No. In the trials that tested it directly, sets stopped one to three reps short of failure built about as much muscle as sets taken all the way52. What seems to matter is getting close, not reaching the rep you cannot finish.
This article covers what "failure" means in these studies, what the trials and meta-analyses found, and where the evidence is thin.
What was studied
Momentary muscular failure is the point in a set where you try for another rep and cannot complete it. The alternative is to stop earlier and count RIR (reps in reserve, how many more reps you could have done). A set "at 2 RIR" ends with two reps left in the tank. For a longer primer, see RIR vs RPE.
Researchers test this in two ways. Some split people into two groups. Others use a within-person design: one leg or arm trains to failure, the other stops short, so genetics, sleep and diet are identical on both sides. Muscle growth is usually measured as muscle thickness by ultrasound, a scan that shows how deep the muscle is at a fixed spot.
| Study | Design | Who | Length |
|---|---|---|---|
| Vasconcelos 2026 | one leg to failure, one leg at 1–3 RIR, leg extension | 19 trained adults (11 men, 8 women) | 8 weeks |
| Refalo 2024 | one leg to failure, one leg at 1–2 RIR, leg press and leg extension | 18 trained adults (12 men, 6 women) | 8 weeks |
| Hermann 2025 | two groups, one set per exercise, to failure or 2 RIR | 42 trained men and women | 8 weeks |
| Refalo 2023 | meta-analysis | 15 studies, healthy adults | pooled |
In plain terms: the best tests put failure and "a few reps short" on two legs of the same person.
What the studies found
The newest direct test comes from Portugal. Nineteen trained adults, average age 22, did leg extensions for eight weeks with one leg taken to failure and the other stopped at 1 to 3 RIR5. The difference in thigh muscle thickness between legs was −0.019 cm, with an interval of −0.094 to 0.054 cm. That is a fraction of a millimetre, and the interval sits on both sides of zero. Strength differed by 0.198 kg (−1.87 to 2.6). The authors called both differences negligible.
An earlier trial with a near-identical design reached the same answer. In 18 trained adults, quadriceps thickness rose 0.181 cm on the failure leg and 0.182 cm on the leg stopped at 1 to 2 RIR over eight weeks2. Training volume was similar between legs.
The 2023 meta-analysis pooled 15 studies in healthy adults of any age and training background1. A meta-analysis combines many trials into one estimate. When "failure" meant true momentary failure, the advantage over stopping short was an effect size of 0.12, with a 95% confidence interval of −0.13 to 0.37. An effect size is the difference in standard units; 0.12 is small, and an interval that crosses zero means "no difference" is well within the plausible range. The review's main analysis, which counted any definition of failure, did find a trivial advantage for failure: 0.19, with an interval from 0.00 to 0.371.
Failure vs stopping short: pooled effect on muscle growth · effect size
One result leans the other way. In a 2025 trial, 42 trained adults did just one set per exercise, twice a week for eight weeks, either to failure or at 2 RIR4. Several muscle measures tended to favour failure, though the absolute differences were modest, and jump height also leaned towards failure. Strength and muscular endurance gains were similar.
In plain terms: at normal training volumes, stopping a couple of reps short matched failure. With a single set, failure may add a little.
Close to failure still matters
"You don't need failure" is not the same as "effort doesn't matter". A 2024 series of meta-regressions estimated RIR for every set across the literature and looked for a dose-response pattern3. Muscle growth increased as sets ended closer to failure. Strength gains were similar across a wide range of RIR.
Two cautions come with that result. The RIR values were estimated by the authors from each paper's methods, not measured. And the authors call the analysis exploratory. Put next to the direct trials, the likely shape is a curve that climbs as you approach failure and flattens over the last rep or two. The 2023 authors describe the same idea as a "potential non-linear relationship"1.
In plain terms: sets that end far from failure seem to grow less muscle; sets that end one to three reps short look about as good as failure. The same pattern shows up with light loads (see light vs heavy weights for muscle growth).
What stopping short saves
Failure is not free. In the 2024 two-leg trial, lifting velocity and repetitions dropped more within sessions on the failure leg than on the RIR leg2. That fatigue can carry into the next sets and sessions (see signs of accumulated training fatigue). The Vasconcelos authors present RIR training as "a practical approach to manage fatigue while maintaining training effectiveness"5.
Can people judge reps in reserve?
Stopping at 2 RIR only works if your guess is close. A 2026 study tested whether that guess improves with practice6. Twenty-six adults with lifting experience but no RIR experience took part: 13 younger (19 to 40) and 13 older (60 to 74). Over six sessions, they called out the rep at which they thought they had 4 and 2 reps left, then kept going to failure, with no feedback.
The average miss shrank with practice. On the leg press at 4 RIR, younger adults' error fell from 4.6 to 2.3 reps and older adults' from 4.6 to 2.56. Errors were smaller at 2 RIR than at 4: guesses got sharper nearer failure. Both groups tended to underestimate how many reps they had left.
Average error guessing 4 reps in reserve, first vs sixth session · reps
Session 1 Session 6
In plain terms: early guesses can be off by several reps. In the study, six sessions in which every set was carried on to failure, so the true count showed, roughly halved the largest errors.
The studies at a glance
| Study | Who was studied | Key result | Year |
|---|---|---|---|
| Refalo et al., *Sports Med*1 | 15 studies, healthy adults | Momentary failure vs stopping short: effect size 0.12 (−0.13 to 0.37), no clear advantage; any failure definition: trivial advantage, 0.19 (0.00 to 0.37) | 2023 |
| Refalo et al., *J Sports Sci*2 | 18 trained adults, one leg each way, 8 weeks | Quadriceps thickness +0.181 cm (failure) vs +0.182 cm (1–2 RIR) | 2024 |
| Robinson et al., *Sports Med*3 | Meta-regressions of the proximity-to-failure literature | Growth rose as estimated RIR fell; strength did not change with RIR | 2024 |
| Hermann et al., *Med Sci Sports Exerc*4 | 42 trained adults, one set per exercise, 8 weeks | Some growth measures tended to favour failure; strength similar | 2025 |
| Vasconcelos et al., *J Strength Cond Res*5 | 19 trained adults, one leg each way, 8 weeks | Difference in thickness −0.019 cm, negligible | 2026 |
| Wiedenmann et al., *BMC* journal6 | 26 adults aged 19 to 74, six sessions | RIR guessing error fell by up to 2.3 reps | 2026 |
Limits worth knowing
- Small samples. The two-leg trials had 19 and 18 people. They are good at showing no large difference, less good at ruling out a small one.
- Mostly legs, mostly young, trained adults. Both direct tests trained the quadriceps; upper-body and older-adult data are thinner.
- Short. Eight weeks is typical. Whether small differences add up over years is untested.
- Estimated RIR. The dose-response analysis estimated RIR from papers; nobody measured it set by set3.
- What "failure" means varies. Many older studies used looser definitions, which is why the 2023 meta-analysis split them out1.
- Abstract only. We read the 2026 two-leg trial, the 2024 two-leg trial, the 2024 meta-regression and the 2025 single-set trial as abstracts. Sets, loads and per-leg gains for the 2026 trial are not in its abstract and are not stated here. The 2023 meta-analysis and the RIR-guessing study were read in full.
Latest research
The newest direct test is the 2026 two-leg leg-extension trial in 19 trained adults, published in August 20265. It adds a second within-person trial to the 2024 one, with the same answer: 1 to 3 RIR matched failure for growth and strength over eight weeks. A same-month study showed that the skill needed to train this way, judging reps in reserve, improves within about four to six sessions in younger and older adults alike6. How these results transfer to the upper body is still open (see how many sets per muscle group per week for the volume side).
The short version
- Failure is not required: sets stopped 1 to 3 reps short grew as much muscle in direct tests.
- Effort still matters: growth tends to rise as sets end closer to failure.
- Failure costs more fatigue within a session than stopping a rep or two short.
- With a single set per exercise, failure may add a little.
- Judging reps in reserve improves with a few weeks of practice.
Educational overview only — not medical advice.
