Sports · Part Four · Recovery and Fueling
Lesson 39 / 64
Heart Rate Variability
The tiny, irregular gaps between an athlete's heartbeats are the clearest readout of how recovered the nervous system really is.
Heart rate variability is the beat to beat variation in the timing between heartbeats, measured in milliseconds and reported in sport as RMSSD from a morning recording. In an athlete it rises under some overload blocks and falls under others, and only that athlete's own trend tells which. The Unified Model of Tone reads the variability of the signal, rather than its average, as where regulation shows itself.
Reading window
Three minutes, supine
Averaging window
Seven days
Phone against ECG
Correlation 1.00
Overreached athletes
RMSSD went up
RMSSD.
The root mean square of successive differences between heartbeats, in milliseconds. It is calculated from consecutive pairs of R-R intervals, the timescale the vagal brake acts on, which is why short field recordings report it.
The rolling baseline.
An athlete's own recent average and spread, against which today's reading is scored. It moves with the season, so the comparison stays current instead of anchoring to a value the athlete produced months ago.
01What the measurements show
The Numbers Behind the Morning Reading
Eight findings on what an athlete's heart rate variability does under training load.
02What the gaps carry
Heart Rate Variability Is the Millisecond Record of the Nervous System Editing the Heart
Heart rate variability is the variation in time between consecutive heartbeats, measured in milliseconds. A heart at 60 beats per minute does not fire exactly once per second. The intervals flex constantly, and that flex is the fingerprint of a living, responsive nervous system. It is also the most direct reading sport has of the system that governs recovery.
The pacemaker is the sinoatrial node, a cluster of specialized cells in the right atrium that sets the rhythm of the heart. The node fires on its own, faster than a resting athlete's pulse. Part of the slow athletic pulse belongs to the node's own remodeled electrophysiology rather than to the vagus, and The Autonomic Nervous System carries the blockade experiment that separates the two.
Two opposing nerve supplies sculpt every interval. The vagus nerve applies a fast parasympathetic brake that slows the heart within a single beat, and the cardiac sympathetic fibers push the rate up more slowly. Healthy variability is the live negotiation between them.
Breathing drives that negotiation on a visible cycle. When an athlete inhales, the vagal brake eases and the heart speeds slightly. When they exhale, the brake returns and the heart slows. The rhythm is called respiratory sinus arrhythmia, and it is the reason there is anything to count.
The circuit behind the number, from cortex to sinoatrial node
Heart rate variability is a brain measurement taken at the chest, because the circuits that govern the heart also govern coordination, focus and the response to threat. The vagus carries parasympathetic command from the brainstem hubs of autonomic regulation down to the sinoatrial node. Sympathetic fibers travel from the upper thoracic spinal cord through the stellate ganglion to the heart.
Sitting above both, the central autonomic network composes what those nerves deliver. It includes the insula, the anterior cingulate and the amygdala, and The Autonomic Nervous System: Circuitry and the Segmental Map carries its anatomy and the exercise imaging. This is where performance neurology meets cardiology. Understanding the reading means understanding all three: the node, the two nerve supplies, and the network above them.
The field version of the instrument holds up against the laboratory version. In 29 subjects at rest, smartphone photoplethysmography and a chest strap were compared with electrocardiography for RMSSD Plews 2017. Average technical error ran 6.35 percent, and correlations with electrocardiography were almost perfect at 1.00. What a squad collects on a phone is the same record.
03Choosing the index
Field Programs Report RMSSD Because the Other Indices Do Not Survive a Short Recording
RMSSD is the metric sport records, and the choice rests on two numbers: its typical error and the recording length it needs. The root mean square of successive differences takes each consecutive pair of intervals between beats and squares the difference. It then averages those squares and takes the root. The result measures how much the timing changed from one beat to the next.
The alternatives appear on every dashboard. SDNN is the standard deviation of all normal intervals, so it captures total variability from every source in the recording. Frequency domain analysis splits the same intervals into bands. High frequency power tracks vagal traffic, and low frequency power blends both branches.
Reliability decides which of them a coach can act on. Fifteen men were recorded on four separate occasions, at seated rest and after submaximal and supramaximal exercise Al Haddad 2011. Typical error ran 4 to 17 percent for time domain indices and 7 to 27 percent for spectral indices. For ratio indices it ran 41 to 82 percent.
Supramaximal work made the spectral indices worse still. An index whose error swamps the change a training block produces cannot report that change, and the ratio scores sit in that category.
How long the recording has to be
Recording length is part of the measurement rather than a detail of it. Across 159 orthostatic tests of 7 minutes supine and 6 minutes standing, the RMSSD analysis window could be cut to the first 2 minutes Bourdillon 2017. Low frequency and total power still needed the full 4 minutes.
Add a settling minute in front and a supine RMSSD recording runs 3 minutes, a frequency domain recording 5. Shorter windows have been tested in athletes too, and The Autonomic Nervous System carries the 60 second comparison against the 5 minute standard.
What a morning protocol holds constant
Posture, the hour and the night before enter the number itself. Supine and standing values are separate measurements, which is why the studies here report them apart rather than pooled. A protocol fixes them so that the difference between two mornings belongs to the athlete.
Interpretation takes the same discipline. A change in any heart rate measure should be read against the error of measurement, the smallest important change for that measure, and the training phase it sits in Buchheit 2014. A number without those three references is a number without meaning.
04The athlete's own baseline
A Heart Rate Variability Reading Carries No Meaning Until It Is Scored Against That Athlete
A value that is high for one athlete is ordinary for the next, and in elite endurance athletes the number can fall while fitness rises. Context decides meaning. The only comparison that carries information is the athlete against their own recent record, and that is what the measurement is.
Most of this literature was built on recreational and well-trained athletes. A review of training adaptation and heart rate variability in elite endurance athletes reports that the small number of studies in elites returned equivocal outcomes Plews 2013b.
Both increases and decreases in variability have been associated with negative adaptation in that group. Signs of positive adaptation, including rising cardiorespiratory fitness, have appeared alongside decreases in variability. Two athletes can move in opposite directions and both be reading correctly.
The review names the mechanism: variability saturates in elite endurance athletes. The index falls even while resting heart rate is also falling. An athlete whose vagal drive already sits near the top of its range has nowhere left to go, and the number starts running backward.
Why a population chart fails an elite athlete
The Unified Model of Tone states the reason directly. Reference ranges describe populations, and an athlete can sit comfortably inside a population range while having drifted far from their own functional baseline. In sport that is the ordinary case.
The authors of the elite review reach the same practical conclusion from the data. Longitudinal monitoring is what reveals an individual athlete's own heart rate variability fingerprint Plews 2013b. A chart of normal values has no way to hold a fingerprint.
The working unit is therefore the rolling baseline: an athlete's own recent mean and spread across seven days. Today's reading is scored against that. The question is whether the number is this athlete's.
05The week beats the day
The Weekly Average Carries the Training Signal That a Single Morning Does Not
Averaging a week of readings moves the correlation with fitness change from trivial to very large. Its size is larger than most of the changes a monitoring program is trying to detect, which makes averaging the first decision rather than a refinement.
Ten runners completed a 9 week intervention and were read both ways Plews 2013a. Change in maximal aerobic speed correlated with single day log RMSSD at r of minus 0.06, which the authors classify as trivial. Against the same index averaged over one week the correlation was 0.72.
Ten kilometer running performance behaved the same way, at minus 0.17 for the isolated day and minus 0.76 for the week. Same athletes, same recordings, one decision about how to read them. The day carried nothing and the week carried almost everything.
The overreaching study fails the same way in reverse. Weekly means detected a progressive rise in parasympathetic modulation in athletes heading into functional overreaching Le Meur 2013. Isolated recordings taken once every seventh day found no clear effect on any variability parameter.
The authors name the cause as wide day to day variability in the measure itself. A program sampling once a week is taking a different measurement, and that measurement detected nothing.
How many mornings a week the reading needs
Five mornings carry the week, and the spread of those five carries more than their average. Nine women from a college soccer team recorded supine and standing readings daily through a phone application across 3 weeks of moderate, high and low training load Flatt 2015.
Means built from 5 days and from 3 days both matched the 7 day criterion, at intraclass correlations of 0.74 to 0.99. Five is the working number because it leaves room for a missed morning. The measure that tracked the changing training load was the coefficient of variation of the supine readings, moderately lower in the low load week than in the high load week.
Effect sizes ran 0.86 to 0.92, with P values of .003 to .045. No significant change appeared in the other measures, including the weekly means. The spread of the week's readings moved while the average of those same readings held still.
06What a drop means
A Falling Morning Reading Reports a Cost, and Does Not Name Its Cause
A drop in morning heart rate variability reports that the athlete's system is paying for something, without naming what. The cleanest overreaching trial on record moved the number the other way entirely.
Twenty-one trained male triathletes were randomly assigned to intensified or normal training across 5 weeks Le Meur 2013. The study ran a 1 week baseline, a 3 week overload and a 1 week taper, with daily recordings throughout. Every athlete in the intensified group lost maximal incremental running performance by the end of the overload.
The loss ran 9.0 plus or minus 2.1 percent of baseline, and performance supercompensated after the taper. That sequence is the definition of functional overreaching. These were athletes measurably in trouble at the moment they were measured.
Their parasympathetic indices climbed. Weekly mean supine log RMSSD carried a 96 percent chance of a positive effect, a 4 percent chance of a trivial one and a zero percent chance of a negative one. Standing values ran 98 percent. During the taper the responses reversed.
So the number can rise in an athlete who is in the middle of failing. The familiar rule that low means fatigue and high means recovery has an exception wide enough to run an entire overload block through.
Fatigue widened the spread fourfold
Fatigue scatters an athlete's readings as well as lowering them. Fifty-seven elite Nordic skiers were surveyed over 4 years, producing 172 recordings taken in a fatigue state and 891 taken outside one Schmitt 2013. In fatigue, supine heart rate ran 6.27 beats per minute higher and log high frequency power 0.46 lower.
Log total power ran 0.36 lower, and standing values moved the same way. In level terms that is the textbook picture, and it points the opposite way from the overreaching trial.
The variance data reconciles them. Intra-individual variance of those parameters was larger in fatigue: 0.32 against 0.08 for log high frequency power supine, and 0.26 against 0.07 for log total power. The authors read the wider variance as reflecting genuinely different fatigue responses.
Fatigue was legible in how widely an athlete's own readings scattered, as well as in how far they fell. Two athletes can carry the same weekly mean while one of them is holding steady and the other is swinging, and the swing is the finding.
What the pooled evidence can and cannot separate
Pooled across the training literature, resting RMSSD rose when training improved performance and rose again when training produced overreaching, which is why the reviewers call for additional measures of training tolerance Bellenger 2016. Tone and the Athlete's Edge carries that meta-analysis in full and the study design it points toward.
07Prescribing from the reading
HRV-Guided Training Moves the Hard Session to the Day the Athlete Can Take It
Heart rate variability guided training replaces a schedule written in advance with a schedule chosen each morning, and it has been tested against written programs in randomized trials. Adaptation is built during recovery, in the window the parasympathetic shift back toward rest reclaims. An athlete who trains heavy on a suppressed reading stacks work onto a system that has not reset.
The first trial set the pattern. Twenty-six moderately fit men were randomized to a predefined program, a guided program or a control group, then ran 40 minute sessions for 4 weeks Kiviniemi 2007. The guided rule left nothing to judgment.
An increase or no change in the morning reading bought a high intensity session that day. The reference value was the 10 day mean minus one standard deviation. A significant drop below it, or a downward trend across 2 days, bought low intensity work or rest.
Peak oxygen uptake in the guided group rose from 56 to 60 mL/kg/min and maximal running velocity from 15.5 to 16.4 km/h. The predefined group improved velocity from 15.1 to 15.7 km/h and did not improve oxygen uptake at all. The velocity gain was significantly larger in the guided group.
The cycling trials, including what they did not show
The cycling trial moved one arm and not the other, and still returned no difference between them. Seventeen well-trained road cyclists spent 4 baseline weeks establishing their resting values, then 8 training weeks split between guided and traditional periodization Javaloyes 2019.
The guided group improved peak power output by 5.1 percent, power at the second ventilatory threshold by 13.9 percent, and a 40 minute time trial by 7.3 percent. The traditional group improved on none of them, and the comparison between the two groups came back empty.
A block periodization comparison in 24 endurance trained men found both groups improving maximal treadmill velocity and 3000 meter running performance Nuuttila 2017. Relative changes in velocity and countermovement jump were greater in the guided group. RMSSD, low frequency power and total power rose significantly only in the guided group.
What the pooled result supports
The meta-analysis sets the size of the effect. Accounting for methodological differences, guided training beat predefined training for vagally mediated variability indices at a standardized mean difference of 0.50, with a confidence interval of 0.09 to 0.91 Manresa-Rocamora 2021.
Everything else came back non-significant. Resting heart rate sat at 0.04, maximal aerobic capacity at 0.20, capacity at the second ventilatory threshold at 0.26, and endurance performance at 0.20. All four confidence intervals crossed zero.
The reviewers summarize their own result this way. Guided training may be more effective for maintaining and improving vagally mediated variability, with less likelihood of negative responses, and any advantage in fitness and performance is small.
Where care sits inside a monitoring program
The spinal column is dense with mechanoreceptors that supply the brainstem its picture of the body, and the cervical and upper thoracic segments sit closest to the pathways serving the heart. That is a claim about where an input lands rather than a claim about what it delivers.
Heart rate variability works the way a strength coach uses bar speed, as honest feedback recorded the same way every morning. It serves as a scoreboard and fails as a target, because a value raised by changing the recording is not a changed athlete.
Care runs alongside the athletic training, strength and medical staff, drug free and compliant with anti-doping rules. Read this way the number becomes a steering wheel rather than a guess, and the season is built on signal instead of ego.
08What we corrected
Five Claims Removed From This Page
This page previously stated that a morning reading below baseline flags incomplete recovery before the athlete feels anything wrong. That lead time could not be traced to a source, so it is gone. What replaces it is the overreaching evidence above, where the weekly reading rose while performance fell Le Meur 2013.
Two figures went with it. The page gave a resting RMSSD of 80 to 100 milliseconds for a young endurance athlete, and an intrinsic sinoatrial rate near 100 beats per minute. Neither could be traced to a source, and neither would change a decision in any case, because the reading is scored against the athlete.
It answered yes to whether care aimed at the nervous system changes an athlete's heart rate variability. Studies measuring the number around spinal adjustment were offered as the evidence, and care was described as reducing the sympathetic drive that flattens it. Those are efficacy claims, and they have been replaced by a statement of where the input lands.
A quotation attributed to Dr. Jason Dulberg was not drawn from anything he said or wrote, and it has been removed. Claims made here are either sourced to the literature or named explicitly as the model's.
09The model's claim
One Instrument Can Carry All Four Readouts the Model Says Move Together
Two layers run through this page. The established science is the overreaching trial, the four year elite fatigue survey, the averaging comparison, the reliability coefficients and the guided training trials. Those results belong to the investigators who recorded the intervals.
The Unified Model of Tone starts with where to look. Tone shows itself wherever a rhythm can be measured for more than its average: in the variability of a signal rather than its mean. Heart rate variability is that instruction turned into an instrument.
It also explains why the training literature scatters. There is no such thing as an input acting upon an empty body. An overload block meets a nervous system that is already organized one way or another, and the direction the reading moves depends on where it started. Thirteen triathletes went up. Fifty-seven skiers went down. The same rule covers both.
The structure of the readings, not their level
The spread of an athlete's week carries information the average does not. Two results already point that way. The coefficient of variation of supine readings tracked changing training load while the weekly means did not Flatt 2015.
And fatigue widened intra-individual variance roughly fourfold in elite skiers, from 0.08 to 0.32 on log high frequency power Schmitt 2013. In both studies the spread carried information the average did not. That is a claim about how performance is organized rather than a claim about what treatment does.
The prediction this page makes
Sport can take all four of the model's readouts from one recording family, which no other single protocol delivers. Variability structure is the seven day coefficient of variation of morning supine RMSSD. Reflex responsiveness shows in the supine to standing change in RMSSD across an orthostatic test.
Coupling is read as high frequency power recorded at a fixed breathing rate, which samples how tightly breath and heartbeat hold step. Recovery time is how long RMSSD takes to return to baseline after a standardized submaximal load test. One morning protocol, one orthostatic test and one load test deliver the set.
Run that on one squad for one season, scoring every value against each athlete's own trend. The model predicts that the seven day coefficient of variation will track training tolerance more closely than the seven day mean does, and will separate productive adaptation from overreaching where the mean cannot.
If the weekly coefficient of variation of RMSSD, the supine to standing change, high frequency power at fixed breathing and RMSSD return time move together across that season, the unification claim is confirmed.
10The tone reading
How Heart Rate Variability Expresses Tone in an Athlete
Three foundations of tone carry the signature in an athlete's morning reading.
Oscillation
A rhythm sets every gap. The beat lengthens on the exhale and shortens on the inhale, and RMSSD counts that swing in milliseconds.
Set point
Elite endurance athletes show variability falling alongside a falling resting pulse, so the value being defended belongs to the athlete rather than to a chart.
Time course
Single day readings correlated at r of minus 0.06 with 9 week fitness change. The same readings averaged over a week correlated at 0.72.
The remaining foundations show up in the same recording. Constraint: overreached triathletes lost 9 percent of their maximal test performance while their parasympathetic indices rose, a system locked in one mode instead of moving freely. Load: a training block is the demand this system is organizing, and its cost showed as a 6.27 beat per minute rise in supine heart rate in fatigued skiers. Input quality: posture, breathing rate and the hour of the recording enter the number, which is why a field protocol fixes all three. Coupling: the beat holds step with the breath, and losing that step flattens the swing RMSSD counts. Gain: how much the vagus lengthens the interval for a given demand. Prediction: the cortical network that composes autonomic output sits above the brainstem, so an anticipated start line changes the intervals before the gun. The full framework is set out in the Unified Model of Tone.
11Where this sits
How This Page Relates to the Rest of the Library
Seven places this argument continues, each with the claim that earns the link.
The branches this instrument samples, the onset handoff in the first 90 seconds of work, and the microneurography showing the trained sympathetic system is not a quiet one.
Vagal anatomy and how fast the brake returns after effort, which is the reactivation kinetics behind every post-exercise reading taken here.
The window in which the morning value is rebuilt, and where a short night is read as an input to tomorrow's recording rather than a verdict on the athlete.
Where functional overreaching separates from overtraining, and where the taper that reversed the triathletes' readings is taught as a dose.
The slower endocrine arm of the same load, which moves on a different clock from the beat to beat record and has to be read alongside it.
The visit where the orthostatic test, the load test and the oculomotor and proprioceptive measures are recorded in one session.
The same number as a measurable state in the general population: what it counts, the transplant experiment, the retired low frequency to high frequency ratio, and what a persistently low value has predicted. Start there for the instrument. Stay here for the training block.
12Questions athletes ask
Questions Athletes Ask
How should an athlete take a morning heart rate variability reading?
Same posture, same hour, on waking, before caffeine. A supine RMSSD recording needs about 3 minutes: one settling minute and a 2 minute analysis window, which is what an analysis of 159 orthostatic tests established. Frequency domain indices need 5 minutes. A chest strap and a validated phone camera both agree with electrocardiography for RMSSD, at an average technical error near 6 percent. Record at least five mornings a week, because five day means match seven day means closely.
Should I change today's session because my reading dropped this morning?
Not on one reading. In 10 runners across 9 weeks, single day values correlated with fitness change at r of minus 0.06, while the same values averaged over a week correlated at 0.72. The trials that succeeded changed sessions by rule, on a threshold set in advance. A decrease below the 10 day mean minus one standard deviation, or a downward trend across 2 days, bought low intensity work or rest. One low morning is information about that morning, and the decision belongs to the week around it.
Does heart rate variability guided training actually beat a written plan?
Modestly. Nine men trained for 4 weeks on sessions chosen from their daily readings, and raised peak oxygen uptake from 56 to 60 mL/kg/min. A matched group on a fixed program did not move it. In 17 well-trained cyclists, the guided arm improved peak power, threshold power and a 40 minute time trial while the traditional arm improved none, though the comparison between arms was not significant. Pooled across trials, guided training won on vagal indices at 0.50 and showed no significant advantage on performance.
My readings went up during a hard block and my legs feel terrible. What is happening?
That combination is documented. Thirteen trained triathletes ran a 3 week overload block, lost 9.0 percent of their maximal incremental running performance, then supercompensated after a taper. That sequence is functional overreaching. Their weekly averaged parasympathetic indices climbed while it was happening, with a 96 percent chance of a positive effect in the supine reading. A rising number can appear in an athlete who is in trouble, so it is read alongside performance, session quality and how widely that week's readings are spread.
Why do performance programs report RMSSD instead of a stress or balance score?
Reliability. Measured on four occasions in 15 men, typical error ran 4 to 17 percent for time domain indices such as RMSSD and 7 to 27 percent for spectral indices. For ratio scores it ran 41 to 82 percent, and supramaximal exercise made the spectral indices worse still. An index whose error is larger than the change a training block produces cannot report that change. RMSSD also survives a short recording, which is what a squad can collect every morning.
How many mornings a week does an athlete need to record?
Five. Nine college soccer players recorded daily through a phone application across 3 weeks of varying training load. Means built from five days matched the seven day criterion at intraclass correlations of 0.74 to 0.99. Sampling less often fails in a specific way. In the overreaching study, weekly means detected a progressive autonomic change while isolated once weekly recordings found nothing at all. Day to day spread is wide enough to swallow a single reading, which is why the weekly value is the unit of decision.
What does the Unified Model of Tone say about an athlete's heart rate variability?
That it is the most direct instrument the model has, because tone shows itself in the variability of a signal rather than its mean. Reference ranges describe populations, and an athlete can sit inside one while having drifted far from their own functional baseline. Every value is therefore scored against that athlete's trend. The model's own prediction is that the weekly spread of an athlete's readings tracks training tolerance more closely than the weekly average does. That is a claim about how performance is organized.
13The sources
References
14 primary sources, each linked to its record. Figures quoted on this page were checked against the published abstract.
Related evidence