Every repetition has two halves, and almost all training attention goes to the first one. The lowering is treated as the bit you get through to arrive back at the start. That is the half where the adaptation is — and for anyone whose limit is energy rather than strength, it is a door most people never notice.
You lift the weight, and you lower it.
The lowering usually happens quickly, with gravity doing most of the work, and nobody counts it. Which is a shame, because there is a whole class of physiological effect sitting in that half of the movement, and some of it is not available any other way.
Concentric — the muscle shortens while producing force. Standing up out of a squat. Curling the weight up.
Isometric — the muscle produces force at constant length. Holding a plank, a dead hang, a wall sit.
Eccentric — the muscle lengthens while producing force. Lowering into the squat under control. Lowering the weight rather than dropping it.
That third one is genuinely unusual. The muscle is being stretched by an external force and resisting it at the same time. It is braking, not driving. And the properties that follow from that are different enough to be worth treating as a separate training modality rather than the back half of a rep.
You can produce more force eccentrically than concentrically — the heaviest thing you can lower under control is heavier than the heaviest thing you can lift. Most training never enters that range at all.
And you do it at a lower metabolic cost. Eccentric contraction consumes substantially less oxygen than concentric contraction for the same external force. That is not a marginal efficiency note. For a person whose limiting factor is energy availability rather than muscular capacity, it is the entire point.
The study people cite here is Chen and colleagues, working with Kazunori Nosaka, published in Medicine & Science in Sports & Exercise in 2017. It is worth describing properly, because this one gets passed through secondary sources and picks up damage on the way.
Thirty women, aged sixty and over, with obesity, split into two groups of fifteen. One group walked down stairs and took the lift up. The other walked up stairs and took the lift down. Twice a week, twelve weeks, repetitions increased gradually. That is the whole intervention. No gym, no equipment, no load beyond bodyweight.
The descending group did better on almost everything measured.
Maximal voluntary isometric contraction strength: +34% in the descending group, against +15% in the ascending group. Knee extensors.
Resting heart rate: −10% descending, −4% ascending.
Systolic blood pressure: −9% descending, −4% ascending.
Bone mineral density: +6%, in the descending group only. The ascending group showed no significant change.
LDL cholesterol: −13% descending, −7% ascending. Triglycerides, total cholesterol, glucose, insulin, HOMA and HbA1c all fell further in the descending group too, and HDL rose further. Oral glucose tolerance improved 12% against 2%.
And they worked less hard doing it. Average heart rate during the sessions was 89 bpm descending versus 114 bpm ascending — roughly 58% of maximum against 74%.
The protocol is worth stating exactly, because it is more modest than people assume. A ten-storey building, 110 steps between the first and sixth floors, walked at about one step per second — roughly two minutes per descent. Week one was two repetitions, about five minutes including lift time. By week twelve it was twenty-four repetitions, around sixty minutes, with a thirty-second rest between each. Twice a week throughout.
One note on sourcing, and one correction.
The 2017 paper itself is paywalled, and its abstract gives the strength, heart rate, blood pressure and bone figures but not the lipid percentages — only that the improvement was greater in the descending group. The 13% is nonetheless properly attributable: the same research group restated it in an open-access follow-up in Scientific Reports in 2021, where they summarise their earlier trial and give LDL as −13% against −7%. Authors citing their own numbers is about as good as secondary sourcing gets.
The correction is to the blood pressure comparison. Several popular write-ups report the ascending group's blood pressure falling by 3%. The paper says 4% — and 4% for resting heart rate in that group as well. A one-point drift, and trivial in itself, but it is exactly the mechanism by which a study becomes a slogan.
Fifteen people per group. One population — sedentary women aged 60 to 82 with body fat above 30%, recruited through district health centres in Taipei. Twelve weeks.
Allocation was quasi-randomised rather than randomised: participants were ranked by baseline knee extensor strength and assigned to groups by lottery within matched pairs, to keep starting strength comparable. That is a defensible design for a small trial and better than the 2017 abstract's phrasing suggests, but it is not the same as full randomisation.
So: a striking result, in a small sample, in a specific group, over one season. That is a reason to be interested and a reason to be careful. A 34% strength gain from walking downstairs twice a week is the sort of finding that deserves independent replication before it becomes a rule.
This is the obvious objection, and it is a good one. Walking downstairs loads the knee extensors. Why would that touch LDL, or fasting insulin, or blood pressure?
The same group went back to the frozen plasma samples and published the answer in 2021. They measured three signalling molecules before and after, and the pattern was striking: only the descending group changed.
C1q fell 51% in the descending group, and not significantly at all in the ascending group. C1q rises with age and is a proposed marker of sarcopenia. Its fall correlated tightly with the strength gain (r = −0.84) and with bone density (r = −0.93).
Apelin rose 23% — a myokine secreted by muscle and adipose tissue, involved in eNOS regulation and nitric oxide production. Its rise correlated with the fall in systolic (r = −0.85) and diastolic pressure.
Adropin more than doubled, rising 127% — a regulator of nitric oxide synthase and of fuel selection in skeletal muscle. Its change correlated with glucose tolerance, HOMA index and cholesterol.
So the answer to the objection is that contracting muscle is an endocrine organ. It secretes signalling molecules that act on distant tissue, and the eccentric group secreted them while the concentric group largely did not — despite performing the same number of contractions, on the same stairs, in the same building.
The authors are appropriately careful about it. Fifteen per group is too small for the multiple regression they would have liked, correlation is not mechanism, and they describe the study as descriptive. But it does point at something specific: the type of contraction, rather than the force or the cardiovascular load, appears to be what drives the signalling. The ascending group worked at a higher heart rate and got less.
Eccentric work produces more delayed-onset muscle soreness than any other kind. It is the standard laboratory method for inducing it deliberately.
Two things worth knowing about that.
First, soreness of this kind is part of an adaptive process rather than a sign of damage in any concerning sense. The muscle rebuilds more resistant.
Second — and this is the practically useful part — the repeated bout effect is more pronounced with eccentric work than with anything else. A single session confers substantial protection against soreness from the next one, and that protection lasts for weeks. Nosaka's group has shown the adaptation appears after remarkably little exposure, and that even low-intensity eccentric work triggers it.
Which means the first session is the worst one by a distance. People who try eccentric training once, feel wrecked for three days and never go back have quit at the single hardest point in the entire process.
So start with less volume than you think you need. Considerably less. Two sets, not five. The urge to be thorough on day one is precisely the thing that stops people returning on day four.
This one sounds like folklore and is not. Train one arm, and the untrained arm gets stronger. It has been documented since 1894.
The current best estimate comes from a 2017 meta-analysis by Manca and colleagues: thirty-one randomised trials, 785 participants, pooling to an 11.9% strength increase in the untrained limb (95% CI 9.1–14.8). Lower limb did better than upper. And the subgroup that matters here — eccentric training produced the largest cross-education effect of any contraction type, at 17.7%.
The mechanism is central rather than muscular. The untrained limb does not grow; the nervous system's capacity to drive it improves. Follow-up work by the same group found the neurophysiological picture is still unresolved, and confirmed the effect is centrally originating without being able to say exactly which adaptation carries it.
The meta-analysis authors also flag a high risk of bias across the included studies, which is worth carrying alongside the number.
A limb in a cast. A post-surgical knee that cannot be loaded. A unilateral injury with weeks of immobilisation ahead. Training the intact side is not a consolation prize — it measurably slows the loss on the immobilised side, and eccentric loading appears to be the most effective way to do it.
My wife is a clinical lead physiotherapist in neurological rehabilitation, and this principle is not news in that world. It is nearly unheard of in general fitness.
When form falls apart at the end of a run, on the last set, or on a long descent, it is usually not the engine giving out. It is the brakes.
Eccentric capacity fatigues before concentric capacity in most untrained people, because it has never been trained. The muscle can still produce force to move you forwards. It can no longer produce force to control you on the way down. That is when landing mechanics deteriorate, joints take load in positions they should not, and something eventually complains.
It shows up in ordinary places. Walking downhill is eccentric quadriceps work. So is descending stairs, and lowering yourself into a chair rather than collapsing into one. Every running stride has an eccentric absorption phase. Catching yourself when you stumble is almost entirely eccentric — whether you recover or fall is decided by whether muscle can produce force fast enough while being lengthened.
That last one is not a small point. Falls are a leading cause of lost independence in older age, and the capacity that prevents them is the capacity nobody trains.
The clinical version I see constantly: the person whose knees hurt going down stairs but not up. That is an eccentric control problem, and it usually resolves with eccentric loading rather than with rest.
Eccentric training gets oversold. Here is where I think the lines actually fall.
On injury prevention specifically, I would be misleading you if I flattened it. The best evidence is for the Nordic hamstring exercise in team sports, where a large meta-analysis pooling over eight thousand athletes found hamstring injuries roughly halved. That finding has since been challenged — a subsequent re-analysis judged the preventive effect inconclusive, largely on study quality and compliance grounds. Both of those things are true at once.
What I can add from thirty-seven years is that eccentric capacity is measurable, trainable, and almost universally neglected, and that I have watched a great many people stop breaking down once they trained it. That is clinical observation, not trial data, and I would rather label it than dress it up.
This is the application that interests me most. The lower oxygen cost means eccentric work can build strength in someone who cannot tolerate much conventional training — the person with a flattened cortisol curve, poor recovery, and a history of every exercise programme making them worse.
The Chen data supports the general shape of this: more adaptation at a lower working heart rate. Whether that translates to someone with genuine HPA axis dysregulation has not, to my knowledge, been tested directly. I use it clinically because the mechanism is sound and the demand is low, not because there is a trial in that population.
Where post-exertional malaise is present, none of the above should be read as an exercise prescription. Post-exertional malaise is not deconditioning and does not respond to graded loading; NICE withdrew its recommendation for graded exercise therapy in ME/CFS in 2021 for good reason.
The lower metabolic cost of eccentric work is a real and interesting property, and it does not exempt anyone from the pacing principle. In these presentations any movement is a careful, individual, monitored conversation — not a protocol, and not something to begin from an article.
The capacity to control a descent and catch a stumble is what keeps people independent, and it declines faster than concentric strength does. This is more directly relevant to how you will be at eighty than almost anything else you could do in a gym.
Deceleration is what fails in runners. And for tendinopathy, eccentric loading is the best-evidenced conservative intervention available.
The principle is simple: slow down the lowering. Four to five seconds, controlled throughout, no collapsing at the end.
Descending stairs. The one with the trial behind it. Lift up, walk down. Start with two or three flights and build. Costs nothing and requires no equipment.
Split squats. Five seconds down, normal speed up. Two sets of six to eight per leg. This will humble most people the first time.
Step-downs. Stand on a step, lower the other foot slowly to the floor without dropping onto it, push back up with the working leg. Directly trains the descending-stairs pattern that so often hurts.
Calf raises, both variants. Straight-leg for gastrocnemius, bent-knee at around thirty degrees to bias soleus. Rise on two legs, lower on one, five seconds down — that is how you load the eccentric phase heavier than you could lift.
Nordic hamstring curls, if someone can hold your feet. Kneel, lower forward as slowly as you can control, catch yourself with your hands. Brutal, and the single most-studied eccentric exercise there is.
Press-up negatives. Lower over five seconds, reset from the knees.
Two sessions a week is enough. The stimulus is potent and the recovery demand is real — this is not something to add on top of an already full week without taking something else out.
Every rep has two halves, and you have been training one of them.
The other one is where the brakes are. The brakes are what fail first, they cost less oxygen to train, they transfer to the limb you are not even using, and the first session is the hardest one you will ever do.
Educational content, not medical advice. If you have an existing injury, a diagnosed tendon problem, or any condition involving post-exertional symptoms, get assessed before starting eccentric training. The loading is genuinely demanding and the first session should be conservative.
If your training keeps breaking you, the question is usually what your capacity actually is — not how hard you are trying.
Ask AIdan → Book a discovery call