Sleep stages without a smartwatch: what's possible, and what isn't

Every sleep app shows the same pretty four-colour chart: awake, light, deep, REM. The problem is that many of them have no way of knowing the fourth one.

Here is what each method can genuinely know. Without the marketing.

What we are trying to tell apart

A night is made of roughly ninety-minute cycles, each passing through several states:

• wake — including micro-awakenings you do not remember;
• light sleep, which occupies the largest share of the night;
• deep sleep, concentrated early, the physically restorative one;
• REM sleep, the dreaming state, where the brain is very active but muscles are relaxed.

These states are defined by brain activity. That is the whole difficulty: no consumer device measures your brain.

What each tool actually measures

The reference: polysomnography

In a lab, electrodes on the scalp measure electrical brain activity, eye movement and muscle tone. It is the only test that determines your sleep stages. Everything else is an approximation of it.

A smartwatch or ring

They measure your heart rate, its variability and your movement, in constant contact with the skin. That trio allows a markedly finer estimate, good enough to distinguish REM sleep — cardiac variation there is characteristic.

It is not a brain measurement, so not absolute truth. But it is the best approximation available outside a lab, and the gap with the reference is reasonable.

A phone on the nightstand

It has two signals: sound and movement.

Sound is richer than people assume. Breathing rhythm changes with sleep depth; snoring appears mostly in certain phases; position changes are audible. Movement completes the picture: a calm night and a restless one do not share a signature.

With those two signals, you can honestly distinguish three zones: awake, light sleep, deep sleep.

You cannot distinguish REM sleep. It makes no particular noise and comes with muscle relaxation, therefore stillness — indistinguishable from calm deep sleep from a nightstand.

The test that exposes a dishonest app

Here is a simple criterion, and it is devastating.

An app that shows REM sleep while you are wearing nothing on your body is inventing data.

It is not making an error: it is fabricating. Usually by applying an average statistical distribution — "REM is around a quarter of the night, mostly later on" — dressed up as a personal measurement. The chart looks more complete, sells better, and has nothing to do with you.

It is also why Apple regulates this kind of claim: an estimate presented as a medical measurement is grounds for App Store rejection.

How to read an estimate usefully

An estimate is not useless — read badly, it is misleading. Three rules are enough.

Follow trends, not nights. An isolated figure means nothing. A curve over three weeks does.

Look at awakenings before stages. The number and timing of interruptions are measured far more reliably than the stage breakdown, and they explain your tiredness better.

Cross-check against how you feel. If the app reports an excellent night and you are exhausted, believe your body. The app is the one that is wrong.

The best of both

If you already own a smartwatch, you do not have to choose. It records your stages in the iPhone Health app; a sound-based app can read them there and cross-reference them with your night's sounds.

That is where it gets interesting, because the question changes in nature. No longer "what were my stages", but: "did my snoring land during deep sleep?" Neither tool can answer that alone.

To remember

• Wearing nothing: three estimated zones, honestly.
• With a watch: four stages, as a solid approximation.
• In a lab: the actual measurement.

A tool that tells you plainly what it does not know beats a complete, invented chart. You will not make better decisions with false precision — just more confident ones, which is worse.