Strength plateau: why a lift stalls and what to change first

Three sessions without a rep is well inside normal variation. Sleep, food, stress, time of day and how hard you judged the last set all move a single session's performance. What distinguishes a stall from noise is a trend at a fixed load: the same weight producing fewer reps than it did a fortnight ago, at the same or higher effort. One bad session is weather; a downward slope at constant load and constant effort is a signal.

This is also why effort has to be part of the judgement. A set of 5 at RPE 7 and a set of 5 at RPE 10 are not the same result, and only one of them means you are out of room.

Weekly working sets per muscle has the best-established relationship with growth of any variable here. Schoenfeld, Ogborn and Krieger's dose–response meta-analysis found gains rising with weekly sets, roughly 0.37% additional muscle per weekly set across the range they examined, while also noting that low volumes of four or fewer weekly sets still produce substantial gains. More is not free, but the direction is consistent.

A newer meta-regression of 67 studies, by Pelland and colleagues, found the same direction for both muscle size and strength, with diminishing returns that were considerably sharper for strength. So extra sets are a weaker fix for a stalled one-rep max than for stalled growth. It also found that counting a set that trains the muscle only indirectly, such as a bench press for the triceps, as half a set predicted results best.

Two failure modes follow, and they look identical from the outside. Too little volume and there is not enough stimulus to drive adaptation. Too much and you accumulate fatigue faster than you recover from it, which suppresses performance while the training itself is still nominally productive. Both present as a lift that stopped moving.

The practical test is arithmetic rather than sensation: count the actual working sets that muscle received in each of the last three weeks. Schoenfeld's group sorted their studies into under 5, 5 to 9 and 10 or more weekly sets per muscle. Average growth was 5.4%, 6.6% and 9.8%, but as groups that was only a trend (P = 0.074), and the 10-plus advantage disappeared when one influential study was removed. The authors still wrote that at least 10 weekly sets "would appear" necessary for the most growth. So treat ten as a rough line, not a tested rule. Well under it, adding sets is a reasonable first try. If the count has crept well past your usual and the lift went backwards, take sets away first.

Robinson and colleagues' series of meta-regressions on estimated proximity to failure found that muscle growth increased as sets ended nearer failure (the confidence intervals for those slopes excluded null), while for strength the intervals contained null, indicating a negligible relationship. The same authors are explicit about the limitation: reps-in-reserve values had to be estimated from study descriptions rather than measured, so the precise dose–response is unclear.

The ceiling arrives quickly, though. Refalo's meta-analysis puts the hypertrophy advantage of taking sets to failure at ES 0.19 (95% CI 0.00–0.37), an interval that touches zero, and reports no advantage at all for momentary muscular failure specifically (ES 0.12, 95% CI −0.13 to 0.37). For strength, Davies found a small pooled effect favouring non-failure training (ES 0.34), worth 0.6–1.3%, which its own authors judged unlikely to be meaningful.

And the extreme is actively counterproductive. In Robinson's four-arm trial in trained men, the arm taking every set to failure gained 0.71 kg on the bench press (95% CI −4.41 to 5.62) against 9.05 kg at 4–6 reps in reserve, and the authors caution that the to-failure arms are hard to interpret because the protocols proved poorly feasible. If the lift stalled while you were grinding every set, that is a plausible cause and a cheap thing to reverse.

There is a subtler version of this. Armes and colleagues found that trained people who stopped at their own self-determined limit still had about two repetitions left (95% CI 0.0 to 4.0). If you believe you are training near failure, you may be further away than you think, which is an argument for checking your effort calibration before concluding you need more of it.

If you are eating in a deficit, expect the scale to move on the plate before it moves on the bar. Murphy and Koehler's meta-analysis found energy deficiency impaired resistance-training gains in lean mass (ES −0.57, p = 0.02) while gains in strength were statistically comparable to eating at maintenance (ES −0.31, p = 0.28). Their meta-regression scaled the penalty with the size of the deficit, and put the point where lean-mass gains stopped at a deficit of about 500 kcal a day.

That asymmetry helps with diagnosis. A stalled lift during a deliberate cut is expected and not evidence that the programme is wrong; visible growth stopping in a deficit is even more expected. Neither is a reason to add volume.

Sleep belongs in this section too, but none of the studies above measured it, so this guide puts no number on it. If you slept badly in the weeks the lift slipped, write that next to the log before you change the programme.

Taking a lighter week is the most popular response to a stall and the least supported by direct evidence. The trial in trained lifters (Coleman and colleagues, 39 participants, a one-week deload at the midpoint of a nine-week programme) found that the continuous group improved more in both isometric and dynamic lower-body strength, with no appreciable difference in muscle size. The authors' conclusion is that a mid-programme deload week appears to negatively influence lower-body strength while doing nothing for hypertrophy, power or muscular endurance.

Coleman's deload was a full week with no training at all. A 2026 trial in 19 untrained men, by Pancar and colleagues, tested a lighter version: in weeks 4 and 8 of an eight-week programme, one leg and one arm trained once a week for 2 sets instead of twice a week for 6 to 8. Growth and 10-rep-max strength were similar with and without those lighter weeks. So one deload cost trained lifters strength, and the other neither helped nor hurt beginners. No trial we found has compared deload sizes against each other, so any figure for how much to take off, half the sets or 10% of the load, is a convention.

"Deload" also spans a wide range of interventions: a full week off is not the same as trimming one exercise's load by 5%, which is what a per-lift adjustment does. But the trials do mean that anyone telling you a deload week is the obvious first response to a stall is stating a practice, not a finding.

1. Confirm the trend. Same load, fewer reps, same-or-higher effort, across two to

three weeks. If not, change nothing.

1. Count the sets. Well under about ten weekly working sets for that muscle, the

rough line from the volume research: add. Well above your own norm with performance falling: subtract.

1. Check your effort calibration before adding any. You are probably further from

failure than you feel, and taking every set to failure is the one pattern with trial evidence against it.

1. Rule out the deficit. Stalling while cutting is the expected result, not a

programming fault.

1. Then, and only then, change the lift: reduce the load slightly and rebuild,

swap to a close variation, or change the rep range.

Change one variable and give it three weeks. Changing three at once means learning nothing from the outcome.

Say you bench once a week and log three working sets at 100 kg (220 lb):

1. The trend is real. Same weight, fewer reps each week, and the effort went up.

That is three weeks of evidence, not one bad day.

1. Count the chest sets. Bench plus three sets of flyes is 6 a week, well under

ten, so two or three more sets is a fair first try. If you already did 12, and that had jumped from 8 the month before, cut back instead.

1. Check the effort. RPE 8 in week 1 should mean two reps left. If those sets were

really grinds, the lift was closer to its limit than the log says.

1. Check the diet. If you started a cut in week 1, this pattern is expected.
2. Then change the lift. For example, drop about 5% to 95 kg (210 lb) and build

the reps back up.

This article cannot tell you which of these applies to you, and it deliberately does not rank them by how likely they are, only by how well-evidenced the response is. The volume figures come from between-study dose–response models, which describe averages across populations and not the count that is right for your shoulder. The proximity-to-failure evidence rests on reps-in-reserve values that were estimated after the fact, which its own authors flag. The deload evidence is two small trials, of 39 and 19 people, each with one specific protocol. Nobody has run the trial that would actually answer the reader's question: take a stalled trained lifter, randomise the response, and measure what recovers the lift fastest.

Shojin flags stalling lifts on the Home screen, under "Needs attention". It uses a version of the trend test described above, not a single disappointing session. A lift is flagged after two stuck sessions in a row, no more than 30 days apart. A stuck session is a top set at RPE 10, or fewer reps at the same load with the top set at RPE 9 or more (or no RPE logged). An RPE 9 top set that kept its reps does not count. On a flagged lift, the progression suggestion stops adding load. It holds the weight, or backs off about 5% on that exercise. Note what that is and isn't: it is a per-lift load adjustment, not the week-long programme deload that Coleman's trial tested and found wanting, and neither result validates the other.

On the bench press example above, Shojin suggests 102.5 kg after week 1 (RPE 8 left room). After week 2, with fewer reps at RPE 9, it suggests 95 kg for 7. After week 3 the lift appears under "Needs attention", with 95 kg for 6.

Shojin does not track sleep hours or calories, so the recovery question above is one it cannot see and cannot answer for you.

Shojin is one subscription with no cap on sets. Shojin Coach is part of it: it writes your program from your own lifts and adjusts it after every session you log.

New subscribers get one month free, then A$4.99 a month or A$34.99 a year, billed by the App Store in your local currency. The subscription renews until you cancel. Cancel any time in Settings. You subscribe in the iPhone app. If you stop, your history stays, and Coach stops adjusting your program.

Every number in this article traces to one of these. Where the evidence is contested or thin, the article says so rather than picking a side.