why did i stop making progress.

Published September 10, 2026·Training · Plateau science·18 min read

Progress at the gym slows on a timetable, not by accident: strength gains follow a curve that flattens after the first year, and when it stops early the cause is one of four measurable things. Here is how to tell a real plateau from a noisy one, which of the four you have, and the four-week protocol that answers it.

let me guess. the first six months were stupid. every week the bar went up, the jeans changed, someone at work asked if you'd been "doing something". you were, and it was working, and you started to believe you were a person progress happens to. then, somewhere around month eight, the log stopped moving. same squat. same press. same body in the same mirror. and you did what everyone does, which is decide the problem was you: not disciplined enough, not eating clean enough, not "pushing" enough. and then you typed it into your phone at one in the morning and every page said the same four words in a different order. overload. adaptation. overtraining. variety. not one of them cited a trial.

i went and read the trials. thirty-two of them, listed at the bottom. the answer is better than "switch up your routine", because it turns out a plateau is not one thing. it is a shape, plus one of four causes, and each cause has a different fix. most people pick the wrong fix because they never diagnosed. this file is the diagnosis. the short version is in the box.

is this normal, or did i break something.

start with the shape, because until you know the shape you cannot tell a plateau from a Tuesday. the biggest dataset on how strength changes over years comes from a 2023 analysis of the training records of 14,690 people, 60 percent of them women, average age 48, who trained once a week with single sets to failure for up to 352 weeks. the curve fits a line against the logarithm of time, which is a maths way of saying it climbs fast and then bends. people gained roughly 30 to 50 percent on their lifts in the first year and were sitting at 50 to 60 percent of baseline six years later. the authors' own word for what the slope does by year one or two is "plateau". sex, bodyweight and age barely moved the curve. one honesty line: this is once-a-week, single-set data from a private company's records, so it says nothing about whether a higher dose bends the curve later. it does say the bend is the default.

the mechanism behind the bend has been measured since 1979. Moritani and deVries trained seven men and eight women for eight weeks and split the strength gain into two parts: the nervous system learning to fire the muscle you already have, and the muscle actually growing. the nervous system did most of the work for the first three to five weeks, then growth took over. that is why beginners add weight every session without looking any different. they are not getting bigger yet. they are getting better at using what they had, and that is a loan the body only makes once.

the growth itself is slower to arrive than the mirror suggests, and the early part of it is fake. Damas 2016 took muscle biopsies from ten men in weeks one, three and ten of a program. in week one, the muscle-building signal, called myofibrillar protein synthesis, which is the rate at which the muscle assembles new contractile protein, was at its highest, and it had nothing to do with how much anyone grew. it was busy repairing damage. real fibre growth showed up only at week ten, and it tracked the week-three and week-ten synthesis, not week one. the same group's 2018 review puts a timeline on it: the size increase you see in the first four sessions is mostly swelling, modest growth arrives around session ten, and the real thing shows after about session eighteen. so the person who "grew so fast at the start and then stopped" often did not grow at the start. they swelled, they learned, and then the slow real process began, which looks like stopping if you were watching the wrong number.

and then there is the number nobody puts on page one, because it is unflattering to everyone selling a program. Ahtiainen 2003 ran eight strength athletes and eight untrained men through the same 21-week program. the untrained men added 20.9 percent to their maximal strength and 5.6 percent to their quad size. the athletes added 3.9 percent and lost 1.8 percent of quad size. same program, same weeks, same supervision. the trained body barely moved, not because it was doing anything wrong, but because the program was written for the other body. this is the whole intermediate problem in one study, and it is sixteen men, so treat the sizes as a picture rather than a prescription. the picture is that the dose that got you here does not get you there.

first: are you actually stuck, or is the log lying.

before you diagnose a plateau, prove it exists. the tool most people use to judge progress is the heaviest lift they can do on a given day, and that tool wobbles. a 2020 review of 32 studies and 1,595 people testing their one-rep max on two separate days found a median coefficient of variation of 4.2 percent, with a range from 0.5 to 12.1. a coefficient of variation is the typical size of the random swing between two tests of the same thing, expressed as a share of the number. on a 60-kilogram squat that is about 2.5 kilograms of noise in either direction, from sleep, stress, cycle phase, what you ate, who was watching. it did not matter whether people were trained, familiarised, women or men. the test itself is that loose.

so a "plateau" of two weeks at the same top set is inside the noise. a real plateau is four or more weeks with no movement in load, reps or the effort it took, on a lift you are logging honestly. the same goes for the mirror and the scale: bodyweight moves by a kilogram or two overnight for reasons that have nothing to do with muscle [the overnight file has the numbers], and visible muscle lags the log by months [the honest timeline]. the log is the only instrument that reads in the right units. if you do not have one, that is not a plateau. that is a guess.

one more thing the log will tell you that nothing else can: whether you are a slow responder rather than a stuck one. in 585 people, 342 of them women, twelve weeks of arm training produced size changes from minus 2 to plus 59 percent and strength changes from zero to plus 250 percent. in 287 untrained adults, 29 percent were low responders for muscle size, but only 7 percent for strength. the ranges are enormous and age and sex did not explain them. a slow responder still responds. the slope is lower, not zero. if your log shows small, steady movement every month, you are on the curve. if it shows none, keep going.

reason one: the stimulus stopped changing.

this is the one page one gets half right. the body adapts to a load, so the load has to keep moving, and if it does not, nothing does. what page one leaves out is that the numbers change depending on who you are. Rhea 2003 pooled 140 studies and 1,433 effect sizes and found untrained people gained the most strength at about 60 percent of their max, three days a week per muscle, while trained people needed 80 percent, two days a week, with four sets per muscle group in both cases. Peterson 2005 extended it to a third group: athletes did best at 85 percent, two days a week, eight sets per muscle group, double the beginner dose. the pattern is not subtle. as you get more trained, the intensity goes up, the sets per muscle go up, and the frequency can come down. if you are eight months in and still running the beginner numbers, you are not plateaued. you are under-dosed for who you have become.

volume on its own has a measurable price list. Schoenfeld 2017 regressed weekly sets against muscle growth across 15 studies and found each extra weekly set per muscle was worth about 0.37 percent more growth, with higher-volume groups beating lower-volume groups by 3.9 percent over a program. Hammarström 2020 did the cleanest version: 34 untrained people, 18 of them women, trained one leg with three sets and the other with one set for twelve weeks. the three-set leg grew 5.2 percent, the one-set leg 3.7 percent, and 13 of the 34 people showed a clear benefit from the higher dose. those are small percentages of a thigh, which is exactly why a plateau feels like nothing is happening: the thing that is happening is 1.5 percent.

and "change" does not have to mean more weight. Plotkin 2022 took 43 trained lifters, 16 of them women, and had half add load with reps held constant and half add reps with load held constant for eight weeks. growth was the same, with the reps group slightly ahead on the rectus femoris, and strength was slightly ahead in the load group, by a difference the authors call "of questionable practical significance". a rep is progress. Fonseca 2014 found something even less intuitive: over twelve weeks, keeping the intensity constant but rotating the exercises produced the biggest strength gains of four schemes, and the varied-exercise groups grew every head of the quad while the fixed-exercise groups missed some. so "muscle confusion" was a crime, but exercise rotation inside a structured plan is not. the difference is whether the rotation is planned and logged or whether it is a saved reel. [when am i supposed to add weight is the file on the mechanics of this.]

reason two: the effort is not where you think it is.

this is the one i see most and the one people are least willing to hear, so i will give you the trials first. Robinson 2024 ran a series of meta-regressions across the whole proximity-to-failure literature, with effort measured as repetitions in reserve, or RIR, which is how many more reps you could have done when you racked it. in every best-fit model, muscle growth increased as sets ended closer to failure, and strength gain did not change across a wide range of RIR. the authors are careful: the RIR in each study was estimated from its methods, and the analysis is exploratory. but the direction held in every model, and it is the direction people get backwards. they push their heavy strength sets to the wall and leave four reps in the tank on the sets that were supposed to grow something.

closer does not mean to the floor. Refalo 2023 pooled 15 studies and found training to set failure beat non-failure for growth by an effect size of 0.19, with the confidence interval starting at zero, and going to true momentary failure beat non-failure by nothing at all. the same group then tested it directly in 18 trained lifters, six of them women, one leg to failure and one leg stopping one to two reps short, for eight weeks. quad thickness grew 0.181 centimetres on the failure leg and 0.182 on the RIR leg. identical. the failure leg just lost more reps set to set and more bar speed, which is fatigue you paid for and did not get growth from. so the zone is one to two reps short, on the sets that count, and honestly most stuck lifters are living at four or five.

how do you know which you are? you test it, once. on your last set of a lift, at the point you would normally stop, keep going until the bar genuinely slows. count the reps you got after "i'm done". if it is three or more, your working sets have not been working sets for a while, and your plateau is an effort plateau. the fix is not more sets and not a harder program. it is the same sets, ended at one to two reps in reserve, which most people find is about two reps past where they have been stopping since March. one caveat the trials support: this applies to your growth sets. your heaviest strength work does not need to chase failure at all, and the deadlift especially should be kept well clear of it, for reasons in the deadlift file.

reason three: recovery is the invoice you have not paid.

this is the one that was mine, so i will start with the number that would have saved me two years. Murphy and Koehler 2022 pooled every randomised trial of lifting in an energy deficit for three weeks or more against lifting fed. lean mass gain was impaired with an effect size of minus 0.57. strength gain was not, minus 0.31 and not significant. their meta-regression put a line through the deficits and found a deficit of about 500 calories a day was enough to prevent lean-mass gain completely. read that as two facts at once: you can get stronger in a deficit for a while, because strength is partly skill and partly nervous system, and you cannot build muscle in a real one. so the lifter who is "eating clean" at a large deficit, hitting the gym five days a week and wondering why the mirror stopped, is not plateaued. she is running the program with the growth switched off.

i trained at 1,200 calories for years and called the result a discipline problem. i cut harder when the log stalled. i still do not have words for how backwards that was, and i am not going to tidy it up here. the workout was never the problem. the deficit was, and every time the log stalled i made the deficit worse. if that is you, the honest fix is the least glamorous sentence in this file: eat at maintenance for eight weeks, keep the protein up, and watch the log start moving again without changing a single set. [the recomposition file is the long version of that sentence.]

protein is the second line item and it has a ceiling, which is useful, because it means you can stop worrying past a number. Morton 2018 pooled 49 trials and 1,863 people and found protein supplementation added 0.30 kilograms of fat-free mass and 2.49 kilograms to a one-rep max over a program, that the effect was bigger in trained lifters, and that above about 1.6 grams per kilogram of bodyweight per day there was no further gain. for a 65-kilogram person that is around 105 grams. if you are under that and stuck, you have found a lever. if you are over it and stuck, protein is not the lever, and buying a different tub will not help. [the protein file does the per-meal maths.]

sleep is the third, and the data here is narrower than the internet pretends, so here is exactly what it shows. Knowles 2018 reviewed 17 studies: one night of no sleep barely dents strength, but consecutive nights of short sleep reduce force output in multi-joint lifts like the squat and press, and not in single-joint ones. that matches the shape of a plateau precisely: your curls are fine, your squat is stuck, and you blame the squat. the mechanism was measured in 13 adults, six of them women: after one night of total sleep loss, muscle protein synthesis fell 18 percent, cortisol rose 21 percent and testosterone fell 24 percent the next day. one night, thirteen people, so it is a mechanism study rather than a verdict. but if you are averaging six hours and your compound lifts have stalled, it is the cheapest experiment on this page. [the sleep file has the full set of numbers.]

and the one everybody reaches for first, overtraining, is the one you are least likely to have. the joint consensus of the European and American sports-medicine colleges separates functional overreaching, a short performance dip that rebounds into a gain after rest, from non-functional overreaching, weeks of decline with mood and sleep disturbance, and from true overtraining syndrome, which is rare and is a diagnosis of exclusion. their list of things to rule out before you call it overtraining reads like a recovery checklist: caloric restriction, insufficient carbohydrate or protein, iron deficiency, magnesium deficiency. in other words, the reasons one to three above. and the "just take a deload" reflex has now been tested: Coleman 2024 put 39 trained lifters, ten of them women, through nine weeks with or without a full week off in the middle. growth, endurance and power were the same. strength gains were smaller in the deload group. nine weeks is short and one week off is one design, but the direction is the opposite of the folk wisdom: if you are stuck and not sleeping and not eating, a week on the couch does not fix the invoice, it just adds a week.

reason four: you keep leaving.

the last cause is not in the gym at all. it is in the gaps between gyms. Bosquet 2013 pooled 103 studies of what happens when people stop lifting: maximal force fell with an effect size of minus 0.46, sub-maximal strength by minus 0.62, and the longer the break, the bigger the loss, with inactive people losing more than recreational athletes. this is the plateau that looks like a plateau but is actually a sawtooth: three good weeks, ten days off for a work trip, two weeks to get back to where you were, one good week, a cold. the log reads flat across six months and every individual month had progress in it. you were not stuck. you were paying the re-entry fee over and over.

the good news is the fee is smaller than it feels, and the maintenance dose is tiny. Bickel 2011 trained 70 adults for 16 weeks and then, for 32 more, dropped them to a third or a ninth of the volume, or nothing. in the young group, one-ninth of the volume held every bit of the muscle, and one-third added more. strength held largely even through full detraining. Ogasawara 2013 ran 14 men for 24 weeks either continuously or in cycles of six weeks on and three weeks off, and both groups finished with the same chest size and bench press, though the continuous group's rate slowed after week six while the cyclers re-gained fast after each break. so a planned break does not cost you the muscle. an unplanned life of breaks costs you the momentum, which is the thing the log measures.

frequency is the lever here, and it cuts both ways. Grgic 2018 found strength effect sizes of 0.74, 0.82, 0.93 and 1.08 for one, two, three and four-plus sessions a week, but the advantage disappeared once total volume was equated. Schoenfeld 2016 found hitting a muscle twice a week grew it more than once, on equal volume. and a 2020 meta-analysis of trained men found a single hard set, one to three times a week, still added about 12 kilograms to a one-rep max. put those together and the rule writes itself: two sessions a week, every week, beats five sessions in the weeks you feel like it. if your plateau is the sawtooth kind, the fix is not a better program. it is a smaller one you will not skip, which is the entire argument of the consistency file.

does this play out differently for women.

the short answer is no, and it is worth saying with numbers because so much of the plateau folklore is men-only data with a women's headline on it. Roberts, Nuckols and Krieger 2020 pooled every study that ran women and men through the same program. muscle growth was the same, effect size 0.07 and not significant, with zero heterogeneity. lower-body strength gain was the same. upper-body relative strength gain actually favoured women, effect size minus 0.60. in the 585-person Hubal study above, women's relative strength gains outpaced men's; men had a 2.5 percent edge in relative size. in the 287-person Finnish cohort, sex did not predict who was a low or high responder. the curve is the curve. the dose rules are the dose rules. the network meta-analysis behind the "80 percent, multiple sets" prescription was 45 to 47 percent women across 178 strength and 119 hypertrophy studies.

where it does differ is in the instruments. the visible-muscle timeline runs longer for a smaller frame and a lower starting muscle mass, so the mirror plateaus before the log does, and the scale plateau has a cycle sitting on top of it. the one honest sex-specific line i can give you is that most of the single trials in this file are men only, and i have said so in the references where that is true. the pooled data, where women are 40 to 60 percent of the sample, says the plateau is not a women's problem. it is a dose, effort, recovery or attendance problem wearing a body.

the protocol: four weeks to a diagnosis.

this is what i would send you if you DMed me the log. it is a diagnosis, not a program, and it takes four weeks because that is how long the noise takes to average out. change nothing for the first two weeks except how you record. then change one thing. one.

so. am i broken.

no. you have a body that did exactly what 14,690 other bodies did, which is climb fast and then bend, and then you did what almost everyone does at the bend, which is treat the shape of the curve as a verdict on your character. it is not. a plateau after year one is the curve. a plateau inside year one is a dose, an effort, an invoice or an attendance problem, and every one of those has a number attached and a fix that takes four weeks to read. the trials do not have a category for "not disciplined enough". they have RIR, sets per week, calories, hours and sessions completed. those are all things you can write down tonight.

the version of me at 1,200 calories would not have believed any of this. she would have added a session. i am not going to land this with a line. i am going to tell you to open a note, write down the last set you did and how many reps you had left, and do that again on Thursday. that is the whole intervention. the rest is reading.

Quick answers

why did i stop making progress at the gym?

because progress bends on a timetable and because one of four things changed. across 14,690 people, strength gains ran 30 to 50 percent in the first year and only reached 50 to 60 percent by year six, so a slowdown after year one is the curve. a full stop inside year one is almost always a dose that no longer matches a trained body, sets that end too far from failure, a calorie or sleep deficit the muscle cannot grow through, or missed sessions that reset the clock. log load, reps and reps-in-reserve for two weeks, then change one variable for four.

how long is a real plateau?

four weeks or more with no change in load, reps or the effort it took, on a lift you are logging honestly. shorter than that is usually noise: the one-rep-max test swings by a median 4.2 percent between sessions across 32 studies, which is about 2.5 kilograms on a 60-kilogram squat. bodyweight moves a kilogram or more overnight and visible muscle lags the log by months, so neither the scale nor the mirror can call a plateau. only the log can.

should i take a deload week to break a plateau?

probably not, unless you are planning one. the one controlled trial, 39 trained lifters over nine weeks, found a mid-program week off produced the same growth, endurance and power as training straight through and smaller strength gains. the sports-medicine consensus on overtraining says to rule out caloric restriction, low protein or carbohydrate, and iron or magnesium deficiency before calling it overtraining. if you are stuck, under-eating and under-sleeping, a week off adds a week; it does not pay the invoice.

do women plateau faster than men?

no. in the meta-analysis that pooled every study running women and men through the same program, muscle growth was the same, lower-body strength gain was the same, and upper-body relative strength gain favoured women. in a 287-person cohort, sex did not predict who responded slowly. what differs is the instrument: a smaller frame makes visible change lag the log for longer, and the scale carries a cycle on top of the noise. the dose, effort, recovery and attendance rules are the same.

keep reading

• when am i supposed to add weight. — the mechanics of reason one, rep by rep
• why can't i stay consistent. — the attendance plateau, and the program small enough to keep
• is bad sleep killing my gains. — the full sleep numbers behind the recovery branch
• how much protein do i actually need. — the 1.6 g/kg ceiling turned into meals
• why don't i look like i lift yet. — why the mirror plateaus months before the log does

---

References

educational content, not medical advice. the numbers above are group results: the long-term curve is once-a-week single-set data from company records, the trained-versus-untrained study is 16 men, the deload and failure trials are 18 to 39 people over eight to nine weeks, the sleep mechanism study is 13 people and one night, and several of the single trials are men only, which the table says where it is true. meta-analyses are cited where they exist. if a plateau comes with a missed period, persistent fatigue, or a pattern of restricting and rebounding, that is a conversation with a doctor or a dietitian, not a protocol table. everything in the tables is sourced.

- MoMo