#654: Dr. Matthew Walker, All Things Sleep Continued — The Hidden Dangers of Melatonin, Tools for Insomnia, Enhancing Learning and Sleep Spindles, The Upsides of Sleep Divorce, How Sleep Impacts Sex (and Vice Versa), Adventures in Lucid Dreaming, The One Clock to Rule Them All, The IP Addresses of Your Memories, and More
The myth-busts, hot takes, explainers, and tools worth keeping.
✶Explainer· 7
✶Explainer0:30
Why one in four couples already sleep apart, and how to propose it without starting a real divorce
Walker opens on the sleep divorce, defined as sleeping in separate locations or at least separate beds, with the diluted Scandinavian method being two beds pushed side by side. Survey data from the US and the UK Sleep Council puts about one in four couples at having had one, rising to almost one in three when asked anonymously, which he attributes to stigma. Objectively, on almost every measure they can quantify, sleep is worse when couples sleep together, though some people report higher subjective satisfaction sharing a bed regardless. He then gives the graded protocol: an open non-defensive conversation, a bounded one or two week trial rather than a permanent arrangement, and a deliberate rebuild of the bookends of sleep so the intimacy survives.
A sleep divorce is separate beds or rooms; the Scandinavian method is the diluted side-by-side version.
One in four couples admit to it openly; almost one in three do anonymously.
Objectively sleep is worse sharing a bed, but subjective satisfaction can run the other way.
The taboo rests on conflating sleeping together with having sex.
Propose it as a one or two week trial, never as permanent.
Rebuild the bookends: the goodnight cuddle and the morning greeting are what people actually miss.
“about one out of every four couples who are surveyed will tell you that they have had a sleep divorce that they will sleep in…”
“what you really miss are the bookends of sleep that sort of getting into bed having a cuddle saying a good night and in the…”
Four or five hours of sleep ages a man's testosterone by a decade, and costs couples their empathy
Walker breaks the sleep-to-sex direction into three mechanisms: hormones, sensitivity, and libido. Men restricted to four or five hours of sleep drop testosterone by roughly the equivalent of ten years of aging, and the same holds for estrogen in women. Lower estrogen reduces vaginal lubrication and therefore sensitivity and pleasure. On libido, an extra hour of sleep produces a 14 percent increase in a woman's reported desire for intimacy, which he benchmarks against FDA-approved libido drugs delivering about 24 percent. He also cites UC Berkeley work showing sleep-deprived couples fight more and resolve conflict worse, because the capacity to empathise degrades and they can now show why in terms of brain networks.
Four to five hours of sleep drops male testosterone by about a decade of aging.
Estrogen falls similarly in women, reducing lubrication, sensitivity, and pleasure.
One extra hour of sleep raises reported desire for intimacy by 14 percent.
FDA-approved libido drugs deliver around 24 percent, so sleep gets over half the benefit drug-free.
Libido scales are self-reported and fungible, much like pain scales.
Poorly slept couples fight more and resolve conflict worse because empathy capacity drops.
“males who if we put them on a diet of maybe four or five hours of sleep they drop their levels of testosterone somewhere by…”
“we've also found that when a woman obtains an extra 1 hour of sleep there is a 14% increase in her desirability to be intimate”
Sex with orgasm buys a 70 percent sleep quality bump, and masturbation gets you 47
Running the other direction, Walker reports that sex associated with orgasm is linked to about a 70 percent increase in reported sleep quality that following night, with faster sleep onset. Masturbation, again conditional on orgasm, produces about a 47 percent increase and roughly halves time to fall asleep. The proposed mechanisms are hormonal and autonomic: oxytocin, a sleep-inducing molecule, rises more in women, while vasopressin appears to do the work in men, acting at the brain stem to shut down wake-promoting regions and at the hypothalamic sleep-wake switch. The autonomic half is that you cannot fall asleep with the sympathetic nervous system switched on, producing the wired-but-tired phenomenon, and orgasm flips you into the parasympathetic state.
Sex with orgasm is associated with about a 70 percent increase in reported sleep quality that night.
Masturbation with orgasm delivers about 47 percent and halves sleep onset time.
Oxytocin is a sleep-inducing molecule and rises more in women; vasopressin appears to do it in men.
Vasopressin may act at the brain stem and at the hypothalamic sleep-wake flip-flop switch.
You cannot fall asleep with the sympathetic branch switched on, which is the wired-but-tired state.
Orgasm removes that block and pushes you into the parasympathetic, the royal road to better sleep.
“sex as long as it accomplishes orgasm is associated with about 70% increase in reported Sleep Quality that following night”
“you cannot fall asleep when your sympathetic nervous system is switched on”
Melatonin times the race but never runs it, which is why it buys you 3.9 minutes
Walker attacks the multi-billion dollar melatonin market at the level of mechanism. Melatonin is a hormone, sometimes called the hormone of darkness, that signals nighttime to the brain and regulates the timing of sleep without participating in the generation of sleep itself. His analogy is the starting official at a 100 metre Olympic race, who brings the racers to the line and begins the race but does not run in it. That predicts the meta-analytic result: melatonin increases sleep onset speed by only 3.9 minutes and sleep efficiency by only 2.2 percent, close to placebo. It earns its place when moving between time zones, not once you are stable in one. Ferriss concedes a walloping dose helps him on long-haul travel, which is the only time he uses it.
Melatonin regulates sleep timing but does not generate sleep.
It is the starting official at the race, not one of the racers.
Five reasons dreaming would get you sectioned if it happened while awake
Walker frames REM dreaming as a nightly psychosis with five diagnostic features: you hallucinate, you are delusional, you are disoriented in time, place, and person, you are affectively labile, and you wake amnesic. Any one of those while awake would prompt psychiatric intervention. Neurologically, scanner work shows vast swaths of visual cortex lighting up along with motor, emotional, and memory regions, with some parts up to 30 percent more active in REM than while awake. The freakish part is the dorsolateral prefrontal cortex, normally the CEO making rational and volitional decisions, which goes offline and is suppressed. That explains why dreams are vivid, motoric, emotional, bizarre, and beyond volitional control. He also covers REM atonia, the brain-stem-driven paralysis of voluntary skeletal muscles.
Dreaming meets five criteria that would be pathological while awake: hallucination, delusion, disorientation, emotional lability, amnesia.
Visual, motor, emotional, and memory regions all activate in REM.
Some brain regions are up to 30 percent more active in REM than in waking.
The dorsolateral prefrontal cortex, the brain's CEO, is suppressed during REM.
That suppression explains the bizarreness and the absence of volitional control.
The brain stem paralyses voluntary skeletal muscles, sparing respiration and heart.
REM sleep paralysis on waking is the same mechanism, and adequately explains most alien abductions.
“when we dream we start to see things which are not there so you're hallucinating you also believe things that couldn't possibly be true so…”
“the prison guards have left the building and the prisoners are are running a mock your prefrontal cortex is down”
Pull an all-nighter and your memory inbox stops accepting mail
Walker covers the sleep-before-learning half of the memory story. Depriving people of sleep leaves them 20 to 40 percent deficient in learning ability the next day. Scanner work shows why: the hippocampus, which sits on both sides of the brain and acts as the informational inbox for new memories, becomes waterlogged and shuts down without sleep, so incoming information bounces. He connects this to what is happening in education populations. He also names the deeper reason dreaming matters for learning, describing the dreaming brain as running a Google search gone wrong that takes new information and drops you at page 20, forging distant non-obvious connections that produce solutions to previously impenetrable problems. That, he argues, is the biological basis of creativity, and why nobody tells you to stay awake on a problem.
Sleep deprivation leaves you 20 to 40 percent deficient in the ability to learn new information.
The hippocampus is the inbox for new memories and becomes waterlogged without sleep.
A full inbox bounces incoming mail, so the experience never commits to memory.
Deep non-REM cements individual memories; REM interconnects them with prior knowledge.
The dreaming brain works like a Google search gone wrong, landing you at page 20.
Non-obvious distant connections are the biological basis of creativity.
You have never been told to stay awake on a problem, for good mechanistic reasons.
“you're about 20 to 40% deficient in your learning ability which I when you have not been getting sufficient sleep”
“it seems to do a Google search gone wrong where it takes the new information it inserts it and it immediately puts you to page…”
Cherry-picking which memories your brain replays overnight using rose perfume and teakettle sounds
Walker traces memory replay from Bruce McNaughton's rat maze recordings at Arizona, where hippocampal cells replayed the maze during deep non-REM at roughly ten times waking speed, etching the memory trace. Researchers then hijacked it. In one design, subjects learned word lists while rose perfume was infused through a mask, bonding the odour to the material, and the odour was re-perfused during deep non-REM, producing a 40 to 50 percent improvement in memory. Ken Paller at Northwestern refined it with congruent sounds on a memory card game, replayed below waking threshold during sleep, allowing selective cherry-picking of which memories get VIP access to overnight replay. Ferriss proposes a cheap green tea incense version; Walker notes it is also a fire hazard.
McNaughton recorded hippocampal cells in rats running a maze and kept recording into sleep.
Memories replay during deep non-REM at about ten times waking speed.
In REM, Matt Wilson at MIT found replay runs at about 0.5 times waking speed.
Rose perfume bonded to learned word lists and re-perfused in deep non-REM raised memory 40 to 50 percent.
Ken Paller used congruent sounds on a card-matching game, replayed below waking threshold.
The sound must be sub-awakening threshold, loud enough to reach the brain but not wake the sleeper.
This gives selective control over which memories get replayed, which Walker finds slightly scary.
“Not only was the memory replayed but it wasn't replayed at waking speed it was replayed almost 10 times faster”
“they started to reperfuse that Rose odor up their nose and as a consequence the next day there was about a 40 to 50% Improvement…”
How researchers proved lucid dreaming was real using eye movements and ejaculation
Walker recounts how lucid dreaming went from charlatan science in the 60s and 70s to settled. The obstacle was that a lucid dreamer is paralysed and cannot report anything, except through the extraocular muscles, which are spared from REM atonia and are why rapid eye movements exist at all. Researchers built an ocular morse code: three deliberate flicks left signals lucidity, circular movements right signal clenching the right hand. The definitive early experiment used orgasm and ejaculation as the test, with subjects signalling the onset and completion of an in-dream orgasm and the experimenters finding proof positive in the room afterwards. Mechanistically, the dorsolateral prefrontal regions appear to ramp back up when lucidity begins, though a recent study argues that signal may be an artifact of eye-movement electrical bleed.
Lucid dreaming was dismissed as charlatan science through the 60s and 70s.
Extraocular muscles escape REM paralysis, which is why rapid eye movements exist.
Researchers built a predefined ocular morse code to communicate from inside a dream.
The definitive proof used signalled in-dream orgasm and ejaculation.
The scientific definition of lucidity is simply being aware you are dreaming while dreaming.
Prefrontal activity appears to ramp back up during lucidity, restoring volition.
A recent study suggests that signal may be an artifact of electrical bleed from eye movements.
“we can now generate essentially a language between me the experimenter and you the Lucid dreamer which is like a an ocular MOS code”
“when you control lucid dreaming by the way the definition is just that you are aware that you're dreaming as you are dreaming”
Drinking on night three after learning still costs you 40 percent of the memory
Walker confirms Ferriss's recollection of two studies. In the first, subjects deprived of sleep the first night after learning, then given full recovery sleep on nights two and three, still showed impaired recall on day five, making sleep for memory an all-or-nothing phenomenon inside the first 24 hours. In the second, by memory researcher Kim Smith, subjects learned an abstract analogical language sober on day one and were tested sober on day seven. Alcohol on night one alone caused a greater than 50 percent memory loss, via acetaldehyde demolishing REM sleep. Alcohol on night three alone, after two clean nights, still caused about a 40 percent deficit. Walker draws the scheduling implication, including having therapy on Monday rather than Thursday if you drink at weekends.
Miss sleep in the first 24 hours after learning and recovery sleep will not salvage it.
Sleep is not a bank; you cannot accumulate the debt and repay it later.
Kim Smith's study used an abstract analogical language with sober learning and sober testing.
Alcohol on night one caused over 50 percent memory loss by day seven.
Alcohol on night three still caused about a 40 percent deficit.
Acetaldehyde is the byproduct that demolishes REM sleep.
Practical rule: have therapy on Monday and not on Thursday if you drink at weekends.
“if you don't sleep within the first 24 hours after learning you lose the chance to consolidate those memories so sleep for memory is an…”
“have your therapy on Monday and not on Thursday if you're going to drink on the weekends”
Exercise buys you deep sleep, but it takes REM sleep as payment
Ferriss asks whether intense pro-inflammatory exercise could simulate the fever-driven deep sleep that produced his outlier exam result. Walker says yes, and that he had just released the first of a three-part series on sleep and exercise. Exercise principally increases deep non-REM sleep and the power of its slow brain waves, driven partly by low-grade inflammation, partly by temperature, and partly by the parasympathetic downshoot after exertion where heart rate and blood pressure dip below baseline. But there is no free lunch: intensive exercise reliably drops REM sleep. Head to head, aerobic work seems to beat resistance training somewhat on deep sleep and sleep consistency, though weights still beat being sedentary. On HIIT versus steady state, Walker says there is not enough data and no randomised trial has been run.
Exercise principally increases deep non-REM sleep and the power of its slow waves.
There is no free lunch: intensive exercise reliably drops REM sleep.
Mechanisms include low-grade inflammation, temperature, and the post-exercise parasympathetic downshoot.
Exercise also boosts sleep continuity, so you wake fewer times in the night.
Aerobic work appears to somewhat outperform resistance training on deep sleep and consistency.
Resistance training still beats being sedentary, which is the worst case.
No randomised controlled trial has put HIIT and steady state head to head on sleep.
“yes you increase the amount of deep sleep but it does seem to come at the cost of REM sleep”
“aerobic seems to potentially outdo if you put them head-to-head in a sort of Coke Pepsi challenge”
The hidden dangers: 478 percent label errors, a 530 percent overdose rise, and shrinking rat testicles
The episode's title topic. Walker classifies 5 or 10 milligram doses as supraphysiological, maybe ten times natural release, against a studied optimum of roughly 0.3 to 0.5 milligram. A study of 30 brands found capsule contents ranging from about 83 percent below to 478 percent above the label, because melatonin is not FDA approved in the US. Reported overdoses to the US Poison Control Center have risen 530 percent in ten years, concentrated in paediatric populations. He then dismantles the reassurance study people cite, a single blind patient dosed for 37 days, whose free-running circadian system makes them a poor control, and notes juvenile male rats show testicular atrophy while high doses block hippocampal synapse formation.
Studied optimal dose is roughly 0.3 to 0.5 milligram; shelf products start at 3 and run to 10.
A 30-brand audit found contents from 83 percent below to 478 percent above the label.
Poison Control reports of melatonin overdose are up 530 percent over ten years, led by paediatric cases.
The reassuring no-feedback-shutdown study is one blind patient dosed for 37 days.
People actually use melatonin for 6 months, 12 months, or 5 years, which nobody has studied.
Juvenile male rats show testicular atrophy; high doses also block hippocampal synapse formation.
Walker's one pro-sleep mechanism candidate is thermoregulatory: melatonin lowers core body temperature.
“there was a report demonstrating that melatonin overdoses that have been reported to the US Poison Control Center they've risen by 530 %c over the…”
“we also know that high doses of melatonin will block the ability of your hippocampus to form new copses”
The paradoxical insomnia treatment: spend less time in bed and sleep more
Walker lays out cognitive behavioural therapy for insomnia as the correct first line, not sleeping pills, and singles out sleep restriction therapy as one of the most potent tools in the box. He argues the name is wrong and it should be called bedtime restriction therapy. The worked case is a patient in bed 8 hours sleeping 4, a sleep efficiency of 50 percent against a healthy target of 85 or greater. Rather than extending time in bed he pushes bedtime from midnight to 2 or 3am while holding the 8am wake time, exploiting accumulating adenosine. Around day four or five the patient sleeps front to back for the first time, then bedtime is walked back in 30-minute steps. He compares it to hitting reset on a Wi-Fi router, and agrees with Ferriss that resetting sleep-related anxiety is central.
CBTI, not sleeping pills, is the current first-line recommendation for insomnia.
Sleep restriction therapy is misnamed; it restricts time in bed, not sleep.
Target sleep efficiency is about 85 percent or greater; the case patient starts at 50.
Restrict at the bedtime end because people find that easier than waking earlier.
Adenosine accumulates across consecutive days of extended wakefulness and forces consolidation.
After four or five days the patient sleeps front to back, then bedtime is walked back gradually.
The durable win is confidence: you control your sleep instead of your sleep controlling you.
“it's actually a poorly termed phrase it should really be called bedtime restriction therapy”
“essentially it's like hitting the reset button on your Wi-Fi router you are retraining the Sleep System”
The two validated induction methods, plus Ferriss's dream journal protocol and wrestling drills
Walker names the two methods with real validity: MILD, the mnemonic induction of lucid dreaming, which is a deliberate pre-bed intention repeated night after night, and the reality testing method popularised by Stephen LaBerge and dramatised in Richard Linklater's Waking Life. Reality testing means checking waking physics until it becomes reflexive, so the habit bleeds into dreams. Ferriss adds the missing selection criterion, that a test you cannot convincingly perform awake will not transfer, ruling out attempting to fly, and offers tile patterns and digital clocks instead. He details step one of LaBerge's protocol, a dream journal written immediately on waking before phone or teeth, in fragments rather than voice memo, mined for recurring motifs to use as triggers. He also describes drilling wrestling in lucid dreams with a coach he had never met.
MILD is a repeated, deliberate pre-bed intention to remember you are dreaming.
Reality testing drills a waking physics check until it becomes reflexive and bleeds into dreams.
Statistical robustness of both methods is modest, though they do reach significance.
A test you cannot convincingly perform awake, such as flying, will not transfer to dreams.
Tile patterns reorient and digital clock readouts change in dreams; both are good tests.
Journal immediately on waking, in fragments, on paper, not by voice memo.
Only 10 to 20 percent of the population are natural lucid dreamers; Walker has been lucid two or three times.
“you create a conscious and very deliberate intention to remember that you're dreaming and usually this is pre-bed”
“immediately upon waking you can't check your phone you can't brush your teeth you need to just roll over and immediately do this”
Nap 15 to 20 minutes, never 30 to 40, and why the lab uses 90
Walker's nap study had subjects learn at midday and again at 6pm, with half napping in between. Nappers showed restored learning ability, non-nappers showed deterioration, and the size of the restoration tracked the number of sleep spindles, a bursting electrical brain wave in non-REM sleep. The lab naps were 90 minutes, matching the human non-REM to REM cycle, so both stages get an equal opportunity to demonstrate a benefit, but Walker says he would not recommend that duration for most people. His practical guidance is 15 to 20 minutes and no longer, because past that you descend into deeper non-REM and waking at 30 or 40 minutes produces sleep inertia, a sleep hangover that leaves you worse than before. Below about 7 to 9 minutes there are no measurable brain or body benefits.
Nappers restored learning ability; non-nappers deteriorated across the same 6-hour window.
More sleep spindles during the nap meant greater recuperation of learning ability.
The human non-REM to REM cycle is about 90 minutes, which is why lab naps are that long.
Walker would not recommend a 90-minute nap for most people.
Optimal practical nap is 15 to 20 minutes and no longer.
Waking at 30 or 40 minutes surfaces you out of deep sleep into sleep inertia.
Below about 9, 8, or 7 minutes there are no measurable brain or body benefits.
“the short and long of it is nap for about 15 to 20 minutes and no longer”
“you have what's called Sleep inera which feels like a sleep hangover and it's even worse when you wake up than it was before you…”