Why We Sleep, Chapter Three: The One That Holds Up

The chapter is called Defining and Generating Sleep: Time Dilation and What We Learned from a Baby in 1952, and it does two jobs. First it answers a question that sounds trivial and is not: how do you know, from the outside, that a person is asleep? Then it explains where our entire picture of the sleeping brain came from, which turns out to be one graduate student, one borrowed machine and one infant.

It is worth being clear about the scope, because readers arriving at a chapter summary often want the wrong chapter. Chapter Three is about what sleep is and how we came to be able to see it. The machinery that generates it - the circadian rhythm on one side and adenosine building sleep pressure across the waking day on the other, with melatonin acting as the timing official rather than the sprinter - is Chapter Two's business, and Matthew Walker treats the consequences of going without in the chapters after this one. If you are looking for the caffeine and adenosine material, or for the sleep deprivation argument that made the book famous, it is not here.

Walker lists the observable tells: the person is horizontal, muscle tone has dropped so they are draped over whatever is holding them up, they do not respond, the state reverses easily, and it is happening on a schedule set by the circadian rhythm. That last one does real work. Reversibility is what separates sleep from a coma, and the schedule is what separates it from fainting. None of this is controversial and all of it is the sort of thing you only notice is load-bearing when someone writes it down.

Then he moves inside the skull. At sleep onset the thalamus, the relay that most sensory traffic has to pass through on its way to the cortex, stops passing it along. Walker calls this a sensory blackout. It is a good phrase for a real thing: the reason a conversation in the next room stops being a conversation and becomes noise is not that your ears switched off, it is that the gate upstream closed.

The historical claim checks out cleanly, which is not something we could say about every anecdote in this book. Eugene Aserinsky, working under Nathaniel Kleitman at Chicago, was recording eye movements during sleep - his own infant son among the early subjects - and found periods where the eyes moved rapidly and repeatedly while the sleeper stayed asleep. The paper is Aserinsky and Kleitman, Regularly Occurring Periods of Eye Motility, and Concomitant Phenomena, During Sleep, published in Science in 1953, and it is two pages long. Everything downstream of it - REM, dream research, the entire architecture of a modern sleep study - starts there.

Walker also explains what verification actually requires, and it is worth repeating because consumer sleep trackers have muddied it. Scoring sleep properly means recording three signals at once: brainwave activity, eye movement and muscle activity. Together they are polysomnography. A wrist device measuring movement and heart rate is inferring from two steps away, which is fine for trends and not fine for the stage-by-stage breakdown it prints for you each morning.

Here is the first genuine complaint. The chapter presents the roughly ninety-minute NREM-REM cycle as the beat the night keeps, and that framing has escaped into every sleep app and calculator on the internet. The number is an average, and the spread around it is wide enough to matter. Feinberg and Floyd, who wrote the modern definition of the cycle in Psychophysiology in 1979, documented systematic changes in cycle length across the night rather than a fixed interval. A 2023 retrospective analysis out of the Centre for Chronobiology in Basel, published in Sleep Health, scored 6,064 polysomnographically recorded cycles from 369 people and put the median at ninety-six minutes - with large variation between individuals, and with age and sex both shifting the balance of NREM and REM within the cycle.

The practical version: if you have ever set an alarm on a ninety-minute multiple to wake between cycles, you were doing arithmetic on a number that is not yours specifically.

The second complaint is vocabulary. The book leans on the older stage-one-through-four description of NREM sleep. The American Academy of Sleep Medicine's 2007 scoring manual folded stages three and four into a single stage, N3, and moved everyone to W, N1, N2, N3 and R. The book is not wrong about the physiology; it is a revision behind on the labels, which matters mainly when a reader takes those stage names to a clinician.

The chapter's own subtitle advertises time dilation, and the claim inside the chapter is milder than the billing: that you cannot consciously track time while asleep. That milder version is uncontroversial. The stronger one, that dreaming stretches time, is where the literature gets interesting and less obliging.

The cleanest evidence comes from signal-verified lucid dreaming, a method Stephen LaBerge developed in which a dreamer who knows they are dreaming marks the start and end of a task with deliberate eye movements that show up on the polygraph. It is a genuine window: the timestamps are objective even though the experience is not. When lucid dreamers count out an interval, the elapsed clock time comes out close to what the same person produces awake. Later work by Erlacher and colleagues found the picture is task-dependent rather than uniform - motor tasks such as squats took roughly forty per cent longer in the dream than awake, while counting did not. So the honest summary is that dream time tracks real time more closely than folklore says, with a slowdown that shows up for some kinds of action and not others.

These are small studies on unusual subjects, because lucid dreamers who can signal reliably in a laboratory are rare. We are flagging it as contested rather than settled, in both directions.

That is the pattern worth taking from this instalment. Chapter Two came apart where the book reached past its evidence toward consequence. Chapter Three stays close to the instruments, and it survives.

One finding in this chapter has a consequence the book does not chase, and it is the one we think about most. Because REM is concentrated in the later cycles, sleep cut short at the end is not a proportional loss. Waking two hours early does not remove a quarter of everything evenly; it removes a disproportionate share of REM specifically, since that is where REM lives. Two hours of sleep deprivation at the front of the night and two hours at the back are not the same injury, and the human sleep literature has been consistent on this for decades.

Now put that next to something from the thermoregulation literature. During REM sleep the body's own temperature defences are largely offline - shivering and sweating responses are suppressed in a way they are not during NREM sleep or wakefulness, a finding reviewed in depth by Cerri and colleagues in Comprehensive Physiology. Stack the two facts and you get a specific, unglamorous conclusion: the hours when you have the least ability to regulate your own temperature are the same hours that carry most of your REM sleep, and they sit at the end of the night when the bed has had seven hours to accumulate your body heat.

That is the honest reason bedding is worth any thought at all, and it is a narrower claim than the industry usually makes. A sheet cannot give you more REM sleep. What it can do is stop being the reason you surface at half past four. We sell sheets, so treat that as an interested party talking - but the mechanism is not ours, it is Parmeggiani's and Cerri's, and it points at breathability and moisture handling in the back half of the night rather than at anything that feels impressive in a shop.

We have written the longer version of the overnight-temperature argument in do bamboo sheets keep you cool, and the fibre-and-weave groundwork sits in bed sheets 101. Nothing in this chapter is medical advice; if you are waking every night and cannot find a reason, that is a conversation for your doctor rather than a shopping problem.

Common questions