Training to failure — why RIR-1 beats hammering every set.

For decades, the dominant cultural story about lifting was that intensity meant grinding every set to absolute failure — the last rep where the bar stops moving no matter how hard you push. Arthur Jones, the founder of Nautilus, built his system around it in the 1970s. Mike Mentzer popularized it as “Heavy Duty” training in the 1980s and 1990s. The bodybuilding culture absorbed the message that effort equals failure, and failure equals growth.

Twenty years of better-controlled research has substantially undermined this. The modern hypertrophy literature, much of it from Brad Schoenfeld’s group, plus the practical coaching syntheses from Eric Helms and Mike Israetel, has converged on a different position: most sets should be taken to about one rep short of failure (RIR-1), with true failure reserved for rare moments. The growth and strength outcomes are equivalent or better than constant-failure training, with substantially lower recovery cost, lower injury risk, and higher sustainable training volumes.

This article walks through what failure actually means, the RIR framework that’s now standard in evidence-based strength coaching, why the modern data favors leaving reps in the tank, and the rare cases where genuine failure still has a place.

What “failure” actually means

The word is used loosely. The distinctions matter.

Concentric failure is the point where you can’t complete another rep with the intended range of motion despite maximum effort. The bar stops moving in the lifting phase. This is what most people mean by “failure” in the gym.

Technical failure is the point where you can’t complete another rep with proper form, even if you might grind out one more with compromised mechanics (lower back rounding on a deadlift, hips shooting up on a squat, lockout collapse on a press). The useful definition, especially on barbell lifts.

Absolute failure is the point where you couldn’t move the weight at all even with any form — including isometric holds and partial reps. Almost never reached in normal training; closer to a powerlifting third-attempt scenario.

The RIR framework operates relative to technical failure. RIR-0 means you’ve hit technical failure on that set. RIR-1 means you stopped with one clean rep left in the tank. RIR-2 means two clean reps left. The honest self-assessment of RIR is a skill that improves with practice; beginners typically underestimate how many reps they actually have left and end sets too early.

Why constant failure underperforms

The case against grinding every set to failure comes from several independent lines of evidence.

Volume drops in subsequent sets. Going to failure on set one means set two produces fewer reps, set three fewer still. Total working volume — which is one of the primary drivers of muscle growth in the modern literature — declines across the session. Stopping at RIR-1 instead preserves rep count across multiple sets and produces more total stimulus over the session as a whole.

Recovery cost is non-linear. The last rep before failure is far more metabolically and neurologically costly than the second-to-last. CNS fatigue from regular failure-training accumulates across the week and reduces session quality on subsequent days. Israetel’s “Stimulus to Fatigue Ratio” framework formalizes this — the goal is the most growth signal per unit of recovery cost, and absolute failure has a poor ratio.

Form degrades on the failure rep. The mechanics that produce a clean rep at 80% of true max start to break down in the final few reps approaching failure. The failure rep itself is almost always the ugliest rep of the set. Repeatedly grooving that ugliest pattern teaches your nervous system the wrong thing.

Injury risk climbs. On compound barbell lifts specifically, the injury risk per session correlates with proximity-to-failure. Most career-shortening injuries in serious lifters come from grinding a failed or near-failed compound rep, not from controlled work in the RIR-1 to RIR-3 range.

The hypertrophy outcomes are equivalent. This is the punch line. Studies comparing failure-training to non-failure-training (stopping 1-3 reps short) consistently show equivalent or better muscle growth from the non-failure groups. The 2022 meta-analysis by Grgic and Schoenfeld pooled the available literature and found no meaningful difference in hypertrophy outcomes — meaning the recovery cost of constant failure was not buying you anything.

The implication: most lifters who grind every set to failure are taking on extra fatigue, extra injury risk, and extra time-in-the-gym for results they could have produced with cleaner, higher-volume non-failure work.

What the RIR framework actually looks like in practice

The practical implementation is straightforward. For a working set:

Pick a weight that lands you at RIR-1 on the prescribed rep range. If the program calls for 3 sets of 8 reps at RIR-1, the weight you choose should be one you could squeeze out a ninth rep on if you absolutely had to — but you stop at eight.

Across multiple sets of the same exercise, the working weight can stay constant. Earlier sets will feel further from failure (RIR-3, then RIR-2), and the last set will land at the prescribed RIR-1. The total volume across the sets is what drives adaptation.

Track honestly. Note in your log what RIR each set actually felt like. If you wrote “8 reps @ RIR-1” but in truth you had three reps left, that’s useful calibration data — load more next session. The RIR estimate gets more accurate over time with deliberate practice.

Progress by load or by reps. Once you hit the top of the prescribed rep range cleanly at RIR-1 across all sets, add weight next session. Most well-designed programs use this double-progression model. The framework keeps you adapting without forcing failure as the trigger.

When failure does still belong

The modern position isn’t “never go to failure.” It’s “failure is a tool with specific uses, not the default mode.” The cases where genuine failure still earns its place:

The last set of an isolation lift. On bicep curls, lateral raises, leg curls, calf raises — movements with low injury risk and small CNS cost — taking the final set to failure provides additional stimulus without meaningful downside. This is the most common useful application.

As a periodic calibration test. Every few weeks, taking a set to genuine failure tells you whether your RIR estimates have been accurate. If you were targeting RIR-1 with 200 pounds for 8 reps and you find you can actually grind out 12 reps to failure, your loads have been too light. Use sparingly — the calibration data is valuable, but the recovery cost still applies.

For advanced lifters with specific goals. Some bodybuilding-oriented protocols use occasional failure-rep work for the metabolic stress component of hypertrophy. This is real but applies to people training six days a week with careful periodization, not to the typical three-times-a-week lifter the rest of this section is written for.

Almost never on barbell compound lifts. Squat, deadlift, bench, overhead press — the risk profile makes the case for staying out of true failure in nearly every situation. RIR-1 or even RIR-2 is the operating range. The strongest lifters in the world almost never train their main lifts to absolute failure; they train hard in the RIR-1 to RIR-3 range and save true failure for competition.

The Mentzer counter-argument, honestly considered

Mike Mentzer is the most cited proponent of the opposite view. His “Heavy Duty” system argued for one set per exercise, taken to absolute failure, with substantial recovery between sessions. He produced strong results in an era when most bodybuilders were doing massive volumes of moderate-intensity work; his system was a corrective overreaction in the other direction.

The honest read: Mentzer’s system works. It’s also not optimal for most people in most situations. The trainees who thrived on it shared certain characteristics — they were often pharmacologically assisted (steroids blunt the recovery cost of failure work), they had unusual recovery capacity, and they had strong baseline form on the lifts they were failing on. For the natural lifter training three times a week, the constant-failure model produces more fatigue than growth signal, and the non-failure RIR-1 model produces better results at lower cost.

Mentzer’s deeper insight — that most lifters train too much volume at insufficient intensity — was correct. His specific prescription, that the answer is always-failure training, was an overcorrection. The modern synthesis (high intensity within a reasonable volume, stopping just short of failure most of the time) splits the difference and outperforms either extreme.

Where to start

A sound working rule is RIR-1 on the last set of every compound lift, with earlier sets at RIR-2 or RIR-3 at the same load. The last working set should feel hard enough that one more rep is visibly possible but would take a deliberate decision to push for — a decision rarely worth making on a barbell compound, where the recovery cost-to-benefit ratio doesn’t justify it.

On isolation work — the last set of curls, lateral raises, calves — taking that final set to true failure occasionally is fine: low injury risk, low CNS cost, modest additional stimulus. Use it where it makes sense; don’t default to it.

Once every six to eight weeks, an honest failure-test set on a single lift is worth doing — pick a weight you’d expect to get 8 reps with and see whether it’s actually 10 or 12. The result calibrates whether the RIR estimates have been honest or whether the training has been too light. Often it’s the latter; that’s useful information.

The unifying principle: failure is data, not the destination. Train hard enough to drive adaptation, stop short of the point where you’re paying disproportionate recovery cost for marginal additional stimulus, and the long arc of training stays productive for decades instead of stalling in burnout after a year.

Books and tools I’d recommend

The case made in this article is laid out in much more detail in a few standard references worth reading.

The Muscle and Strength Pyramid by Eric Helms is the most accessible serious introduction to evidence-based hypertrophy and strength training, including the RIR/RPE framework. Two volumes (Training and Nutrition); the Training volume is the one most relevant here. Reads cleanly without requiring a background in exercise science.

Scientific Principles of Hypertrophy Training by Israetel, Hoffmann, Smith, and Feather is the heavier, more technical treatment. The Stimulus to Fatigue Ratio framework, MEV/MAV/MRV volume landmarks, and the structured programming templates that have shaped modern coaching all come from this lineage. If you’re going to design your own training, this is the textbook.

Heavy Duty II by Mike Mentzer is worth reading even though I disagree with the specific prescription — Mentzer’s critique of the volume-heavy culture of his era still lands, and understanding the original case for failure-training helps you understand why the modern position landed where it did. Short, opinionated, occasionally infuriating.

A simple training notebook is the cheapest piece of equipment in this entire section that produces an outsized return. Record sets, reps, load, and RIR for every session. Review weekly. The data is what makes the RIR framework work — without honest logging you’re guessing at your own training. A Moleskine works fine; a notes app on your phone works fine; pick the format you’ll actually use.

Lifting straps are worth mentioning because grip failure on heavy pulls (deadlifts, rows) often arrives before back or leg failure does — meaning you stop the set because the bar slipped out of your hands, not because the target muscles actually hit RIR-1. Straps let the working muscles get the stimulus you came for. Versa Gripps PRO are the standard recommendation; cheaper canvas straps work too.

The bottom line

Training to absolute failure on every set is the most common avoidable mistake in serious recreational lifting. The modern hypertrophy and strength literature consistently shows that stopping at RIR-1 produces equivalent or better growth and strength outcomes at substantially lower recovery cost, lower injury risk, and higher sustainable training volumes.

The practical rule: stop your working sets with one clean rep left in the tank. Reserve true failure for isolation lifts where the calculus favors it and for occasional calibration tests. Almost never grind a compound barbell lift to failure. Track your RIR honestly; let the data improve your estimates over time. Progress by load or by reps within the framework.

Done this way, lifting becomes a practice that compounds for decades instead of a sprint that burns out in a year. That’s the difference between training the way the body adapts and training the way ego demands.