Wearable sleep trackers report a detailed breakdown of the night, including how long was spent in each stage. Understanding what the device actually records explains which of those numbers deserve attention.
Clinical staging depends on brain activity
The reference method for classifying sleep records electrical activity at the scalp alongside eye movement and muscle tone, and a trained scorer assigns each short segment of the night to a stage.
The stages are defined by specific electrical patterns, which is why the classification requires electrodes rather than any indirect signal.
Even between trained human scorers, agreement on stage boundaries is good but not perfect, which sets a practical ceiling on how precise any staging can be.
Wearables measure something adjacent
A wrist device typically records movement through an accelerometer and blood flow through an optical sensor at the skin, from which it derives heart rate and beat-to-beat variation.
Those signals do correlate with sleep stage, since heart rate variability and movement genuinely differ between deep sleep, dreaming sleep and quiet wakefulness.
The device then runs a model that converts those correlations into a stage estimate, so the output is an inference from proxy signals rather than an observation of sleep itself.
Totals hold up better than breakdowns
Validation studies consistently find that modern wearables identify total sleep time reasonably well, because the distinction between sleeping and being awake and moving is a comparatively easy one.
Stage-level agreement is markedly weaker, particularly in separating light sleep from deep sleep, where the proxy signals overlap considerably.
The practical reading is that trends in total duration and timing across weeks carry real information, while a single night's deep sleep percentage carries much less than its precision implies.
Lying still is not the same as sleeping
Movement-based estimation has a known failure mode: someone lying awake and motionless is frequently scored as asleep.
This matters most for the people most likely to buy a tracker, since lying still and awake is the defining experience of insomnia.
The result can be a device reporting seven hours to somebody who knows they were awake for two of them, which does nothing for confidence in the rest of the report.
Measurement can become the problem
Clinicians have described a pattern in which anxious attention to sleep scores worsens the sleep being scored, since preoccupation with sleeping is itself one of the maintaining factors in insomnia.
The recommended handling is to use the data at weekly resolution for timing and duration, and to stop treating a nightly score as a verdict on the night.
Where sleep is persistently poor, assessment addresses causes such as apnoea or insomnia directly, and no consumer device is designed or validated to identify those.