The HRV-Sleep Connection: What Heart Rate Variability During Sleep Actually Tells You

The Most Useful Number You Probably Misunderstand

Heart rate variability is one of the most useful and most misunderstood numbers your wearable shows you.

Open your Whoop, Oura, or Apple Watch and you'll see a number between roughly 20 and 100 milliseconds, usually labeled "recovery" or "readiness." Most people glance at it, decide it's good or bad, and move on.

That's a missed opportunity. HRV during sleep is one of the cleanest windows into your autonomic nervous system you'll ever get. It also has more measurement quirks and bad consumer-app interpretations than almost any other biometric. Let's get specific.

Key Facts

• HRV is the variation in time between successive heartbeats, measured in milliseconds, and reflects autonomic nervous system balance (Task Force of the European Society of Cardiology, 1996).¹
• Lower nocturnal HRV is associated with elevated cardiovascular risk. A meta-analysis of 21 studies found that low HRV predicted a roughly 32-45% increase in first cardiovascular event risk (Hillebrand et al., 2013).²
• HRV varies dramatically across sleep stages. Parasympathetic tone, and HRV, peak during slow-wave sleep, while REM is marked by sympathetic surges (Vanoli et al., 1995).³
• RMSSD is the time-domain metric most consumer apps use because it's robust over short windows and primarily reflects vagal (parasympathetic) tone (Shaffer & Ginsberg, 2017).⁴
• Wrist-based PPG sensors typically sample at 25-100 Hz, well below the 250-1000 Hz that ECG uses for clinical HRV, which introduces meaningful error (Nelson et al., 2020).⁵
• Komori does not measure HRV. Komori reports position changes, gross movement, bed presence, and bed exit — it does not measure heart rate, HRV, or breathing rate.

What HRV Actually Measures

Your heart doesn't beat like a metronome. Even when you feel completely still, the time between beats changes by milliseconds. Sometimes the gap is 850 ms, sometimes 920 ms, sometimes 790 ms. That variation is HRV.

The variation isn't random. It's controlled by the autonomic nervous system, which has two branches:

• Sympathetic ("fight or flight"). Speeds the heart up, makes the rhythm more uniform.
• Parasympathetic ("rest and digest"). Slows the heart, allows more variability between beats.

When parasympathetic activity dominates, beat-to-beat intervals stretch and contract more freely and HRV goes up. When sympathetic activity dominates, HRV drops.

HRV isn't really about the heart. It's a proxy for autonomic balance, which reflects how stressed, recovered, sick, or rested your body is right now (Tobaldini et al., 2013).⁶

Why Nocturnal HRV Is More Useful Than Daytime HRV

Here's the problem with daytime HRV: almost everything changes it. Coffee, a flight of stairs, a tense email, standing up, speaking on Zoom. Daytime HRV is so noisy that interpreting it requires standardized conditions you'll never actually replicate.

Nocturnal HRV, especially during deep sleep, gives you something cleaner. You're not eating, not moving much, not thinking about your boss. What's left is closer to your autonomic baseline.

This is why every serious recovery platform built its scoring around overnight HRV. Whoop's "recovery" score uses HRV from the last slow-wave period. Oura averages across the night with a heavy emphasis on early morning. The signal-to-noise ratio is just better at night (Stein & Pu, 2012).⁷

HRV Across Sleep Stages

Sleep isn't a uniform state, and your autonomic nervous system knows it.

N1 and N2 (light NREM). Parasympathetic tone starts to rise, sympathetic falls. HRV climbs above wake levels.

N3 (slow-wave / deep sleep). Maximum parasympathetic dominance. HRV is at its highest of the entire 24-hour cycle. This is the window most associated with cardiovascular recovery (Vanoli et al., 1995).³

REM sleep. Things flip. Despite muscle paralysis, the brain is highly active and sympathetic activity spikes. Heart rate becomes more variable in a different way, with sudden surges and drops. Some researchers call REM "autonomic storm" sleep. It's also the window when most cardiovascular events that happen during sleep tend to cluster (Bonnet & Arand, 1997).⁸

This is why averaging HRV across the entire night can hide more than it reveals. A night with strong, long deep-sleep periods produces a totally different HRV profile than a night dominated by REM and light sleep, even if the average looks identical.

The Metrics That Matter

You'll see five HRV metrics most often. Some are useful for sleep, some less so.

Time-domain metrics

RMSSD (Root Mean Square of Successive Differences). Take the difference between each pair of consecutive beats, square it, average those squares, take the square root. RMSSD captures fast beat-to-beat changes, which are almost entirely vagal (parasympathetic). For sleep recovery, RMSSD is the metric most worth paying attention to. It's robust on short recordings, less polluted by respiration than spectral measures, and what most consumer wearables actually report under the label "HRV" (Shaffer & Ginsberg, 2017).⁴

SDNN (Standard Deviation of NN intervals). Captures the total variability across a recording, including slow drifts. SDNN over a full 24-hour ECG is a strong long-term cardiovascular risk marker. SDNN over 5 minutes is much weaker. For nightly tracking, SDNN tells you less than RMSSD about acute parasympathetic state (Task Force, 1996).¹

Frequency-domain metrics

HF (High Frequency, 0.15-0.4 Hz). Power in this band is driven by respiration-linked vagal modulation. HF is a direct readout of parasympathetic activity. Useful, but only if your tracker has a clean enough signal to compute spectra reliably.

LF (Low Frequency, 0.04-0.15 Hz). Power in this band reflects a mix of sympathetic and parasympathetic activity, plus baroreflex influence. LF is harder to interpret cleanly. It's not a pure sympathetic marker, despite being called that for decades.

LF/HF ratio. Marketed as a "sympathovagal balance" number. Most physiologists are skeptical. The math behind it assumes things about the autonomic system that don't hold up. Treat LF/HF with caution. It's a popular number, not a clean one (Billman, 2013).⁹

If you're ranking these for tracking sleep recovery on a consumer device, RMSSD belongs at the top, HF second, and the rest a distant tie.

How Consumer Trackers Measure HRV (And Where They Go Wrong)

Almost every consumer wearable uses photoplethysmography (PPG), an LED that bounces light off your skin to track blood volume changes with each pulse. From the waveform, the device infers each beat's timing and computes HRV from those intervals.

PPG is not ECG. ECG measures the heart's electrical activity directly, and the QRS complex gives a sharp timing reference. PPG measures a downstream pressure wave, smeared by vessel elasticity. The PPG pulse peak arrives hundreds of milliseconds after the electrical beat, and that delay shifts with blood pressure.

Three things that make consumer HRV worse than people think:

1. Sample rate. Clinical ECG samples at 250-1000 Hz. Consumer PPG often samples at 25-100 Hz. At 25 Hz, each sample is 40 ms apart, and RMSSD for healthy adults is often in that same 30-50 ms range. Studies validating wrist PPG against ECG have found errors of 5-15 ms in RMSSD even under ideal conditions (Nelson et al., 2020).⁵

2. Motion artifacts. A wrist moves, even during sleep. Every micro-movement disturbs the optical path. Devices reject windows with motion, but the rejected windows cluster during REM and lighter sleep, exactly the stages where HRV behavior is most interesting.

3. Fixed sampling windows. Most wearables don't track HRV continuously. They sample a few minutes during presumed deep sleep, or average the last hour before wake. The number you see is a snapshot, not your real overnight HRV.

Finger PPG (Oura) tends to be cleaner than wrist PPG (Apple Watch, Fitbit, Garmin) because the finger has a stronger pulse signal and less motion. Chest-strap ECG (Polar H10) is the consumer gold standard (Pietilä et al., 2018).¹⁰

What HRV Trends Actually Predict

A single night's HRV doesn't mean much. The trend is where the signal lives.

Overtraining. A sustained drop in nocturnal RMSSD is one of the earliest markers that a hard training block is overshooting recovery. Endurance athletes have used HRV trending for decades for exactly this reason (Plews et al., 2013).¹¹

Illness onset. Several studies have found that nocturnal HRV starts dropping 1-3 days before symptoms of viral illness appear. Whoop's COVID research showed this pattern across their user base (Mishra et al., 2020).¹²

Sleep quality. Lower nocturnal HRV correlates with fragmented sleep, more light stages, and more arousals. Higher HRV nights tend to map onto deeper, more consolidated sleep.

Cardiovascular risk. This one is for the long view. Hillebrand et al.'s meta-analysis found low HRV predicted a 32-45% increase in first cardiovascular event risk, even after adjusting for traditional factors.² Persistently low nocturnal HRV (over months, not days) is a signal worth raising with your cardiologist.

Stress and life load. Travel, alcohol, late meals, and emotional stress all compress HRV. If your trend has been low for two weeks and you've been drinking more wine and arguing with your in-laws, that's not a coincidence.

Where Komori Stands on HRV

Time for the honest part.

Komori does not measure HRV. It can't. Our 60 GHz radar tracks position, gross motion, and bed presence. It resolves whether a body is in the bed, what posture it's in, and when it shifts. It cannot resolve the sub-millimeter chest-wall pulse from each heartbeat with the precision needed for valid beat-to-beat intervals.

Research-grade radar systems do attempt cardiac BCG-style measurement, and the results are mixed. None currently produce HRV numbers we'd trust enough to put on a consumer product. So we don't. (A separate research-grade Pro sensor suite — under research/IRB protocol, with its own FDA dialogue — is the channel where cardiorespiratory metrics may eventually be evaluated. That is not the consumer Komori product.)

If you want HRV data, a finger ring (Oura), a chest strap (Polar), or an under-mattress BCG sensor will give you the cardiac signal. Komori is designed to handle the contactless side — position changes, gross movement, true bed exit — without touching the wrist or finger. The two layers complement rather than replace each other.

How HRV Measurement Methods Compare

The Bottom Line on HRV

HRV is a great signal if you trust the source.

If you're using a wrist PPG that samples at 25 Hz and only snapshots HRV for a few minutes a night, treat the number as a rough trend, not as truth. If you're using a finger ring or a chest strap, you can take the trend more seriously and use it to inform training, recovery, and sleep choices.

Whatever tracker you use, look for trends, not single nights. Compare yourself to your own seven- and thirty-day baselines, not to internet averages. If your nocturnal HRV is consistently low and dropping, especially in combination with rising resting heart rate and worse sleep, that's worth a conversation with your cardiologist, not a new app.

And if your HRV tracker is telling you you slept badly, but you don't know whether you spent the night on your back snoring or whether you got out of bed at 3 AM, you're missing context. That's the gap a contactless monitor like Komori fills. We've written more about why movement during sleep matters, why you wake up at 3 AM, and what your body is actually doing during the night in our overview of sleep stages.

For more on the data behind contactless monitoring, see our research page and how Komori works.

---

Footnotes

1. Task Force of the European Society of Cardiology and the North American Society of Pacing and Electrophysiology. "Heart rate variability: standards of measurement, physiological interpretation, and clinical use." Circulation 93, no. 5 (1996): 1043-1065. https://doi.org/10.1161/01.CIR.93.5.1043 ↩ ↩²

2. Hillebrand, S. et al. "Heart rate variability and first cardiovascular event in populations without known cardiovascular disease: meta-analysis and dose-response meta-regression." Europace 15, no. 5 (2013): 742-749. https://doi.org/10.1093/europace/eus341 ↩ ↩²

3. Vanoli, E. et al. "Heart rate variability during specific sleep stages: a comparison of healthy subjects with patients after myocardial infarction." Circulation 91, no. 7 (1995): 1918-1922. https://doi.org/10.1161/01.CIR.91.7.1918 ↩ ↩²

4. Shaffer, F. and Ginsberg, J.P. "An overview of heart rate variability metrics and norms." Frontiers in Public Health 5 (2017): 258. https://pmc.ncbi.nlm.nih.gov/articles/PMC5624990/ ↩ ↩²

5. Nelson, B.W. et al. "Accuracy of consumer wearable heart rate measurement during an ecologically valid 24-hour period: intraindividual validation study." JMIR mHealth and uHealth 8, no. 3 (2020): e10828. https://pmc.ncbi.nlm.nih.gov/articles/PMC6431828/ ↩ ↩²

6. Tobaldini, E. et al. "Heart rate variability in normal and pathological sleep." Frontiers in Physiology 4 (2013): 294. https://pmc.ncbi.nlm.nih.gov/articles/PMC3797399/ ↩

7. Stein, P.K. and Pu, Y. "Heart rate variability, sleep and sleep disorders." Sleep Medicine Reviews 16, no. 1 (2012): 47-66. https://doi.org/10.1016/j.smrv.2011.02.005 ↩

8. Bonnet, M.H. and Arand, D.L. "Heart rate variability: sleep stage, time of night, and arousal influences." Electroencephalography and Clinical Neurophysiology 102, no. 5 (1997): 390-396. https://doi.org/10.1016/S0921-884X(96)96070-1 ↩

9. Billman, G.E. "The LF/HF ratio does not accurately measure cardiac sympatho-vagal balance." Frontiers in Physiology 4 (2013): 26. https://pmc.ncbi.nlm.nih.gov/articles/PMC3576706/ ↩

10. Pietilä, J. et al. "Evaluation of the accuracy and reliability for photoplethysmography based heart rate and beat-to-beat detection during daily activities." EMBEC & NBC 2017 (2018): 145-148. https://doi.org/10.1007/978-981-10-5122-7_37 ↩

11. Plews, D.J. et al. "Training adaptation and heart rate variability in elite endurance athletes: opening the door to effective monitoring." Sports Medicine 43, no. 9 (2013): 773-781. https://doi.org/10.1007/s40279-013-0071-8 ↩

12. Mishra, T. et al. "Pre-symptomatic detection of COVID-19 from smartwatch data." Nature Biomedical Engineering 4, no. 12 (2020): 1208-1220. https://doi.org/10.1038/s41551-020-00640-6 ↩