How Hard Should It Feel? The Case for the Middle Setting
There are two pieces of advice in circulation and they contradict each other. One says push harder, the session only counts if it hurts. The other says be gentle with yourself, anything is better than nothing.
Both are half right, and a small randomised trial published in August 2026 is one of the few to test them against each other in the same design.
This is the intensity question. The volume question is in how much exercise you need, and the separate matter of matching effort to the day in front of you is covered in training with your energy. Whether a session filmed for a general audience can calibrate to you at all is a related problem, taken up in expert-designed versus random YouTube workouts. Whether the hard part belongs in its own session or inside a mixed one is covered in do you need separate cardio and gym days.
Testing both errors at once
Thirty healthy older adults were randomly assigned to dynamic balance training at one of three difficulty settings. Optimal, where task demand was matched to each person’s measured ability. Or one of two suboptimal settings: too hard, or too easy.
That design is rarer than it sounds. Both comparison arms were still training, so the question was never whether tailoring beats doing nothing, but whether getting the difficulty right beats getting it wrong in either direction.
The matched group showed superior performance gains in three of six sessions, with the advantage clearest under high task demands, and greater improvement on closely related movement tasks both during and after the intervention. The benefit did not carry over to the thinking tests, which the authors say plainly (Experimental Gerontology, PMID 42562172, August 2026).
The limits: about ten people per arm, healthy older adults, and a balance task rather than resistance or cardio work. Three of six sessions is a partial result rather than a clean sweep. This is direction, not proof.
But the direction is useful, because it says something most gentle-fitness content avoids: if you want to get better at a movement, too easy is a real error, not a safe one.
But the claim only goes so far. This trial measured whether people improved at a task, not whether they were healthier for doing it. The health floor is a different question, and it sits much lower: in the 87,340-person Norwegian cohort covered in how much exercise you need, the lowest activity tier explicitly included taking it easy without breaking into a sweat, and it still tracked with 9% lower stroke risk and 11% lower mortality. Easy effort counts for your health. What it does not buy is getting better at the movement.
Harder is not the lever either
The other half of the advice fails for a different reason.
Researchers trained both legs of 25 resistance-trained adults twice weekly for eight weeks. One leg got 120% more volume than habitual, the other 20% more. Each person was their own control. Muscle cross-sectional area rose in both legs, with no difference between them. Under the microscope there was no meaningful change in either leg, and the growth signals the researchers tracked looked much the same on both sides (J Appl Physiol, PMID 42461790, July 2026).
Eight weeks, already-trained adults aged 18 to 35, and a null is not proof of equivalence. Still, the participants who did more than five times the extra work got nothing extra for it.
Put the balance trial and the two-legs trial together and the shape emerges. Under-loading costs you progress. Over-loading does not buy you more of it. The useful range for getting better at something is wider than the “no pain no gain” school implies, and it has a lower edge as well as an upper one. The lower edge is the half gentle-fitness writing tends to skip.
How the middle feels
Nothing above gives you a number, and a number is what people want. For the most part the research points away from one, and the single exception is narrow.
Nineteen already-active adults walked on a treadmill at about 3 km/h in four 15-minute conditions: a metronome set to their own preferred stride frequency, the same metronome 10% faster, 10% slower, and one condition with no metronome at all. Free walking, with no cue at all, cost about 4% less oxygen and was rated slightly easier and slightly more pleasant than being paced, and that held even when the metronome was set to the walker’s own cadence. Pacing 10% either side of their own stride frequency, about four or five strides a minute, was no worse than pacing at exactly the right number, so it is the external cue rather than the wrong number that carries the cost (PLoS One, PMID 42555564, August 2026). At n=19 and a single session, that is a small acute study. The physiological half was already well established; the affective half is the newer part.
Which suggests a rule of thumb rather than a target: cue the intent, not the count. “Lower under control” travels better than a ticking metronome, because a prescribed rhythm can degrade how a session feels even when the prescribed number is the right one.
The one tempo cue with a trial behind it is worth knowing, because it pulls against that. Thirty sports-science students trained squats for eight weeks at different lowering speeds. The two-second lowering group beat both the four-second group and a self-selected group on jump and sprint. On maximal strength all three groups improved by a similar amount with no significant difference between them, though the two-second group showed the largest effect size. The four-second group declined on jump and sprint (PLoS One, PMID 42507730, July 2026). The two trials are not scoring the same thing. The treadmill study measured oxygen cost, perceived exertion and how the walking felt, and there your own rhythm won. The squat study measured maximal strength, jump and sprint, and there a two-second lowering won, with the self-selected group second on the power measures. Winning on how the work feels does not mean winning on power, and the reverse holds too. At roughly ten per group in trained students with power-biased outcomes, the direction is more trustworthy than the magnitudes, and “slow and controlled” still has a form and safety rationale for a true beginner that this trial did not test. For a beginner the feel result decides more, because the session that felt bad is the one you do not come back to.
A usable version
If you want something to carry into your next session:
A set should end because the quality dropped, not because a number arrived. That is the closest thing to a calibrated difficulty you can self-assess.
If every session finishes with plenty left, that is information. Not necessarily a signal to add weight. Range, control or stability are all variables you can move first. It is not a signal to add sessions either, as how much strength training a week is enough sets out.
If a session leaves you unable to do the next one, the setting was wrong. In the two-legs study the much bigger workload added no extra muscle that the scan could pick up. But both legs belonged to the same person, so the study could not show what the extra sets cost in recovery.
Structure the hard part as short efforts. Nineteen inactive women, average age 53, did three treadmill interval sessions of differing length and intensity. The four four-minute blocks produced the highest heart rate and perceived exertion and the lowest enjoyment. Ten one-minute efforts felt harder than eight fifteen-second efforts but were enjoyed just as much, which is why the authors landed on the one-minute format as the best trade between work done and willingness to repeat it (Applied Physiology, Nutrition, and Metabolism, PMID 42561455, August 2026). Small and acute, so a default rather than a law.
ZeroFit asks how you’re doing and how long you’ve got, then sets the dose from the answers. One less thing to judge at the start. Get beta access.
ZeroFit content is educational and not medical advice. If you have an injury, a health condition, or you’re pregnant, check with a qualified professional before starting any new movement routine.