HRV

Heart rate variability (HRV) is the variation in time between consecutive heartbeats, measured in milliseconds. At rest, most of the fast beat-to-beat variation comes from the vagus nerve, so HRV at or above your own usual level generally signals good recovery, and a drop below it often follows hard training, short sleep, alcohol or the start of an illness.

HRV is one of the few recovery signals a watch can read every night without you doing anything. It is also one of the easiest to misread. The number depends on which formula produced it, how long the recording ran, and what your body was doing at the time. A reading of 35 ms can be healthy for one person and a warning for another. This entry covers what the number measures, how SDNN and RMSSD differ, how Apple Watch records HRV and how accurate it is, what counts as normal, and why Titan scores your HRV against your own nights instead of a population chart. The wearable data quality guide compares HRV accuracy with every other wearable metric.

01What the number measures

At 60 beats per minute your heart does not fire exactly every 1,000 ms. The gaps might run 990 ms, then 1,040 ms, then 1,010 ms. Your heart has a built-in pacemaker, and two branches of the autonomic nervous system adjust its timing. The parasympathetic branch works through the vagus nerve and slows the heart within a single beat. The sympathetic branch speeds it up, more slowly. When you breathe in, vagal braking eases and the next few beats come slightly faster. When you breathe out, the brake returns. That rhythm is called respiratory sinus arrhythmia, and it accounts for much of the fast variation in a resting recording.

HRV is a statistic computed on that series of beat-to-beat intervals. Shaffer and Ginsberg's 2017 review in Frontiers in Public Health calls RMSSD, one of those statistics, "the primary time-domain measure used to estimate the vagally mediated changes reflected in HRV." Laborde and colleagues' 2017 methods guide describes HRV as an index of cardiac vagal tone, the parasympathetic contribution to heart rate control. A well-rested, aerobically fit person at rest usually has strong vagal activity and more variable beat timing. Stress, fatigue and illness shift control toward the sympathetic side, and the intervals become more uniform.

HRV is a readout of how the nervous system is steering the heart. Fitness, stress and recovery all change that steering, which is why HRV tracks them. It is also why a dozen unrelated things can move the number.

02SDNN and RMSSD

HRV is a family of statistics. Two of them account for nearly every number you will see on a wearable. Both start from the same list of intervals between normal beats.

SDNN is the standard deviation of those intervals. It asks how spread out the intervals are around their average, so it captures fast beat-to-beat changes and slow drifts together. Apple Health stores HRV as SDNN.

RMSSD is the root mean square of successive differences. Subtract each interval from the next, square the differences, average them and take the square root. It asks how much each beat differs from the one before it, which isolates the fast, mostly vagal component. Oura, WHOOP, Garmin and Polar report RMSSD-based HRV, and so does most research on monitoring athletes.

The two respond to different kinds of variation. Take two six-beat series with nearly the same average interval.

Illustrative values computed from the intervals shown.

SDNN rates the steady drift as slightly more variable than the back-and-forth series, 37.4 ms against 33.3 ms. RMSSD rates it three times less variable, because each beat differs from its neighbor by only 20 ms. The first series looks like a relaxed heart following the breath. The second looks like heart rate settling after you sit down.

Recording length also affects the two formulas differently. SDNN grows as a recording gets longer, because a longer window catches more slow drift from posture, temperature regulation and the day-night cycle. Shaffer and Ginsberg state that "it is inappropriate to compare metrics like SDNN when they are calculated from epochs of different length." RMSSD holds up far better in short windows. In 3,387 adults, Munoz and colleagues (2015) found that RMSSD from a single 10-second ECG correlated with a 4 to 5 minute reference at r = 0.85 to 0.86 with little bias. SDNN from the same 10 seconds correlated at r = 0.76 and carried a larger bias. At every recording length they tested, RMSSD outperformed SDNN.

No fixed factor converts SDNN into RMSSD, because the ratio depends on how much of your variation is fast and how much is slow. Pick one, and compare it only with itself.

03How Apple Watch records HRV

Apple Watch reads your pulse optically. Green LEDs shine into the wrist, photodiodes measure the reflected light, and the watch times the gaps between pulse peaks. HRV is computed from that series of pulse-to-pulse intervals. Movement smears the pulse wave, so the watch waits for stillness. Apple's support page says HRV "is sampled while you're still, so the number of measurements you see will vary according to your activity level." You do not choose when most readings happen.

Apple's HealthKit documentation is explicit about the formula. "While there are multiple ways of computing HRV, HealthKit uses SDNN heart rate variability, which uses the standard deviation of the inter-beat (RR) intervals between normal heartbeats." Health also stores heartbeat series, the beat-to-beat timing behind a reading, which lets an app such as Titan compute RMSSD from the same beats. iOS 27 adds a separate heartRateVariabilityRMSSD type to HealthKit. Apple's September 2026 announcement says Apple Watch Series 12 and Ultra 4 measure HRV "as often as every five minutes" and report two variants. Recovery HRV is meant for daily stress and recovery, and overall HRV for broader cardiovascular health.

A reading taken during sleep and one taken at your desk after coffee capture different autonomic states. The Apple Watch HRV guide covers the three kinds of moments the watch records HRV and why the daytime numbers jump around.

How accurate the watch is

Accuracy depends on the conditions. Hernando and colleagues (2018) compared Apple Watch with a Polar H7 chest strap in 20 healthy volunteers who sat still for two 5-minute stages, one relaxed and one during a Stroop test. Across 12,109 paired intervals, concordance was 0.989 at rest and 0.977 under mild stress. About 10 percent of the watch's intervals were missing, in gaps averaging 6 seconds, yet SDNN and RMSSD did not differ significantly from the strap.

Across a full night the picture is looser. Miller and colleagues (2022) put six wearables on 53 adults in a sleep laboratory with ECG as the reference. The Apple Watch Series 6, read through a third-party sleep app, underestimated RMSSD by 9.6 ms on average. Its absolute error averaged 22.5 ms, its 95 percent limits of agreement spanned ±55.2 ms, and its intraclass correlation was 0.67, which the authors rated good. The error was proportional. The watch read high for people with low ECG values and low for people with high values. WHOOP 3.0, the one device that supplied raw beat intervals, agreed almost perfectly, with limits of ±7.6 ms. Kinnunen and colleagues (2020) found similar near-perfect nightly agreement for the Oura ring against ECG, r² = 0.980 with a mean bias of −1.2 ms.

Two practical points follow. An Apple Watch is good enough to follow your own HRV from night to night, provided the conditions stay the same. Its absolute values do not transfer to another device, so a switch from ring to watch, or from one watch to another, means building a new baseline.

04Why Titan's Recovery uses night HRV

Titan's Recovery score reads HRV from one window only, the night that ended this morning. Sleep is the closest thing to a controlled recording that everyday life offers. Posture stays constant for hours, breathing is unforced, and there is no coffee, conversation or commute in the signal. The research supports the choice. Nuuttila and colleagues (2022) found nocturnal LnRMSSD highly reliable between comparable nights in recreational runners, with intraclass correlations of 0.92 to 0.97, and a maximal 3,000 m run lowered it most consistently in the full-night and first-4-hour segments. In a 2024 follow-up, the same group found nocturnal and morning HRV correlated at baseline but responded differently to two weeks of overload. The nocturnal responses were associated with changes in running performance.

This is how Titan builds each night's HRV, from the current iOS code.

1. The window is your main sleep session that ended this morning. If Health has no sleep session, Titan reads 22:00 the evening before through 10:00.
2. Titan drops readings outside 3 to 300 ms and exact duplicates, then keeps one device's readings. It prefers the device that recorded your sleep.
3. Titan averages the logarithm of each reading, which gives the geometric mean. HRV distributions are skewed, and one unusually high reading pulls an ordinary average up more than it should. Researchers log-transform RMSSD for the same reason (Plews et al., 2012).
4. Titan pairs that value with your sleeping heart rate, the lowest 30-minute average of your heart rate during the same window.

Your baseline is the median and median absolute deviation of your measured nights in your Recovery window, 60 nights by default. Last night is excluded. A baseline needs at least 7 measured nights covering at least 25% of the window, so 15 nights on the default window. It then expresses last night's HRV and sleeping heart rate as distances from those medians, in units of your own typical night-to-night spread, and combines them on a fixed curve. A night exactly at both baselines scores 56. HRV one typical step above baseline scores 85, and one step below scores 22. Sleeping heart rate counts about half as much, in the opposite direction. The same night always gets the same score, and the score freezes at noon. Low is 0 to 39, Moderate 40 to 74 and High 75 to 100.

Only readings inside that overnight window reach Recovery, so daytime readings never do. A reading from a Mindfulness session gets no special treatment and counts only if it falls inside the window, as one reading among many. The first reading after you wake can nudge an estimated score only on mornings with no usable overnight data. If your watch recorded sleep but no HRV, Titan marks the score Estimated and carries HRV forward from your last 7 measured nights. Estimated nights never enter your baseline. Titan uses SDNN by default. Use RMSSD for HRV, under Settings > Recovery & Sleep, computes RMSSD from the beat-to-beat heartbeat series instead, and Titan keeps a separate history for each method so the two scales never mix. The Recovery score and baselines help articles cover the screens.

05Normal HRV ranges

Population data exist, but each table belongs to one metric, one recording length and one population. The most cited short-term reference is Nunan and colleagues' 2010 systematic review of 44 studies and 21,438 healthy adults, as tabulated by Shaffer and Ginsberg.

Age moves HRV more than almost anything else. Tegegne and colleagues (2020) derived reference values for RMSSD from 10-second resting ECGs in 84,772 healthy Dutch adults and adolescents.

Source: Tegegne et al. (2020), Table 1.

The median falls by about 60 percent between the early twenties and the early sixties. Women aged 20 to 45 had medians about 5 ms higher than men. The healthy range inside a single age band is enormous. For men aged 40 to 44 it runs from 8.1 to 105.5 ms, a thirteenfold spread.

The cutoffs you see online, SDNN below 50 ms unhealthy, 50 to 100 ms compromised and above 100 ms healthy, come from 24-hour Holter recordings. Kleiger and colleagues (1987) found that among 808 heart attack survivors, those with 24-hour SDNN below 50 ms had 5.3 times the mortality of those above 100 ms. A 24-hour SDNN includes a full day and night of heart rate swings. A watch reading covers a short window at rest. Holding a 40 ms watch reading against a Holter cutoff tells you nothing about your heart. The HRV by age article has the full tables, the sex differences and the data on master athletes.

06Why your own baseline matters more

Population tables tell you where you sit among strangers measured a certain way. They cannot tell you whether last night was a good night for you. Two healthy people of the same age can differ tenfold, and neither number is a problem. What carries information is a departure from your own usual level.

Plews and colleagues (2013) reviewed HRV monitoring in elite endurance athletes and concluded that longitudinal monitoring is required to understand each athlete's "unique individual HRV fingerprint." They recommended averaging over several days instead of reacting to single readings. In an earlier case study of two elite triathletes, the 7-day rolling average of log RMSSD in the athlete who became non-functionally overreached fell steadily toward the race where that athlete performed poorly, while the other athlete's average held flat (Plews et al., 2012).

Using the baseline to steer training has some support. Kiviniemi and colleagues (2007) randomized 26 moderately fit men to a fixed plan or to HRV-guided training, where a morning HRV below the athlete's 10-day mean minus one standard deviation, or falling for two days, meant an easy day or rest. Peak oxygen uptake rose from 56 to 60 ml/kg/min in the HRV-guided group and did not change significantly in the fixed-plan group. A 2021 meta-analysis by Manresa-Rocamora and colleagues put the broader evidence in proportion. HRV-guided training improved vagal HRV more than predefined training, and its advantages for aerobic fitness and endurance performance were small and not statistically significant.

Buchheit (2014) adds the rule that makes a baseline useful. Judge each change against the measure's normal error and the smallest change that matters for you. Titan's median absolute deviation does exactly that. A 5 ms drop is a large move for someone whose HRV rarely varies and ordinary for someone whose HRV swings every night. The HRV and training readiness guide covers how to turn that comparison into training decisions.

07What moves HRV

Most of the forces that lower HRV are ones you can see coming.

Hard training. A maximal effort lowers HRV the following night. In Nuuttila's 2022 study, LnRMSSD dropped after a 3,000 m time trial compared with the night before. A few days of reduced HRV after a hard block are expected. A downward trend that continues through easier days can be an early sign of overreaching.

Alcohol. Pietilä and colleagues (2018) compared nights with and without drinking in 4,098 Finnish employees. An HRV-derived recovery index for the first three hours of sleep fell by 9.3, 24.0 and 39.2 percentage points after low, moderate and high doses. Being young or physically active did not protect against the effect.

The menstrual cycle. A meta-analysis of 37 studies found vagal HRV falls from the follicular to the luteal phase, with a standardized effect of d = −0.39 (Schmalenberger et al., 2019). The largest drops appeared in the days before menstruation.

Age. RMSSD falls steadily until about 60 and then levels off (Tegegne et al., 2020). It changes over years, so it does not explain a bad week.

Illness, heat, travel and short sleep. These shift autonomic balance toward sympathetic control and often lower HRV. Resting heart rate usually rises at the same time, which is why Titan scores the two together.

A few habits keep the measurement itself from causing false alarms. Wear the watch snugly to bed, keep the same device, and stay on one formula for the life of your baseline. When HRV drops for a single night, check for an obvious cause before changing your training. When it stays below baseline for several nights while sleeping heart rate climbs, take the signal seriously.

08References

1. Apple (2026, September 9). Apple advances health and fitness capabilities using Apple Intelligence. Apple Newsroom. apple.com/newsroom
2. Apple. heartRateVariabilityRMSSD. HealthKit developer documentation. developer.apple.com
3. Apple. heartRateVariabilitySDNN. HealthKit developer documentation. developer.apple.com
4. Apple. Monitor your heart rate with Apple Watch. Apple Support. support.apple.com/en-us/HT204666
5. Buchheit M (2014). Monitoring training status with HR measures: do all roads lead to Rome? Frontiers in Physiology 5:73. doi.org/10.3389/fphys.2014.00073
6. Hernando D, et al. (2018). Validation of the Apple Watch for heart rate variability measurements during relax and mental stress in healthy subjects. Sensors 18(8):2619. doi.org/10.3390/s18082619
7. Kinnunen H, et al. (2020). Feasible assessment of recovery and cardiovascular health: accuracy of nocturnal HR and HRV assessed via ring PPG in comparison to medical grade ECG. Physiological Measurement 41(4):04NT01. doi.org/10.1088/1361-6579/ab840a
8. Kiviniemi AM, et al. (2007). Endurance training guided individually by daily heart rate variability measurements. European Journal of Applied Physiology 101(6):743-751. doi.org/10.1007/s00421-007-0552-2
9. Kleiger RE, et al. (1987). Decreased heart rate variability and its association with increased mortality after acute myocardial infarction. American Journal of Cardiology 59(4):256-262. doi.org/10.1016/0002-9149(87)90795-8
10. Laborde S, Mosley E, Thayer JF (2017). Heart rate variability and cardiac vagal tone in psychophysiological research: recommendations for experiment planning, data analysis, and data reporting. Frontiers in Psychology 8:213. doi.org/10.3389/fpsyg.2017.00213
11. Manresa-Rocamora A, et al. (2021). Heart rate variability-guided training for enhancing cardiac-vagal modulation, aerobic fitness, and endurance performance: a methodological systematic review with meta-analysis. International Journal of Environmental Research and Public Health 18(19):10299. doi.org/10.3390/ijerph181910299
12. Miller DJ, Sargent C, Roach GD (2022). A validation of six wearable devices for estimating sleep, heart rate and heart rate variability in healthy adults. Sensors 22(16):6317. doi.org/10.3390/s22166317
13. Munoz ML, et al. (2015). Validity of (ultra-)short recordings for heart rate variability measurements. PLOS ONE 10(9):e0138921. doi.org/10.1371/journal.pone.0138921
14. Nunan D, Sandercock GR, Brodie DA (2010). A quantitative systematic review of normal values for short-term heart rate variability in healthy adults. Pacing and Clinical Electrophysiology 33(11):1407-1417. doi.org/10.1111/j.1540-8159.2010.02841.x
15. Nuuttila OP, et al. (2022). Reliability and sensitivity of nocturnal heart rate and heart-rate variability in monitoring individual responses to training load. International Journal of Sports Physiology and Performance 17(8):1296-1303. doi.org/10.1123/ijspp.2022-0145
16. Nuuttila OP, et al. (2024). Morning versus nocturnal heart rate and heart rate variability responses to intensified training in recreational runners. Sports Medicine - Open 10:120. doi.org/10.1186/s40798-024-00779-5
17. Pietilä J, et al. (2018). Acute effect of alcohol intake on cardiovascular autonomic regulation during the first hours of sleep in a large real-world sample of Finnish employees: observational study. JMIR Mental Health 5(1):e23. doi.org/10.2196/mental.9519
18. Plews DJ, et al. (2012). Heart rate variability in elite triathletes, is variation in variability the key to effective training? A case comparison. European Journal of Applied Physiology 112(11):3729-3741. doi.org/10.1007/s00421-012-2354-4
19. Plews DJ, et al. (2013). Training adaptation and heart rate variability in elite endurance athletes: opening the door to effective monitoring. Sports Medicine 43(9):773-781. doi.org/10.1007/s40279-013-0071-8
20. Schmalenberger KM, et al. (2019). A systematic review and meta-analysis of within-person changes in cardiac vagal activity across the menstrual cycle: implications for female health and future studies. Journal of Clinical Medicine 8(11):1946. doi.org/10.3390/jcm8111946
21. Shaffer F, Ginsberg JP (2017). An overview of heart rate variability metrics and norms. Frontiers in Public Health 5:258. doi.org/10.3389/fpubh.2017.00258
22. Tegegne BS, et al. (2020). Reference values of heart rate variability from 10-second resting electrocardiograms: the Lifelines Cohort Study. European Journal of Preventive Cardiology 27(19):2191-2194. doi.org/10.1177/2047487319872567