No, you do not need to train to failure to build muscle, but you do need to get close. The best current evidence shows that stopping one to three reps short of failure produces about the same growth as grinding every set to the last possible rep, with less fatigue.
The nuance is that “close” matters. Sets stopped far from failure grow less muscle, especially with light loads. So the practical goal is not to avoid failure or chase it, but to control proximity to failure: hard enough to stimulate growth, measured enough to recover and keep progressing.
What the research says
- Failure is not required for hypertrophy. Across 15 studies, momentary failure showed no significant advantage over non-failure training (effect size 0.12, confidence interval crossing zero).1 Strong evidence
- Closer to failure tends to mean more growth. A 2024 series of meta-regressions found muscle size gains increased as sets ended with fewer estimated reps in reserve, while strength gains did not depend on it.2 Moderate evidence
- 1–2 reps in reserve can match failure in trained lifters. In an 8-week within-subject trial, quadriceps thickness rose 0.181 cm with failure and 0.182 cm with 1–2 RIR.3 Moderate evidence
- Failure costs more recovery. Sets to failure slowed recovery of neuromuscular performance by up to 24–48 hours versus stopping halfway at equal volume.4 Moderate evidence
- Your RIR estimate is probably off by about one rep. Lifters underpredicted reps left by 0.95 reps on average, with bigger errors on high-rep sets.5 Moderate evidence
Evidence grades reflect the number and quality of studies (meta-analyses > single RCTs > acute/mechanistic data).
What does “training to failure” actually mean?
Researchers use the term inconsistently, which explains a lot of the confusion online. The strict definition is momentary muscular failure: you attempt another rep with good form and cannot complete it. Other studies use looser definitions, such as stopping when form breaks down or when the lifter says they are done.
The alternative is to stop short of failure, usually described in reps in reserve (RIR). A set ended at 2 RIR means you believe you could have completed two more clean reps. A third approach, common in velocity-based training research, stops a set once bar speed has dropped by a set percentage (for example 20% or 40%). Larger velocity loss means you finished closer to failure.
If you are new to rating effort, start with our guide on calibrating RPE and RIR. Everything below assumes you can estimate RIR within a rep or two.
Does training to failure build more muscle?
Three meta-analyses form the core of the evidence, and they agree on the big picture while differing in the details.
Refalo et al. (2023) pooled 15 studies that measured hypertrophy. Using any definition of failure, there was a trivial advantage for failure (effect size 0.19, 95% CI 0.00 to 0.37). When they restricted the comparison to strict momentary muscular failure, the advantage disappeared (0.12, CI −0.13 to 0.37). Higher velocity-loss thresholds (over 25%) were not better than moderate ones (20–25%) either. The authors concluded the relationship between proximity to failure and growth is likely non-linear: getting reasonably close matters, but the final reps add little.1
Robinson et al. (2024) took a different approach. Instead of failure versus not, they estimated the RIR of each training group across the literature and ran dose-response meta-regressions. For hypertrophy, the slopes were consistently negative: the fewer reps left in the tank, the more muscle growth. For strength, RIR had a negligible relationship with gains.2 The important caveat is that RIR was estimated by the authors from study descriptions, not measured, and the analysis is explicitly exploratory.
Grgic et al. (2022) also included 15 studies. Overall, there was no significant difference for hypertrophy (effect size 0.22, CI −0.11 to 0.55) or strength. One subgroup is worth knowing: in resistance-trained participants, failure showed a small significant hypertrophy advantage (effect size 0.15, CI 0.03 to 0.26).6 That subgroup result is one reason some coaches still prescribe failure for experienced lifters.
Put together: training closer to failure probably helps up to a point, but the jump from 1–2 RIR to true failure appears small and uncertain. None of these analyses supports the idea that failure is required.
What happens in trained lifters?
Most early studies used untrained participants, who grow from almost any stimulus. Two within-subject trials in trained people are especially useful because each participant trained one leg each way, which removes differences in genetics, diet, and sleep.
Key study
Failure vs 1–2 RIR in trained lifters (Refalo et al., 2024)
- Design
- 8-week randomized within-subject trial, two sessions per week
- Who
- 18 resistance-trained adults (12 men, 6 women)
- Compared
- Leg press and leg extension to momentary failure on one leg vs a perceived 2 RIR (leg press) and 1 RIR (leg extension) on the other
- Result
- Quadriceps thickness increased 0.181 cm (failure) vs 0.182 cm (RIR). Volume was similar; failure caused consistently greater velocity and rep loss.
Why it matters: Stopping 1–2 reps short matched failure for growth while producing less acute fatigue. Caveats: small sample, one muscle group, ultrasound muscle thickness rather than MRI volume, and only eight weeks.3
Santanielo et al. (2020) ran a similar 10-week within-subject study in 14 trained individuals. Vastus lateralis cross-sectional area grew 13.5% on the failure leg and 18.1% on the non-failure leg, a difference that was not statistically significant. Leg press and leg extension 1RM gains were also similar.7
Neither study shows non-failure training is better. They show that, in people who already lift, ending sets close to failure is enough.
Light loads are the exception: there, effort matters more
Proximity to failure matters more when the weight is light. Lasevicius et al. trained untrained men on the knee extension at either 30% or 80% of 1RM, to failure or not, with total volume equated. Quadriceps cross-sectional area grew about 8% in both heavy conditions and in the light-to-failure condition, but only 2.8% (not significant) when light sets were stopped short of failure.8
This matches a simple idea: with a heavy weight, even reps well before failure require high muscle activation. With a light weight, you only reach that level of effort near the end of the set. If you use high reps, push them close to failure. For more on load choice, see the best rep range for muscle growth.
The “effective reps” idea: a useful hypothesis, not a fact
A popular explanation for why proximity to failure matters is the effective reps hypothesis. It builds on a well-established principle of motor control: muscles recruit motor units in order of size, small ones first. As a set progresses and fibres fatigue, the nervous system recruits larger, high-threshold motor units to keep the weight moving, and the bar slows down. The hypothesis proposes that the reps performed under that high recruitment and slow, involuntary bar speed, often said to be the last five or so reps before failure, provide most of the growth stimulus.
It is a reasonable model, and it fits the light-load data above. But it has not been directly tested as stated, the exact number of “effective” reps is an assumption, and it predicts that failure should beat 1–2 RIR more clearly than the trials actually show. Treat it as a mental model for why you should not stop sets far from failure, not as a counting system.
What does failure cost? Fatigue and recovery
If failure adds little growth, the deciding factor becomes cost. Here the evidence is clearer.
Velocity-loss studies. Pareja-Blanco et al. (2017) had young men squat for 8 weeks stopping each set at either 20% or 40% velocity loss. The 40% group (closer to failure) did more reps and gained more vastus lateralis and intermedius hypertrophy on MRI, but the 20% group achieved similar squat strength gains, larger jump improvements (9.5% vs 3.5%), and did about 40% fewer reps. The 40% group also showed a drop in the proportion of fast type IIX myosin, which the 20% group preserved.9 A larger 2020 follow-up in 64 trained men compared 0%, 10%, 20%, and 40% velocity loss: only the 20% and 40% groups showed significant hypertrophy, but 40% produced negative neuromuscular changes such as a slower early rate of force development.10
Recovery time course. Morán-Navarro et al. (2017) had 10 trained men perform bench press and squat as 3×10 to failure, 3×5 with the same load (half the possible reps), or 6×5, which matched the failure session’s total volume. Failure caused bigger acute drops in jump height and bar speed, and recovery of neuromuscular function was significantly slower between 24 and 48 hours afterward, even compared with the volume-matched 6×5.4
Key study
Recovery after sets to failure (Morán-Navarro et al., 2017)
- Design
- Acute crossover study, measurements from 12 h before to 72 h after
- Who
- 10 resistance-trained men
- Compared
- 3×10 to failure vs 3×5 and 6×5 with the same 10RM load (bench press and full squat)
- Result
- Failure produced larger drops in jump height and bar velocity; recovery of neuromuscular function and of metabolic and hormonal markers was slower, by up to 24–48 h
Why it matters: The cost of failure is not just the set itself. It can carry into your next session for the same muscles. Caveat: a single session in 10 men, measuring performance and blood markers rather than long-term growth.
Translation: failure on every set of every exercise, especially heavy compounds, eats into the recovery you need to train the muscle again soon and to keep adding productive weekly sets. If your training frequency is high, that matters more; see how often to train each muscle.
How accurate is your RIR, really?
All of this relies on knowing how close to failure you are. Halperin et al. (2022) pooled 12 studies with 414 participants in which people predicted their remaining reps and then went to failure. On average, people underpredicted their reps by 0.95 (95% CI 0.17 to 1.73), meaning they had about one more rep in them than they thought. Prediction errors were much larger on sets of more than 12 reps, and accuracy improved closer to failure, in later sets, and with heavier loads. Training experience did not clearly improve accuracy.5
So when you think you are stopping at 2 RIR, you may be at 3. On a set of 20, you may be off by more. Two practical fixes:
- Calibrate occasionally. On a safe exercise, rate your RIR, then keep going to failure and compare. Do it on a machine or isolation lift, never a heavy barbell squat without safeties.
- Log RIR with every working set. Over a few weeks, you can see whether your “2 RIR” sets lead to steady rep and load progress or whether you are sandbagging.
Which exercises should you take to failure?
Because failure seems to offer a small upside at most and has real costs, use it where the cost is lowest. The cost depends on exercise type: how much muscle mass is involved, how technical the lift is, and what happens if a rep fails.
| Exercise type | Examples | Suggested RIR | Failure? |
|---|---|---|---|
| Heavy barbell compounds | Squat, deadlift, bench press, overhead press | 1–3 RIR | Rarely; only with safeties, and not as routine |
| Supported compounds and machines | Leg press, hack squat, chest-supported row, machine press | 0–2 RIR | Occasionally on the last set |
| Isolation and cable exercises | Curls, lateral raises, leg extensions, cable flyes, calf raises | 0–1 RIR | Fine on the last set, especially in higher rep ranges |
| Light, high-rep sets (about 20+ reps) | Pump work, light-load blocks | 0–1 RIR | Get very close; light sets far from failure grow little |
A simple pattern that works for most intermediate lifters: keep the first sets of an exercise at 1–3 RIR and allow the final set of isolation or machine work to go to (technical) failure. This keeps your RIR honest, limits fatigue on the big lifts, and still captures any small benefit of failure where it is cheap.
Adjust around context. In a deload week, stay well away from failure (see how to run a deload). Older lifters or those doing many weekly sets may need more reps in reserve. When you are short on time and doing fewer sets, pushing those few sets closer to failure is a reasonable trade.
A sample week, applied
Here is how a lower-body session might look with this approach:
- Back squat: 3 sets of 6–8 at 2 RIR. Add weight when all sets hit 8 at 2 RIR.
- Romanian deadlift: 3 sets of 8–10 at 2–3 RIR.
- Leg press: 3 sets of 10–12, first two at 1–2 RIR, last at 0–1 RIR.
- Leg extension: 2 sets of 12–15, last set to failure.
- Seated calf raise: 3 sets of 12–20, last set to failure.
Progress is then judged by whether reps and load climb at the same RIR over weeks, which is the signal that matters. For more ways to measure that, see progressive overload beyond weight and reps, and for how this fits into the bigger picture, our hub on the science of muscle growth.
Limits of the evidence
- Most trials last 6–12 weeks. Whether small differences compound over years is unknown.
- Many studies use untrained participants and lower-body exercises, often only the quadriceps.
- Hypertrophy is often measured with ultrasound muscle thickness at one site, which can miss regional differences.
- Definitions of failure and the accuracy of RIR differ between studies, which blurs comparisons.
- The dose-response analysis estimated RIR from study descriptions, so the exact curve is uncertain.
The practical answer
Train hard, but not recklessly. Most working sets should end within about 0–3 reps of failure, with heavy compounds at the conservative end and isolation work closer to zero. Use failure as a tool on the last set of low-risk exercises and to check your RIR, not as a requirement for growth. Then track your sets, reps, and RIR so you know whether the effort is turning into progress.
Frequently asked questions
Do you have to train to failure to build muscle?
No. Meta-analyses comparing momentary muscular failure with non-failure training find no clear hypertrophy advantage for failure, and trained lifters stopping 1 to 2 reps short have grown as much as when they went to failure. What matters is that sets end reasonably close to failure, roughly 0 to 3 reps in reserve.
How many reps in reserve is best for hypertrophy?
Most hypertrophy work should end around 0 to 3 reps in reserve. Dose-response analyses suggest growth tends to increase as sets end closer to failure, but the difference between 1 and 2 reps in reserve and true failure appears small, while the fatigue cost of failure is larger.
Which exercises should you take to failure?
Failure is safest and cheapest on isolation and machine exercises such as curls, lateral raises, leg extensions, and cable work, ideally on the last set. Keep heavy compound lifts like squats, deadlifts, and barbell presses at about 1 to 3 reps in reserve, where technique breakdown and recovery cost are higher.
Is training to failure bad for recovery?
It is more costly. In trained men, sets taken to failure caused larger drops in performance and slowed recovery of neuromuscular function by up to 24 to 48 hours compared with stopping halfway, even at the same total volume. That is a reason to use failure selectively, not to avoid it entirely.
Are lifters accurate at estimating reps in reserve?
Only roughly. A meta-analysis of 12 studies found people underestimated how many reps they had left by about one rep on average, with larger errors on high-rep sets. Accuracy improves closer to failure, with heavier loads, and in later sets, so occasionally testing failure on safe exercises helps calibrate.
References
- Refalo MC, Helms ER, Trexler ET, Hamilton DL, Fyfe JJ. Influence of resistance training proximity-to-failure on skeletal muscle hypertrophy: a systematic review with meta-analysis. Sports Med. 2023;53(3):649-665. PubMed
- Robinson ZP, Pelland JC, Remmert JF, Refalo MC, Jukic I, Steele J, Zourdos MC. Exploring the dose-response relationship between estimated resistance training proximity to failure, strength gain, and muscle hypertrophy: a series of meta-regressions. Sports Med. 2024;54(9):2209-2231. PubMed
- Refalo MC, Helms ER, Robinson ZP, Hamilton DL, Fyfe JJ. Similar muscle hypertrophy following eight weeks of resistance training to momentary muscular failure or with repetitions-in-reserve in resistance-trained individuals. J Sports Sci. 2024;42(1):85-101. PubMed
- Morán-Navarro R, Pérez CE, Mora-Rodríguez R, et al. Time course of recovery following resistance training leading or not to failure. Eur J Appl Physiol. 2017;117(12):2387-2399. PubMed
- Halperin I, Malleron T, Har-Nir I, Androulakis-Korakakis P, Wolf M, Fisher J, Steele J. Accuracy in predicting repetitions to task failure in resistance exercise: a scoping review and exploratory meta-analysis. Sports Med. 2022;52(2):377-390. PubMed
- Grgic J, Schoenfeld BJ, Orazem J, Sabol F. Effects of resistance training performed to repetition failure or non-failure on muscular strength and hypertrophy: a systematic review and meta-analysis. J Sport Health Sci. 2022;11(2):202-211. PubMed
- Santanielo N, Nóbrega SR, Scarpelli MC, Alvarez IF, Otoboni GB, Pintanel L, Libardi CA. Effect of resistance training to muscle failure vs non-failure on strength, hypertrophy and muscle architecture in trained individuals. Biol Sport. 2020;37(4):333-341. PubMed
- Lasevicius T, Schoenfeld BJ, Silva-Batista C, et al. Muscle failure promotes greater muscle hypertrophy in low-load but not in high-load resistance training. J Strength Cond Res. 2022;36(2):346-351. PubMed
- Pareja-Blanco F, Rodríguez-Rosell D, Sánchez-Medina L, et al. Effects of velocity loss during resistance training on athletic performance, strength gains and muscle adaptations. Scand J Med Sci Sports. 2017;27(7):724-735. PubMed
- Pareja-Blanco F, Alcazar J, Sánchez-Valdepeñas J, et al. Velocity loss as a critical variable determining the adaptations to strength training. Med Sci Sports Exerc. 2020;52(8):1752-1762. PubMed
