Proximity to failure: RIR and intensity for hypertrophy

How many reps you leave in the tank changes how much a set builds muscle, but not in a straight line. What the failure trials, the meta-regressions and the RPE work say, and how to turn it into a set-by-set rule.

The short answer is that a hypertrophy set should end close to failure, not at it. Most of the muscle-building effect of effort is captured when a set stops with about one to three repetitions still available. Grinding every set to the last possible rep adds little growth and a lot of fatigue; stopping five or more reps early leaves growth on the table. The rest of this page explains where that rule comes from and where its edges are.

What “close to failure” means in numbers

Effort in a set is easiest to describe as distance from the end. Momentary muscular failure is the rep you try and cannot complete with acceptable form. Reps in reserve (RIR) counts how many more you could have done when you racked the weight. A set ended at 3 RIR stopped three reps short of failure; 0 RIR is failure itself.

The RPE scale used in strength training is the same idea expressed on a 10-point dial. Helms and colleagues (2016) anchored it to RIR so that RPE 10 means no reps left, RPE 9 means one left, RPE 8 means two, and half-points cover the cases where a lifter is unsure between two numbers. Coaches write programs as “3 × 8 @ RPE 8”; lifters read that as eight reps with roughly two in the tank.

Two related terms show up in the research and are worth separating. Volitional failure is when the lifter decides to stop because the next rep feels impossible; momentary failure is when the next rep actually fails. Some studies used one, some the other, and a few used bar-speed drop-off as a proxy for effort. That mix is part of why the literature looks noisier than it is.

The mechanism that makes effort matter

Muscle fibers are recruited in order of size. Small, low-threshold motor units fire first; as they fatigue during a set, the nervous system brings in larger, higher-threshold units to keep the bar moving. The biggest units, which control the fibers with the most growth potential, are reliably active only once the set is hard. That is the physiological reason a set of 12 with a light weight can still build muscle, provided it is taken close enough to failure that the late recruits are working.

From that comes the popular “effective reps” idea: only the last handful of reps before failure count. It is a useful mental model but an unproven one. No trial has isolated the final five reps as the sole source of growth, and the meta-analytic data below suggest the relationship is graded rather than a switch that flips five reps from the end.

Evidence, line by line

Binary comparisons: failure versus non-failure. Grgic, Schoenfeld, Orazem and Sabol (2022) pooled 15 studies of young adults that assigned one group to train to failure and another to stop short. Across all of them there was no significant difference for either strength or hypertrophy. Two subgroup results add texture. In resistance-trained participants, failure training produced a small but significant hypertrophy edge. In studies that did not equate volume between groups, the non-failure groups actually gained more strength, which is what you would expect if failure sets cut the total work a lifter could do.

Refining the comparison. Refalo and colleagues (2023) also reviewed 15 hypertrophy studies but separated momentary failure, set failure and velocity-loss thresholds. Their headline: momentary failure was not superior to non-failure training, and higher velocity-loss thresholds (which imply sets ending nearer failure) did not always produce more growth. The authors read this as a sign that the relationship between proximity and hypertrophy is probably non-linear, with a flat region near failure.

Treating effort as a dial. Robinson, Pelland, Remmert, Refalo and colleagues (2024) took a different route. Instead of grouping studies as “failure” or “not failure”, they estimated the RIR each protocol produced and ran a series of meta-regressions adjusted for load, volume equating, duration and training status. For hypertrophy the slope was negative with confidence intervals that excluded zero: the closer to failure, the more growth. For strength the slope was indistinguishable from zero. Put together with the two reviews above, the picture is a curve that rises as RIR falls and flattens at the failure end.

One controlled trial worth knowing. Lacerda et al. (2020) had young men train for 14 weeks on a failure protocol and a non-failure protocol with total volume matched. Muscle growth was similar. The practical detail is that the non-failure group needed more sets per session to reach the same volume. That is the trade-off in one sentence: closer to failure buys more stimulus per set, further from failure buys more sets.

Turning RIR into a load you can actually lift

A fixed percentage of your one-rep max does not produce the same effort every day. Sleep, stress, food and accumulated training all shift how strong you are, so 75% of 1RM for ten reps might land at RPE 7 on Monday and RPE 10 on Thursday. Prescribing by RPE flips the logic: the target is the effort, and the load is whatever produces it today.

Helms et al. (2018) tested this in trained lifters by running two periodized programs matched for sets and reps, one loaded by percentage and one by RPE. Muscle growth was similar between groups, and strength outcomes did not favor percentage loading. RPE is therefore at least a workable replacement for percentages, with the advantage that it adjusts itself to a bad day.

The weak point is rating accuracy. Research on RIR estimation from Zourdos, Helms and colleagues shows estimates are least reliable when the set is far from failure and when the lifter is inexperienced; they tighten as the set approaches failure. Beginners tend to call a set “two reps left” when four or five were available. A simple fix used by many coaches is to take the last set of a safe exercise (a machine or a cable movement) to genuine failure every week or two and compare it with the reps you predicted.

Where to place effort across a session

Two caveats sit under the table. First, effort interacts with volume: a lifter running high weekly set counts (see the volume dose-response analysis and our page on training volume for hypertrophy) usually needs to stay at the higher end of each RIR range, or recovery fails before the week ends. Second, effort interacts with frequency. Training a muscle three times a week leaves less time to recover from failure sets than training it twice, which is one reason the frequency evidence is hard to read without knowing how hard the sets were.

The costs that rise near failure

The final reps of a set are the most productive per rep and the most expensive per rep. Three costs climb together:

• Within-session fatigue. A failure set on the first exercise lowers the reps you can complete on later sets of the same muscle, so the session total can drop even as one set improves.
• Between-session recovery. Higher-effort sessions take longer to recover from. If the next session for that muscle arrives before recovery is done, the extra stimulus is paid back in lost performance.
• Technique drift. The reps closest to failure are where form tends to slip, especially on free-weight compound lifts. The load shifts to other muscles and joints at exactly the moment it is heaviest relative to capacity.

None of this makes failure dangerous in itself. It makes it a tool with a price, which is why programs that work tend to use it selectively.

Effort and food are linked

Hard sets only become muscle if the recovery inputs are there. Protein intake near the meta-analytic plateau and an energy intake at or slightly above maintenance determine whether high-effort training is converted into tissue or just fatigue. The details are in our page on nutrition for hypertrophy and in the protein intake review.

What remains unsettled

The trials behind these conclusions mostly ran 6 to 16 weeks, most participants were young men, and hypertrophy was measured with ultrasound or DXA, tools that struggle to detect small differences. Nobody has tested what happens over years of training at a fixed RIR. It is also unclear whether the flat region near failure is the same for every muscle and exercise; a curl to failure and a squat to failure are very different events. Until those gaps close, the defensible rule is the one at the top: most sets at one to three reps in reserve, failure used on purpose, and effort rated honestly enough that the numbers mean something.