What the Ergometer Actually Trains, and What It Does Not

The indoor rower gets described as a simulator, and nearly every claim about indoor training inherits that description without testing it. It is worth being precise about what the machine is. It is a flywheel on a slide in a room. It loads your legs, it makes you breathe hard, and it reports a number to two decimal places. What it does not do is reproduce the thing you are training for.

The machine and the boat are not related the way you think

The most useful thing anyone has measured about this is how badly raw erg speed predicts boat speed. Forty-nine junior elite male rowers, a national training camp, both tests inside the same period: a 2,000 m Concept2 Model C and a 2,000 m single-scull on the water. Water speed was 3.66 m/s against an erg equivalent of 4.96 m/s, and the correlation between them explained 28.9% of the variance ([1] Nevill).

Read that number carefully, because it is doing more work than it appears to. Twenty-eight point nine per cent is not "the erg is a bad test". It is "the erg measures something that overlaps with boat speed by about a third, and the other two-thirds is something else". The two-thirds is not noise to be averaged away. It is the water.

And the reason is mechanical, not psychological. In the same study, body mass predicted erg performance strongly (r = 0.68) but predicted water performance not at all (r = 0.039, non-significant), and modelling showed body mass contributing negatively to single-scull speed once erg speed was accounted for. That is drag. A heavier rower is faster on the erg because mass helps push a flywheel, and slower on the water because mass has to be pushed through it.

So the fix is arithmetic, and it is not small. Dividing erg speed by body mass raised the explanatory power from 28.2% to 59.2% ([1] Nevill). The machine was never a bad proxy. It was a proxy on the wrong scale, and club rowers have been comparing raw times to raw times for decades.

What the stroke actually asks of you

Physiologically, the erg is a long aerobic piece with a real anaerobic tail, and the proportions are not in serious dispute. The foundational measurement work put men's VO2max in elite oarsmen at 6.1 ± 0.6 L/min, energy cost at 36 kcal per minute - among the highest reported for any predominantly aerobic sport - and mechanical efficiency at 20% ([3] Hagerman). Aerobic and anaerobic calculation put the aerobic contribution over the men's 2,000 m at 70 to 75%, with the remainder anaerobic ([3] Hagerman).

That is a description of a cardiovascular event. The broader physiology literature points the same way and explains why: rowing involves almost all muscles during the stroke, and competition demands a large oxygen uptake that is itself limited by pulmonary diffusion capacity and by blood flow to working muscle ([4] Volianitis). The legs do the work during the relatively slow phase of the stroke, which is precisely what makes the recruitment slow-twitch dominant - and the training response is hypertrophy of exactly those fibres, alongside cardiac enlargement ([4] Volianitis).

The more recent synthesis has moved on from reporting a single percentage to asking how the three energy systems actually interact across durations and measurement methods, and the honest answer is that a clean 70/30 split is a simplification of something continuous ([5] Gastin). What has not changed is the shape: a 2K is a long aerobic piece with a short anaerobic tail, not a sprint with a long warm-up.

What actually sets the number

If the erg is mostly a cardiovascular test, the obvious follow-up is which cardiovascular variable. The best current answer is not a laboratory one.

Twenty-two national-level rowers, nine of them female, were profiled with an incremental erg test alongside strength and power testing, dual-energy X-ray absorptiometry for lean body mass, MRI for muscle volume, diffusion tensor imaging for muscle architecture and MR spectroscopy for estimated fibre typing. Hierarchical regression, with sex and lean body mass as covariates: VO2peak carried the largest independent association with final-stage mean power, and the estimated proportion of type II fibres came second. Most other muscle characteristics added almost nothing once those were in the model ([2] Bellinger).

Two things follow. The first is that sex and lean body mass explain most of what looks like "fitness" in a raw 2K comparison, which is the same finding as the drag problem above arriving from a different direction. The second is that once you have accounted for the body you have, what is left that you can move is mostly aerobic.

Local measurements say the same thing at the muscle. Near-infrared spectroscopy over biceps and vastus lateralis during incremental erg work, a training session and a 2K race found vastus lateralis saturation falling from 73.7% at rest to 7.9% at maximal power, while biceps fell from 68.2% to 28.9% - the legs are being emptied of oxygen far more completely than the arms, which is exactly what a leg-driven machine should do ([10] Yan).

The practical consequence is in what predicts a 2K from the outside. In 40 NCAA Division I women, a 60-second maximal effort was the single best predictor of 2K time (R² = 0.611), ahead of the 20-second test (R² = 0.536) and the 7-stroke peak power test (R² = 0.417) ([9] DeLeo). The 2K is well predicted by a one-minute all-out effort. That is a test of anaerobic capacity doing most of the explaining, in an event that is roughly three-quarters aerobic - which says the tail matters more than the headline percentage suggests.

Rate is not an economy variable

A common belief is that a higher stroke rate makes you more efficient. The evidence is that it does not, and this is one of the better-designed studies in the field.

Seventeen well-trained female rowers rowed at 28, 34 and 40 strokes per minute on a modified ergometer with power transfer to the flywheel held constant and the respiratory exchange ratio at 1, so the comparison was between rates at identical work. Gross efficiency did not differ significantly between the three rates ([7] Hofmijster). Internal power losses were not a function of cycle frequency across the range used in competition.

This matters because it changes what rate is for. If rate did not change efficiency, then a higher rate is a pacing and technique decision, not an economy one - you are choosing how to spend the stroke, not how much energy the stroke costs. The manufacturer's own material makes the same separation, holding rate and pace as independent levers ([17] Concept2), and its material on long strokes at a given pace is a useful corrective to the assumption that more strokes is always more efficient ([19] Concept2).

Damper setting is the other machine variable, and it is the one that changes the load curve without changing the work done ([14] Concept2). It is also the variable where the "simulator" framing breaks most visibly, because damper selection is a judgement about what you want to feel, and the manufacturer is explicit that the choice differs between a steady-state piece and a race ([18] Concept2). Which is the point: on the erg the damper is chosen against a number, and on the water it is chosen against conditions the machine cannot report. The manufacturer's own account of damper choice in competition makes the same separation ([20] Concept2).

The specific things the machine teaches you to get wrong

The strongest specific criticism available is about sweep rowers, and it is measured rather than argued.

Twenty-five male rowers - nine scullers, sixteen sweep - were filmed from behind at 30, 60 and 90 minutes of a 90-minute endurance erg session. Scullers stayed centred on the seat, drifting no more than 0.6 cm. Port-side sweep rowers sat 3.0 to 3.6 cm to the left, starboard sweep rowers 3.1 to 3.9 cm to the right, and the drift was consistently to the side opposite the side they row in the boat ([11] Cerasola).

That is not a rounding error. A sustained asymmetric load on a machine, in a sport where side-to-side imbalance is already a known problem, is the erg actively reinforcing a postural habit. And it is a habit the machine does not warn you about, because the PM5 reports pace, rate, drive length and drive ratio - the machine's own view of stroke efficiency ([15] Concept2) - and has nothing to say about where you are sitting.

This is not the same as saying the erg cannot teach technique. The manufacturer's stroke sequence is a real reference ([16] Concept2). But the evidence for how technique responds to training is more interesting than the folk version, and it is measurable.

Seven junior rowers were tracked across two mesocycles with a sensor on the upper back. Overall stroke duration and its variability stayed stable across the whole period. The durations of the two phases of the stroke - the forward drive and the backward recovery - changed systematically, significantly in 10 of 12 paired comparisons, as the training emphasis shifted from strength to intensive rowing ([12] Anpilogov). The stroke's internal structure is sensitive to what you have been training even when its overall tempo is unchanged. Which means technique specificity is real, measurable, and not something you can assume you get from time on the machine.

What the machine is genuinely good for

None of the above means the erg is a poor tool. It means it is a tool for something specific.

Across four intensity domains in 24 trained rowers, the erg was sensitive enough to separate heavyweights from lightweights on absolute anaerobic energy production, peak metabolic power and total energy expenditure, while the lightweights matched them on relative cardiorespiratory response ([6] Rios). A protocol that distinguishes absolute from relative load in one session is doing something a boat test cannot.

It is also the only place most rowers will ever do a repeatable maximal effort. In 51 single scullers sorted into elite, sub-elite and developmental tiers, a three-part erg protocol separated all three on anaerobic power, on aerobic profile and on the ability to hold power under fatigue simultaneously - tier alone accounted for 72% of the variance in the aerobic measures ([8] Gong). If you want to know whether a training block moved you, this is close to the most reliable instrument available.

And it is where you can find out whether a particular distance suits you at all. Rowers who did relatively better over 1,500 m than 2,000 m were distinguishable by their power reserve ratio, and by their force-velocity slope once a single outlier was removed ([13] Baumann). The machine will not tell you how to race, but it will tell you which distance your physiology currently rewards.

What is still uncertain

The honest limits of all this are worth stating, because the field is thinner than the volume of rowing physiology writing suggests.

The specificity question rests substantially on one study. The 28.9% figure and the body-mass correction come from a single investigation of 49 junior elite males in one national camp, and its own recommended model is a power-to-weight ratio rather than a validated formula you can apply to any rower ([1] Nevill). The direction of the effect is not in doubt - mass helps on the machine and hurts on the water - but the exact exponent should not be treated as settled.

Sample sizes in the newer work are small too. The muscle-profiling study had 22 participants ([2] Bellinger), the force-velocity work 9 ([13] Baumann), the technique-stability pilot 7 ([12] Anpilogov), and the sweep-asymmetry case study 25 of whom only 16 were sweep rowers ([11] Cerasola). These are honest and careful studies, and none of them is large enough to be the last word on anything.

The clearest gap is that nobody has good longitudinal evidence on how much of what the erg builds transfers to boat speed over a season. The cross-sectional correlation is weak, the mechanism for why is well understood, and the longitudinal question - does a winter of erg work make you faster on the water - appears to be largely unmeasured. If you are making a training decision and the erg is a large part of your winter, that is the question most worth asking of the literature, and the literature is thin.