Mind

How Work Stress Shows Up in Your HRV

Cameron Usher · Founder, PX

10 September 2026 · 9 min read

12 years in competitive sport; trained alongside the Australian Olympic squad; coached 100+ people one-on-one.

It is Thursday. You have not trained since Monday. You slept seven hours. And your heart rate variability is down, your resting heart rate is up a couple of beats, and the app you pay for is suggesting you take it easy, which makes no sense at all because you have done nothing.

You have done a great deal. It just did not happen in a gym.

Stress affects HRV in a way that is measurable, substantial, and completely invisible on a training log. Your body does not keep a separate ledger for work. This post covers how much of your recovery a bad week actually spends, why your HRV cannot tell you where the cost came from, and what a training plan should do about it.

How much a difficult hour actually costs

Start with the measurement, because it is bigger than most people assume.

Researchers pooled 12 studies covering 758 people who were put under acute mental stress in controlled conditions. RMSSD, the most stable of the vagal heart rate variability measures, fell by 12.03 milliseconds. High-frequency power fell by 359.7 ms². The interbeat interval shortened by 142 milliseconds.

For scale: pooled across 44 studies and 21,438 healthy adults, the population mean for short-term RMSSD is 42 milliseconds, with a standard deviation of 15. A laboratory stressor lasting minutes moved the number by roughly a third of a population mean. That comparison is our arithmetic across two papers rather than a published finding, so treat it as an order-of-magnitude point rather than a precise one. But the order of magnitude is the point.

A related study is worth noting because of what it ranked. Thirty adults were exposed to three different stressors within the same session. Mental arithmetic dropped RMSSD by 9.2 milliseconds. A cold pressor task left RMSSD statistically unchanged. Noise produced no global change at all. The cognitive stressor hit vagal tone harder than the physical one. Small sample, laboratory setting, so hold it loosely. But it points the right way: the difficult conversation is not a lesser stressor than the cold.

The study that should change how you read a bad week

Measurement is one thing. Consequence is another. This is the consequence.

Researchers took 31 participants through a resistance exercise protocol and measured how quickly maximal force came back afterwards, split by how much life stress the participants were carrying.

Twenty minutes after the session, the low-stress group had recovered 38.0% of maximal force. The high-stress group, 21.7%. At sixty minutes, low-stress had recovered 60.3%. High-stress, 38.2%. Life-event stress moderated recovery significantly, and explained 9.2% of the variance in recovery trajectories.

Now the detail that makes this the most important study in this post. Stress had no significant effect on self-reported energy, fatigue or soreness. The high-stress group recovered at roughly two-thirds the rate and could not tell.

Read that again if you have ever justified a hard session on the grounds that you felt fine. Feeling fine was not informative. It was the one thing the study specifically checked and specifically found unhelpful.

The same research group followed the recovery out to 96 hours and found the moderating effect persisted across four days. So this is not a phenomenon of the hour after training. It is a phenomenon of the week.

And it comes for the plan from the other side too

If stress only degraded recovery, you could work around it. It also reduces how much you do.

A systematic review examined 168 studies, analysing 55 prospective ones in depth. In 76.4% of them, psychological stress predicted reduced physical activity or increased sedentary behaviour. Among the higher-quality studies, that rose to 85.7%. Six of seven studies using objective stress markers agreed.

There is one nuance the authors add that is worth carrying, because it is hopeful and it is specific: habitually active people tended to exercise more under stress. It is newer exercisers who drop off. If you have been training for years, your risk is training through something you should have adjusted. If you are in your first months, your risk is disappearing. Different failure modes, same cause.

Separately, a 12-week study of 135 people doing a structured weight-training programme found the low-stress participants gained significantly more on bench press and squat than the high-stress ones. Same programme, same weeks, different outcome, and the variable was what was happening in the rest of their lives.

So a hard fortnight at work attacks from three directions at once. It slows your recovery from the work you do. It reduces the work you do. And it blunts what you get from the work you complete anyway.

Why your HRV cannot tell you the reason

Here is the honest limit, and it matters more than any of the numbers above.

Your HRV does not know where the stress came from. It cannot distinguish a hard Tuesday at work from a hard Tuesday of intervals, and you should not expect it to, because it cannot even reliably distinguish good training from bad training.

That last claim has a specific source. A meta-analysis of 27 studies looked at how HRV responds to positive training adaptation versus overreaching. Post-exercise RMSSD rose with positive adaptation, at a standardised effect of 0.60. It also rose with overreaching, at 0.64. Resting high-frequency power, meanwhile, was essentially unmoved by overreaching at 0.04, not significant. The authors concluded that additional measures of training tolerance may be required.

HRV reports the state of your autonomic nervous system. It does not report the cause. It is a thermometer, not a diagnosis.

Which is exactly the argument for reading it alongside everything else. Your training load says what you did. Your sleep says what you recovered. Your calendar says what kind of week it was. Your own report says how it felt. Any one of those alone will mislead you, and the combination will not. That is not a marketing line, it is the direct implication of a meta-analysis saying the signal is ambiguous on its own.

Why one morning reading is close to noise

Before you act on today's number, one more thing worth knowing, and it is the single most useful statistic about HRV in existence.

Researchers compared how well HRV tracked changes in 10 km running performance depending on how it was averaged. Weekly-averaged HRV correlated with performance change at r = −0.76. Single-day values correlated at r = −0.17.

The same work reports that even the best-behaved index carries a day-to-day coefficient of variation of about 12.3%. That is the floor of the noise, in the most stable measure available.

Two consequences follow, and they are both practical.

First, today's number on its own is not actionable. A trend across days is. If your morning reading is down and you have not been able to explain it, wait for the second and third day before concluding anything. One low morning is weather.

Second, do not compare your number to anyone else's. Across 44 studies of healthy adults, RMSSD ranged from 19 to 75 milliseconds depending on the population, with between-study variation running as high as 118%. There is no normal to be above or below. There is only your own rolling baseline, which is why PX builds every signal against 14 days of your own history rather than against a table of averages. A number compared to strangers is describing the strangers.

And a caution that runs the other way: higher is not automatically better. Vagal HRV measures fall at both extremes of vagal tone, so a drop has to be read alongside your resting heart rate rather than in isolation. Anyone selling you a simple up-is-good dial is simplifying something that does not simplify.

What actually helps, and for how long

Two honest answers.

Slow breathing works while you do it, and much less afterwards. A meta-analysis of 31 studies covering 1,133 people found slow-paced breathing raised RMSSD with a standardised effect of 0.90 during the practice, falling to 0.37 after it, with high-frequency HRV showing no significant post-effect at all. So five minutes of slow breathing before a difficult meeting is doing something real and measurable in the moment. It is not banking recovery for the afternoon.

A holiday raises it, and for most people it does not stick. In a randomised study using continuous ECG monitoring across a six-day resort stay, HRV rose during the stay in both the structured and unstructured groups. A month later, at the population level, it had not persisted. Among those with complete individual data, 18.2% were long-term responders and 27.3% went straight back to baseline.

Neither of those is a reason not to do them. They are a reason not to expect them to solve a structural problem. If your recovery is being spent by your working week, the recovery intervention is not the lever. The training load is.

What the plan should do about it

This is the part almost nothing else does, and it is why the ambiguity of the signal is an argument for PX rather than against it.

If your recovery signals are suppressed and your training log is quiet, a single-metric app has two options: tell you your recovery is 61, or tell you nothing. Neither is a decision. The number is real and the instruction is missing, which is the entire gap this company exists in.

What should happen instead is that the plan reads the suppression, notices that it is not explained by training load, and adjusts anyway, with the reason stated. In PX that is a cross-domain override: when the signals say the system is carrying more than the training log accounts for, the prescription eases and the score is capped, and you get a sentence rather than a number. Something like your recovery markers have been below your own baseline for three days without a matching training load, so today is easy and the intervals move to Saturday.

You do not have to explain your week to it. That is the part people find strange at first. The plan does not need to know you had a board meeting. It can see the cost, and the cost is what it needs to size the session.

Two things this specifically is not. It is not a claim that PX detects stress, because your HRV cannot tell PX what the cause was and we are not going to pretend otherwise. And it is not a health assessment: nothing here is about your cardiovascular system or your long-term anything. It is a training decision made with better information about what your body is already carrying.

The mirror image of all this is worth knowing too. A hard week at work makes an evening session more expensive, which affects whether it should happen at that hour at all, covered in why you cannot sleep after an evening workout. And how you feel is itself a signal worth reading rather than overriding, which is whether mood affects workout performance. Together with the physical signals in biometric training, that is the full picture, and it is why the system that keeps training consistent has to read all of it or none of it.

Let the plan carry the week you actually had. PX reads what you already track and decides your day before it starts, with the reason on every call. Try PX.