Wearable sleep trackers and dreams: what they measure, and what they miss

How does a sleep tracker know you are asleep?

Movement first

At the heart of every watch, band or ring is an accelerometer, a sensor that records movement in three directions. The logic is simple: when you stay still for a long time, you are probably asleep; when you move, you are probably awake. Sleep researchers have used this method, called actigraphy, for decades, long before consumer wearables existed.

Its weak spot is just as simple. If you lie still in the dark, wide awake and thinking about tomorrow, the sensor has little way of knowing.

Then your heart

Most current devices add an optical sensor on the underside, called photoplethysmography (PPG). Its green light follows blood flow under the skin to estimate heart rate and heart rate variability (HRV). Both change across the night: the heart tends to slow down and steady in deep sleep, and to become more irregular during REM sleep. By combining movement and heart signals, the algorithm makes an educated guess about your sleep stage.

Some models also track skin temperature or blood oxygen (SpO2). None of these sensors measures brain activity. Yet in a sleep lab, stages are defined from brain waves recorded by EEG, as part of a full sleep study called polysomnography (PSG).

What do the numbers actually mean?

Total sleep time: the most solid figure

How long you slept, and the share of your time in bed spent asleep (sleep efficiency), are the most reliable figures on your screen. They rest mainly on the stillness-versus-movement signal, which is the easiest one to read.

Sleep stages: an estimate

"Light," "deep" and "REM" sleep roughly match the clinical stages N1 and N2, N3 and REM. On a wearable, though, they are inferred from heart rate and movement patterns, not measured. Read that colored graph as a likely sketch, not as a recording of your brain.

Sleep scores: a brand's recipe

A "sleep score" or "readiness score" blends duration, estimated stages, heart rate and HRV into one number. Each company uses its own formula, rarely published. A score of 85 on one brand does not mean the same thing as 85 on another, and it does not tell you how rested you feel.

How accurate are they?

What the research shows

In a 2021 study, Chinoy and colleagues tested seven consumer devices against polysomnography in a sleep lab. All of them detected sleep very well, but they often classified time spent awake as sleep. Their sleep-stage results were mixed, and performance varied a lot from one device to another.

A meta-analysis of Fitbit models by Haghayegh and colleagues (2019) points the same way. Older models, which relied on movement alone, tended to overestimate total sleep time and underestimate time awake during the night. Newer models that also use heart rate did better, but the authors stress that they are not a substitute for polysomnography.

Reviews of the field, such as de Zambotti and colleagues (2019), call for caution, especially with sleep stages. And because results differ so much between studies, Menghini and colleagues proposed a standard method for testing each new device against polysomnography. In practice, a trend over several weeks tells you more than any single night.

What they cannot do

No wearable sleep tracker can detect a dream. Dreams occur most often in REM sleep, but also in other stages. Even perfect REM detection would only tell you that your brain was in a state where dreaming is likely. It would not tell you whether you dreamed, what happened in the dream, how it felt, or whether you will remember it.

When the score keeps you awake

In 2017, Kelly Glazer Baron and colleagues described patients whose effort to get a "perfect" score had made their insomnia worse. They called it orthosomnia. It is not a formal diagnosis, but it is a useful warning sign. If checking your numbers each morning makes you anxious, look only at weekly averages, or hide the score for a while.

When to talk to a doctor

The American Academy of Sleep Medicine states that consumer sleep devices cannot be used to diagnose or treat sleep disorders, although their data can support a conversation with a clinician. Some watches now show alerts about possible breathing problems; only a medical evaluation can confirm them. See a doctor if you snore loudly, if someone notices pauses in your breathing, if you are very sleepy during the day despite enough hours in bed, if insomnia lasts, or if frequent nightmares are disrupting your nights.

In short: good at telling sleep from movement, less good at spotting quiet wakefulness, inconsistent on stages, and blind to dreams. Read them over weeks, not nights.

Tracker and dream journal: how to use both

Two different kinds of information

Your tracker gives you numbers: duration, estimated stages, heart rate, restless periods. A dream journal keeps what you lived: the scenes, the emotions, how clear the memory is, how you feel on waking. Neither replaces the other.

A journal also catches things no sensor can see: a recurring dream that comes back every week, a shift toward more anxious dreams before a stressful deadline, or very vivid dreams returning after a few nights of catch-up sleep, which researchers link to REM rebound after sleep debt.

Cross them over weeks, not nights

Since a single night's data is uncertain, compare the two over several weeks. A few questions to keep in mind: do I remember more dreams on nights when the tracker shows awakenings late in the night? Are my most unpleasant dreams on short nights? Does my "good score" match how I actually feel? The answers will be personal, and that is the point.

A tracker and journal example

Fictional example: this entry shows how to keep the data, the dream and the context side by side without mixing them up.

• Tracker: "6 h 10, score 68, woke up three times, the last one at 5:50."
• Dream: "I kept missing a train. Every time I reached the platform, it changed."
• Emotion: "Irritation, then relief when I woke up."
• Recent context: "Worked late, coffee at 5 p.m."
• Question to keep: "Do I remember my dreams better after nights with early-morning awakenings?"

One entry proves nothing. Ten entries over a month start to show whether a pattern holds.

What is coming next in sleep tech?

Headbands that read brain activity

EEG headbands place electrodes on the forehead and record brain activity directly, the same kind of signal used in a sleep lab, with fewer electrodes. In principle, that brings them closer to polysomnography than a wrist device. They are less comfortable to wear, though, and each model still needs independent validation.

Sensors you do not wear

Under-mattress pads and bedside radar devices track breathing, heart rate and movement without touching you. You simply sleep as usual. Like wrist devices, they estimate stages indirectly.

AI, and the limit it will not cross

Researchers expect machine learning trained on large datasets to improve what consumer sensors can estimate, as a 2020 review by Perez-Pozuelo and colleagues describes. Even if a device one day reliably signals that you are probably dreaming, knowing what you dream will still depend on what you remember and tell.

Who sees your sleep data?

Sleep data is health data

Heart rate, HRV, blood oxygen, bedtimes and wake times say a lot about your health, your stress and your routine. Most wearable brands store this data on their servers under their own privacy policy. In the United States, consumer apps are generally not covered by HIPAA, the law that protects medical records.

Questions to ask before you choose

Can I export and delete my data? Is it shared with third parties, and for what purpose? What happens to it if the company is sold or shuts down? Night after night for years, sleep data becomes an unusually intimate record.

The same questions apply to a dream journal. Noctalia hosts journal data in the EU, uses the service providers it needs for AI, transcription and subscriptions, does not sell personal data and does not run targeted advertising. It is not an offline-only journal, so what matters is transparency and control. The details are in our privacy policy, and our guide to AI dream journal privacy explains what to check.