Training to failure builds a little more muscle, but not more strength

Whether you grind every set to the point where another repetition is impossible barely changes how strong you get, and adds only a little more muscle than stopping a rep or two short [s1]. That is the picture from the most complete analysis of the question to date, and two controlled trials since have found the muscle difference small enough to be hard to see [s2][s3].

The variable in question is "proximity to failure," usually quantified as repetitions in reserve (RIR) — how many more reps you could have done when you racked the weight. Zero RIR is a set taken to momentary muscular failure; two RIR means you stopped with two in the tank. The training culture treats hard failure as the price of results. The evidence treats it as one dial among several, with a smaller effect than the effort it costs.

What the meta-regression found

In 2024, researchers led by Zac Robinson and Jacob Pelland pooled the resistance-training literature and ran a series of meta-regressions relating estimated RIR to two outcomes, strength and muscle growth, while adjusting for load, how volume was equated, intervention length and training status [s1]. The two outcomes came apart. For strength, the best-fitting models put the relationship with RIR at essentially nothing: the confidence intervals around the slope contained a null value, meaning strength gains were similar across a wide range of stopping points [s1]. For muscle hypertrophy, the slope was negative and its confidence interval excluded null — growth increased as sets were terminated closer to failure [s1].

That is the headline, and it is worth stating precisely: closer to failure helped size, not strength. The authors were also candid about the limits. The RIR for each study was estimated from written descriptions rather than measured, the overall model fit was modest, and they framed the work as exploratory, cautioning against reading exact prescriptions into it [s1]. It narrows the question rather than closing it.

Two trials, the same answer

Controlled trials that pit failure against a fixed reps-in-reserve target land in the same place. A 2024 study at Deakin University randomised the two legs of each of 18 resistance-trained adults (12 men and 6 women) so that one leg trained the leg press and leg extension to failure and the other stopped shy of it, twice weekly for eight weeks [s3]. Quadriceps thickness rose almost identically: 0.181 cm in the failure condition (95% highest-density interval 0.119 to 0.243) and 0.182 cm with reps in reserve (0.115 to 0.247) [s3]. Because sets taken to failure lost more reps across a session, the reps-in-reserve legs accumulated similar total work with less fatigue [s3].

A 2025 trial from Brad Schoenfeld's group took a different design — 42 trained men and women doing a single set of nine exercises twice a week for eight weeks, one group to failure and one holding two reps in reserve — and reached a compatible verdict [s2]. Several measures of muscle thickness tended to favour the failure group, but the absolute differences were generally modest, while gains in strength and muscular endurance were similar between the groups [s2]. The trial's other finding is practical: a single set per exercise was enough to drive gains even in trained lifters transitioning from higher-volume routines [s2].

What it means, and doesn't

None of this says failure is useless. It says the return on the last, hardest reps is specific and small: a modest edge for muscle size, little to nothing for strength [s1][s2]. For a reader whose goal is strength, the evidence gives permission to leave a rep or two in reserve and gain about the same, with less fatigue to recover from [s1][s3]. For someone chasing maximum size, training near failure appears to help at the margin, but "near" is doing real work in that sentence — the trials above used two reps in reserve, not failure, and still grew muscle [s2][s3].

The honest caveats are in the sign of the effect and the size of the studies. The trials ran eight weeks with a few dozen participants each; the meta-regression rests on estimated rather than measured effort [s1][s2][s3]. What has moved is the framing. Failure is no longer the gate every set must pass through — it is a lever with a known, limited effect, and one that costs recovery to pull. That fits the broader hypertrophy picture, where weekly volume does more of the work than any single set variable, and it bears on how much soreness and fatigue a programme generates, a subject where the popular fixes turn out to be poorly evidenced. For lifters weighing effort against payoff over a career, the dose-response for resistance training and health sits alongside these findings as a reminder that consistency, not maximal strain, is what most of the evidence rewards.