Loading fundamentals 04: Load and repetition range

Short answer. Light loads built as much muscle as heavy ones, but only in trials that took every set to failure, and heavy loads built more strength regardless. Our position, an inference no trial has tested, is that load matters more the further from failure you stop. Light, high-repetition sets such as bodyweight and small-muscle isolation work should therefore be taken closer to failure.

---

What it is

You already know the correction. The old three-range prescription is dead, load does not determine hypertrophy, you can grow with light weight.

That finding is real and it came attached to a condition that gets dropped almost every time it is quoted. The condition is sitting in the inclusion criteria of both meta-analyses, and it is the one this publication has spent an entire entry arguing against.

The proposed mechanism

The reason load was thought to matter is the size principle. Motor units are recruited in order of size, smallest first, as force demand rises. A heavy lift demands high force immediately and therefore recruits the large, high-threshold units from the first repetition. A light lift does not, so the large units are never called on and never adapt.

That account is correct as far as it goes, and it omits fatigue. As a light set continues, the small units already working begin to fatigue and can no longer produce the required force. To keep the bar moving, the nervous system recruits progressively larger units. By the final repetitions of a set taken to failure, a light load is recruiting the same high-threshold motor units a heavy load recruited at the start.

This is the mechanism behind the modern claim, and it also contains its limit. A light set reaches full recruitment only near the end. A heavy set has it throughout. The two are equivalent at failure and diverge before it.

What has actually been tested

The headline meta-analysis, and the criterion inside it. Twenty-one studies comparing loads at or below 60% of one-rep max against loads above it. One-rep max strength gains were significantly greater with heavy load. Isometric strength showed no significant difference. Hypertrophy was similar across the loading spectrum. The authors concluded that maximal strength benefits come from heavy loads while hypertrophy can be equally achieved across a range of loads (Schoenfeld et al., 2017).

Now the inclusion criterion. Studies were eligible only if all sets in both protocols were performed to momentary muscular failure. That is not a methodological footnote. It is the condition under which the finding was generated, and the finding has not been established outside it.

A network meta-analysis reached the same place with the same restriction. It compared low (above 15RM, a load you could lift more than fifteen times), moderate (9 to 15RM) and high (8RM or below) loads, again including only sets taken to volitional failure. Hypertrophy improvements appeared load independent while strength gains were superior with high loads (Lopez et al., 2021).

Before failure was standardized, the picture leaned heavier. An earlier meta-analysis in largely untrained participants found trends favoring high load for both outcomes, neither reaching significance. The strength difference was 1.07 ± 0.60 (95% confidence interval, CI, -0.18 to 2.32, p = 0.09) and hypertrophy difference 0.43 ± 0.24 (95% CI -0.05 to 0.92, p = 0.076) (Schoenfeld et al., 2016).

The strength result is not a small one. Pooled data have put the effect favoring loads above 60% for strength gains at around 0.58, a moderate-to-large difference. Load independence applies to size. It does not apply to how much you can lift.

Fiber type has been examined and does not rescue the old prescription. Comparisons of low against high load found no significant differences in type I or type II fiber hypertrophy, with very wide confidence and prediction intervals. The idea that light loads preferentially build slow fibers and heavy loads fast ones is not supported at the resolution currently available.

Our reading

The load-independence finding is real and it is conditional, and the condition is one this publication has argued against elsewhere.

Every trial in both meta-analyses took every set to momentary failure. Our reading of the proximity-to-failure literature is that there is no good reason to train that way as a default, because failure roughly doubles within-session capacity loss without producing more hypertrophy in trained people. (→ Loading fundamentals 02: Reps in reserve)

Put those together and a specific problem appears. The condition that makes light loads equivalent is the same condition we recommend against. The last few repetitions are where a light set does its work, and stopping short removes exactly those. Heavy loads do not have this problem: full recruitment is present from the first repetition, so stopping short costs proportionally less of the stimulus (Schoenfeld et al., 2017).

So our position is that load matters more the further you train from failure, and matters less the closer you get. Nobody has tested that directly. It follows from combining the recruitment account with the inclusion criteria of the meta-analyses, and it explains something the pure load-independence reading does not: why the moderate range has survived in practice despite decades of evidence that light loads work.

The practical consequence is not about picking a load. It is about the places where you cannot pick one.

Most trainees are not choosing between 30% and 80% of their maximum. They are working in a moderate range where this problem does not arise, and the argument above is academic. It stops being academic in two situations, and both are common.

The first is bodyweight training. The load is whatever you weigh, and for many movements that lands well below what you would select with a barbell. You cannot add plates to a push-up. The second is high-repetition isolation work on small muscles, where sets of fifteen or more on something like a lateral raise are standard because the alternative is unmanageable technique.

In both cases the load is fixed at the light end and the rep count is high, which is exactly the region where the stimulus concentrates in the final repetitions. A set of fifteen stopped at 3 RIR (three reps in reserve, three short of failure) has discarded a larger share of its useful work than a set of eight stopped at the same point.

Light, high-repetition sets deliver a smaller fraction of their stimulus before the final repetitions, so stopping early costs more there than it does on a heavy set. That is an argument for varying the set endpoint by exercise rather than applying one number to a session. (→ Loading fundamentals 02: Reps in reserve)

Where the goals diverge

Maximal strength. Heavy loads, and this is the least ambiguous finding in the branch. Strength is specific to the load you train and both meta-analyses agree. (→ Loading fundamentals 05: Progressive overload)

Size, training near failure. A free choice across the spectrum, though taking everything to failure has costs of its own. (→ Loading fundamentals 02: Reps in reserve; Loading fundamentals sub-hub)

Size, training short of failure. The choice is not free. Moderate to heavy loads preserve the stimulus at a given RIR; light, high-repetition sets may not.

Choosing which movement to load. Load and repetition range say nothing about which exercise carries them, and that choice has its own consequences. (→ Loading fundamentals 06: Exercise selection)

Training with bodyweight only. You cannot adjust the load, so the adjustment has to happen at the set endpoint. Expect to train these closer to failure than you would a loaded compound lift.

Working around joint pain or an injury. Light load is the tool and it is well supported. The price is the same one: those sets need to go closer to failure, which is more uncomfortable than most people expect.

The case against

Our central claim is an inference. No trial has crossed load with proximity to failure and measured hypertrophy. We have combined a mechanism with a set of inclusion criteria and drawn a conclusion neither was designed to support.

The recruitment account is a model. Motor unit recruitment during fatiguing contractions is inferred largely from surface electromyography, which is an imperfect instrument for this question, and the "effective repetitions" framing built on it is contested.

The earlier meta-analysis trended heavier for hypertrophy too. Before failure-only inclusion, the point estimate favored heavy load at p = 0.076. That is not nothing, and it is compatible with load mattering somewhat even at failure.

Most of this literature is untrained or recreationally trained participants over six to twelve weeks. The population that cares most about this question is the least represented in it.

What is not known

Whether load independence survives when sets stop short of failure. This is the most important gap in this piece and the whole practical section rests on it.

Whether very light loads recruit fully even at failure, or whether there is a floor below which the account stops working.

Whether trained lifters behave like the untrained participants who supplied most of these data.