Critical speed for runners, explained

critical speed
D prime
critical power
marathon pacing
running physiology
evidence

Critical speed for runners, explained

A runner in black shorts mid-stride along the inside lane of a red athletics track on a sunny day

The speed where running changes character

Run a little below a certain speed and your body settles: oxygen uptake levels off, blood lactate stabilises, and the effort, while hard, holds steady. Run a little above it and nothing settles. Oxygen uptake climbs towards its maximum and exhaustion arrives on a fairly predictable schedule. Physiologists call that boundary critical speed (CS), and it can be worked out from a few best efforts.

For nearly all runners, the biggest gains come from running more, consistently, over months and years. Critical speed is a second-order lens on top of that: it helps explain what your running has built and how to spend it on race day, and it is largely the product of that running in the first place.

Key takeaways

  • Critical speed marks the boundary between running you can sustain in a physiological steady state and running that leads steadily towards exhaustion. D′ (“D prime”) is the fixed distance you can cover above it.

  • Both come from a straight-line fit of distance against time for three or more all-out efforts, typically lasting about 2 to 15 minutes.

  • In 12 elite male marathoners (bests of 2:03:38 to 2:08:21), marathon pace averaged 96% of CS calculated from their shorter personal bests.

  • In more than 25,000 recreational marathoners averaging about 3:50, marathon pace averaged about 85% of CS: 93% for runners finishing in about 2:30, 79% for those slower than 5 hours.

  • Recreational runners who ran the early kilometres faster than 94% of CS were more likely to slow by over 25% late on. The data are observational and don’t make a pacing rule.

Where the idea came from

Critical power: Monod and Scherrer, 1965

The modern concept starts with H. Monod and J. Scherrer at the CNRS work physiology laboratory in Paris. Their 1965 paper in Ergonomics studied single muscle groups and found that the total work done before exhaustion rose in a straight line with time to exhaustion. They called the slope the critical power, the maximum rate a muscle group “can keep up for a very long time without fatigue”, and the intercept a finite reserve of work available above it.

Later research extended the model to whole-body exercise, including treadmill running by the mid-1980s, and revised the “very long time”: work at critical power is typically sustainable for no more than about 20 to 30 minutes. In running, the same quantities are called critical speed (metres per second) and D′ (metres).

Andrew Jones and the lab work

Much of the running-specific evidence comes from Andrew Jones, now at the University of Exeter. In 2001, Clare Smith and Jones tested eight trained male runners (average VO2max 54.9 ml/kg/min) on a treadmill. CS (14.4 km/h) averaged close to maximal lactate steady state (13.8 km/h) and the lactate turn point (13.7 km/h), but individual differences were too large to swap one for another in a given runner. In 2019, Jones and colleagues argued that critical power, and by extension CS, is the best marker of the maximal metabolic steady state, a position still debated by researchers who favour lactate-based measures.

Tinman’s critical velocity

Runners may also know the term from US coach Tom “Tinman” Schwartz, who coached Drew Hunter as a high schooler. His “Critical Velocity” is a different construct. Writing in 2016, Schwartz said he created it in 1989–90 while coaching at the University of Wisconsin–La Crosse, defined it as about 90% of VO2max (a pace serious runners hold for roughly 30–35 minutes), and hadn’t heard of the scientific term when he named it. Like any fixed percentage, it is generic. kaizen is hyper-personalised: it uses the overall intensity of the running you actually do, from pace and elevation, to set your training target for the week, and recalculates as you run.

How critical speed and D′ are calculated

The classic method is three to five all-out efforts on separate days, each lasting about 2 to 15 minutes, with distance plotted against time. The points fall on a nearly perfect straight line. The slope is CS; the intercept is D′.

A made-up example shows the arithmetic. Suppose a runner covers 3,000 m in 11:20 (680 seconds) and 5,000 m in 19:20 (1,160 seconds). The extra 2,000 m took 480 seconds, so CS is about 4.17 m/s, or 4:00 per kilometre. D′ is the distance left over: 5,000 m minus CS × 1,160 s, about 167 m. Researchers use three or more efforts, since two points always make a perfect line.

Above CS you spend D′, and the further above it you go, the faster it runs out. Jones and Vanhatalo’s 2017 review in Sports Medicine used personal bests from 2 km to 15 km to put Haile Gebrselassie’s CS at 5.91 m/s (about 2:49 per kilometre), with a D′ of 351 m. The review places CS at roughly 70–90% of VO2max depending on training status, above the lactate threshold.

What critical speed says about marathon pacing

Elite marathoners: about 96%

In the same review, Jones and Vanhatalo took personal bests between 1,500 m and 15 km from 12 elite male marathoners, at least four distances each, and compared the resulting CS with each runner’s marathon best. Marathon pace averaged 96% (±2%) of CS.

The authors flagged the weaknesses: the shorter bests often came years before the marathon best, distances differed between athletes, and CS may fall during a long race. The sample is 12 men, with no women or sub-elites.

Recreational marathoners: about 85%

Barry Smyth and Daniel Muniz-Pumares (Medicine & Science in Sports & Exercise, 2020) used training logged on an online platform by 31,190 runners before the Dublin, London and New York marathons between 2014 and 2017, with more than 25,000 included in the analysis. These were recreational runners: average age around 40, average finish about 230 minutes, and mean logged running of roughly 34–47 km a week over three to four runs, depending on sex, city and year.

For each runner they took the fastest grade-adjusted 400 m to 5,000 m segments from the 16 weeks before the race and calculated CS. The best model predicted marathon time with an average error of about 7.7%, roughly 18 minutes for a 230-minute runner. They found:

  • Runners raced at 84.8% of CS on average.

  • Runners finishing in about 150 minutes raced at 93.0% of CS; those slower than 300 minutes at 78.9%.

  • Running from 2 km to 16 km faster than 94% of CS went with a higher chance of slowing by more than 25% over the final 12.2 km: 26.0% versus 20.5% of men, and 15.6% versus 9.6% of women.

On the made-up 4:00-per-kilometre CS above, 96% works out at about 2:56 for the marathon and 85% at about 3:19. A fraction borrowed from elite data can set a very different target from what a recreational runner can hold.

Reading these numbers carefully

The authors are candid about the limits of this observational data.

  • Efforts weren’t verified as maximal. A runner who never ran hard over short distances would have CS underestimated, which may explain why 0.82% of runners raced faster than their supposed CS.

  • The logs are incomplete. Only logged running counted, and anonymised data can’t show whether one runner appears in several races.

  • The pacing link is an association. A runner who starts fast might slow because of the fast start, or because of heat, illness or an over-ambitious goal that also drove the fast start.

  • Volume is a likely confounder. Faster finishers were probably fitter, higher-volume runners, and fitness may protect CS from falling over a long race, one explanation the authors offer. Weekly volume was reported but not adjusted for in the pacing analysis.

The authors suggest staying within about 94% of CS early on. That is a hypothesis drawn from population averages, and percentage-of-CS pace bands are generic in the same way. kaizen calculates your fitness from your recent running and past races, and the weekly training target it sets comes from the pace and elevation of what you actually run, recalculated as each run arrives.

What critical speed is good for

CS is a clear way to think about intensity: easy and steady running sit below it, hard intervals and short races above it. Jones and Vanhatalo’s review extends the model to intervals, tracking how D′ is spent during hard reps and partly restored in recoveries, though recovery rates vary widely between individuals. What raises CS is endurance training, and for most runners the most reliable way to build that is more running, week after week. That last point is kaizen’s interpretation of the evidence above.

Sources and uncertainty

  • Monod and Scherrer 1965, Ergonomics. Origin of critical power, from single muscle groups; not a running study.

  • Smith and Jones 2001, European Journal of Applied Physiology. Eight trained men on a treadmill; poor individual agreement with lactate markers.

  • Jones and Vanhatalo 2017, Sports Medicine. Review plus 12 elite male marathoners’ public personal bests; 96 ± 2% of CS.

  • Smyth and Muniz-Pumares 2020, Medicine & Science in Sports & Exercise. More than 25,000 recreational marathoners (average about 3:50); CS from unverified training efforts; observational.

  • Jones et al. 2019, Physiological Reports. A debated position paper.

  • Schwartz 2016, LetsRun.com. A coach’s own account.

No study here tested whether pacing by CS improves marathon times, and the elite data cover only men.

The bigger picture

Critical speed explains why some paces feel stable and others don’t, and why a fast start costs so much later. The number itself comes from the running you have done. For almost every runner, more running, done consistently, is what raises it, and pacing by percentages of CS is a second-order refinement that only pays off once that running is in place.

Where kaizen fits

If you’re spending months chasing seconds, kaizen calculates your fitness from your recent running and past races, then sets the weekly training target your goal requires, built from the pace and elevation of the running you actually do. It recalculates as each run arrives, and you choose the sessions. See your fitness in kaizen and build the consistent weeks that move every number in this post.