What is sleep cycle?

Sleep is conventionally divided into non-REM, itself split into three stages, and REM. A cycle moves through them in a rough sequence rather than a strict one, and it is normal to move backwards between stages within a cycle.

Stage one is the brief transition into sleep, lasting only a few minutes, and is easy to be woken from. Stage two is light sleep and accounts for the largest share of the night. Stage three is deep, slow-wave sleep, the hardest to be woken from and the stage associated with physical recovery. REM follows, with brain activity resembling wakefulness and most vivid dreaming.

The proportion of each changes as the night progresses. Deep sleep dominates the early cycles and can disappear almost entirely from the last ones; REM periods start short and lengthen. A cycle at 11pm and a cycle at 5am are not the same thing.

Ninety minutes is the figure everyone quotes, and it is a reasonable average rather than a rule. Actual cycle length varies between individuals and between nights for the same person, commonly falling somewhere between roughly seventy and one hundred and twenty minutes.

It also varies within a single night. Early cycles tend to be shorter than later ones, so even for a person whose average is exactly ninety minutes, the cycles are not ninety, ninety, ninety.

This is the detail that undermines the arithmetic built on top of it, and it is usually left out of the advice.

The popular version says to set your alarm for a multiple of ninety minutes after falling asleep, so that it lands between cycles and you wake from light sleep rather than deep sleep. The underlying physiology is sound: waking from deep sleep really does produce worse sleep inertia than waking from light sleep.

The arithmetic is where it falls apart. It requires knowing your own cycle length, which varies; knowing exactly when you fell asleep, which you do not; and assuming the cycles are uniform, which they are not. By the fourth or fifth cycle, an error of fifteen minutes per cycle has accumulated into an hour, which is longer than the window the calculation was trying to hit.

It is also worth noting the pattern the advice produces in practice: people set the alarm earlier than they otherwise would, sleeping less in the hope of waking better. Total sleep has a larger and better-established effect on how you feel than the stage you woke from, so that trade tends to run the wrong way.

The version of this idea that does hold up is much simpler. Keep a consistent schedule, get enough total sleep, and the cycle structure sorts itself out without any arithmetic.

Clinical sleep staging uses polysomnography, which measures brain activity, eye movement and muscle tone directly. That is how stages are actually identified.

A phone on a mattress or a watch on a wrist infers stages from movement and sometimes heart rate. That can approximate broad patterns over weeks, and is genuinely useful for spotting trends in how long you slept and how disturbed it was. It cannot reliably tell you which stage you are in at a given moment.

That limitation matters most for exactly the feature people buy these devices for. A smart alarm claiming to wake you in a light-sleep window is acting on an inference it cannot make precisely, which is why the results feel inconsistent.

Alcohol changes the shape of the night rather than simply reducing it. It shortens the time to fall asleep and suppresses REM early, then produces a rebound in the second half with lighter, more broken sleep. Total time can look normal while the structure is substantially altered.

Caffeine has a longer reach than most people account for. Its half-life is several hours, so an afternoon coffee is still measurably present at bedtime, and it reduces deep sleep even when it does not prevent sleep onset. People frequently conclude it does not affect them because they fell asleep fine.

Age changes the structure on its own. Deep slow-wave sleep declines gradually from early adulthood, and sleep becomes lighter and more easily interrupted. That is a normal change rather than a problem to be fixed, though it does mean the same total sleep can feel less restorative later in life.

And anything that fragments sleep repeatedly, including sleep apnoea, a noisy environment, or a snooze chain in the morning, prevents cycles completing. Fragmented sleep of a given length is meaningfully less restorative than continuous sleep of the same length, which is the main reason the snooze button is worth avoiding.

Cycle structure is the reason nap advice is so specific about length, and the specificity is genuinely justified rather than arbitrary.

A nap of roughly ten to twenty minutes stays in light sleep. You wake easily, with little grogginess, and get a real if modest improvement in alertness. This is the version most nap advice recommends, and it is the one that works reliably for most people.

Between about thirty and sixty minutes you are likely to enter deep slow-wave sleep and be woken part-way through it, which produces the worst sleep inertia available. The nap can leave you feeling considerably worse than before you lay down, and the effect can last half an hour or more. This is the range worth avoiding.

Around ninety minutes gives a reasonable chance of completing a full cycle and waking from lighter sleep at the end of it, which is why the longer nap is often suggested as the alternative to the short one. Cycle length varies, so this is a probability rather than a guarantee.

One caution that applies to all of them: napping late in the day reduces the sleep pressure that helps you fall asleep at night. A nap that fixes the afternoon at the cost of the following night is not a good trade if the mornings are already difficult.