You can track REM sleep using wearable devices like smartwatches, under-mattress sensors, smartphone apps, or clinical sleep studies. Each method works differently and varies widely in accuracy. For most people, a wrist-worn tracker that measures heart rate is the most practical starting point, though no consumer device matches the precision of a clinical sleep lab.
What REM Sleep Looks Like to a Sensor
REM sleep has a distinct biological signature. Your brain produces low-voltage, mixed-frequency electrical activity (similar to when you’re awake), your eyes dart rapidly beneath your eyelids, and the muscles in your chin and body go nearly limp. In a clinical sleep study, technicians identify REM by reading all three of these signals simultaneously: brain waves from electrodes on the scalp, eye movements from sensors near the eyes, and muscle tone from a chin sensor.
Your heart also behaves differently during REM. Heart rate becomes more variable and slightly faster compared to deep sleep, with a specific shift in the rhythm patterns that sensors can detect. This cardiac signature is the key that unlocks REM tracking for consumer devices, since most of them can’t read brain waves or eye movements directly.
How Wrist-Worn Trackers Estimate REM
Smartwatches and fitness bands from companies like Fitbit, Garmin, Apple, and Oura use a small optical sensor pressed against your skin. This sensor shines light into your wrist and measures how blood flow changes with each heartbeat. From that pulse signal, the device calculates your heart rate and heart rate variability throughout the night.
Because REM sleep shifts your heart rate variability in a predictable direction (more low-frequency fluctuation, less high-frequency), algorithms can use these patterns to estimate when you’re in REM versus deep or light sleep. The device also picks up wrist movement through a built-in accelerometer, which helps distinguish sleep from wakefulness. Together, these two data streams feed into a classification algorithm that assigns a sleep stage to each 30-second window of your night.
The advantage of wrist-worn trackers is convenience. You charge them, wear them, and check your results in the morning. They require no setup and work anywhere you sleep.
How Accurate Consumer Trackers Really Are
Consumer sleep trackers are good at detecting when you’re asleep but less reliable at telling you exactly what stage you’re in. A study comparing seven popular devices against clinical polysomnography found that all of them correctly identified sleep epochs at least 93% of the time. But their ability to detect wake periods was much weaker, with specificity ranging from just 18% to 54% depending on the device. The Garmin models tested performed worst at detecting wakefulness (18-19% specificity), while the Fitbit Alta HR scored highest at 54%.
Sleep stage classification (separating REM from light sleep from deep sleep) was described as “mixed” across devices. Performance also dropped on nights with poor or disrupted sleep, which is exactly when accurate tracking matters most. The takeaway: your tracker’s REM numbers are useful for spotting broad trends over weeks and months, but treating any single night’s data as precise is a mistake. If your tracker says you got 1 hour and 42 minutes of REM, the real number could be meaningfully different.
Under-Mattress and Bedside Sensors
If you don’t want to wear anything to bed, pad-style sensors that slide under your mattress offer another option. These devices use a technique called ballistocardiography: they detect the tiny vibrations your body makes with each heartbeat as blood pulses through your vessels. From those micro-movements, the sensor extracts your heart rate and heart rate variability, then runs the same kind of staging algorithm that wrist devices use.
Products like the Withings Sleep Tracking Mat and Eight Sleep work this way. Some bed-leg sensors can even pick up these cardiac vibrations through the bed frame itself. The accuracy is broadly comparable to wrist-worn devices, since both rely on the same underlying cardiac signals. The main advantage is comfort: nothing on your body, nothing to charge and strap on each night.
Smartphone Apps and Their Limits
Several apps claim to track your sleep stages using just your phone. They work through three different approaches, and the differences in accuracy are dramatic.
Audio-based apps like SleepRoutine use your phone’s microphone to analyze breathing sounds throughout the night. Changes in breathing regularity, pitch, and amplitude reflect shifts in your nervous system that correlate with sleep stages. In a validation study against clinical polysomnography, SleepRoutine achieved the highest REM detection accuracy among all consumer trackers tested, with an F1 score of 0.76 (where 1.0 would be perfect). It also performed best at detecting wakefulness.
Sonar-based apps like SleepScore emit ultrasonic pulses from your phone’s speaker and track chest movements to measure breathing effort. This approach scored considerably lower for REM detection, with an F1 score of just 0.34, and showed significant bias in estimating how long it took users to fall asleep.
Movement-based apps like Pillow use your phone’s accelerometer (placed on the mattress) to detect body movements. This was the least accurate method by a wide margin, with a REM F1 score of only 0.14. The app was heavily biased toward labeling everything as deep sleep, because motion data alone simply can’t distinguish between sleep stages with any reliability. If your app only uses movement, its REM numbers are close to meaningless.
What Healthy REM Sleep Looks Like
REM sleep makes up about 20 to 25% of total sleep time in adults. For someone sleeping seven to eight hours, that works out to roughly 90 to 120 minutes of REM per night. But REM doesn’t arrive in one block. It cycles throughout the night, alternating with lighter and deeper stages in roughly 90-to-120-minute cycles.
Your first REM period of the night is typically the shortest, sometimes lasting only 1 to 5 minutes. Each subsequent REM period gets longer, with the final ones lasting up to an hour. This is why cutting your sleep short by even an hour tends to disproportionately reduce REM: most of it is concentrated in the last third of the night. If your tracker consistently shows REM below 15% of total sleep time, that pattern is worth paying attention to.
Getting the Most From Your Tracker
Whichever device you use, a few practices will improve the quality of your data. Wear wrist trackers snug enough that the optical sensor maintains good skin contact, but not so tight that it’s uncomfortable. A loose band lets light leak in and degrades the heart rate signal. Keep the sensor clean, since sweat and grime interfere with the optical reading.
For under-mattress sensors, placement matters. Position the pad near your chest, since that’s where cardiac vibrations are strongest. If you share a bed, some pads may pick up your partner’s movements, which adds noise to the data.
For phone-based apps, audio-based options outperform motion-based ones by a significant margin. If you go this route, place your phone on the nightstand near your head (not under your pillow) and make sure the microphone isn’t muffled. Background noise from fans or white noise machines can interfere with breathing analysis.
Most importantly, focus on trends rather than individual nights. A single night’s REM reading from any consumer device carries substantial error. But if your average REM percentage drops noticeably over two or three weeks, or shifts after a lifestyle change, that pattern is more likely to reflect something real. The value of tracking isn’t in the precision of any one number; it’s in watching how your sleep architecture responds over time to how you live.