Training to Failure for Hypertrophy: Is It Necessary?

No. Training to failure is not necessary for hypertrophy. Every 2021-2024 review and trial agrees: stop 1-3 reps short and you build the same muscle as going to failure, with less fatigue. The catch: most lifters who think they are 2 reps out are really 3 or 4 away. Riven is the Apple Watch app that scores muscle failure.

Muscle failure in the gym is the point in a set where you physically cannot complete another rep with good form — the bar stalls, the muscle gives out (the full definition, with the concentric/technical/absolute distinctions, is in what muscle failure actually is). RIR in lifting ("reps in reserve") is how many more reps you had left when you stopped — reps in reserve, explained if you want the long version. A set taken to true failure is 0 RIR. A set stopped two reps early is 2 RIR.

What to do this week

Five rules. The evidence behind each one is further down the page.

• Heavy barbell compounds (squat, bench, deadlift, row): 1-3 RIR. Leave a rep or two. The growth stimulus is already banked by then, and the last grinding rep is where form goes.
• Machine compounds (leg press, chest press, hack squat): 0-2 RIR. Lower risk, so push harder.
• Isolation and cable work (curls, lateral raises, leg extensions, pushdowns): 0-1 RIR, with the last set of the exercise taken to failure.
• Once this week, on a machine, keep going past "done" and count the extra reps. Pick a set you would normally stop at "2 left" and instead keep repping until the weight will not move. The number of extra reps you got is your error, and it is usually bigger than you expect. (The other way to check is to read how much your reps slow down across the set: can an Apple Watch detect muscle failure?)
• If you train alone, pick lifts that fail safely. Machines, cables and dumbbells can be taken to failure with nobody watching; a heavy barbell cannot. How to train to failure without a spotter has the list.

Is training to failure necessary for hypertrophy?

Training to failure is not strictly necessary for hypertrophy. When researchers match volume and stop sets near failure, going to absolute failure adds little. A meta-analysis of 15 studies by Refalo et al. (2023) found only a trivial advantage for training to set failure versus stopping short — a standardized effect size of 0.19 (confidence interval 0.00 to 0.37). Trivial. The growth stimulus is mostly captured before that last grinding rep.

Here's how I coach it in practice: failure is a tool, not a requirement. You can grow plenty without ever touching it, as long as the effort is honest and the volume is there.

Is 1-3 reps in reserve close enough? (what "not necessary" actually means)

Yes — 1-3 RIR is close enough to capture essentially all of the hypertrophy stimulus, and that is exactly what "training to failure is not necessary" means in the reviews. It does not mean effort doesn't matter. Sets stopped more than ~5 reps short consistently underperform — that's where junk volume lives. Robinson's slope says closer is better; Refalo's and Ruple's trials say 1-2 RIR and 0-1 RIR match failure; the fatigue data say the last rep costs the most. Put together, the target is a narrow band just shy of failure. The 1-3 / 0-2 / 0-1 bands at the top of this post are that band, split by how much a failed rep costs on each kind of lift. A fuller breakdown by goal — including where strength work should sit — is in how close to failure you should train.

Now the trap. You almost certainly leave more in the tank than you think. A scoping review and meta-analysis summarized by Stronger By Science (Halperin et al., 12 studies, 414 participants) found lifters underpredict reps to failure by about 0.95 reps on average. When you call it "2 reps left," roughly 3 are usually there. Accuracy is worse when you're fresh, worse on early sets, and worse the further you are from failure — it only tightens up as you close in. And that average hides a steep experience gradient. In Steele et al. (2017), 141 trainees predicted how many reps they had left and were then taken to momentary failure on six exercises: lifters with under six months of training underpredicted by about 3.6-4.7 reps, against roughly 1.5 for lifters with 1-3 years of experience and 0.8 for those with 3 or more.

Do the math on that. A lifter aiming for "3 RIR" who undershoots by even one rep is really training at 4 RIR, and on a fresh first set maybe 5 — right at the edge of the underperforming zone. The plan said "close to failure." The execution said "junk volume." This happens constantly, and the lifter never knows. That is why I care less about the "failure vs 2 RIR" debate than about how to measure reps in reserve at all.

Riven is the Apple Watch app that scores muscle failure. It reads your wrist motion and stays quiet during the set; about three seconds after you rack the weight you get the exercise, the rep count and a 0-100 score. Reps slow down as a muscle nears failure, and that is what the score is built on. No barbell clip, no camera, no extra hardware. It shows you which set was the real one.

So: failure isn't necessary. Training close enough is. The hard part was never the rule — it was knowing when you've hit it.

Should you train to failure every set?

No. Keep most sets at 0-3 RIR and reserve true failure for the last set of an exercise, lighter isolation work, or the final week of a block — not the whole session. The answer also splits by goal, because the Robinson meta-regressions found that hypertrophy responds to proximity while strength does not. A powerlifter and a physique competitor should not be programming failure the same way. Here is the practical table I give my own lifters:

Failure backfires through fatigue you pay after the set, not during it. In the Refalo et al. (2023) crossover trial, a set of bench press at 75% 1RM taken to failure left lifting velocity down 25% four minutes later; stopping at 1 RIR cost 13%, and 3 RIR only 8%. Failure roughly doubled the neuromuscular fatigue of 1 RIR and tripled that of 3 RIR, for no extra growth. Everything recovered by 48 hours, but the cost lands inside the same session and across the week: that extra fatigue eats into the volume you can do on your next sets, and since accumulated hard volume is the real growth driver, chronic failure training can quietly lower your total weekly stimulus. You spend everything on set one and limp through the rest.

Then there is the part nobody programs for: how much it hurts. Refalo et al. (2025) found that training to failure pushed session RPE up by 1.1 points (5.4 vs 4.3), raised perceived discomfort, and produced markedly worse post-set feelings. Over months that is an adherence tax. Near-failure work you keep doing for years beats "optimal" failure work you quit by week six.

Does training to failure build more muscle?

Marginally, sometimes, under specific conditions — and the biggest one is load. Failure matters most when the weight is light. Lasevicius et al. (2019) showed that at 30% of 1RM, training to failure produced 7.8% hypertrophy versus 2.8% for non-failure — a real gap. But at 80% of 1RM, failure added no benefit at all.

The mechanism is intuitive once you see it. With a light load, the early reps barely recruit your high-threshold motor units. You have to fight through fatigue to get there — effort substitutes for tension. With a heavy load, you're recruiting those big fibers from rep one, so you don't need to crawl to failure to stimulate them. This is the physiology behind the effective reps idea: the reps that count are the slow ones near the end, and a heavy load gets you there sooner.

So "does failure build more muscle?" splits by context:

• Light loads / high reps (cables, machines, blood-flow-restriction-style work): yes, push close to or to failure — effort is doing the work that load isn't.
• Heavy compound barbell work (~80% 1RM): no meaningful added benefit, and the technique-breakdown risk goes up.

There's also a population wrinkle. Grgic's subgroup analysis found a small, significant edge for failure training in resistance-trained lifters (ES 0.15) and none in untrained people. The honest answer to "is failure necessary" is load- and population-dependent, not universal: a beginner on heavy barbell work gains nothing from it; an advanced lifter on light isolation work probably gains a little.

And for strength specifically? The Robinson meta-regressions found strength gains flat across a wide RIR range, and Grgic's pooled strength comparison was a null (ES -0.09) — in fact the studies that didn't equate volume slightly favoured non-failure for strength (ES -0.32), presumably because failure training cost sets. Failure's modest edge is a size thing, not a strength thing. Layne Norton's review lands the same way.

The fatigue trade-off

This is where I get opinionated. Equal per-set growth does not mean equal weekly growth — because failure is expensive, and the fatigue cost is now quantified.

Vieira's meta-analysis put numbers on it: across 12 pooled crossover studies, sets taken to failure produced a large drop in acute performance (standardized mean difference -0.96 for velocity, power and jump height), about 4.5 mmol/L more lactate, more creatine-kinase-measured muscle damage (SMD 0.76) and a much higher RPE (SMD 1.93) than the same work stopped short. Refalo's 2023 crossover trial shows the gradient within a single set type: bench press to failure left lifters down 25% in velocity four minutes later, 1 RIR down 13%, and 3 RIR only down 8%. Everyone had recovered by 48 hours — but those first hours are the rest of your session.

Now play it forward. More fatigue per set means fewer quality sets afterward. A messier next exercise. Slower recovery before your next session. If chasing failure on set one tanks the quality of sets two through five — or forces you to cut a session short — you can end up with less total quality volume for the week. Less weekly stimulus. Less growth. (Whether to take every set there is its own question — see "Should you train to failure every set?" above — and the answer is no.)

That's the trade-off most "go hard or go home" lifters never account for. They optimize the single set and quietly sabotage the week.

Failure also isn't free on the injury and form ledger. The rep where you fail a heavy squat or deadlift is the rep where your spine rounds and your knees cave. The fatigue cost is most contained — and the upside most present — on stable, low-risk movements: machines, cables, isolation. Save failure for the last set of those. Almost never bring it to a heavy multi-joint barbell lift, and if you train to failure without a spotter, never.

Is closer to failure always better? (the dose-response)

No — the dose-response flattens. People assume "closer to failure = more growth, always," a clean linear ramp. The data don't show that. In the Refalo meta-analysis, when you isolate momentary muscular failure vs non-failure specifically (5 studies), the effect size drops to 0.12 (confidence interval -0.13 to 0.37, p = 0.343) — statistically nonsignificant. And when the same review graded proximity by velocity loss, hypertrophy effect sizes looked like this:

Notice what happens. The curve climbs from low to moderate effort, then flattens. Going from "high effort" to "actual failure" barely moves the needle (0.42 to 0.41). This is a diminishing-returns relationship, not a dose-dependent one. The last rep or two is mostly cost, little benefit. (If you want to use those velocity-loss bands as actual stop rules, velocity-loss stop thresholds walks through the numbers.)

The strongest single piece of evidence is a direct trial. Refalo et al. (2024) took 18 resistance-trained adults, trained them twice a week for eight weeks, and compared a 2-RIR/1-RIR group against a to-failure group. Quadriceps thickness gains: 0.182 cm for the RIR group, 0.181 cm for the failure group. Identical. But the failure group ate consistently greater lifting-velocity loss and more per-set fatigue — for zero extra growth.

That's the punchline of the modern literature: same result, more wear.

What do the systematic reviews and meta-analyses actually say? (2021-2024)

They agree more than the internet arguments suggest: failure is not required for growth, it does nothing extra for strength, and it costs real fatigue. Here is every major review and the key trials people search for, with the headline number from each abstract and a link to the paper itself — no paraphrase in between.

The one sentence from that literature that gets quoted most — and that people type into Google verbatim — is the conclusion of the Robinson meta-regressions:

"Strength gains were similar across a wide range of RIR, while muscle hypertrophy improves as sets are terminated closer to failure." — Robinson, Pelland, Remmert, Refalo, Jukic, Steele & Zourdos, Sports Medicine, 2024 (PubMed 38970765)

In plain English: if you want to get stronger, it doesn't much matter whether you stop 1 rep or 4 reps short — heavy load and practice do that job. If you want to get bigger, closer is better, on average, but the authors themselves flag that the RIR in most of those studies was estimated from training descriptions, that the models fit only modestly, and that "closer is better" is a slope across a range, not proof that the last rep matters. Read alongside the Refalo and Ruple trials, the honest summary is: closer is better, but 1-3 RIR gets you essentially all of it.

Which paper should you read first? (a reading list)

If you only have time for one or two, pick by the question you actually have:

1. "Is failure necessary for growth?" — read Refalo et al. 2023, Sports Medicine: the 15-study meta-analysis with the 0.19 / 0.12 effect sizes and the velocity-loss dose-response table.
2. "Does it differ for strength vs size?" — read Robinson et al. 2024, Sports Medicine: the meta-regressions behind "strength gains were similar across a wide range of RIR."
3. "Show me a head-to-head trial, not a pooled estimate" — read Refalo et al. 2024, J Sports Sci (0.182 vs 0.181 cm) and Ruple et al. 2023, Physiol Rep (0-1 vs 4-6 RIR).
4. "What does failure cost me?" — read Vieira et al. 2021, Sports Medicine for the pooled fatigue numbers, Refalo et al. 2023, Sports Med - Open for the -25 / -13 / -8% velocity gradient, and Refalo et al. 2025, Eur J Sport Sci for how much worse it feels.
5. "Does it depend on load?" — read Lasevicius et al. 2019: failure matters at 30% 1RM, not at 80%.
6. "How wrong is my own RIR guess?" — read Steele et al. 2017, PeerJ: 141 trainees, with novices underpredicting reps to failure by 3.6-4.7 reps.

FAQ

Is training to failure necessary for hypertrophy?

No. With matched volume and sets taken close to failure (about 0-3 RIR), stopping short builds essentially the same muscle as going to failure — the 2024 Refalo RCT measured 0.182 cm versus 0.181 cm of quad growth — while generating much less fatigue.

Should you train to failure every set?

No. Keep most sets at 0-3 RIR. Reserve true failure for the last set of an exercise, lighter isolation work, or the final week of a training block. Going to failure every set adds fatigue without proportional growth.

Is there a systematic review showing training to failure is not necessary for hypertrophy?

Yes, several. Grgic et al. (2021/22, 15 studies) found no overall hypertrophy difference between failure and non-failure (ES 0.22, CI crossing zero); Refalo et al. (2023, 15 studies) found only a trivial edge for failure (ES 0.19) that disappeared for momentary failure specifically (ES 0.12, p = 0.343). Both are open access on PMC and linked in the table above.

What does "strength gains were similar across a wide range of RIR" mean?

It is the conclusion of Robinson et al.'s 2024 meta-regressions in Sports Medicine: how many reps you leave in reserve had no measurable relationship with strength gain, while muscle growth did improve as sets got closer to failure. Practically — load and practice build strength; proximity to failure nudges size.

Do the 2023-2024 reviews agree with each other?

Broadly, yes. Refalo 2023 (categorical) says failure adds little for size; Robinson 2024 (continuous) says closer is somewhat better for size but irrelevant for strength; the Refalo 2024 and Ruple 2023 trials show 1-2 RIR and 0-1 RIR matching failure; Vieira shows failure costs more fatigue. The consensus is "train close, not necessarily to, failure."

Does training to failure build more muscle than stopping short?

Only trivially when matched, and mostly with light loads. At 30% 1RM, failure beat non-failure (7.8% vs 2.8% hypertrophy); at 80% 1RM it added nothing. Heavy loads recruit the big fibers without needing failure.

Is closer to failure always better for growth?

No — the dose-response plateaus. Effect sizes climb from low to moderate effort, then flatten (about 0.39 to 0.42 from ~20% velocity loss up to failure). The last rep or two adds little once you're already close.

How many reps in reserve should I leave?

Roughly 0-3 RIR for hypertrophy. I'd use 1-3 RIR on heavy barbell lifts, 0-2 on machines, and 0-1 on isolation work. Just remember most people undershoot true RIR by about a rep, so your "3" may really be a "4."

Does failure help strength?

Not meaningfully. Meta-regression shows strength gains are roughly flat across a wide RIR range, and Grgic's pooled comparison was a null. Failure's small edge is specific to muscle size, not strength.

Sources

• Grgic, Schoenfeld, Orazem & Sabol (2021/2022), Effects of resistance training performed to repetition failure or non-failure on muscular strength and hypertrophy: A systematic review and meta-analysis, Journal of Sport and Health Science — https://pmc.ncbi.nlm.nih.gov/articles/PMC9068575/
• Vieira et al. (2021/2022), Effects of Resistance Training to Muscle Failure on Acute Fatigue: A Systematic Review and Meta-Analysis, Sports Medicine — https://pubmed.ncbi.nlm.nih.gov/34881412/
• Refalo et al. (2023), Influence of Resistance Training Proximity-to-Failure on Skeletal Muscle Hypertrophy: A Systematic Review with Meta-analysis, Sports Medicine — https://pmc.ncbi.nlm.nih.gov/articles/PMC9935748/
• Refalo, Helms, Hamilton & Fyfe (2023), Influence of resistance training proximity-to-failure, determined by repetitions-in-reserve, on neuromuscular fatigue in resistance-trained males and females, Sports Medicine - Open — https://pmc.ncbi.nlm.nih.gov/articles/PMC9908800/
• Ruple et al. (2023), The effects of resistance training to near failure on strength, hypertrophy, and motor unit adaptations in previously trained adults, Physiological Reports — https://pmc.ncbi.nlm.nih.gov/articles/PMC10161210/
• Refalo et al. (2024), Similar muscle hypertrophy following eight weeks of resistance training to momentary muscular failure or with repetitions-in-reserve, Journal of Sports Sciences — https://pubmed.ncbi.nlm.nih.gov/38393985/
• Robinson, Pelland, Remmert, Refalo, Jukic, Steele & Zourdos (2024), Exploring the Dose-Response Relationship Between Estimated Resistance Training Proximity to Failure, Strength Gain, and Muscle Hypertrophy: A Series of Meta-Regressions, Sports Medicine 54(9):2209-2231 — https://pubmed.ncbi.nlm.nih.gov/38970765/ (preprint: https://sportrxiv.org/index.php/server/preprint/view/295)
• Refalo et al. (2025), The Effect of Proximity-To-Failure on Perceptual Responses to Resistance Training, European Journal of Sport Science — https://pmc.ncbi.nlm.nih.gov/articles/PMC11832030/
• Lasevicius et al. (2019), Muscle Failure Promotes Greater Muscle Hypertrophy in Low-Load but Not in High-Load Resistance Training — https://pubmed.ncbi.nlm.nih.gov/31895290/
• Steele et al. (2017), on the accuracy of predicting repetitions to momentary failure across resistance-training experience, PeerJ — https://pmc.ncbi.nlm.nih.gov/articles/PMC5712461/
• Biolayne (Layne Norton) Research Review, The Failure Factor — https://biolayne.com/reps/issue-32/the-failure-factor-the-relationship-between-proximity-to-failure-and-strength-and-hypertrophy/
• Stronger By Science (Greg Nuckols), summary of Halperin et al. on predicting repetitions to failure — https://www.strongerbyscience.com/reps-in-reserve/