Blog post by Marco Altini.
In the last few posts on best practices for HRV measurement I covered: context & measurement time of the day, measurement duration and measurement type (lying, standing or orthostatic). To sum up, by looking at published literature and performing some additional analysis using data from HRV4Training users, we found that:
Now it's time to look at another important aspect of HRV measurements, breathing. More specifically, in this post I will first briefly highlight what is known from literature, then move on to some self-experimentation and finally use once again HRV4Training users data.
In particular, I will focus on two aspects:
Why are these two aspects important? One of the main reasons behind using paced breathing is that it is supposed to make the measurement more reliable and improve repeatability, so I will look at what current literature says about this. Secondly, the impact on HRV baseline values is relevant in the context of longitudinal changes and comparisons. If deeper breaths increase your HRV, and you change your breathing rate settings at a certain point in time, data might not be comparable with what you had before (while still valid for day to day guidance). The same holds for comparisons between individuals.
background on hrv and paced breathing: literature
Paced breathing is one of those aspects where there is much controversy. In , HRV (SDNN and frequency power) was recorded during paced breathing at the relatively high frequencies of 15 and 18 breaths per minute. Features were the same compared to spontaneous breathing for 15 breaths per minute, and reduced for 18 breaths per minute. Thus, according to this study HRV baseline should not be affected by paced breathing or be reduced in case of very high rates. This is what I would expect for breathing rates that are relatively high. Especially because during spontaneous breathing we tend to breathe at rates that are anyway > 10 breaths per minute.
Normally, we expect increases in HRV when we measure at much lower rates, for example close to 0.1 Hz (or 6 breaths per minute), which is what is typically done during biofeedback exercises, to get the heart "in synch" with breathing (more on this later).
In  we have a meta analysis of many different HRV studies. According to the authors, paced breathing is one of the reasons behind different HRV values between studies, together with physical activity level, age, errors in correcting RR intervals, and many other factors. The point here being that rMSSD, our most relevant metric to monitor training load, was slightly higher during paced breathing. This is important because it means we might have baseline changes if we change breathing frequency or start doing paced breathing after measurements where we were breathing spontaneously. Similarly,  reports higher HRV (frequency domain features) with reduced breathing rates, analyzing breathing rates over a wide range (3 to 14 breaths per minute). The same is shown in .
On the contrary, in  the authors report pretty much the same rMSSD values across paced breathing rates between 6 and 15 per minute.  reports very small improvements in reliability when using paced breathing, while  says paced breathing is not necessary, as long as subjects are reminded to avoid irregular respiration (so maybe it is necessary?).
Based on literature, while results are definitely conflicting, I would derive the following:
self-experimentation, n = 1
Here are some results from tests I ran to determine the effect of paced breathing at different breathing rates on my HRV. I will look at:
For the spot checks, measurements where taken with Camera HRV, since the app lets you configure breathing rates between 6 and 14 breaths per minute and take multiple measures any time you want and for how long you want. Camera HRV is a handy tool if you know what you are doing or want to play around. I also recorded one measurement using HRV Logger and a Polar H7, for the ones that are skeptical about the camera.
For the longitudinal data, I've been taking double tests in the morning, one using HRV4Training and one using StayFit. StayFit is an app used to quantify fitness level (similarly to VO2max, by trying to address some of the limitations of VO2max, as you can read here). The app measures HR over 30 seconds and uses paced breathing at 10 breaths per minute.
Let's start with the spot checks.
HRV spot checks
breathing rate, hrv and RESPIRATORY SINUS ARHYTHMIA
Respiratory sinus arrhythmia (RSA) refers to getting the heart rate in sync with respiration. Typically this is the case when we breathe at lower rates (deep breaths), however in general when we breathe regularly we can see this kind of synchrony even at higher breathing rates. Basically, breathing in will shorten RR intervals and therefore raise instantaneous heart rate, while breathing out will do the opposite, prolong RR intervals and decrease HR.
Here is an example using data I recorded using Camera HRV:
We can see how the RSA is showing up clearly at 6 breaths per minute, which is probably the ideal frequency for this exercise (at least for most individuals). At 8 and 10 we can still see clearly the changes in RR intervals due to breathing, we can even count them, we should have 18 swings at 6 per minute and 30 at 10 per minute, since we have 3 minutes of data for each measurement. At 14, things get a bit more messy.
What about spontaneous breathing?
Here we have two recordings of 6 and 8 minutes, in which I first followed the breathing bar for paced breathing and then was breathing spontaneously, or the other way around. We can see RSA present only during the paced breathing parts.
breathing rate and hrv (BASELINE) changes
We've seen from the previous plots that there is a clear effect of paced breathing and breathing frequency on the RR intervals. Now the more interesting question is, how does this impact HRV, and in particular rMSSD? The recordings above where I was breathing between 6 and 12 breaths per minute were taken consecutively. I repeated the measurements twice. Let's look at some data.
Here we have boxplots for HR, rMSSD and LF over the different breathing rates:
This is the same data shown above for RR intervals in the context of RSA. Since we have 3 minutes of data, features are computed 3 times (every 60 seconds). The HR is quite constant between breathing rates, please note the breathing rates are not in order, sorry about that. The sequence is 12, 10, 6 and 8.
For rMSSD, we see a clear effect of breathing rate in my data. Differences are not huge, but definitely consistent between what seems to be a threshold between the two breathing rates that involve deep breathing for me (6 and 8 per minute) and the two higher breathing rates (10 and 12 per minute), shallow breathing.
I also plotted LF since this is an interesting feature in the context of RSA. We expect to have the peak of the frequency spectrum around 0.1 Hz or 6 breaths per minute, and LF covers frequencies between 0.04 and 0.15 Hz, so RSA reflects on this feature more than in others. Looking at the third plot we see indeed that for 6 breaths per minute we have the highest LF, as expected.
Then, I took the same measurements once again:
Pretty much the same story. Similar HR, even if this time it seems higher for low breathing rates. Definitely higher rMSSD for lower breathing rates and LF peaking at 6 breaths per minute.
All the above recordings were taken using Camera HRV. Then I also took another measurement using the HRV Logger and measuring with paced breathing at 6 breaths per minute, spontaneous breathing and paced breathing at 10 breaths per minute. Here is the data:
Here once again we can see the impact of breathing for the first few minutes (RSA), when I was following the 6 breaths per minute paced breathing, and then the effect disappears during spontaneous breathing. The final part breathing at 10 breaths per minute seems a bit more regular than the central part of the recording. Let's look at HRV features for this segment:
Spontaneous breathing behaves similarly to breathing at 10 breaths per minute, so shallow breathing. Once again we have HR similar across conditions, rMSSD and LF much higher for breathing at 6 breaths per minute.
What can we derive from this data?
Something that might be more interesting than spot checks is longitudinal data. Here the data I have is a bit more limited, but still interesting. As I was mentioning at the beginning of the post, I've been taking double tests in the morning, one using HRV4Training and one using StayFit. StayFit measures HR over 30 seconds and uses paced breathing at 10 breaths per minute. In the background I also compute HRV, even though it is not reported by the app, so here is what I have:
Since rMSSD and HR are pretty much independent of the time window used to compute them, and the recordings are always taken one after the other, I can assume differences in values, if any, are mainly due to the different breathing rates (this is not necessarily true, since other factors might play a role, for example I could get impatient during the second measurement, or bored or simply have more noise in one or the other recording, however we need to make some assumptions).
Here is the distribution of HR and rMSSD values for a period of about 5 months in which I took about 90 recordings with each app:
This data is consistent with all our previous findings. HR is very similar, even though it seems slightly higher for HRV4Training, i.e. for lower breathing rates. As a matter of fact, the very first reason why I started recording with both apps, was that I had the feeling breathing at lower breathing rates (or more deeply), increases my HR. While there is much overlap between distributions, the peaks are quite different.
For HRV data, again the rMSSD values are higher for data acquired using HRV4Training, possibly due to the lower breathing rate (8 vs 10).
The spot checks measurements already speak to this point as well. All the Camera HRV measurements were taken consecutively, and therefore the changes show that changing breathing rate, something we might do even without thinking about it if we go for spontaneous breathing, provokes changes in rMSSD.
On the other hand, results at the same breathing frequency are quite consistent. This is a point most people getting started with HRV typically struggle with. You take a measurement, then take another measurement, supposedly nothing changed, but the results are different. HRV values are much less "stable" than HR measurements, and that's what makes them much more interesting. The boxplots above where I compared all the different breathing rates show that there is variability in my HRV results, since I recorded for three minutes each condition, and therefore for each condition I obtained 3 values (60 seconds computation window). What is interesting, is that the difference due to breathing is much higher than the natural variability occurring in HRV recordings, and this should give some extra "trust" in the validity of these measures.
For the small tests I've done, it doesn't seem that spontaneous breathing creates more problems in terms of repeatability, but I haven't extensively validated this.
cross-sectional hrv4training data
To conclude, here we are looking at HRV4Training users data cross-sectionally. This means that there might be many other factors we are not considering that have an impact on HRV, and therefore we cannot really derive any conclusions. However, it still makes sense to look at HRV values for users using different breathing rates, to see if lower breathing rates correspond in general to higher HRV values.
The data is part of the new dataset I've been collecting with the latest HRV4Training version, since May 2015. After filtering out users with less than 1 full month of measurements and outliers, I am left with data from 848 users and 20 798 HRV measurements.
Breathing rates are stored for each measurements, so here is the data for HR and rMSSD across breathing rates for all users part of this dataset (for each user I first computed the mean HR and rMSSD, so that each data point is a user, not a measurement):
From these plots it seems there is no difference in HRV at the cross sectional due to breathing rate. Different users using breathing rates of 6, 8 or 10 breaths per minute have similar distributions of rMSSD.
Given the nature of this data, we cannot really derive any conclusions, apart from the fact that breathing rate does not seem to be one of the major factors influencing HRV at the cross-sectional level. However, we can think about some hypothesis that could be further validated, for example:
As a side note, I also built a simple regression model to analyze the factors influencing rMSSD and included as parameters together with breathing rate also age, gender, number of trainings per week and BMI. However, the relation between rMSSD and breathing rate stayed pretty much the same.
As a suggestion for your measurements, I would advise to use paced breathing at a breathing rate that is comfortable for you, i.e. not too fast or too slow, and stick to that for all your measurements. For me, that breathing rate is 8 breaths per minute.
 Bernardi, Luciano, et al. "Effects of controlled breathing, mental activity and mental stress with or without verbalization on heart rate variability." Journal of the American College of Cardiology 35.6 (2000): 1462-1469.
 Nunan, David, Gavin RH Sandercock, and David A. Brodie. "A Quantitative Systematic Review of Normal Values for Short‐Term Heart Rate Variability in Healthy Adults." Pacing and Clinical Electrophysiology 33.11 (2010): 1407-1417.
 Song, Hye-Sue, and Paul M. Lehrer. "The effects of specific respiratory rates on heart rate and heart rate variability." Applied psychophysiology and biofeedback28.1 (2003): 13-23.
 Yildiz, M., and Y. Z. Ider. "Model based and experimental investigation of respiratory effect on the HRV power spectrum." Physiological Measurement27.10 (2006): 973.
 Guzik, Przemyslaw, et al. "Correlations between the Poincare plot and conventional heart rate variability parameters assessed during paced breathing." The Journal of Physiological Sciences 57.1 (2007): 63-71.
 Ginsburg, P., Bartur, G., Peleg, S., Vatine, J. J., & Katz-Leurer, M. (2011). Reproducibility of heart rate variability during rest, paced breathing and light-to-moderate intense exercise in patients one month after stroke. European Neurology, 66(2), 117-122
 Kobayashi, Hiromitsu. "Does paced breathing improve the reproducibility of heart rate variability measurements?." Journal of physiological anthropology28.5 (2009): 225-230.
Register to the mailing list
and try the HRV4Training app!
1. Intro to HRV
2. How to use HRV, the basics
3. HRV guided training
4. The big picture
5. HRV and training load
6. HRV, strength & power
7. Overview in HRV4Training Pro
1. Context & Time of the Day
3. Paced breathing
4. Orthostatic Test
5. Slides HRV overview
6. rMSSD vs SDNN
7. Normal values and historical data
1a. Acute Changes in HRV
1b. Acute Changes in HRV (population level)
1c. Acute Changes in HRV & measurement consistency
1d. Acute Changes in HRV in endurance and power sports
2a. Interpreting HRV Trends
2b. HRV Baseline Trends & CV
3. Tags & Correlations
4. Ectopic beats & motion artifacts
5. HRV4Training Insights
6. HRV4Training & Sports Science
7. HRV & fitness / training load
8. HRV & performance
9. VO2max models
10. Repeated HRV measurements
11. VO2max and performance
12. HR, HRV and performance
13. Training intensity & performance
14. Publication: VO2max & running performance
15. Estimating running performance
16. Coefficient of Variation
17. More on CV and the big picture
18. Case study marathon training
Camera & Sensors
1. ECG vs Polar & Mio Alpha
2a. Camera vs Polar
2b. Camera vs Polar iOS10
2c. iPhone 7+ vs Polar
2d. Comparison of PPG sensors
3. Camera measurement guidelines
4. Validation paper
5. Android camera vs Chest strap
6. Zoom HRV vs Polar
7. Apple Watch and HRV
8. Scosche Rhythm24
9. Apple Watch
1. Features and Recovery Points
2. Daily advice
3. HRV4Training insights
4. Sleep tracking
5. Training load analysis
6a. Integration with Strava
6b. Integration with TrainingPeaks
6c. Integration with SportTracks
6d. Integration with Genetrainer
6e. Integration with Apple Health
6f. Integration with Todays Plan
7. HRV4T Coach advanced view
8. Acute HRV changes by sport
9. Remote tags in HRV4T Coach
10. VO2max Estimation
11. Acute stressors analysis
12. Training Polarization
13. Custom desirable range / SWC
14. Lactate Threshold Estimation
15. Functional Threshold Power(FTP) Estimation for cyclists
16. Aerobic Endurance analysis
17. Intervals Analysis
18. Training Planning
19. Integration with Oura
20. Aerobic efficiency and cardiac decoupling
1. HRV normal values
2. HRV by sport
3. HRV normalization by HR
4. HRV 101