Watches Are Only So-So at Guessing Sleep Stages
All six devices caught over 90% of sleep, yet wake detection ran only 29% to 52%. Stages are the weak part of the chart.
Evidence rating for the proposition “Consumer wearables measure sleep parameters with acceptable agreement against polysomnography”: Uncertain · KnowYourPrime Evidence Registry, Finding KYP-ER-0007
Most wrist and ring wearables estimate sleep stages from movement and pulse, not brain activity. Agreement with polysomnography is fair to moderate. In a six-device night-lab comparison, Cohen's kappa ranged from 0.21 to 0.53. All six devices detected more than 90% of sleep epochs, but specificity for wake was only 29% to 52%. Total sleep time comes closer to the lab than stage minutes do.
Do wearables measure brain activity to score sleep stages?
Most do not. A wrist or ring device reads pulse and motion. An algorithm then infers the stage those signals most likely belong to.
The stage labels come from a lab standard built for brainwave recordings. The N1 to N3 and REM categories are defined by EEG, as described in what REM sleep looks like on an EEG. A pulse sensor is reconstructing those labels from a thinner signal.
EEG-based wearables are a separate and much smaller category. A systematic review covered 34 of them across 60 papers. Its 42 validation studies reported consistently high accuracy in sleep staging. That result describes headbands and similar EEG devices, not watches and rings. Apple Watch uses its own names for these stages, Core for N1 and N2 and Deep for N3, set out in Core vs Deep sleep.
How close are wearable sleep stages to a sleep lab?
A 2025 study put six devices through a night in a sleep lab. The devices were the Fitbit Charge 5, Fitbit Sense, Withings ScanWatch, Garmin Vivosmart 4, Whoop 4.0 and Apple Watch Series 8.
Every device detected more than 90% of sleep epochs. Specificity was lower, at 29.39% to 52.15%. That means each device often called real wake time sleep.
Cohen's kappa ranged from 0.21 to 0.53. That is fair to moderate agreement. The Apple Watch Series 8 scored 0.53, the Fitbit Sense 0.42 and the Fitbit Charge 5 0.41.
2 studies
- Six consumer wearables (Fitbit Charge 5, Fitbit Sense, Withings ScanWatch, Garmin Vivosmart 4, Whoop 4.0, Apple Watch Series 8) all detected more than 90% of sleep epochs, with specificity of 29.39% to 52.15% and Cohen's kappa of 0.21 to 0.53 against polysomnography.
- In 35 adults aged 20 to 50 without a sleep disorder, sensitivity for sleep versus wake was 95% or higher for the Oura Ring Gen3, Fitbit Sense 2 and Apple Watch Series 8. Stage sensitivity ranged from 50% to 86%.
Is total sleep time accurate even when stages are not?
It is closer, though not exact. A meta-analysis of 24 studies and 798 participants compared consumer wrist devices with polysomnography. Total sleep time differed significantly, by about 17 minutes. Sleep efficiency, sleep latency and time awake after sleep onset also differed.
Subgroups varied. In the same analysis, Fitbit did not differ from polysomnography on time awake after sleep onset. A separate meta-analysis of Fitbit models without a staging algorithm found overestimates of total sleep time. The range was roughly 7 to 67 minutes depending on the study. It found underestimates of wake after sleep onset of about 6 to 44 minutes.
Fitbit models that do stage sleep did better. That meta-analysis found no significant difference from polysomnography for total sleep time, time awake after sleep onset or sleep efficiency. Those models underestimated sleep onset latency. Sensitivity was 0.95 to 0.96 and specificity 0.58 to 0.69. The authors call them not a substitute for polysomnography.
Do newer algorithms close the gap?
Some do, on some measures. The Oura Ring Gen3 with its second-generation staging algorithm was tested in 96 healthy Japanese adults aged 20 to 70. It did not differ significantly from polysomnography for total sleep time, wake after sleep onset, light sleep or deep sleep. It underestimated sleep efficiency by 1.1% to 1.5% and REM sleep by 4.1 to 5.6 minutes. Stage accuracy ran from 75.5% for light sleep to 90.6% for REM sleep.
In the 35-adult comparison, the Oura Ring did not differ from polysomnography for wake, light, deep or REM sleep. The Fitbit Sense 2 overestimated light sleep by 18 minutes and underestimated deep sleep by 15 minutes. The Apple Watch Series 8 underestimated deep sleep by 43 minutes and overestimated light sleep by 45 minutes.
Results differ by study. The Apple Watch Series 8 had the highest kappa in the six-device comparison and the largest deep sleep error in the three-device one. Both facts come from single-night lab studies with different participants.
Does stage accuracy matter beyond the nightly score?
It could. A program tracking 6,785 people with commercial wearables for a median of 4.5 years found deep sleep and REM sleep inversely associated with incident atrial fibrillation. Irregular sleep went with higher odds of obesity, hypertension, depression and anxiety.
That is an association in long-term wearable data. It does not show that any device measured stages correctly. It does show why the accuracy question gets asked. If stage proportions relate to health outcomes, then estimating them badly costs more than miscounting wake.
For amounts by stage, see the research on deep sleep. For the brain-health question, see the research on sleep and dementia risk.
What do these studies not settle?
Each result describes one device, one algorithm version and one group of people. The Oura study enrolled healthy adults in one country. The three-device study enrolled 35 people for a single night.
None of it shows how these algorithms perform in people with sleep disorders, in older adults, or in children. Firmware also changes. A study of a past-generation model is not a study of the one on a wrist today.
The Fitbit meta-analysis covers models studied up to about 2019. Newer devices and firmware are not represented.
The Oura Gen3 study enrolled 96 healthy adults aged 20 to 70 in one country, and the three-device study enrolled 35 healthy adults for one night. Neither establishes how these algorithms perform in people with sleep disorders, in children, or across repeated nights at home.
Common questions
How accurate are wearable sleep stages?
Fair to moderate against polysomnography. In a six-device night-lab study, Cohen's kappa ranged from 0.21 to 0.53. All devices detected more than 90% of sleep epochs, but specificity for wake ranged from 29.39% to 52.15%.
Do wearables measure brain waves?
Most wrist and ring devices do not. They estimate stages from pulse and motion. EEG-based wearables are a separate category, and a review of 42 validation studies reported consistently high staging accuracy for them.
Is total sleep time more accurate than sleep stages?
Generally closer. A meta-analysis of 24 studies found total sleep time from consumer wrist devices still differed significantly from polysomnography, by about 17 minutes. Fitbit models without a staging algorithm overestimated it by roughly 7 to 67 minutes, while Fitbit models with staging showed no significant difference.
Which wearable is closest to a sleep lab?
No single answer holds across studies. In one 35-adult night study the Oura Ring did not differ from polysomnography on wake, light, deep or REM sleep. In a six-device study the Apple Watch Series 8 had the highest kappa at 0.53.
Do REM and deep sleep on a wearable predict health outcomes?
In a long-term wearable dataset of 6,785 people, REM and deep sleep were inversely associated with incident atrial fibrillation. That is an association, and it does not show the stage estimates were accurate.
KnowYourPrime Evidence Registry: Finding KYP-ER-0007: Wearable Sleep Tracking Accuracy Against a Sleep Lab · Finding KYP-ER-0016: Wearable Sleep Stage Accuracy
Sources
- A performance validation of six commercial wrist-worn wearable sleep-tracking devices for sleep stage scoring compared to polysomnography.
- Accuracy of Three Commercial Wearable Devices for Sleep Tracking in Healthy Adults. [funding listed from Oura]
- Validity and reliability of the Oura Ring Generation 3 (Gen3) with Oura sleep staging algorithm 2.0 (OSSA 2.0) when compared to multi-night ambulatory polysomnography. [conflict statement: three authors report financial support from Oura]
- Performance of consumer wrist-worn sleep tracking devices compared to polysomnography: a meta-analysis.
- Accuracy of Wristband Fitbit Models in Assessing Sleep: Systematic Review and Meta-Analysis.
- Sleep assessment using EEG-based wearables - A systematic review.
- Sleep patterns and risk of chronic disease as measured by long-term monitoring with commercial wearable devices in the All of Us Research Program. [author affiliations include Google; funding listed from Google]
- The visual scoring of sleep in adults.