.

The condition-specific literature is genuinely large. Reviews cover the neurological range [71], with dedicated work on traumatic brain injury and tau [72], stroke [73], the role of cerebrospinal fluid influx in acute ischaemic swelling [74], small vessel disease [75] [76], Parkinson's disease [77] [78], and brain tumours [79]. Whether all of it survives the methodological questions above is a fair thing to wonder.

The Things the Internet Says That the Papers Do Not

This topic has a peculiar problem. It is scientifically unsettled and commercially attractive at the same time, which is the exact recipe for confident nonsense.

Here are the claims you will meet, and what the underlying research actually shows.

The side-sleeping claim deserves a closer look because it is the most viral. The study it comes from is real work: Hedok Lee and colleagues used contrast MRI and kinetic modelling to compare glymphatic transport in anaesthetised rodents lying supine, prone or on their side, validated against fluorescence microscopy and radiotracers [80]. Lateral came out most efficient. That is the entire basis.

Anaesthetised rodents. Not humans. Not left versus right. Not a trial of anything.

Somewhere between that paper and your search results, it became a percentage and a mattress recommendation.

Voluntary exercise increased amyloid clearance in mice [81], and rhythmic physical activity has been examined in the same frame [82]. Whether any deliberate intervention can be monitored or enhanced in a person is being asked seriously as of 2026, and the honest answer in that literature is that we are not there yet [83].

What Is Actually Settled, and What Is Not

You have now read a lot of qualification, so let us be concrete about which is which.

Notice how much of the left column is settled. This is not a field built on sand. The anatomy is solid, the fluid movement is solid, and the drainage routes are solid. What is unsettled is the part everybody built a wellness industry on.

What Would Actually Settle It

It is worth asking what evidence would end the argument, because that tells you what to watch for over the next few years.

Three things would do it.

A human clearance measurement that does not require a needle in your spine. Right now the choice is between an invasive method with tiny samples and a non-invasive proxy with questions attached. Something that measures actual solute removal in an ordinary person, repeatedly, would change everything. Work on monitoring approaches is under way [83].

A standardised protocol that removes the anaesthesia problem. The 2018 five-lab paper already identified anaesthetic, animal age and tracer delivery as the variables producing opposite results. Fixing those by convention rather than by argument would let studies be compared directly.

An intervention trial. If glymphatic failure really is a route to dementia, then improving clearance should change an outcome. Nobody has shown that. It is the difference between a mechanism and a treatment, and the field is currently full of the former.

Until those exist, this remains a strong hypothesis with excellent anatomy underneath it and a contested physiology on top. Recent reviews of the mechanisms [84] and of the disease links [85] [86] [87] read very differently depending on which they emphasise, which is itself a signal.

What You Can Take From This Tonight

Not a routine. A way of reading.

The next time you meet a confident claim about your brain cleaning itself, ask four questions. Which species was it measured in. How many participants if it was human. Was it flow or was it clearance. And does the page mention that anybody disagrees.

Those four questions will separate almost every accurate article on this topic from almost every inaccurate one. They would have caught the sixty percent misquote, the side-sleeping percentage, and the silent omission of the 2024 result, all without needing a biology degree.

And the underlying science, stripped of the marketing, is genuinely remarkable. Your brain is threaded with fluid-filled sleeves around every artery. Those arteries do not merely pulse with your heartbeat, they oscillate slowly on their own, and that slow oscillation appears to work as a pump. Something in your arousal chemistry sets the rhythm. Waste protein comes out, eventually, through lymphatic vessels in a membrane that anatomists somehow missed until 2015.

Whether sleep makes that go faster is, right now, an open question in the literature. That is not a disappointing ending. It is what an active field looks like from the inside, and you are reading about it while it is still being decided.

If you want the neighbouring pieces, our articles on what the brain is actually doing across the stages of sleep, how neurons communicate across the spaces this fluid fills, and how sleep consolidates what you studied approach the same night from three other directions.

Sleep well. The reasons are still being counted.