✕Sleep gives the brain a bath / deep sleep is the only time the brain gets cleaned
Xie 2013 is a mouse study. Miao 2024 (male mice, a different method) found clearance was actually lower during sleep (about ¼), and about ½ under anesthesia; Hauglund 2025 challenged that method. No human study has measured faster clearance during sleep.
More accurateDuring sleep, fluid really does flow through the brain on a large scale; whether sleep is a "brain bath" is something scientists are still arguing about.
✕Xie 2013: "brain cells shrink by 60%", "clearance is 10 times faster during sleep"
What grows by about 60% is the space between cells (14% → 23%); injected Aβ is cleared about 2 times faster; tracer inflow is about 95% lower during wakefulness.
More accurateWhen mice fall asleep, the space between their brain cells grows by about 60 percent.
✕Sleep clears Alzheimer's proteins "twice as fast" (stated as a fact about humans)
It was injected Aβ, in mice, measured at different times of day.
More accurateAlways make clear that it was a mouse study.
✕With every slow wave, a surge of cerebrospinal fluid rushes into the brain
Fultz 2019 (n=13): CSF waves came roughly every 20 s, a few seconds after slow-wave power rose; the measurement site was the fourth ventricle; clearance was not measured.
More accurateAbout every twenty seconds the brain's slow waves grow stronger, and a few seconds later a big surge of cerebrospinal fluid flows in.
✕Deep sleep (delta waves) drives brain cleaning
Hauglund 2025: delta power does not predict clearance (norepinephrine oscillations do); Miao 2024: the more delta, the less clearance.
More accurateWhile you sleep, your blood vessels widen and narrow to the rhythm of norepinephrine.
✕One all-nighter will give you dementia / one sleepless night grows plaques in your brain
Aβ PET +5% (n=20; may reflect changes in blood flow, and cannot distinguish soluble Aβ from plaques); CSF tau +51.5% (n=6); both are short-term levels, and causation is unproven; Lucey 2019 is a cross-sectional study.
More accurateAfter one sleepless night these proteins rise for a short while, but that doesn't mean you're more likely to get dementia.
✕Taking sleeping pills stops the brain from cleaning itself and raises dementia risk
Only a single dose of zolpidem in mice; the effect on clearance was borderline (one-tailed P=0.065).
✕Sleeping on it makes painful emotions fade ("overnight therapy" treated as settled)
Lipinska 2022 (24 studies): pooled physiological reactivity did not change; self-rated arousal was actually higher after sleep. Baran 2012: reactivity was preserved after sleep.
More accurateSleep changes how we remember emotional events, but the evidence that sleeping on it makes the pain fade is still mixed.
✕You only dream in REM sleep / REM = dreaming
Siclari 2017: people also report dreams from non-REM sleep; dream experience corresponds to a posterior "hot zone", not to a sleep stage.
More accurateThe most vivid dreams come in REM sleep, but we also dream often in non-REM sleep.
✕Most dreams happen in REM sleep
Siclari gives no such proportion; its argument is the opposite. Another source is needed.
More accurateThe most vivid, story-like dreams tend to come in REM sleep.
✕Sleep paralysis passes within tens of seconds to a few minutes and is completely harmless (stated as a sourced fact)
Sharpless 2011 gives only prevalence (7.6% / 28.3%); it says nothing about duration and does not call it harmless; fear is common.
More accurateCite a clinical source before giving a duration; until then, say: it's very common and usually passes quickly.
✕Near dawn you're dreaming almost the whole time
Cajochen 2024: in the 97-minute 4th cycle, REM took up about 33 minutes (roughly a third).
More accurateThe later in the night it gets, the more of your sleep is spent dreaming.
✕A night usually has four to five cycles
Cajochen 2024 (8 h lab nights): about 53% of nights completed 4 full cycles, about 26% had 3, and about 17% had 5.
More accurateA night usually has three to five cycles, most often four.
✕As long as you sleep well in the first cycle (the golden 90 minutes), sleeping less doesn't matter
The first cycle is the shortest (about 76 minutes); cycles 2–4 contain most of the REM; N2 also helps memory.
More accurateProtect the first half of the night, but don't cut the second half either.
✕Newborns spend about 90% of their sleep in REM
Newborns: about 50% (active sleep).
✕Norepinephrine is released during REM sleep and strengthens REM
Norepinephrine is at its lowest level during REM (this is the basis of the "overnight therapy" model).
More accurateIn REM sleep, norepinephrine drops almost to its lowest point.
✕The whole cortex rises and falls together, with tens of billions of neurons firing at once
Slow waves are often local; only about 15–20% are global (literature cited by Staresina); about 16 billion is the total number of cortical neurons, not the number firing at the same time.
More accurateNeurons across large stretches of the cortex fire together and fall silent together.
✕The more precisely the three rhythms lock together, the better you remember the next day
Helfrich 2018 measured only the timing between slow waves and spindles (no ripples) and is a correlational study; Staresina 2015 did not measure memory.
More accuratePeople whose slow waves and spindles line up more precisely remember better the next day.
✕Playing recordings while you sleep lets you learn new things / sleep-learning devices
Targeted memory reactivation (TMR) can only strengthen, in lab tasks, what was already learned while awake; g≈0.29, about 0.13–0.18 after bias correction; no effect during REM or wakefulness.
More accurateSounds played during sleep can only strengthen what you learned during the day, and the effect is small.
✕TMR: "you remember better the next day" (Rudoy 2009)
Rudoy 2009 used an afternoon nap and tested shortly after waking.
More accurateYou remember better after you wake up.
✕Pulling an all-nighter cuts learning ability by 40% / makes the amygdala 60% more reactive (as firm numbers)
The 40% cannot be found in Krause 2017; the 60% is just one contrast from a single small fMRI study.
✕Sleeping well the night after learning helps more than staying up to review once more (as a proven comparison)
None of the papers we read directly compares sleeping with studying a bit longer.
More accurateAdd a direct source, or say: staying up to review costs you both consolidation and the next day's ability to learn.
✕Sleep deletes "unimportant memories" (synaptic homeostasis presented as settled)
SHY is a hypothesis; the pattern is that "less-used connections are weakened more"; de Vivo's 18% refers to axon-spine contact area in mice.
More accurateOne leading hypothesis is that sleep weakens the connections we use less.
✕Drinking makes you fall asleep faster (at normal amounts)
Gardiner 2025: people fall asleep faster only from about 5 standard drinks; at typical drinking levels there is no reliable change.
More accurateAlcohol makes you feel sleepier, but normal amounts won't make you fall asleep noticeably faster.
✕Even if you get enough hours after drinking, you're still tired the next day (citing Gardiner 2025 as evidence)
Gardiner 2025 did not measure any next-day outcomes; the change in total sleep time was not reliable either.
More accurateWhen you drink, the structure of your sleep changes.
✕Cutting the last two hours of sleep costs you X% of your REM (a specific number)
This is only an estimate from the structure of sleep cycles; in our worked example, sleeping 25% less time loses about 40-45%.
More accurateBased on how sleep is structured, the share of dreaming time you lose is much larger.
✕People spend about six years of their lives dreaming
We could not find a research source. It is extrapolated from REM sleep time: about 1.5-2 hours a night × a lifetime ≈ 5-6 years. But REM ≠ dreaming (15-20% of REM awakenings report no experience, while 40-60% of NREM awakenings report content). It is also often confused with a 2015 UK commercial poll claiming people spend "about six years of their lives daydreaming".
More accurateOver a lifetime, we spend about five or six years in REM sleep, the stage where vivid dreams are most likely (an estimate).
✕Dreams are made up in the instant you wake up
The estimated length of a dream correlates with how long REM lasted before waking; in patients who act out their dreams, their movements match the dream content.
More accurateDreams are real experiences that happen during sleep, not stories made up on waking.
✕Dreams replay the day's events / dreaming is just replaying memories
About two-thirds of dreams contain fragments of the day, but only 0.5-2% replay events as they happened (Fosse 2003, Malinowski 2014); even the people who "dreamed of the maze" did not dream of the route itself.
More accurateDreams borrow people and events from the day, but almost never replay them as they happened.
✕Dreams are just the brain's haphazard interpretation of random noise and mean nothing at all
The original hypothesis has been revised by its own author; people also dream in NREM, damage to the forebrain alone can abolish dreaming, and dream content is selective and biased. Whether dreams have a function is still unsettled, but "completely random" does not hold up.
More accurateOne school of thought holds that dreams are the brain weaving scattered signals into a story, but dream content is not completely random.
✕Dreaming about what you learned makes you learn it better / dreaming makes you smarter
In the original study only 4 people dreamed of the maze, and it was an observational comparison; a 2019 replication compared 8 vs 9 people and the result was not significant after controlling for baseline, and other labs have not replicated it; the authors themselves say dreams do not cause consolidation.
More accuratePeople who dream about what they just learned improve more, but that may simply be because the brain is processing it; we can't say dreams make you smarter.
✕Not remembering your dreams means you didn't dream / slept badly
Everyone dreams; the difference is whether you remember. People who recall more dreams mainly spend longer awake during the night; they show no differences in memory, imagination or personality. Walker also says how many dreams you recall has no clear relationship with your amount of REM or how you feel the next day.
More accurateNot remembering your dreams is normal; people who recall lots of dreams often just spend a bit longer awake at night.
✕Science has proven what dreams are for (emotion / memory / threat rehearsal...)
The conclusion of a 2025 review of the field: many hypotheses have been proposed, with no settled answer; each camp has one line of evidence and one big gap. [dream-15 to dream-22]
More accurateScientists still don't know what dreams are for; there are several hypotheses, each with some evidence.
✕Most dreams are nightmares / 80% of dream emotion is negative
That figure is the "emotions mentioned" in home dream reports from US college students in 1966; one study found the overall mood is more often positive, 25-30% of REM dreams have no emotion at all, and negative dreams are also more likely to be remembered.
More accurateThe emotions people mention in dreams lean somewhat negative, but many dreams have no emotion at all.
✕Adrenaline / noradrenaline in the brain drops to zero while dreaming (presented as a measurement in humans)
Noradrenaline and serotonin neurons almost stop firing during REM; this comes from single-cell recordings in cats and rats. It has not been measured directly in the human brain, and "adrenaline" is a different hormone.
More accurateIn animals, the neurons that release noradrenaline go almost completely quiet during dreaming.
