What Sleep Trackers Actually Measure

What a Real Sleep Study Actually Measures

The gold standard for sleep measurement is polysomnography, usually shortened to PSG.

That is the full sleep-lab setup. Electrodes on the scalp. Sensors near the eyes. Muscle sensors. Heart rhythm. Breathing. Oxygen. The whole thing.

The key piece is EEG, which measures electrical activity in the brain. That is what lets a sleep study directly score sleep stages. If someone is in light sleep, deep sleep, REM, or awake, PSG is looking at the brain activity behind that state.

It also captures things your wearable cannot fully see. Breathing patterns. Airflow. Oxygen drops. Limb movement. Muscle tone. Eye movement. That is why a real sleep study can help diagnose sleep apnea, REM behavior disorder, periodic limb movement, and other sleep problems.

I have seen what a full PSG setup looks like. It is not subtle.

There is a reason it is the gold standard, and there is also a reason nobody wants to sleep like that every night.

That is where wearables come in. They are not trying to be a full sleep lab. They are trying to give you a useful signal every night without wires glued to your head.

What Your Wearable Is Actually Measuring

Most consumer sleep trackers use a few basic signals.

The biggest one is optical heart rate, usually called PPG. A small light shines into your skin and reads changes in blood volume as your pulse moves through the tissue. From that, the device estimates heart rate and sometimes heart rate variability.

Then there is movement. An accelerometer tracks whether you are still, restless, rolling around, or getting up. Stillness usually means you are probably asleep. Movement usually means you may be awake or transitioning between stages.

Some devices add skin temperature. That can help with longer-term patterns, illness detection, menstrual cycle tracking, and sometimes sleep-stage estimates.

Many devices also include SpO2, or blood oxygen saturation. That can be useful if oxygen is dipping overnight, especially if there are possible sleep apnea concerns.

But here is what is missing:

No EEG.

No direct brain-wave measurement.

No EMG muscle sensor like a sleep lab.

No airflow sensor.

No full respiratory belt.

So when your watch says you got 42 minutes of deep sleep, it did not directly observe deep sleep. It inferred deep sleep from signals that tend to correlate with deep sleep.

That is not a scam. It is just what the technology is.

These are proxies. Some are good proxies. But they are still proxies.

And once you understand that, you stop treating the device like a sleep doctor and start treating it like a trend tool.

What the Validation Research Actually Shows

Stone and colleagues published a 2025 validation study comparing six consumer devices against polysomnography. That is exactly the kind of study you want — wrist-worn devices measured against the actual sleep-lab standard.

The main finding was clear.

The devices were good at detecting sleep. Across all six devices, sleep-detection sensitivity was above 91%. If you were asleep, the device usually knew you were asleep.

The problem was wake detection.

Specificity was low, ranging from 29% to 52%. When people were actually awake, the devices were much less reliable at recognizing it. They tended to call quiet wakefulness sleep.

That matters. If you lie in bed awake but still, your tracker may think you are sleeping. That is one reason wearables can overestimate total sleep time.

The bigger drop-off comes with sleep stages.

Stone 2025 reported Cohen's kappa scores for sleep-stage agreement. Kappa measures how much the device agrees with the sleep lab beyond random chance. Higher is better. Moderate is not the same as excellent.

In that study:

• Apple Watch Series 8 had the best sleep-stage agreement of the group, with κ=0.53
• Fitbit Sense was κ=0.42
• Fitbit Charge 5 was κ=0.41
• Whoop 4.0 was κ=0.37
• Withings Scanwatch was κ=0.22
• Garmin Vivosmart 4 was κ=0.21

My honest read: an Apple Watch getting moderate agreement with a sleep lab is not bad for a wrist device with no brain sensor. It is also not good enough to obsess over a single night's deep sleep number.

Both things can be true.

Chinoy and colleagues' 2022 validation work supports the same general picture from an earlier generation of devices. This is not a one-study fluke. It is the recurring pattern with consumer sleep tracking.

Wearables are decent at sleep versus wake.

They are weaker at wake detection when you are lying still.

They are much less reliable for exact sleep stages.

What Trackers Are Actually Good For

Some people act like wearables are basically sleep labs on your wrist. They are not.

Other people see the limitations and decide the whole category is useless. I do not think that is right either.

Trackers are genuinely useful if you use them for what they are good at.

The first is trend tracking. If your average sleep duration is moving from six hours to seven hours over several weeks, that is real information. If your bedtime is getting more consistent, that matters. If your resting heart rate is creeping up and HRV is trending down, that is a signal worth noticing.

Not panicking over. Not diagnosing yourself with. But noticing.

The second is behavioral feedback. If seeing the data makes you go to bed earlier, drink less, stop crushing caffeine late in the day, or realize that late-night scrolling is wrecking your sleep timing, the tracker is doing something valuable. Even if the deep sleep number is not perfect.

The third is outlier detection. If your normal sleep score is in the 80s and suddenly you get a night in the 40s, something changed. Maybe alcohol. Maybe illness. Maybe stress. Maybe a hard workout. Maybe travel.

The exact REM number may be wrong, but the “something was off” signal may still be useful.

That is how my wife uses her Oura. She is not waking up and trying to optimize her deep sleep percentage like it is a video game stat. She is looking at trends. She notices when recovery is off. She pays attention to whether changes she makes actually move the needle over time.

A tracker should make you more aware, not more neurotic.

The Orthosomnia Problem

There is a term for sleep-tracker anxiety: orthosomnia.

Baron and colleagues used it to describe people becoming so focused on perfecting their sleep data that the tracking itself starts making sleep worse.

That is not hard to imagine.

You wake up. Your tracker says your sleep was bad. Now you feel bad before the day even starts. Then at night you lie there thinking about whether you are getting enough REM. Now you are awake because you are worried about the device that is supposed to help you sleep.

That is the point where the tool has turned against you.

Tracking sleep is not the same thing as improving sleep.

If the data helps you make better choices, use it. If the data makes you anxious, obsessive, or convinced you are broken because your watch says you got 38 minutes of deep sleep, that is a feedback loop. And not a good one.

HRV: Useful, But Easy to Misuse

HRV stands for heart rate variability.

It measures the tiny variation in time between heartbeats. More variation usually suggests more parasympathetic activity — the recovery side of the nervous system. Less variation often suggests more sympathetic activity — stress, strain, illness, alcohol, hard training, poor sleep, or some combination of those.

Biohackers love HRV because it feels like a recovery score hiding inside the body. And honestly, I get the appeal.

Deep sleep is usually where nighttime HRV reaches its highest point. That makes sense — deep sleep is when the nervous system should be shifting into a more recovered, parasympathetic state.

But HRV is not magic.

A single-night HRV drop does not automatically mean something is wrong. It could be noise. It could be a hard workout. It could be dehydration. It could be a bad sensor read. It could be that you ate late or had a drink or fought off a virus before you felt sick.

The trend matters more than the single number.

Seven nights tells you more than one night. Thirty nights tells you more than seven.

Your baseline matters more than anyone else's number. Comparing your HRV to someone else's is mostly a waste of time. The better question is: what is normal for you, and is that changing?

If your HRV is dropping for several nights without an obvious reason, that is worth paying attention to. It is not a diagnosis. It is not a command.

Signals, not verdicts.

What This Means Before You Buy a Tracker

The right question is not which tracker is perfectly accurate. None of them are. Every consumer wearable has the same fundamental limitation: no EEG, no direct brain measurement.

The better question is what you actually want to learn.

If you want to know whether your sleep schedule is consistent, a wearable can help.

If you want to see how alcohol, late meals, hard training, stress, travel, or illness affect your recovery trends, a wearable can help.

If you want a nudge that gets you to take sleep more seriously, a wearable can help.

If you want a medical-grade answer about sleep apnea, REM behavior disorder, or exactly how much deep sleep you got last night, a consumer tracker is the wrong tool.

That distinction is not a knock on the devices. It is just honest.

The form factor matters too. Some people hate sleeping with a watch. Some people do not want a ring. Some people care about battery life. Some people want the simplest possible recovery score.

Those are real differences. The device choice should come after the measurement question, not before it.

What do you want to track? How will you use the data? Will it change your behavior?

A sleep tracker does not need to be perfect to be useful. But you need to know what it is actually measuring.

Related Reading