A comprehensive guide to the science of cadence for runners

Your cadence is the number of steps you take per minute while running. Simple measurement, right? But there are many questions surrounding it, including how it differs across runners, how it changes as you run faster, whether a higher cadence is more efficient, whether a lower cadence causes injury, and whether you should aim for a specific number.

It almost goes without saying that there is an enormous amount of erroneous information circulating around when it comes to cadence, to such an extent that—with a few exceptions—I’m mostly going to avoid trying to refute everything wrong about cadence you may have seen elsewhere, and instead focus on the (correct) biomechanics of cadence in running.

Cadence is squarely in my scientific wheelhouse; I spent most of my time during my PhD collecting and analyzing biomechanical data. One of the chapters of my dissertation is even focused on the merits of cadence (and other gait metrics) as a predictor of biomechanical loading. So, buckle up—we’ll be covering everything you need to know about cadence.

Defining cadence and its various synonyms (step rate, stride rate, stride frequency, etc.)

What is one stride? And what is one step? These are not hard questions to answer, but you do need to make some choices to standardize things. In biomechanics, one “stride” is a full gait cycle—starting when your right foot first makes contact with the ground, and ending the instant before your right foot makes contact with the ground again.^[1]

One “step” is just the right-leg, or left-leg, “half” of a gait cycle: a right step occurs from initial contact of your right foot with the ground, up until initial contact of your left foot with the ground. As such, a gait cycle is composed of a right step, and a left step (and nothing else).^[2]

So, your stride time is the amount of time, in seconds, that it takes for you to complete one full gait cycle (Right-Left-Right). Note that for stride time, it doesn’t matter which side you “start” on—your average stride time measured Right-Left-Right will equal your average stride time measured Left-Right-Left, even if you have an asymmetric step time.

Suppose you record yourself running with slow-motion video and find that it takes you 700 milliseconds (0.7 sec) to cover one full gait cycle. That’s your stride time. How many strides per second would you take? Well, it’s a simple matter of doing 1 ÷ 0.7 = 1.43 strides per second. Want that in strides per minute? Just multiply by 60, and you get 85.7 strides per minute.^[3] That is your stride rate—strides taken per unit time. Equivalently, you might call it your “stride frequency”—the frequency at which strides happen—same thing.

If you had a perfectly symmetric stride, your step rate—individual steps taken per second—would be equal to your stride rate on each side, i.e. 85.7 left steps per minute and 85.7 right steps per minute, for a total step rate of 171.4 steps per minute (also, a “step frequency”).

In reality, left/right asymmetries are pretty common, so your left-step rate is not necessarily going to be equal to your right-step rate. However, since one gait cycle is always a left step plus a right step, your two individual step rates always add up to twice your stride rate (171.4 in our example above)—if you were strict with terminology you might call this your average step rate.

In practice, when you hear runners quoting cadence numbers like “160” or “180,” they are always providing this “average step rate,” i.e. double your true stride rate.

This is in fact how watches and heart rate monitors work under the hood: when you dig into the raw data, you will find that what’s recorded is stride rate; when it’s displayed the recorded number is simply multiplied by two.^[4] The units of steps per minute are often abbreviated as “spm,” analogous to heart rate data in bpm (beats per minute).

I use “cadence” for convenience and simplicity

I always follow the intuitive common convention among runners: when I say “cadence” I mean your average step rate, in steps per minute (spm), with the understanding that this number is an average of your left and right steps, and conveys no information about left/right asymmetry.

Scientists can’t agree about the right terminology for cadence

I included the whole mess of cadence terminology above—stride time, step time, stride rate, stride frequency, step rate, step frequency— because you will encounter all of them in various scientific papers that deal with cadence. Here’s one on step rate, here’s another one on stride frequency, here’s one that prefers step frequency, this one uses cadence, and this one uses stride length which (as we’ll see below) is really just cadence in disguise.^[5] Here’s one on step length just for good measure!

So, it can be quite frustrating to try to follow the literature, since you need to simultaneously track seven different names for the same thing!

Speed equals cadence multiplied by stride length

Now, I mentioned above that stride length is just cadence in disguise. What do I mean? Let’s continue with our example above, with a runner taking 85.7 full strides per minute (= 171.4 spm). Suppose this runner is running at 6:42/mi or 4:10/km, which is a speed of 4.0 meters per second. Without getting out a measuring tape or opening any biomechanical software, I can guarantee that this runner has a stride length of exactly 2.80 meters.

How do I know that? Well, let’s quickly define stride length as we did stride rate. Your stride length is the distance you cover from one right-foot contact to the next, measured heel-to-heel.

Concretely: imagine I have you run through a puddle, then continue running on flat pavement, leaving wet footprints. I take out a measuring tape and measure the distance from the back of one right footprint to the next. In our example above, that distance will be exactly 2.80 meters (9’2”).

I know this is the case because speed, stride rate, and stride length are mathematically linked: speed = stride rate × stride length. This is one way of defining how fast a runner is moving—it’s not an approximation, it’s an exact relationship. 

It is a fun proof to show that the units cancel out appropriately:

   

Note that I’ve converted 85.7 strides per minute back to 1.43 strides per second; it’s a simple unit conversion to get this to any units you prefer, including pace as opposed to speed.

Because of this relationship, if I know how fast you’re running, and I know your cadence, then I automatically know your stride length (and vice versa; give me speed and stride length and I know your cadence too). And the same is true for step length—which is half your stride length—subject to the same “average step length & stride asymmetry” provisos as earlier.

Modern GPS watches and heart rate monitors can measure cadence very accurately

One confusing aspect of modern wearable tech for runners is that it is not always clear which metrics are being measured accurately. Issues related to GPS accuracy in downtown areas, dense forests, and deep canyons or canals are well-known, and heart rate data from some watches are notoriously prone to issues like cadence lock.

Fortunately, cadence data are very accurate, even from lower-end equipment. The reason is that the signal is just very, very strong. Here’s what raw acceleration data from the wrist looks like at 8:00/mi pace:

The step-to-step acceleration jumps out clear as day—it’s pretty hard to mess up on the data processing side. Maybe pushing a stroller, or carrying a heavy water bottle might screw it up? But with normal running, no problem.

Likewise, stride length data are also nearly as accurate if you have accurate speed data. That’s a big if, of course; you should not trust stride length data on a treadmill or in situations where you have bad GPS data.

The reason stride length data are nearly as accurate as cadence data is that GPS watches and heart rate straps don’t even measure stride length—they just calculate it from speed and cadence.^[6] Notice how calculated stride length (speed times cadence) matches up essentially perfectly with the device-reported stride length in the data below, from a Garmin Forerunner 245 and Garmin HRM-Run:

It does not make sense to talk about cadence without talking about speed

I just made the case that stride length is just cadence in disguise—as long as you know cadence and speed, you also know stride length.

“Ok, but you might not know what speed I’m going!”, you might object. And this is true! But it is also true of cadence, and that leads me to my next point, which is that it makes no sense to talk about cadence without also talking about speed.

Since stride length and cadence determine your speed, it’s clear that to run faster, you have to change at least one of those two variables. In practice, runners change both, but to different degrees.

Let’s look at a few examples. Here’s a plot of my own cadence data, collected across a range of speeds, with a fitted curve showing my speed-cadence strategy:^[7]

It’s clear to see that I have my own characteristic speed–cadence relationship. Asking “what’s your cadence” is an ill-posed question; it depends a lot on how fast I’m going.

The same is true for other runners. Here are a few other examples of speed–cadence plots from various other athletes (using the open-access data from this study of mine):

And you can see that (a) everyone has their own characteristic speed–cadence strategy, and also (b) everyone increases their cadence as a function of speed, albeit to different degrees.

So, to emphasize: cadence data are meaningless without speed data. Is 170 steps per minute low? It is for 5:00/mi (3:07/km) pace. But that’s a fairly high cadence if we’re talking about 9:00/mi (5:36/km).

How cadence and stride length differ across runners and across speeds

Taking the curves from above and putting them all on the same plot shows how much variation there is in the speed-cadence strategies used by different runners:

To get an even better sense for the range of variation, here's a fitted curve (sans the raw data) for all 49 runners in the dataset. I’ve also plotted, with a thicker purple line, which is the speed–cadence curve for an average runner across this sample of runners.^[8]

From the plot, three things are clear:

1. At a given speed, there is a wide range of cadences across different runners
2. As speed increases, all runners increase their cadence to some degree
3. Runners also differ in how much they increase their cadence as a function of speed

Let’s consider the implications of each of these three factors in turn.

Runners use a wide range of cadences at the same speed

If you get a large group of runners together and have them run a constant speed—say, 8:00/mi (5:00/km)—you’ll observe a wide range of different cadence values. In our runners from above, cadence varies from about 145–195 spm at that pace!

Some—but not all—of these differences in cadence are be explained by factors like height, weight, and leg length, but these relationships are much weaker than you’d think. One study found that, as common wisdom holds, runners with longer legs have a lower cadence, but leg length only explained about 40% of the variation in cadence across runners. Other work pegs the variance explained by leg length even lower: just 20% according to this study.

In the runners from the data above, height (which is obviously closely related to leg length) explains only 24% of the variation in cadence, and as you can see, there’s a wide range of cadence values at 8:00/mi (5:00/km) even among runners of the same height:

Body weight is even less informative: body weight explains only 8% of the variation across runners in cadence at 8:00/mi!^[9]

The real truth is that your naturally-chosen cadence at a given speed probably depends on many “invisible” factors like the stiffness of your tendons, the length of your muscle fibers, and the strength of your various muscle groups—though that’s mostly speculation on my part.

Whatever the cause, it’s clear that runners naturally choose a very wide range of cadence values at a given speed, and basic things like height, weight, and leg length does not explain most of this variation.

All runners increase their cadence and stride length as they run faster

Another thing that’s eminently clear in the speed–cadence plot from earlier is that all runners increase their cadence as they go faster.

Since speed is the product of cadence and stride length, it’s clear that to run faster, at a minimum you must increase cadence or stride length, if not both. In practice, all runners use a combined strategy of increasing their cadence and increasing their stride length, albeit to different degrees.

Runners differ in terms of whether they rely more on cadence or more on stride length to run faster

One more subtle difference across runners is that some athletes use a “cadence dominant” strategy for running faster, while others use a “stride length dominant” strategy. 

The plot above shows how two runners with the same cadence at 9:00/mi (5:50/km) can diverge by nearly 10 spm at faster speeds.

It is not correct to think of “cadence dominant” versus “stride length dominant” as a dichotomy—it is a continuous spectrum, with most runners clustered somewhere in the middle. Indeed, you can reasonably model the speed–cadence curves from earlier by using a Gaussian (“normal”) distribution to describe how runners vary from the typical cadence-vs-stride-length strategy:^[10]

A quick aside: running shoe nerds might notice the connection here to ASICS’ marketing campaigns that center around pitching different shoes for different “styles” of running.

While I’m open to the idea that gait characteristics are predictive of how much you’ll benefit from a given shoe, I have not seen any data that would convince me (either way) that the current incarnation of this idea actually has scientific merit.^[11]

If a cadence of 180 spm is optimal, almost nobody uses it

You knew this part was coming, so let’s not hold back any longer. In the plots above, you’ll notice that the magic “180 spm” number is quite rare among runners until you get to rather high speeds.

The idea that 180 spm is an “optimal” cadence, for performance or for injury, traces its roots to a sidebar included in Daniels’ Running Formula—presented as almost an offhand comment about noticing that runners at the Olympics seemed to have a cadence of at least 180 strides per minute.

This magic number, and the various refutations of it, have taken on lives of their own, so much so that some sectors of running-related social media are an eternal pendulum swinging between “you need to hit 180 spm” and “no one should ever change their cadence.” This “180 spm” magic number idea became so persistent it even made its way into Daniels’ New York Times obituary!

But, as we’ve just seen, it does not make any sense to talk about cadence without talking about speed. And in the plots above, you’ll see that few runners get close to 180 spm until speed creeps up to 7:00/mi (4:20 min/km) or faster. And do you know who tends to do a lot of running at faster than those kinds of speeds? Runners at the Olympics.

Runners (probably) choose the most energetically optimal cadence

Now, to the nearly as bandied-about “refutation” of the magic 180 number: the claim that your body naturally chooses the right cadence, and you should never worry about it or monitor it at all.

The first part of this claim is mostly true—as best we can tell, runners do naturally select a cadence that is energetically optimal.

In experimental studies, when you have people increase or decrease their cadence by a substantial amount (say, 10% or more: asking a runner at 160 spm to increase their cadence to 176 spm), what happens is that they consume more oxygen. In other words, their running economy gets worse.

Figure from Hunter and Smith 2007

There’s even stronger evidence for this “energetically optimal” theory, which is the following: in a long steady-state race like a half marathon, your cadence often decreases as you get fatigued (even when you maintain the same speed).

What happens if you try to bring your cadence back up to your “fresh” preferred cadence that you were using earlier in the race? A study from 2007 has the answer: your running economy gets worse!

These findings suggest that your body is adjusting your cadence (and other aspects of your biomechanics) dynamically, in real time, to find the most energetically optimal way to run, under the current constraints your body is facing.

So, when you are fatigued late in a race, your muscles cannot produce the same forces with the same energetic costs that they could when you were fresh. As a result, the optimal way to run changes, and your body shifts your cadence in its attempt to find the new optimum.

However, you ought to exercise caution when interpreting cadence data from a race (or workout): by far the most common reason for cadence to drop later in a workout or a race is simply that you’re slowing down! Like we saw earlier, it still makes no sense to talk about cadence without also talking about speed.

One last point regarding the degree to which changing your cadence affects your running economy: almost all research to date has focused on acute changes in running economy, i.e. right now in the lab today. Relatively little research has brought people in, measured their running economy, initiated a multi-week cadence retraining protocol, then brought them back in to see whether their economy improved.

I strongly suspect that the negative effects of changing cadence would be at least mitigated, if not entirely eliminated, after several weeks of retraining.

Could a higher cadence be more efficient for running?

Interestingly, a recent meta-analysis found that across different runners, having a higher cadence was associated with slightly better running economy—though the relationship was relatively weak (r = 0.2), meaning differences in cadence explain only about 4% of the variation in running economy across different runners.

That’s on par with another study that found that even considering all aspects of running gait, you can only explain 30–40% of the variation in running economy (the rest being, presumably, hard-to-observe factors like muscle fiber composition and neural pattern generation in the brain). So, it would make sense that cadence is only a small factor in the grand scheme of things.

There’s a few ways to square this finding with the “runners choose their optimal cadence” data from above. One explanation might be that a higher cadence is just a marker of some other factor that’s associated with better running economy—certain body types, more efficient musculature, stiffer tendons, etc. Under this view, there’s no contradiction between the within-person vs. across-person differences.

Another explanation could be that acute changes in cadence hurt your running economy, but over a period of weeks or months, you could learn to more effectively make use of a higher cadence.

To resolve this question, what I would like to see is longitudinal data on cadence at different speeds as runners go from novice to intermediate levels, ideally across a span of a few years.

I have some suspicion that what you’d see—especially in the runners who are running a lot of volume—is a gradual increase in their preferred cadence at a given speed. And probably an improvement in running economy too! But I’d need to see the data first.

Now let’s deal with the second part of the “your body chooses the right cadence automatically” argument—the implicit (or explicit) exhortation that you never need to worry about cadence. I do not think the evidence supports that position. Because…

Energetically optimal cadence does not mean injury optimal cadence

Many people implicitly assume that “optimal cadence” means “ideal for all aspects of running.” But that’s not true. If you are optimizing your cadence for energetic efficiency that necessarily means that you are not optimizing for low biomechanical loading on your body. So, it’s quite possible for your “ideal” running gait to be doing quite a bit of damage to your body, compared with other ways to run.

There is some (weak) evidence that a lower-than-average cadence is a risk factor for injury

I’ll be honest, as a biomechanist I am more persuaded by the biomechanical case for increasing cadence than the epidemiological case, simply because many very obvious risk factors for injury—like weekly mileage—are quite resistant to “showing up” in epidemiological studies on running injury, because of things like selection bias (e.g. who runs high mileage? Injury-resistant runners!).

Research on cadence suffers from many of the same issues. One example: a military study that measured stride length, cadence, and baseline fitness in over 800 military recruits. Those who got injured tended to have shorter stride lengths than those who stayed healthy—but no effect on cadence. How can this be, given that cadence and stride length are mathematically linked?

Well, clearly, the later-injured recruits were running slower (same cadence, shorter strides)—and the study collected data at “preferred running speed,” not a standardized speed. So, the slower runners were at greater risk of injury. Why? Well, a good bet is that they were not very fit, because they had not been running much before basic training. And that’s a major risk factor for injury!

A similarly sized study on recreational runners came to similar conclusions: this study on over 800 recreational runners found no association between cadence and injury risk, though again it suffers from the same “preferred speed” problem (and is also an example of the paradoxical “mileage is not a risk factor” findings).^[12]

The two more promising studies regarding cadence and injury risk both come from more advanced and more homogenous groups: one study on high school runners found that the runners with the lowest cadence (at a fixed speed) were more likely to sustain knee and shin injuries. Another study on collegiate runners found that a higher cadence was associated with a lower risk of bone stress injuries (i.e. stress fractures).

Though both of these studies were smaller, they do have the benefit of being on far more homogenous groups than the large population-based studies above.

One of the issues with broad-based study populations is that cadence could be confounded with any number of other variables: height, weight, age, training experience, etc. In a homogenous group—like a high school or college team—you at least know that you’re dealing with runners of a similar age, fitness level, and training approach.

Nevertheless, the weight of evidence right now is not especially strong in terms of cadence as a prospective predictor of overall running injury risk.

But again, you can say the same thing about weekly mileage, and I still wouldn’t conclude from that result that everyone should start running 100 miles per week since “mileage isn’t associated with injury risk.”

A higher cadence usually means lower biomechanical loading on your body

If you’ve read my article on biomechanical training load, you’ll know that injuries are the product of three things: the number of steps you take while running, the amount of force per step experienced by a given biological tissue (e.g. your Achilles tendon), and the structural integrity of that biological tissue (e.g. Achilles tendon stiffness).

Cadence manipulates two of these variables: total steps and force per step in the tissue. Clearly, if you increase your cadence while maintaining the same speed, it’ll take you more steps—and hence more loading cycles—to cover a given distance (or a given time, of course).

But, in most cases, increasing your cadence while maintaining the same speed leads to lower biomechanical forces in your body. There’s good research showing this is the case for your patellofemoral joint, Achilles tendon, and tibia (which are the #1, #2, and #3 most-common locations of running injury), and it’s plausible for most other load-bearing tissue in the lower body as well.

As we saw in my biomechanical training load article, it’s generally a good trade to take more steps with a lower load per step, since small increases in tissue load can lead to very large increases in damage per step, whereas total damage done is merely a linear function of steps taken.  

The most notable exception to the “cadence lowers tissue damage” argument may be the hip joint. Some research has found increasing cadence leads to more hip loading per step; I’ve also seen the same thing in some data I have on hip flexor forces.

Since hamstring loads are often quite large during the swing phase, it’s also plausible that the hamstrings might be in this category as well (and they are hip extensors as well as knee flexors, after all).

However, the hip is a complex joint and forces in, say, the femoral neck may not react the same way as forces in the hip flexors or the various hip extensor muscles, and I’m not extremely confident in any of these findings.

One other ambiguous case worth mentioning is the metatarsals; this study found that metatarsal strain per step goes down but the actual accumulation of damage is unchanged when changing your cadence.

On the whole, though, many of the “usual suspects”—patella pain, shin pain, Achilles pain, frequent lower-leg stress fractures—are good potential targets for cadence retraining, especially in runners who have very low cadence at their usual training speeds.

When does it make sense to change your cadence?

Changing your cadence shouldn’t be a first-line treatment for most injuries, but it can make sense if you’ve had a string of knee, shin, or Achilles issues, especially if you already have a lower-than-average cadence at your typical training paces.

Very rarely, I could see a case to be made for changing your cadence if you didn’t have a history of injury but had a profoundly low cadence compared with other runners—and you aren’t very fast (since you have less to lose from a running economy perspective).

In these cases, a cadence change would be “prophylactic”—essentially a bet that you’d run into injury troubles in the future.

But let’s focus on the most common case: a chronically injured runner with lower than average cadence at their usual training paces.

Designing a modern cadence retraining protocol for runners

Gait retraining has been a pretty hot topic in biomechanics for the last decade or so, and as a result there’s enough research to get a good picture of the kinds of protocols used to change a runner’s cadence.

Cadence retraining is a fairly involved process that takes several weeks, but the good news is that it’s quite cheap—maybe even free if your watch already has cadence data (and nearly all do).

The first question is “how much should you increase your cadence”? Research provides a pretty solid answer: aim for an increase of 5–10%. Various studies have used 5%, 7.5%, and 10%, with the motivation being that changes above 10% produce unacceptably large increases in the metabolic cost of running.

Broadly, you have three tools to use for implementing a cadence retraining program:

• Real-time feedback from the cadence data on your watch (and your pace, too)
• External cadence cues in the form of a digital metronome or smartphone app
• How and whether you use a “faded feedback” schedule that “fades out” real-time feedback and external cues to help internalize your new cadence strategy

Here’s how modern research approaches each of these tools.

You need real-time feedback on both cadence and pace

One point that’s constantly emphasized in the scientific literature is that you must also monitor your pace—if you just tell runners to “increase their cadence” without any pacing feedback, they often just run faster (for reasons which should be clear from the many plots above!). So, you need real-time feedback for your cadence and your pace. Fortunately, on a modern watch it’s easy to set up a watch screen showing both pace and cadence.

External cues with a metronome might be less important than feedback

Early work on cadence retraining made heavy use of metronomes as an external cue for targeting your cadence. Partially this was just because watches that measured cadence were not very common circa 2010. Reviewing the literature now, in 2026, there’s less usage of “external” cueing, at least when cadence retraining is the goal.

Some papers literally just give runners a watch with a speed + cadence display and tell them to check it often, and this approach works pretty well. Other studies use the “alert” functionality on modern watches to make the watch beep if your cadence strays by more than 5 spm from the target.

My recommendation is to start with some metronome usage (say, the first 10 minutes of a run), then transition to more casual checks of your watch, perhaps assisted by a ±5 spm cadence alert. That’s what this 2019 study did.

Fading out feedback helps but you don’t need to be strict about it

Early research on gait retraining used very strict “feedback fading.” That structure would look something like this, for a series of 30 min runs:

Session	Structure
1	Cadence feedback for all 30 min
2	6 × (cadence feedback for 4 min, no feedback for 1 min)
3	6 × (cadence feedback for 3 min, no feedback for 2 min)
4	6 × (cadence feedback for 2 min, no feedback for 3 min)
5	6 × (cadence feedback for 1 min, no feedback for 4 min)
6	30 min with no feedback

By “feedback” it just means “looking at your watch often to check your cadence and pace.”^[13] And perhaps using a cadence alert as well, if your watch supports it.

Later studies have tended towards just telling people to pay attention to their cadence on certain runs, and not pay attention to it on others. For example, here’s the protocol from a 2015 study:

A faded feedback schedule was also used such that [the runners] were provided with real-time feedback on [cadence], via the [Garmin watch], during runs 1–3, 5, and 7, but did not receive feedback on runs 4, 6, and 8. This faded feedback design was utilized to encourage internalization of the new running pattern.

The “internalization” idea here comes from research into motor learning which (maybe?) suggests that having some practice sessions where you don’t get external feedback helps make the new movement pattern what you default to when you aren’t constantly being reminded to check your cadence.

In practice, I recommend something closer to the quote above: check your cadence often for a few runs, then gradually fade out how often you check up on it. Super strict feedback fading does not seem to be an important ingredient for success.

Special considerations on cadence retraining for competitive runners

Most research on cadence retraining focuses on recreational runners. These people basically just run the same speed all the time—they just “go running” at one set speed every day. So, programming their cadence retraining is easy: just increase it 10% and you’re done.

If you are a competitive runner, though, a better strategy is to increase your cadence on easy runs by more like 7.5%, then taper it down so you leave your cadence alone when running at race-specific paces. Let’s take an example from the data earlier. Let’s say you’re a 10k runner with a PR of 37:30 (which is 6:00/mi or 3:45/km) and your normal easy runs are 8:30–9:00/mi (5:20–5:40/km).

If you have the speed–cadence curve seen below, the idea is that you’d aim for about a 7.5% increase in cadence at your easy run pace, but no change at your 10k race pace, with the change tapering off as you go faster.  So, you’d have a different target cadence for different workouts:

Cadence is not the only gait metric you can retrain to reduce injury risk

Some runners find that attempting to change their cadence, even by only 5%, feels extremely awkward—they never get used to it, and they just end up feeling sore, inefficient, and slow.

Fortunately, changing your cadence is not the only way to reduce the biomechanical loading on your body. Some other options include: increasing your step width to decrease iliotibial band forces, increasing your trunk lean to reduce patellofemoral joint forces, and decreasing vertical oscillation or vertical ratio to reduce Achilles and patellofemoral joint forces (and perhaps reduce biomechanical loads more broadly, as increasing cadence seems to do).^[14]

As is the case with cadence, this is a topic best discussed with a good physical therapist or physiotherapist who’s familiar with the latest science on running injuries.

Recap: the science of cadence for runners

Your cadence is the number of steps you take per minute of running. You can count it manually if you want, but cadence data from modern GPS watches is very accurate.

Cadence is mathematically linked with your speed and stride length: if you change your cadence at the same speed, your stride length will necessarily change (and vice versa). In other words, stride length is just cadence in disguise.

It doesn’t make sense to talk about cadence without talking about speed. That’s because all runners increase their cadence to some degree as they get faster. However, there are a few layers of individual variation to keep in mind.

First, at a given speed, there is a wide range of cadence values across different runners. Second, as these runners get faster, you’ll also see wide variation in the strategy runners use to increase their speed.

Some runners use a cadence-dominant strategy to run faster, while others use a stride-length-dominant strategy (and the majority of runners use an intermediate approach, relying on both to similar degrees). 

Your cadence is only weakly predictive of your running economy, and there’s no one “magic cadence” to shoot for. But having a lower-than-average cadence at your usual training pace may be a risk factor for injury, since you take fewer steps with higher-magnitude biomechanical loads, which is an unfavorable trade from an injury perspective.

Changing your cadence is not something that most runners should pursue—changes to cadence are best reserved for chronically injury-prone runners, especially those with frequent injuries to the Achilles, knee, and shin.

If you do decide to increase your cadence, the usual protocol is to aim for a 5–10% increase from your normal cadence, with the additional caveat that competitive runners should not aim to change their cadence at race-specific speeds.

Finally, if you find that increasing your cadence feels too awkward, you may want to look into alternative gait retraining targets, such as step width, vertical oscillation, or trunk lean. I'll be writing more about those in the future!

Learn more about the biomechanics of running

If you enjoyed this article, subscribe to my email list below! It’s the best way to find out when I’ve got a new article on training, a new comprehensive review of the science of running, or a new web app coming out.

To support my work, check out my new book: Marathon Excellence for Everyone: it is the comprehensive guide to marathon training.

Learn more about Marathon Excellence here, or get the book now on Amazon. If you live outside of the United States, Marathon Excellence is also available in a Metric Edition with all workouts in kilometers!

If you aren't up for the marathon just yet, I published a shorter book in 2013 on a simple scientifically based approach to 800m to 10k training, Modern Training and Physiology. Check it out!

Footnotes

---

^[1] Or you can define a gait cycle as starting on the left side, it doesn’t matter. There’s no one standard in biomechanics. Because biomechanical data is collected in discrete time samples (e.g. one ‘frame’ of data every 0.01 second) it is easy to define the gait cycle as an inclusive-exclusive range (so, from the first frame of foot-ground contact on the right side, up until the last frame before the next time your right foot touches the ground).

^[2] Runners commonly use “step” to colloquially mean “when my foot is on the ground”; in biomechanics this is specifically referred to as the “ground contact phase” or “stance time.” This distinction becomes important in walking, when both feet are on the ground at the same time, so a “right step” includes ground contact on both feet at various points.  

^[3] Note that, by measuring stride time, then converting to stride rate, we ended up with a decimal value, and there’s nothing wrong with that—it’s quite easy to take, say, one and a half strides per second.

^[4]  Some older devices actually recorded stride rate as an integer number of strides per minute, so when you look at cadence data on those devices you’ll see that it always takes on an even number: 170, 172, 174. Newer devices have a data field called “Fractional Cadence” that is either 0 or 0.5; in this case the displayed cadence number that you see on the watch is 2*(Cadence + Fractional Cadence). 

^[5] Some of the choices in the biomechanics literature are slightly less arbitrary than you might think; sometimes researchers will only put reflective markers on a single leg, and when you do that you really ought to talk about the step rate for that side since it could differ from the opposite side. I would still prefer they just say “cadence” and then clarify it in the methods section, though.

^[6] One notable exception: the Stryd pod, which actually calculates speed from stride length, as opposed to the other way around.

^[7] Throughout this post I’m plotting speed on the X axis, but showing the axis tick marks as paces (min/mi, min/km) since it’s much more intuitive to grasp. So, that’s why the pace tick marks get “compressed” – the axis is moving in equal increments of speed, not pace.

^[8] This group average curve is actually a pretty sophisticated statistical construction—it’s not simply a pointwise average across runners. The group average comes from a semiparametric mixed model, which uses a global smooth plus individual-level smooths for each runner, which share a common smoothing parameter. This sort of model is described by Gavin Simpson in this StackOverflow answer. Whether or not it makes sense to share a common smoothing parameter is up for debate, but it does help as a prior against noisy GPS speed data.

^[9] This correlation is still statistically significant, at p = 0.048, though that statistical significance disappears if you use a multi-variable model using height and weight together. That’s starting to ask a lot out of a dataset of 49 people though; presumably the “real” effect size of weight alone is some single digit percentage of variance explained.

^[10] The distribution below comes from a different model from the one in the footnote above; this model uses a global smooth plus Gaussian random subject-level intercepts and (uncorrelated) Gaussian random slopes to model deviations from the global smooth. The fit is almost as good as the full-strength subject-specific random smooth model and the clean parameterization allows for the “cadence strategy as Gaussian” interpretation. Setting the random intercepts to be uncorrelated with the random slopes is a limitation of mgcv but you can mitigate the possible downsides by centering the speed data.

^[11] It is much harder to establish that a given trait (like a stride length dominant strategy for running faster) is predictive of increased benefit from a super shoe, versus just establishing that the super shoe on average has an energetic benefit. When you are just trying to show that a shoe is beneficial over some control shoe, each runner can act as their own “control subject”—swap the shoes and you’ve got a perfect counterfactual runner, who is similar in every aspect except their footwear. But when you try to study “subject-level” characteristics like body weight, leg length, or speed–cadence strategy, now you need a much larger sample size to find an equal effect size: you essentially need, in statistical expectation, for there to be a “control subject” who is more or less the same as your long-legged runner, except that they have short legs instead. That can blow up your necessary sample size from a dozen or so subjects (the original “Vaporfly 4% study” had 18) to 100 subjects or more. 

^[12] This Malisoux study did conduct a secondary analysis that was adjusted for speed, but that’s still not quite the analysis I’d like.

^[13] These early studies on gait retraining were often done in research labs, and so you really could turn on / off the computer monitor showing the runner their real-time cadence data. Modern retraining studies using more sophisticated biomechanical metrics like hip angles still use this kind of structured fading.

^[14] One thing to be aware of with these other gait metrics is that they are not measured as accurately as cadence. Vertical oscillation and vertical ratio data from watches are, as far as I can tell, pretty useless, but the accuracy is not awful if you’re using a chest strap or a foot pod (and chest straps are somewhat more accurate than foot pods). No devices exist for measuring step width, as far as I know—you have to eyeball it by running on a painted line, or by using a high speed camera.

Related articles