Each cycle lasts about 90 minutes, but the mix shifts across the night. Wake two hours early and you lose a disproportionate share of REM; go to bed late and you lose deep sleep.
The stages
The diagram’s rows, top to bottom. Shares are rough figures for a healthy young adult; deep sleep shrinks and light sleep grows with age.
Awake
Brief awakenings between cycles are normal; most last seconds and are never remembered.
linked to fragmentation when there are many of them — alcohol, sleep apnoea, noise, a warm room, a full bladder. Fragmented sleep leaves you less rested even if total hours look fine.
Light — N1
The doorway into sleep: a few minutes of drifting, easy to wake from. Hypnic jerks — the falling sensation — happen here.
linked to little on its own. A lot of N1 is a sign of broken-up sleep rather than a benefit.
Light+ — N2
True sleep. The brain produces sleep spindles (short bursts of fast activity) and K-complexes, which help keep you asleep through noise.
linked to memory consolidation, especially motor skills — spindle density tracks overnight improvement on learned tasks. Spindle-rich N2 is concentrated in the last hours, another cost of cutting the night short.
■ Deep — N3
Slow-wave sleep: large, synchronised brainwaves, the hardest stage to wake from. Grogginess when woken from it is “sleep inertia”.
linked to moving facts and events from the hippocampus into long-term storage; the night’s main release of growth hormone; physical recovery and immune function. Poor deep sleep is associated with beta-amyloid and tau build-up (see Alzheimer’s below). The idea that deep sleep “washes” waste from the brain comes mostly from mouse studies and is still debated. Caffeine reduces it; it declines steeply with age.
■ REM
Rapid eye movement sleep: vivid dreaming, brain activity close to waking, and the body’s muscles temporarily paralysed so you don’t act dreams out.
linked to emotional processing — taking the sting out of difficult memories — plus creative, associative thinking and reading other people’s emotions accurately. Suppressed by alcohol, cannabis and several antidepressants, and lost first when you wake early.
The four pillars
Walker’s framing in the Huberman Lab series: good sleep is not one number, it is four things at once.
Your day, hour by hour
Set your wake and bed times; the cut-offs move with them.
The 12 rules
The appendix of Why We Sleep, adapted from the US National Institutes of Health list. Tick what you already do — it’s saved in this browser only.
When you can’t sleep
Lying awake
If you’ve been awake for about 20 minutes, get up, go to another dimly lit room, and read or do something calm until you feel sleepy. The brain learns associations; lying in bed awake teaches it that bed is a place for being awake. Don’t check the clock.
A racing mind
Walker suggests a “mental walk”: picture a familiar route in vivid detail — every door, turn and object. It occupies the mind enough to stop rumination. Writing a to-do list or worry list earlier in the evening also helps offload. Meditation and non-sleep deep rest are reasonable wind-down tools.
The day after a bad night
Don’t compensate. No lie-in, no nap, no extra caffeine, no early bedtime. Each of those drains sleep pressure and makes the next night worse too. Keep your normal schedule and let the built-up pressure carry you into a better night.
Chronic insomnia
Cognitive behavioural therapy for insomnia (CBT-I) is the first-line treatment, ahead of pills. It typically combines sleep restriction, stimulus control and work on sleep anxiety, and its benefits persist after treatment ends. In Denmark, ask your GP for a referral.
Wearables
Walker uses and promotes trackers, but warns against obsessing over the scores (“orthosomnia”). Use them for trends, not nightly verdicts — consumer devices are decent at total sleep and poor at staging.
Substances
Caffeine
It blocks adenosine, the sleep-pressure molecule. Half-life is roughly 5–6 hours, quarter-life 10–12, so an afternoon coffee is still partly active at bedtime. Even if you fall asleep fine, it reduces deep sleep. Walker’s rule of thumb: none after midday, and decaf still contains some.
Alcohol
A sedative, not a sleep aid: sedation is not natural sleep. It fragments the night with brief wake-ups you don’t remember and suppresses REM. Walker’s honest answer to “how much is fine”: none is best; if you drink, earlier and with food is less bad.
Cannabis
THC can speed sleep onset but suppresses REM, and stopping brings rebound insomnia and intense dreams. Evidence on CBD for sleep is thin.
Melatonin
A timing signal (“it’s night”), not a sleep generator. Useful for jet lag and possibly in older adults whose own production has dropped; little help for most insomnia. Over-the-counter products in some markets vary widely in actual dose.
Sleeping pills
Older sedative hypnotics produce sedation rather than natural sleep architecture and are linked to next-day impairment. Walker is strongly critical of long-term use; that link to mortality is associational, not proven causal. Never stop prescribed medication without your doctor.
Why it matters
Graded by how well the evidence holds up: [STRONG] replicated and broadly accepted · [MODERATE] real but smaller, fewer or narrower studies · [CONTESTED] Walker states it more confidently than the data allow.
Memory consolidation [STRONG]
Deep NREM (slow waves and sleep spindles) moves new memories from the hippocampus into long-term cortical storage. Sleep before learning also restores capacity to take in new information.
Emotional regulation [STRONG]
After a night without sleep, the amygdala reacted about 60% more strongly to negative images in Walker’s 2007 study. REM sleep appears to strip the emotional charge from difficult memories — his “overnight therapy” idea.
Attention and driving [STRONG]
Chronic short sleep degrades attention, and people consistently underestimate how impaired they are. Drowsy driving is a major cause of crashes.
Blood sugar and appetite [MODERATE]
Short sleep reduces insulin sensitivity and shifts hunger hormones toward eating more. Walker’s “prediabetic after a week” line is based on small lab studies.
Heart health [MODERATE]
Short sleep is associated with higher blood pressure and cardiovascular risk. Heart attacks rise the Monday after the spring clock change, a natural experiment he often cites.
Testosterone [MODERATE]
A week of 5-hour nights lowered testosterone in young men by an amount Walker equates to roughly ten years of ageing. Real effect, but from a small study.
Social behaviour [MODERATE]
His lab found sleep loss makes people lonelier, more socially withdrawn and less willing to help others.
Alzheimer’s risk [MODERATE]
Poor deep sleep is linked to beta-amyloid and tau build-up, and the relationship seems to run both ways. Promising and mechanistically plausible; not yet proof that better sleep prevents dementia.
Immune collapse and cancer [CONTESTED]
The famous “70% drop in natural killer cells after one 4-hour night” comes from older, small studies; the cancer framing goes well beyond what they show.
“Shorter sleep, shorter life” [CONTESTED]
The mortality data are U-shaped and observational — long sleepers also die earlier, likely because illness causes long sleep. Causation is not established.
Life stages
Teenagers
Their circadian rhythm shifts about 1–3 hours later. Early school start times force them to sleep against their biology — not laziness.
Adults
Need doesn’t shrink with age through adulthood. Night owls and larks are genetically real; regularity matters more than forcing an early schedule.
Older adults
The need for sleep doesn’t decline — the ability to generate deep sleep does. Treat sleep complaints in older people seriously rather than as normal ageing.
Newer research
Walker left Berkeley at the end of 2025 and now directs the Sleep Innovation Laboratories at UT Dallas’s Center for BrainHealth.
In September 2026 his group published work in Nature Neuroscience showing that in healthy older adults, tau build-up in the frontal cortex tracks shortened, isolated slow waves during deep sleep, and those altered waves track memory loss. A 2025 paper found slow-wave synchrony during NREM sleep tracks cognitive impairment in early Alzheimer’s.
The direction of travel: using deep-sleep brainwaves as an early biomarker of brain ageing, and testing whether boosting slow waves (for example with timed sound or stimulation) protects memory.
Caveats
In 2019 Alexey Guzey published a detailed critique of Why We Sleep, documenting overstated claims and factual errors — including a “sleep-loss epidemic” attributed to the WHO that the WHO never declared, and a graph with a data bar omitted. Walker acknowledged some errors. Read the book’s mechanisms as solid and its scarier headline numbers as advocacy.
Walker is also a co-founder of Somnee, a commercial sleep-stimulation headband. That doesn’t invalidate his research, but weigh his statements about sleep technology accordingly.
None of the practical advice here depends on the contested claims. The core habits — regularity, darkness, cool room, no late caffeine or alcohol, get up if you can’t sleep — are well supported regardless.
Sources
- Walker, M. Why We Sleep (2017), incl. appendix “Twelve tips for healthy sleep”.
- Huberman Lab Guest Series with Dr. Matthew Walker, 6 episodes (2024), and earlier Huberman Lab appearances.
- The Matt Walker Podcast (2021–); TED talk “Sleep is your superpower” (2019); MasterClass on sleep.
- Yoo, Gujar, Hu, Jolesz & Walker (2007). The human emotional brain without sleep. Current Biology.
- Walker & van der Helm (2009). Overnight therapy? Psychological Bulletin.
- Mander et al. (2015). β-amyloid disrupts human NREM slow waves and memory consolidation. Nature Neuroscience.
- Ben Simon & Walker (2018). Sleep loss causes social withdrawal and loneliness. Nature Communications.
- Winer et al. (2019). Sleep as a potential biomarker of tau and β-amyloid burden. Journal of Neuroscience.
- Walker (2021). Sleep essentialism. Brain.
- Ben Simon et al. (2022). Sleep loss leads to the withdrawal of human helping. PLOS Biology.
- Sharon et al. (2025). Slow wave synchrony during NREM sleep tracks cognitive impairment in prodromal Alzheimer’s. Alzheimer’s & Dementia.
- Sharon et al. (2026). Tau and isolated slow waves in human memory impairment. Nature Neuroscience.
- Guzey, A. (2019). Matthew Walker’s “Why We Sleep” is riddled with scientific and factual errors.
Not medical advice. Persistent insomnia, loud snoring or gasping, or daytime sleepiness that affects driving warrant a doctor — sleep apnoea is common and treatable. Spotted an error? get in touch.