Why We Sleep, Chapter Two: The Chapter Nobody Fact-Checked

In 2019, Alexey Guzey published a line-by-line audit of Why We Sleep that became the standard citation for anyone sceptical of Matthew Walker's book. What most people who cite it have not noticed is how little of the book it covers. Guzey checked Chapter 1. He spent roughly 130 hours doing it, and appended a note explaining that auditing the whole book at that rate would take him something like 3,000 hours. He stopped.

We covered that first chapter in the previous instalment of this series. Chapter 2 is where the book stops making claims about how important sleep is and starts making claims about how sleep is generated - the machinery. It is also, for most readers, the chapter with the highest ratio of practical advice to prose. And it sits in genuinely unexamined territory. Guzey's silence on it is not a clean bill of health. It is an absence of data.

One note on scope, because chapter summaries online tend to blur together: this is not the REM sleep chapter and not the sleep deprivation chapter. Dreaming, the function of dreams, and the health consequences of sleep loss all come later in the book. Chapter 2 is narrower and more mechanical. It is about the biological timekeeping that decides when you go to bed, and how hard it is to stay awake once you are there.

The chapter's full title is Caffeine, Jet Lag, and Melatonin: Losing and Gaining Control of Your Sleep Rhythm, which is a promise about control, and the mechanism section is where Matthew Walker earns it. Its central idea - that your sleep is governed by two independent systems rather than one - is standard chronobiology and it survives contact with the literature intact. One is the circadian rhythm, a roughly 24-hour cycle run by the suprachiasmatic nucleus, which decides when you should be awake. The other is sleep pressure: adenosine, a byproduct that accumulates in the brain for every hour you are awake and dissipates throughout the hours you sleep. That one decides how badly you need to. Adenosine does not care what time it is; it cares how long you have been awake, and it is cleared by an hour of sleep, not by an hour of rest.

Walker's point, and it is a good one, is that these two systems are not connected to each other. You can be flooded with adenosine at 4 p.m. and still not fall asleep because your circadian signal is peaking. You can be circadian-ready at 11 p.m. and lie there awake because you napped. Most bad sleep nights are one of those two mismatches, and knowing which one you have is the whole diagnosis. The first is a pressure problem and the fix is behavioural. The second is a schedule problem and the fix is light. Treating one as the other is the most common self-help error in this entire field.

That last row is the chapter's most useful sentence and it is worth restating plainly: caffeine does not remove adenosine. It occupies the receptors that would otherwise report it. The adenosine keeps piling up behind the blockade, which is why the crash when caffeine clears is worse than the tiredness you started with. We put numbers on the clearance in the caffeine half-life piece.

Three paragraphs into Chapter 2, Walker writes that every living creature on the planet with a life span of more than several days generates this natural circadian cycle. This is the single Chapter 2 claim Guzey checked, and it is false as written. Brewer's yeast lives well past several days and is not known to run one - Guzey cites Fabrizio and Longo's 2003 paper in Aging Cell on yeast lifespan alongside Wildenberg and Murray's 2014 paper in eLife, whose entire project was to evolve a 24-hour oscillator in budding yeast, on the premise that it does not natively have one.

In fairness, this corner of the biology is itself contested - a 2010 paper titled A circadian clock in Saccharomyces cerevisiae argued the opposite. Which is rather the point. It is not a settled enough area of research to make an every living creature claim about.

Is that pedantry? Partly. Walker plainly meant animals, and no reader's behaviour changes because of it. But it is a universality claim, and a universality claim is refuted by one counterexample, so the sentence should not have survived editing. It also points at something the chapter does elsewhere: reaching for every and always where the evidence supports most. Worth noting too that circadian dominance itself is contested - Guzey points to work in PLOS Biology and Proceedings of the Royal Society B describing animals whose activity runs around the clock with no overt rhythm. Treat the strong version of the claim as unsettled.

The chapter's origin story is Nathaniel Kleitman and Bruce Richardson descending into Mammoth Cave, Kentucky, in June 1938 and staying 32 days, 140 feet down, in constant darkness at a steady 54 degrees. They were testing whether a human body could be talked into a 28-hour day. Richardson, 25, adapted. Kleitman, 43, did not - his body temperature cycle stubbornly held near 24 hours. That difference is the finding: something internal was keeping time without any external cue to keep it by.

The modern number is more precise than the cave could ever have been. Czeisler and colleagues, working under carefully controlled light and publishing in Science in 1999, put the intrinsic human circadian period at 24.18 hours - about 24 hours and 11 minutes - with a tight distribution and no meaningful difference between younger and older adults. Chapter 2's account of this is sound, and the practical consequence is the part worth keeping: your clock runs slightly long, so it drifts later unless something resets it every day. Morning light is that something. No supplement substitutes for it. The drift is small - roughly eleven minutes a day - but it compounds throughout a week, which is why an irregular schedule costs a healthy adult more sleep than any single late night does.

Here the chapter gets the biology right and most of its readers still get the wrong idea. Walker's framing is that melatonin signals darkness and times the onset of sleep - it does not generate sleep itself. That is correct, and it is the opposite of how melatonin is sold.

The numbers are worth having. The most-cited meta-analysis of melatonin for primary sleep disorders - Ferracioli-Oda, Qawasmi and Bloch, PLoS ONE, 2013 - pooled 19 randomised placebo-controlled studies and found melatonin reduced the time to fall asleep by 7.06 minutes, with total sleep time and sleep quality improving modestly as well. The authors were candid that the effects are small, though they noted the useful detail that they do not appear to fade with continued use.

There is a second problem the chapter does not cover, and it is arguably the bigger one. When Erland and Saxena analysed 31 melatonin products for the Journal of Clinical Sleep Medicine in 2017, melatonin content ranged from 83% below the labelled dose to 478% above it. More than 71% of the supplements missed their own label by more than 10%, lot-to-lot variation within a single product reached 465%, and 26% of them contained serotonin. So the honest summary is: a small real effect, delivered by a bottle whose contents you cannot verify. Melatonin is a medication question, not a shopping question - talk to your doctor before starting it, particularly if you take anything else.

The jet lag section is where Chapter 2 is at its most alarming and its least well supported. The centrepiece is Cho's 2001 study in Nature Neuroscience, which compared cabin crew on short recovery intervals against crew on long ones and found the short-recovery group had smaller right temporal lobe volume, worse spatial memory and higher salivary cortisol.

It is a real finding in a real journal and we are not waving it away. But it is ten people per group - twenty in total - and it is cross-sectional, which means it shows an association at one moment in time and cannot tell you which way the arrow points. A popular science book leaning that hard on an n of 20 is exactly the pattern Guzey criticised in Chapter 1, and it recurs here without anyone having said so. Read the jet lag pages as a hypothesis worth taking seriously, not as a settled result about what flying does to a human brain. Walker's broader point - that modern society schedules long-haul travel and shift work as though the body clock were negotiable - rests on much firmer ground than this one small study, and it does not need the brain-imaging result to stand.

Strip Chapter 2 to its actionable core and you get three levers: light timing, adenosine timing, and body temperature. We sell sheets, so we have an obvious interest in the third one, and we would rather state that plainly than bury it. Falling asleep requires your core body temperature to drop, and the bedding that sits against you for eight hours either helps that or fights it - we laid out the thermostat side of this in the bedroom temperature protocol, and the fabric side in whether bamboo sheets actually keep you cool.

What we will not tell you is that sheets fix a circadian problem. If your clock is drifting because you get no morning light, no fabric on earth corrects it, and any company that implies otherwise is selling you the wrong lever. In the real world the major levers are the free ones: light early, a consistent start time, less caffeine after lunch. Bedding is a thermal tool. That is a modest claim, and it is the true one.

The questions people also ask.