Six elite road cyclists sat on a metabolic cart, spinning at four fixed cadences — 60, 80, 100, and 120 rpm — while researchers dialled the resistance from nothing up to 350 watts and measured exactly how much oxygen it cost them to keep the pedals turning (Foss & Hallén 2004). Most riders settle on a cadence early — a number that feels smooth on a Sunday spin — and defend it everywhere: easy day, group ride, race day, all the same rpm. The data says that instinct is only half right. The cadence that costs your body the least oxygen isn't a fixed personal setting. It moves, and it moves with how hard you're pushing.

This matters more than it sounds like it should. Cadence is one of the few training variables a rider controls on every single pedal stroke, all ride, every ride — unlike power, which is dictated by the terrain and the effort you're chasing, cadence is a live, constant choice. Get the relationship between cadence and workload wrong and you're not just leaving a small amount of oxygen on the table on one interval. You're doing it on every interval, every week, for as long as the habit holds.

Section 01What the data actually shows

At zero load — freewheeling resistance — the most economical cadence in Foss and Hallén's trained cyclists sat at 60 rpm (Foss & Hallén 2004). As the researchers raised the load to 350 watts, that same "cheapest" cadence had climbed to 80 rpm — a 20 rpm shift, measured across fixed test points from 0 to 350 W (Foss & Hallén 2004). Maximal oxygen uptake didn't change across cadences — riders reached the same ceiling however they pedalled — but at the higher power outputs, the 80 rpm cadence also produced the better output over roughly five-minute efforts (Foss & Hallén 2004). Grinding a "smooth" 60 rpm at threshold wasn't just costing oxygen. It was costing performance.

This isn't an isolated result. Marsh and Martin measured freely preferred cadence against the metabolically cheapest cadence in well-trained cyclists and found the two numbers don't live in the same neighbourhood — the most economical cadence sat between 53 and 60 rpm across the group, while well-trained riders' preferred cadence held at 90–100 rpm regardless of the power they were putting out (Marsh & Martin 1997). That's a 30–45 rpm gap between what the body "should" prefer on an oxygen-cost basis, and what trained legs actually choose. Less-trained riders showed a different pattern — their preferred cadence dropped as power climbed, from around 80 rpm at 75 W down to roughly 65 rpm at 175 W, tracking closer to the economical number as the effort got harder relative to their fitness (Marsh & Martin 1997).

Then there's the field data. Lucía, Hoyos, and Chicharro tracked seven professional riders across three-week Grand Tours and logged their actual cadence across three race contexts — high mountain stages, flat group stages, and individual time trials (Lucía et al. 2001). On climbs, mean cadence dropped to 71 rpm. On flat group stages it rose to 89 rpm. In individual time trials — the discipline where every watt is measured and pacing is deliberate, not tactical — mean cadence peaked at 92.4 rpm (Lucía et al. 2001). The pattern lines up with the lab: as intensity and sustained power output rise, so does the cadence that goes with it. The climbs are the exception, and they're the exception for a mechanical reason, not a metabolic one.

Section 02Why the number moves

The mechanism is about how the body recruits muscle, not about lung capacity. At low power, a slow cadence keeps force-per-pedal-stroke modest and lets slow-twitch, fatigue-resistant fibres do most of the work — cheap, in oxygen terms. As power output rises, each pedal stroke at a slow cadence demands a bigger force spike from the same muscle mass, which drags in more fast-twitch fibre recruitment — fibres that are metabolically expensive to run, because they rely more heavily on anaerobic pathways and fatigue faster under repeated load. Spinning faster spreads that same power across more, smaller force pulses per second, keeping each individual pulse below the threshold that would force those costlier fibres into play. Above a certain load, it's cheaper to turn the pedals faster than to push them harder. That crossover point is exactly what shifted from 60 to 80 rpm in the lab data (Foss & Hallén 2004). It's also why the crossover isn't a single universal number — it depends on your own muscle-fibre mix, your gearing, and how much of your ride is spent above versus below that load threshold.

Maximal aerobic capacity being cadence-blind (Foss & Hallén 2004) matters here too — this isn't a story about lungs or heart-rate ceilings. It's a story about the most efficient way to route a given power demand through the muscle you have available, at the load you're asking it to handle. There's also a real gap between what's economical and what trained riders actually choose: preferred cadence in well-trained cyclists holds at 90–100 rpm regardless of power output, against a most-economical range of just 53–60 rpm (Marsh & Martin 1997). Tellingly, Marsh and Martin found that cycling experience itself wasn't what predicted a rider's preferred number — something else in a trained rider's neuromuscular makeup is driving that choice, not simply years on the bike or a conscious effort to minimise oxygen cost. That's a genuinely open question in the literature, and it's part of why "just copy the economical number" is bad advice: the body isn't optimising for oxygen cost alone when it picks a cadence, and neither should you.

Section 03The protocol

1. Test, don't assume

   Ride the same climb or trainer segment at a fixed power three times, at 70, 85, and 100 rpm. Watch heart rate at matched power, not perceived smoothness — that's your real economy signal, not the cadence that feels familiar.
2. Let cadence rise with intensity

   Endurance rides: comfortable-low, wherever legs settle without strain, generally 75–85 rpm. Tempo and threshold work: elite riders spontaneously settle around 90 rpm during their hardest sustained road efforts — flat stages and time trials, specifically (Lucía et al. 2001) — so there's little reason to defend a much lower number once you're working that hard yourself.
3. Keep climbing separate

   Climbing cadence runs lower for real mechanical reasons — standing leverage, gradient, gearing limits — not because low cadence is suddenly "efficient" again. Pros in three-week tours held roughly 71 rpm on climbs versus 89–92 rpm on flat and TT stages (Lucía et al. 2001). Don't force a flat-road number onto a wall of a climb.
4. Build the pattern, don't force the number

   Marsh and Martin measured a 30–45 rpm gap between economical and preferred cadence in trained riders (Marsh & Martin 1997). Don't read that gap as something to close by force — treat it as the direction months of high-cadence riding move you, not a number to fight on a single ride. Sudden cadence changes on race day without that base behind them won't replicate it.
5. Check your gearing before you blame your legs

   A cadence target is only reachable if your gearing lets you hit it at the power and gradient in front of you. If you're grinding sub-70 rpm on every climb regardless of what the protocol above says, the fix is often a compact chainring or a wider cassette, not more willpower.

Section 04Where riders get this wrong

Section 05Applying it with ULTRA

Section 06Glossary

Section 07Bottom line

There is no single number worth defending across every ride. The cadence that costs your body the least oxygen rises as your power output rises — roughly 60 rpm near zero load, climbing to 80 rpm by 350 watts in the lab data (Foss & Hallén 2004). Trained legs already drift toward a higher cadence than the lab-optimal number, and pros in real racing hold 89–92 rpm on flat roads and time trials, dropping to around 71 rpm on climbs where the demands are different (Lucía et al. 2001). Test your own economy at matched power across a few cadences instead of copying a number off a chart — and let the cadence rise as the effort does.