The entire purpose of running by heart rate zones is to create a specific pattern of physiological response in your body when running in each zone. “Zone 2 training,” for example, has specific (purported) fitness benefits that are connected with the physiological responses seen below LT1, the first lactate threshold.
There is a similar connection between “Zone 4 training” and LT2: the purpose is to be close to, but not above, LT2, the second lactate threshold.
In Part I of this series on heart rate zones, we covered the science of maximal heart rate (HRmax), %HRmax, heart rate reserve (HRR), and %HRR. We also covered the science of LT1 and LT2 (which is really just a method of estimating SSmax), and why LT1 and LT2 must be at the foundation of any scientifically based zone training model.
(Note that reading Part I is not a strict pre-requisite; if you are familiar with the basics of %HRmax, %HRreserve, LT1, and LT2 you should be able to follow along just fine).
We concluded Part I with two simple questions: (1) what heart rates correspond to LT1 and LT2? and (2) can specific heart rate zones correctly position you below or above these key thresholds?
Our purpose in this article is to answer these questions. More specifically, we are interested in the individual variation in the heart rates that correspond to LT1 and LT2. It would be wonderful—and suspiciously convenient—if LT1 and LT2 corresponded to exactly 70% and 90% of HRmax, as some zone models claim. Unfortunately, heart rate zones are neither this convenient nor this consistent.
In short, LT1 and LT2 can occur at a shockingly broad range of heart rate values, even when expressed as %HRmax or %HRreserve. Here's the key data:
| Threshold | Maximal heart rate (HRmax): 90% range | Heart rate reserve (HRR): 90% range |
|---|---|---|
| LT1 | 69 – 94% HRmax | 65 – 90% HRR |
| LT2 | 80 – 98% HRmax | 73 – 95% HRR |
...and those are 90% ranges, meaning that one in ten runners will be outside these ranges!
If Zone 2 training is important to you, you need to be running at less than 70% HRmax, or less than 65% HRR to be 95% sure you are in Zone 2.[1]
These numbers come from a meta-analysis of 412 runners from 25 different scientific studies, the results of which I present below.
If these ranges seem broad, and this zone 2 cutoff seems very low, that’s because it is—the real story about the heart rates that correspond to LT1 and LT2 is that there is massive individual variation, even among runners of similar fitness levels.
Indeed, one of the goals of training is to push LT1 and especially LT2 to a very high percentage of HRmax, so when viewed in that light, this individual variation is not so surprising.[2]
So, let’s dive in and see why the variation in LT1 and LT2 is so large and what, if anything, can be done about it.
Gathering data on heart rates at LT1 and LT2
In an ideal world, there would be one massive, carefully-controlled study on heart rates associated with LT1 and LT2 in a large, representative sample of athletes.
Amazingly, there is just such a study…but it’s in cyclists! It’s not a safe move to assume that L1 and LT2 occur at the same heart rates in cycling as they do in running; however, we should already be prepared to see a wide range of individual variation in runners, given that the study linked above found that LT1 can occur at anywhere from 60%-90% HRmax in cycling!
To get reliable numbers for runners, though, we need the right data. I spent several days chasing down every study I could find that met the following criteria:[3]
- Studied runners (not cyclists, sedentary people, etc)
- Measured LT1 and/or LT2 using blood lactate data (not gas exchange-based estimates like VT1)
- Reported LT1 and/or LT2 as a percentage of VO2max, HRmax, or HRreserve
This search turned up 25 studies, totaling 412 different runners.
Sources of variation in heart rate, LT1, and LT2
One nice thing about doing this large-scale meta-analysis across many different studies, in comparison to having just one large study like the Ianetta et al. study above, is that pooling data allows us to understand and decompose several different sources of variation in the heart rates that correspond to LT1 and LT2.
These sources of variation include:
- The true individual variation in the heart rate (%HRmax or %HRR) that corresponds to LT1 and LT2, among runners of the same fitness level
- The variation in LT1 and LT2 heart rates across runners of different fitness levels (or who differ in age, sex, etc).
- The variation from one lab to the next in the definition and measurement of heart rate, LT1 and LT2.[4]
- The variation from one study to the next in terms of the population of runners that tend to participate in their studies.
Now, we can’t fully disentangle all of these levels of variation, but modern meta-analysis tools give us some pretty powerful ways to parse out these sources of variation.[5]
Lab testing as the “gold standard” for LT1 and LT2 heart rates
Every study used some variant of the following protocol for determining LT1, LT2, and the heart rates associated with them:
- Set the treadmill to a 1% grade to account for the lack of air resistance when running indoors
- Do a very easy warmup on the treadmill, and optionally, take a baseline lactate sample afterwards
- Run for 3–5 minutes at a constant speed, then take a blood lactate sample during a short (~30 sec) rest period
- Increase the treadmill speed by ~0.5 mph and repeat step 2
- Continue until the runner is too exhausted to complete a full stage.
During each stage, heart rate is recorded using a chest strap, and the average heart rate over the last ~30 seconds of each stage is taken as the heart rate associated with that speed.[6] The highest heart rate reached before exhaustion is HRmax.
If you have a few hundred dollars burning a hole in your pocket, you can get your own in-lab lactate testing done, which will (hopefully!) follow a very similar procedure.
Uncertainty in true LT1 and LT2 pace
Results from an in-lab lactate test look like this:
And the semi-well-accepted descriptive definitions of LT1 and LT2 are:
LT1: the speed associated with the first significant increase in blood lactate above the baseline value
LT2: the speed associated with a sudden and sustained increase in blood lactate
Now, what exactly counts as a “significant increase,” and what counts as a “sudden and sustained increase”?
Well, let’s consult the scientific literature. Surely there are standardized, objective methods for defining LT1 and LT2 when given a nice, clean lactate curve, right?[7]
Why yes, there are at least 14 competing standards:
Some of these definitions get extremely complicated, but two reasonably reliable heuristics (according to this study) are as follows:
LT1: First stage with a blood lactate level ≥ 0.5 mM above the baseline value
LT2: First stage with a blood lactate level ≥ 1.5 mM above the baseline value
These heuristics are also nice because, unlike more complicated curve-fitting approaches, they are amenable to use “in the field” when checking blood lactate levels during a workout with a lactate meter.
Applying these heuristics to our sample data above gives the following values for LT1 and LT2:
However, a reasonable lab applying different standards for LT1 and LT2 might come up with different values, even given the exact same data. So, this explains one of the intrinsic sources of variation in heart rate at LT1 and LT2: different labs will come to different conclusions on what speeds correspond to LT1 and LT2.
In the example above, for instance, a subjective analysis (not using a strict cut-off value like baseline + 1.5 mM) would probably label 10.7 mph (5:37/mi) as LT2. In fact, these data are from an athlete I work with, and based on his workouts, 5:37 was a much better estimate for his LT2 at the time of this test!
This variation is important, because what we’re trying to reconstruct is the range of values that would be actually measured for your LT1 and LT2 heart rates, which are a function of both true individual variability from one person to the next, and variation in how LT1 and LT2 are defined from one lab to the next.
In other words, even if you do shell out for personalized in-lab testing, you still need to account for the fact that you might’ve gotten different results had you gone to a different lab.
Is lab testing really the gold standard?
I covered this in Part I of this series, but one point worth reiterating is that LT2, as determined by a single-visit treadmill protocol like the one above, is not the gold standard for estimating steady-state max (SSmax)—the current gold standard is critical speed testing.[8]
In fact, a standard LT2 test only estimates your true SSmax within 5–10% or so. So, that’s another area of variation that won’t show up in our current analysis.
Fortunately, single-visit lab testing is the current gold standard for LT1, though it may not stay that way forever.[9]
Heart rates as %HRmax at LT1 and LT2
Now we’re finally at the point where we can dive into the data. Here are the rapid-fire results, expressed as 90% ranges (i.e. a range that is valid for nine out of ten runners):
LT1 can occur at 69–94% HRmax
Seven studies on a total of 140 runners measured LT1 as a percentage of HRmax. Here’s what the individual variation looks like:
About half of the variation in LT1 comes from true individual variation—the other half comes from variation from one lab to the next in terms of how LT1 is defined.[10] The relatively large amount of lab-to-lab variation in LT1 is not too surprising, since LT1 seems particularly hard to define (what exactly is a “significant” rise in lactate over baseline?).
LT2 can occur anywhere from 80–98% HRmax
Nine studies on a total of 160 runners measured LT2 as a percentage of HRmax. Here’s what the individual variation looks like:
LT2 is a little bit easier to define (and more studies included subgroups), so for LT2, more of this variation is “true” individual-level variation. Only about a quarter of the variation in heart rates associated with LT2 is attributable to differences from one lab to the next in the definition of LT2.
One runner’s LT1 heart rate can be above another runner’s LT2 heart rate
One other useful comparison comes from superimposing the LT1 and LT2 heart rate distributions. This plot illustrates an important point: there is substantial overlap in heart rates at LT1 and LT2 across different runners.
In other words, it’s perfectly possible for one runner to be running at 85% HRmax and be below LT1, while another runner can be running at 85% HRmax and be above LT2!
Heart rate at %HRreserve at LT1 and LT2
LT1 can occur anywhere from 70–91% HRR (or maybe 65–90% HRR)
Disappointingly, only one study reports LT1 as a percentage of heart rate reserve—a 1990 paper on 31 male runners. This study tended to come up with estimates that were quite high across the board (for LT1 and LT2, expressed relative to VO2max, HRmax, and HRR).
I’ll present those results first, then leverage the %HRR–%VO2max correspondence to come up with a (potentially) better estimate, since nine studies on 159 runners reported LT1 as a percentage of VO2max.
The story is more or less the same when using VO2max percentages as a proxy for %HRReserve:
Happily, the VO2max-based results line up reasonably well, suggesting that %HRR is indeed a pretty good estimate for %VO2max. Even so, we still have the same issue: massive individual variation!
One advantage to the %VO2max data is we can break down the sources of variation. In this case, the majority of it is coming from the true individual variation in LT1, not lab-to-lab variation in LT1 definitions.[11]
If you’re serious about Zone 2, you need to be running at <70% HRmax or <65% HRR
From the LT1 data for HRmax and HRreserve, one useful fact we can derive is a lower 95% limit: your heart rate needs to be <70% HRmax or <65% HRR for you to have a 95% chance of being below LT1 (i.e. within “Zone 2”).
You might find that this is absurdly low, but that’s the entire point here—the individual variation in heart rates, even in relative terms, is so large that the value of the entire endeavor becomes questionable without individualized data.
LT2 can occur anywhere from 73–95%% HRR (or maybe 79–98% HRR)
We have the same issue as above with LT2—that same 1990 study on 31 male runners is the only one to explicitly report LT2 as a percentage of heart rate reserve.
That study finds LT2 occurs at 79–98%. The upper range seems a little high to me, but the %VO2max data (18 studies, 293 runners!) gives an estimate that's not too far off: 73–95% HRR.
Breaking down the sources of variation, we again find that individual variation dominates, though with more lab-to-lab effects in this case:
Lastly, the same LT1/LT2 overlap issue persists. This is the case for the true HRR data:
And also the much more numerous %VO2max data:
In other words, even when using heart rate reserve, it’s quite easy for one runner to be below LT1, and another runner to be above LT2, at the same percentage of heart rate reserve—or, for the physiologists out there, %VO2max!
Where does the individual variation in LT1 and LT2 heart rates come from?
To my surprise, none of the “usual suspects” were reliable predictors of where LT1 and LT2 occur relative to VO2max. Age, VO2max, proportion of male subjects, and subjective performance level (recreational, trained, well-trained, elite) were not significant moderators of LT2 as a percentage of VO2max.
Part of the reason for this lack of a moderating effects might just be range restriction—the lowest group VO2max was 54.9, still nearly double the average sedentary person’s VO2max. Ditto for age: with a range of 21.6–39.9 years, we’re not exactly pushing the limits in either direction.
This large study on a very diverse group of subjects indicates that LT1 (estimated using ventilatory threshold on a bike) does increase as a percentage of VO2max in better-trained individuals, for example.
Another point on this front is that 18 studies on 293 runners just isn’t that much data when you’re doing meta-regression. So I don’t want to make too much of these findings.
Lastly, it's also possible for true individual effects to not show up in a group-level analysis; this is the so-called "ecological fallacy."
These findings do, however, suggest that a large proportion of the individual variation in LT1 and LT2 is truly individual, explained by differences in intrinsic physiology and training habits—not “demographic data” like age, sex, etc.
Are these results plausible?
Before getting too excited, let’s take a step back and look at some of the benefits and limitations of this analysis.
Some of the high LT1 values from some studies seem suspicious to me: my guess is that those labs use a working definition of LT1 that’s actually closer to LT2. In a large study on cycling, for example, LT1 ranged from 45–74% VO2max (vs. our range here of 69-94%). Some research in triathletes suggests that LT1 (as %VO2max) can indeed be ~12 percentage points higher in running than in cycling LT1; this is why I didn't just generalize from the cycling study cited below.
With LT2, I’m more confident in the results. It’s clear that some athletes really do have an LT2 in the upper 90% range: the average runner in Jones 2021—the Breaking2 study—had an LT2 at 94% of HRmax!
Likewise, given the wide-ranging results for LT1 in the general-population study from earlier, and similar findings in trained cyclists (70–96% VO2max for MLSS), the wide range of LT2 values is totally plausible.
Even though seeing lab-to-lab variation is great, I would still love for “one study to rule them all,” like Iannetta et al.’s cycling study mentioned earlier. Individual-level data from a very large study would also help pin down some of the sources of individual variation better than a meta-regression.
Recap: Your heart rate zones are probably wrong
Hopefully, the results above are enough to convince you that there is massive individual variation in the heart rates that correspond to LT1 and LT2, even if you set aside things like lab-to-lab variation in the criteria for LT1 and LT2.
Given the width of the heart rate ranges, it would be futile to go through every possible definition of heart rate zones and evaluate them—the conclusion would be that all heart rate zone models are going to be wrong for a significant fraction of runners.
For nine out of ten runners, their LT1 and LT2 will fall within the following ranges:
LT1: 69–94% HRmax or 65–90% HRR (alternatively, 70–91%% HRR, see above)
LT2: 80–98% HRmax or 73–95% HRR (alternatively, 79–98% HRR, see above)
To be 95% sure you are in Zone 2, you need to be running at below 70% HRmax or below 65% HRR.
And even these ranges will be wrong for about one in ten runners!
What’s to be done about it? Well, long-time readers are going to know what I recommend: if you insist on physiologically based training, just use 90% 5k pace as your LT2 estimate, or use my handy threshold, CV, and VO2max pace calculator. But if you’re dead-set on using heart rate instead of pace, the following strategy is much better:
- Do a stage-wise test (as above) to determine your HRmax, while also recording your heart rate as a function of speed.
- Plot your heart rate as a function of speed to find out the (nominal) heart rate associated with any given speed
- Do critical speed testing (or estimate it from a race result), then use your heart rate – speed plot to infer the correct heart rate associated with your steady-state max.
- Remember that heart rate will drift upwards by up to 10–15% over the course of 10–20 minutes of continuous running, even in cool conditions
What about LT1? Well, you can use the threshold above (70% HRmax or 65% HRR) if you insist on a conservative estimate. Beyond that, though, you’re out of luck for now: there are no valid field tests for LT1 heart rate—so far, at least.[12]
The good news is that I am working on a mathematical technique for estimating LT1 from performance data, and should have news on that front in the relatively near future.
This new technique should allow heart rate devotees to follow a similar procedure as above: find LT1 pace in ideal conditions, then use your heart rate – speed plot to infer your LT1 pace. Join my email list to be the first to find out when my LT1 prediction model is finished.
Supporting data and R code for this post are freely available on my GitHub.
Learn more about the science of running
If you enjoyed this article, subscribe to my email list! It’s the best way to find out when I’ve got a new article on training, exercise science, or when I have a new tool ready like my upcoming LT1 calculator.
I also have a book, Modern Training and Physiology for Middle and Long-Distance Runners, that focuses on the science of performance and training for events from 800m to the 10k. Check it out!
Footnotes
[1] Using the most common terminology of a five-zone system. Physiologically speaking, there are actually only three defensible “zones,” in the sense of intensities that produce a qualitatively similar response from the body, but that’s a story for another day. In a five-zone model (as used by Garmin watches, for example), LT1 defines the upper end of Zone 2 and LT2 defines the upper end of Zone 4. The same applies to Coros zones, which uses a 6-zone model. More physiologically based heart rate training often uses a three-zone model, in which LT1 defines the upper end of Zone 1, and LT2 defines the upper end of Zone 2.
Also note that using the lower end of a (centered) 90% range as a boundary means there is a 95% chance your LT1 lies above (e.g.) 70% HRmax since the 90% range excludes 5% symmetrically on either side.
[2] As we’ll see later in this article, across all the studies I reviewed, the athletes with the highest LT2 as a percentage of HRmax were…16 of the greatest marathoners in the world. Given the goals of their training, it’s not surprising that these athletes are able to sustain a metabolic steady-state at an exceptionally high percentage of HRmax.
[3] This was not a systematic review in the sense of searching Pubmed + Web of Science + … for a well-defined search term, mostly because doing a real systematic review is a giant pain. All the meta-analysis and meta-regression I did is 100% by-the-books science though. My guess is that I captured ~80% of the articles I would’ve gotten had I done a systematic search. At some point I might write this up as a real systematic review with meta-analysis and meta-regression and publish it in a peer-reviewed journal, but I doubt the conclusions will change very much.
[4] Lab-to-lab variation accounts for differences in the nominal definition of LT1 and LT2 (e.g. one lab using 4.0 mM vs. another lab using baseline + 1.5 mM), and also differences in the equipment and experimental protocols. As an example, two of the most popular benchtop lactate analyzers actually assess blood lactate differently: one measures it in whole blood, while the other measures it in blood plasma; some papers suggest the same blood can turn up different lactate readings on these two different types of machines.
[5] On a technical level, I’m doing a random effects meta-analysis (using R package metafor) using a second level of within-study random effects for studies with multiple groups (e.g. Støa et al 2020, who report separate results for recreational, national, and elite athletes). After fitting the meta-analysis or meta-regression models (when testing for moderators), I use the pooled-variance strategy from Siegel et al 2021 to estimate the individual-level variance; in other words, the LT1 and LT2 ranges are “90% reference ranges.” The variance of the reference range is just the sum of the tau-squared parameter of the meta-regression model (or tau_1 and tau_2, for multilevel versions) and the pooled variance; I use tau_1 (the between-papers variance) as an estimate of the variance in LT1 and LT2 from lab to lab. One possible interpretation of tau_2—the variance from group to group within the same study—is that it represents variation at the level of different groups, e.g. between recreational and well-trained runners, or between the different pools of subjects that each group draws upon. For instance, in a study on recreational runners from a local run club and well-trained runners from the university track team, there are other relevant differences aside from performance level between the pool of subjects for each group.
(I have a full technical write-up here if you are interested)
[6] Careful readers will notice a flaw here: if you continue running for 10–20 minutes at a given speed, heart rate will continue to drift upward. This “slow component” of heart rate, sometimes also (and maybe incorrectly) called “cardiac drift,” adds another complicating factor to heart rate training that will not show up in a progressive treadmill protocol. I have an entire article on cardiac drift and the slow component of heart rate; join my email list to find out when that comes out.
[7] Incredibly, many physiology labs don’t even apply objective criteria to define LT1 and LT2—they just eyeball it! (ahem, sorry, they are “determined by visual inspection”).
[8] Some percentage of physiologists argue that critical speed testing should not be considered the gold standard; they typically advocate for maximal lactate steady-state (MLSS) as the gold standard for SSmax instead.
[9] You might imagine an MLSS-style testing protocol that aims to determine the minimum speed that generates a VO2 slow component, for example.
[10] And also perhaps what type of runners they recruit for their studies—no studies reported sub-group data for LT1 as a percentage of HRmax, so some group-by-group variation “leaks” into the lab-to-lab variation.
[11] I’m not sure lab-to-lab variation is so much smaller for LT1 as a %VO2max vs. LT1 as a %HRmax. There might be systematic differences between the studies (or subject populations) who participated in one group of studies vs. the other, but beyond that I'm stumped!
[12] I occasionally get questions about the “one hour heart rate drift test” which is popular among some advocates of heart rate training. There is no research on the validity of this method, and other heart rate based field tests, like the Conconi test, do not have a good track record of holding up against gold-standard estimates. I’ll hold off on a full analysis until my heart rate drift article comes out, but the short version is that I have several reasons to believe that the one hour heart rate drift test will not accurately predict LT1. This paper is excellent and apropos.
Related articles
New web app: Predicting LT1 pace and Zone 2 pace from 5k time
Designing a plyometrics program for improving bone strength in young runners
Three theories of tissue damage accumulation during running
Lecture: The science behind modern marathon training
What assumptions are baked into your race prediction model?
Tendons do not store energy for free
In windy conditions, running at a constant effort is usually better than running at a constant speed
A comprehensive guide to the science of cadence for runners
Understanding tissue loading, tissue damage, and running injuries
Noisy data and noisy methods give bad conclusions. I don't see anything in Jones 2021 as evidence against the null hypothesis which should be that VT2 ~= OBLA (the actual metabolic threshold of interest) is always very close to 87.5% HRR in elite runners, just like in amateur runners. The authors' weird use of LTP, sometimes at >6 mM, and averaging of extremely noisy individual curves, just adds to the conclusion that studies that look at blood lactate curves are unreliable both from measurement and methodological standpoint. VT1 on the other hand will of course vary dramatically.
Hi John,
First of, really appreciate all your work, thanks for all the resources here and elsewhere. I really like your in depth approach and sharing your scientific insights with us.
Reading your approach to LT1 and LT2 and reading other sources there is one thing I cannot seem to reconcile.
From several of your blogs on the subject you seem to take the position: the observed variability in heartrates from runners at actual LT2 is rather wide. So you conclude, don't start from heartrate as a means to estimate LT, too unreliable and variable.
I think a lot of runners who like to read about the stuff and try to setup their own schedules will also come across the Joe Friel method of using LTHR and the accompanying 30 minute field test as a means to start 'zone based training'. Trying to validate that method I found: A COMPARISON OF METHODS FOR ESTIMATING THE LACTATE THRESHOLD (Mcgehee, Tanner & Houmard (2005). They find that the 30 minute field test provides a relatively accurate way to estimate LT2 as compared to actually measuring it.
Am I missing something here? How do you reconcile your research with the study above?
I think your view on this might help others as well. At least I find that the LTHR approach usefull because all sportswatches allow you to measure and set it. So it provides a cheap way to make sure you are not overdoing it, even when following your pace based Full spectrum approach to training.
Curious about your thoughts on this.
Laurens
I personally am not a huge fan of the 30 min LTHR test BUT even if you are, I think the two views are reconcilable: my conclusion from this LT1/LT2 analysis is that there is a lot of *individual variability* in LT1 and LT2 as percentage of HRmax (and also HRreserve, and also VO2max) so attempting to use a static percentage of HRmax for everyone (say, "LT1 = 80% HRmax" or whatever) you aren't going to get good results. However, if you do an individualized test (maybe LTHR, or maybe LT1 test in the lab using lactate meter testing) and you get an *individualized* estimate -- "my LT1 is 84% HRmax" -- then that can work much better!