Labs Score REM From Brain Waves, Wearables Estimate It

REM sleep is scored in a lab from brain waves, eye movements and chin muscle tone, not from dreaming. Wearables estimate it from movement and pulse. Against polysomnography, the Oura Ring Gen3 underestimated REM by 4.1 to 5.6 minutes in 96 adults, with 90.6% accuracy on REM epochs. A review of eight studies found WHOOP differed from the lab on REM by 21.0 minutes and Fitbit Charge 4 by 4.0. Across six devices, agreement on all stages ran 50% to 65%.

What is REM sleep?

A stage scored from body signals, not from dream reports. The American Academy of Sleep Medicine scoring rules recommend three EEG derivations, two electro-oculography derivations and one chin EMG. Scoring uses 30-second epochs.

That means brain waves, eye movements and muscle tone. The task force defined the phenomena of REM sleep and simplified the rules for its onset and end. It noted that studies are needed to test the reliability of the new rules.

A consumer device records none of these signals in the same way. Most wrist and ring devices estimate the stage from movement and pulse. See how wearables track sleep stages.

What is REM sleep proposed to do?

Balance the work of deep sleep in memory. A review describes memory replay during slow-wave sleep, which favors turning episodic memory into longer-lasting cortical memory. REM sleep that follows might balance local synaptic rescaling with a global synaptic renormalization process.

This is a proposed mechanism. The review gives no minutes target for REM. For deep sleep amounts, see how many hours of deep sleep are needed.

How close is a wearable's REM estimate to a lab?

It depends on the device. The Oura Ring Gen3 was tested in 96 healthy adults across up to three polysomnography nights. It underestimated REM sleep by 4.1 to 5.6 minutes. Sleep staging accuracy ranged from 75.5% for light sleep to 90.6% for REM.

In a single night with 35 adults, the Oura Ring did not differ from polysomnography for REM sleep. The Fitbit Sense 2 and Apple Watch Series 8 differed on light and deep sleep instead.

A review of eight studies of recent wearables found WHOOP closest on total sleep time and deep sleep but furthest on REM, at 21.0 minutes. Fitbit Charge 4 was closest on REM, at 4.0 minutes, with 86.5% sensitivity for REM.

Six devices were tested in one lab night with 53 adults. Agreement on all stages combined was 61% for Oura, 60% for WHOOP, 53% for Apple Watch and 50% for Garmin. That is not REM accuracy alone.

Do devices that read brain waves do better?

In one review, they did. A systematic review covered 60 papers on 34 EEG-based wearables. Its 42 validation studies were conducted across diverse populations and showed consistently high accuracy in sleep staging.

These are headbands and similar devices that record EEG. They are a different category from watches and rings.

What does this not settle?

The wearable results come from healthy adults over one to three lab nights. They say little about how a device scores REM in people with sleep disorders or at home over months. Each result is tied to one device and firmware version.

The lab standard is itself imperfect. The scoring task force noted the reliability of the rules still needed testing.

Common questions