Individual variation in aerobic and anaerobic energy contributions to different events

Many books on training and exercise physiology (including mine) begin with a discussion of aerobic and anaerobic energy production. The next logical question to ask is how much these two energy systems contribute to the various running events—how “aerobic” is, say, the 800m? Such a question is sure to spark heated debate online, but I have always been skeptical of such figures, for two reasons.

First, some of the methods that were traditionally used to come up with the aerobic vs. anaerobic contribution to a particular event have not held up to scrutiny.

Take the concept of “oxygen debt”—the idea that you could measure anaerobic energy expenditure by measuring post-exercise oxygen consumption, as your body “paid” for its anaerobic energy production. This concept was known to be shaky as early as 1984, and though later research has improved on the accuracy of these metrics considerably, many of the papers that coaches and training books cite do not use these modern, improved methods.

Second, I have a broader disagreement with the search for one number: to say, for example that the 800m is “65% aerobic” (as Joe Vigil’s book does) or ~50% (as Coe/Martin’s book does) glosses over the fact that there is tremendous individual variation in how much aerobic versus anaerobic energy goes into a given event.^[1]

In fact, that range of variation looks like this:

Event	% Anaerobic (range for nine out of ten runners)
800m	15–33%
1000m	10–25%
1500m / mile	8–20%
3000m / 2 miles	4–10%
5000m	3–6%
10,000m	2–4%

Where do these numbers come from? Read on to find out!

Individual variation matters for training

To illustrate the range of individual variation among runners in the same event, let's look at an example.

At random, I pulled up profiles from two American college runners who both ran 1:46 last year in the 800m. Their 1500m times? 3:36 and 3:47.^[2]

At 10k cross-country, the difference is even more dramatic: 29:45 vs. 32:00. One of these two runners (the 32-min 10k runner) has also competed at 200m. His time? A 22.2!

Clearly, if these runners were to compete side-by-side, finishing within a few hundredths of one another in the 800m, there is no way their aerobic and anaerobic contributions would be comparable.

The next step in reasoning here should also be obvious: it would be absurd to train these two runners in an identical way. Understanding the range of individual variation in a given event matters quite a bit when it comes to individualizing training.

The critical speed model as a way to characterize variation in energy system contributions

I’ve written extensively about the critical speed model for runners, but something important clicked for me when I was building my threshold, CV, and VO2max pace calculator: if you have multiple performances for a given runner, you can use the critical speed model to estimate the fraction of their energy production that’s aerobic versus anaerobic—with a few caveats, of course.

As a quick refresher, the critical speed model posits that exercise performance in events lasting ~2–20 minutes is well-described by a hyperbolic function:

This hyperbolic function has two parameters: the titular “critical speed,” which is the gold standard estimate for steady-state max (SSmax)—the fastest speed at which you can maintain a metabolic steady state; and D’ (pronounced “D-prime”), which represents your finite anaerobic energy reserves.^[3]

The D in D’ stands for distance, because (1) that’s how the math works out, and (2) if you assume that running economy is constant across different speeds (which is almost true^[4]), then distance is the same thing as energetic expenditure: a mile burns the same amount of calories, whether you run it fast or slow.

One way of thinking about what D’ represents is as follows: imagine you were racing against an identical twin of yourself, with the same level of fitness. Instead of running all-out, though, your twin runs at SSmax pace, i.e. the fastest pace that produces a metabolic steady-state—“threshold pace,” if you like. If you run the race all-out, D’ is how much you win the race by, in meters.

This interpretation of D’ as anaerobic energy gives rise to a simple rule: your anaerobic contribution to any given race distance is your D’ divided by the race distance, in meters.

Applying this rule to a large dataset of track times (with corresponding critical speed data) would allow us to measure the individual variation in aerobic and anaerobic energy contribution to a given event.

Fortunately, I have just such a dataset from the high school and college track times I used to construct my threshold/CV/VO2max pace calculator (and you can download that dataset here if you want).

Applying this D’-over-race-distance rule allows us to analyze the individual variation in anaerobic energy contribution to every event from 800m to the 10k.

Here's what that looks like:

Aerobic and anaerobic contributions to middle-distance and long-distance races

Here’s a breakdown of each event, with the middle 90% range highlighted.^[5] The quoted figures are correct for nine out of ten runners. Keep in mind that one in ten people will fall outside these ranges!

The key fact to appreciate here is the individual variation in shorter events: there is a wide range of physiological profiles, even among athletes running the same times.

800m: 15–33% anaerobic

The 800m is a massively variable event: you can easily compete against a runner who’s relying on three times as much anaerobic energy as you are! The wide variation in anaerobic contributions to the 800m helps explain why there’s so much variation in the type of runner who performs well in the event, and why training for the 800m needs to be highly individualized.

1000m: 10–25% anaerobic

It’s informative to contrast the 1000m with the 800m and 1500m / mile: even though it’s closer in distance to the 800m, distributionally it looks more similar to the mile. This matches up well with advice that my college coach would give our middle distance runners—the 1000m is more like a “short mile,” not a “long 800.”

1500m, 1600m, and mile: 8–20% anaerobic

The mile has quite a lot of individual variation as well—that’s why you see a wide variety of training strategies, such as the lower-mileage, higher-intensity approaches used by Bernard Lagat, Seb Coe, and Andy Powell, as well as higher-mileage and more aerobically-oriented approach exemplified by the “Norwegian method” (Jakob Ingebrigtsen and Narve Gilje Nordås, and others).

One thing is clear, though: if you are on the high end of this spectrum, it would be a big mistake to train with an extreme focus on classical aerobic workouts at the expense of faster interval sessions. The converse is also true: if you’re an aerobic miler, you’re unlikely to reach your full potential without doing some 5k/10k-style training at some point during your career.^[6] 

3000m and 3200m: 4–10% anaerobic

Right around two miles is where performance becomes truly dominated by aerobic fitness. The enormous aerobic contribution to the 3k and 3200m should match up well with what most high school and college coaches already know: a great cross-country runner is almost guaranteed to be a great two-miler, but a great miler is not guaranteed to be a great cross-country runner.

5000m: 3–6% anaerobic

The 5k follows the same pattern as the 3k, with aerobic energy dominating the equation. While (relatively) lower mileage approaches can still work for the 5k, they will still need to be aerobically-oriented, no matter the athlete.

10,000m: 2–4% anaerobic

The nearly uniform distribution of aerobic and anaerobic contribution in the 10k is a good explanation for why so many 10k runners train alike: high mileage, long tempo runs, long repeats on the track, etc. While you can still individualize training for the 10k, that individualization always needs to respect the fact that >95% of the energy you need in the 10k is produced by your body’s steady-state aerobic capabilities.

Fast-twitch and slow-twitch is a false dichotomy: physiological profiles are a continuous spectrum

There’s one other point to make regarding the aerobic/anaerobic distributions above. It’s so obvious that you might’ve just accepted it without realizing it: physiological profiles are a smooth, continuous distribution: we see a nice, bell-curve-like shape, not a sharp “bimodal” distribution.

What’s a bimodal distribution look like? Well, here’s an example:

That’s what you get when you look at the raw distribution of 800m performances among college athletes. Bimodal distributions occur when you have one powerful variable that exerts a large effect on the outcome. In this case, it’s obvious: the two sub-distributions are male and female:

But the aerobic/anaerobic contribution plot shows no such bimodal shape. In fact, men and women do not even differ in their D’ values:

So, notions about being a “fast-twitch” or “slow-twitch” runner are an oversimplification: fast-twitch and slow-twitch are two extremes of a continuous distribution, with most runners falling in the middle.

Dichotomization is necessary sometimes: an athlete needs to train with the mid-distance group or the long-distance group, for example.

But keeping this continuous distribution in mind can still help tweak and individualize workouts: if your mid-distance group is doing 16x200m, a long sprinter might do fewer repeats but finish with the final four reps faster, while a true miler might do more repeats or some of them together (e.g. 5x(200-200-300m) vs. 16x200m).

Caveats and limitations

There are three big limitations to this analysis. I’ve touched on them briefly already, but they’re worth elaborating on.

Aerobic energy production continues to increase above SSmax

First, physiologists will rightly point out that there is additional aerobic energy production that happens above SSmax. In a strict exercise biochemistry sense, this aerobic energy should count in the “aerobic” column, not the “anaerobic” column.

But from my perspective as a coach, I would rather put that unsustainable aerobic energy in a different category, since it’s clearly not something we have direct access to via classical “aerobic” training components like high mileage, long runs, threshold repeats, etc.^[7]

Running economy is (probably) not constant at very high speeds

Second, one of the crucial assumptions behind this use of the critical speed model is that running economy is constant across different speeds. This is not a bad assumption across most speeds, but when we’re talking about high-level middle distance running, the speeds involved are so fast that it’s implausible to think that efficiency stays the same. 

Nobody talks about this much, since it’s not possible to experimentally measure running economy at speeds above SSmax—the anaerobic contribution is “invisible” to a metabolic analyzer.

However, biomechanically we know that high speeds involve heavy recruitment of fast-twitch fibers, which are inherently less efficient in an energy-to-force sense.

Moreover, the high absolute muscle fiber velocities at very fast paces will further increase the energy cost of running, since energy-to-force efficiency decreases dramatically at very high muscle fiber velocities.

On balance, these effects mean the critical speed model will underestimate the anaerobic contribution to shorter distances, especially in high-performance runners who are running faster in an absolute sense.^[8] Likewise, this analysis will not work at all for the 400m and below—it’s far too short to reasonably apply the critical speed model.

High school and college track times are susceptible to selection bias

Another reason to believe my anaerobic contribution estimates are in fact underestimates, especially for the 800m and 1500m, comes from a more subtle issue that’s particular to the dataset I’m using.

To be able to fit a critical speed model at all, I needed race times from three different distances (from 800m to 5000m), from the same athlete during the same season.

The issue, though, is that many middle-distance runners never compete in longer events. For example, many 800m specialists only run the 400m, 800m, and (indoors) the 1000m. These runners are exactly the kind of athlete who you’d expect to have a large anaerobic contribution to their event, and they’re systematically less likely to be included in the data since they don’t have enough performances to fit a critical speed model. So, there is a selection bias issue.

…or at least the potential for selection bias. While writing this article it occurred to me that I could empirically check whether there was in fact selection bias on at least one dimension: among the ~6,000 800m performances, was there a trend for faster or slower runners to be excluded from the critical speed analysis?

Turns out the answer is no, not to a statistically significant effect (p=0.13 in case you were wondering). I was very surprised by this finding. Now, it doesn’t prove that no selection bias exists: fast-twitch runners at all levels of performance might all be less likely to be included in the critical speed model, and that’s something I can’t test for.

---

In any case, these limitations only reinforce the overarching message: there is an enormous amount of individual variation in aerobic and anaerobic energy contribution to the same events, and you need to appreciate this individual variation if you want to understand training.

Conclusion: individual variation matters for aerobic and anaerobic training

For a given running event, there is a significant amount of individual variation in the anaerobic contribution to performance—especially for middle-distance events.

The 800m can be considered 15—33% anaerobic, but even with this broad range, one in ten runners will fall outside it! The same is true for the 1500m and the mile: 8—20% is a good starting point, but only for the middle 90% of runners.

At 3000m (or 3200m), the range is 4–10% aerobic, and things are more stable for 5000m (<6%) and up.

When training middle-distance runners, keeping this individual variation in mind is important.

Two athletes, even at the same level of performance, can have a very different physiological profile. As a result, they will need different training (especially during the “general” or base phase of training)^[9] to reach their potential.

So, the next time you hear someone quote a figure like “the 800m is 75% aerobic,” remind them that what matters is not the average—what matters is the value for the individual.

There's one last questions, which is: how do you know what your own physiological profile is?

The answer is simple, conceptually at least: you fit a critical speed model to your own performances to find your D-prime value, then apply it to your race distance of choice. That's something I will incorporate into my upcoming critical speed calculator (and I'll update this post when I'm done with it!).

Learn more about training for middle and long-distance events

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 available like my threshold and CV pace 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] There is actually a third reason I’m skeptical of aerobic/anaerobic contribution estimates, which has to do with what I call “unsustainable aerobic energy production”—basically, the oxygen consumption that happens in between steady-state max (SSmax) and VO2max. It’s not obvious to me that this portion of aerobic energy production is “aerobic” in the same sense as energy production below SSmax. That’s a topic for another day, though.

^[2] I’m rounding off decimals and changing the times a bit here because I don’t want to single out or use any particular runner as an exemplar; people run slow times for all sorts of reasons, and I randomly picked these two guys from a performance lists, so I don’t know the particulars. But in general it is trivial to find runners equally matched at 800m or 1500m who are separated by a vast gulf at 400m or 5000m. 

^[3] It’s not quite right to call these energy reserves “anaerobic,” since they do have some aerobic components as well. Evidence on this front comes from two facts: At extremely high altitudes, both critical speed and D’ decrease, and the opposite is true when you give people extra oxygen using an oxygen mask. The fact that some of this supposed “anaerobic” energy is in fact coming from aerobic energy sources ties in with the idea of “unsustainable aerobic energy” that I mentioned in footnote #1, but for training purposes I think it makes more sense to think of this unsustainable yet aerobic energy as something that’s distinct from the usual “aerobic” fitness that we associate with threshold, long runs, etc.

^[4] There is in fact a subtle U-shaped efficiency curve at speeds below SSmax, but the differences are only ~5% or so.

^[5] i.e. these are the (0.05,0.95) quantiles of the distribution.

^[6] You can periodize your “style” of training throughout the season, or throughout your career—Joe Rubio’s 1500m guide is a good example of the former strategy; fall training under his model strongly resembles classic 5k/10k training, with “miler stuff” getting phased in over the winter and becoming the focus in the spring.

^[7] Another possible aspect to consider here might be individual variation in LT1—the first lactate threshold. LT1 dictates when you start relying on oxidative energy from fast-twitch fibers to a greater degree (and when you start losing efficiency from slow-twitch fibers), and although I’ve never heard anyone talk about it in this manner, it makes sense to me that you might also want to individualize training based on where someone’s LT1 is relative to their SSmax, since that would dictate “what kind” of aerobic energy they’re relying on: traditional fat/carb-mixture energy from slow-twitch fibers, or a more carb and lactate-focused mixture that increasingly relies on fast-oxidative fibers too.

^[8] Indeed, the critical speed model performs poorly at predicting performances in the 200m, 400m, and 600m: as we’d expect from a worsening of running economy, it predicts speeds that are too fast.

^[9] Individual variation matters less as you get closer to race-specific training. Whether you are a fast-twitch or a slow-twitch miler, you need to get to the point where you can do (as an example) 5 x 600m at mile pace with 3–4 min recovery. How you build up to that point is the part that gets influenced more strongly by the individual’s physiological profile.

Related articles