Adding exercise buys about 30 percent of what the arithmetic promises#
The additive model of human energy expenditure is the one everybody carries around: your body has a baseline cost, and exercise piles calories on top of it. The constrained model says something else — that raising physical activity causes the body to spend less on other things, so total daily expenditure rises far less than the addition predicts. Pooling human aerobic exercise interventions, one recent review puts a number on the gap: total daily energy expenditure increased by only about 30 percent of the change an additive model expects1. Burn 300 calories on a bike and the day ends up perhaps 100 calories bigger than it would have been.
That is the finding underneath why exercise burns fewer calories than you think, and this article is about the model rather than the consequence — because the model is genuinely contested, and the contest has produced something more useful than either side's headline. The short version of what follows: a ceiling on human energy expenditure certainly exists, it sits much higher than most training ever reaches, the compensation people actually experience is more behavioral than metabolic, and it gets substantially stronger when you are dieting at the same time. That last clause is the one that matters if you are reading this while in a deficit.
What the two models actually predict#
State the disagreement precisely, because "exercise doesn't burn much" is a conclusion both camps reach for different reasons.
| Additive model | Constrained model | |
|---|---|---|
| Effect of +300 kcal of exercise | Day is +300 kcal | Day is less than +300 kcal; the rest is reallocated |
| Where the difference goes | Nowhere — nothing is reallocated | Basal metabolism, immune and reproductive function, other movement |
| Prediction for very active groups | Much higher total expenditure | Total expenditure similar to sedentary groups |
| Prediction across activity range | Straight line | A curve that flattens |
The strongest observational support is the flattening: measured across five populations, total expenditure rose with activity and then bent (Pontzer et al., 2016, handled in full by the pillar). Animals compensate almost completely, at around 100 percent, and reductions in basal and sleeping metabolic rate account for part of the human effect but not all of it1 — an admission worth noting: the mechanism has a hole in it.
The critics found a hole in the statistics, not just the story#
The most substantive objection is not "exercise obviously works." It is technical, and once you see it you cannot unsee it.
A 2023 perspective in Advances in Nutrition laid out four problems with the evidence base. Much of it is cross-sectional, comparing populations that differ in body size, composition and lifestyle at once. Slopes below 1 — the standard criterion for declaring compensation — are also produced by regression dilution, meaning ordinary measurement error attenuates the slope whether or not any biology is happening. The compensatory reduction has generally been inferred rather than observed. And the sharpest point: activity energy expenditure is usually computed as total minus basal expenditure, so a negative correlation between the two "can be obtained simply because BEE is 1 of the variables in the correlation but is also a negative term in the calculation of the other variable"2. The trade-off can appear in data generated by no trade-off at all.
Their verdict is deliberately unexciting: insufficient evidence to fully support either model, and "an upper limit of TEE probably exists, but this is likely irrelevant for most people." One disclosure belongs on the page, applied as this blog applies it in the other direction: two of those authors report consultancy and research funding from food and beverage companies including PepsiCo, Unilever and Sugar Nutrition UK, and "exercise really does add calories" has long been a beverage-industry-friendly conclusion. The Pontzer and Trexler review declares no competing interests. Neither fact settles anything; both belong in view.
The trial that measured compensation directly — and found it in the fridge#
If compensation is metabolic, a controlled trial with good instruments should catch a metabolism slowing down. One went looking.
E-MECHANIC randomized 198 adults with overweight or obesity to no exercise, 8 kcal/kg/week of supervised exercise, or 20 kcal/kg/week for 24 weeks, measuring intake by doubly labeled water, activity by accelerometer and resting metabolic rate by indirect calorimetry. Compensation was real and dose-dependent: participants ended up 1.5 kg (95% CI 0.9–2.2) and 2.7 kg (95% CI 2.0–3.5) heavier than the exercise dose predicted. But the mechanism was not the one the constrained model nominates. Intake rose 90.7 kcal/day (95% CI 35.1–146.4) and 123.6 kcal/day (95% CI 64.5–182.7) in the two exercise groups against essentially zero in controls, while resting metabolic rate and non-exercise physical activity did not differentially change between the three groups. The authors' conclusion is blunt: "Compensation resulted from increased energy intake and concomitant increases in appetite... Compensation was not due to activity or metabolic changes"3.
That is a serious result. It does not disprove metabolic constraint — six months is not a lifetime, and a slowing basal rate is a small signal to detect — but it means the compensation most exercisers meet runs through appetite, which is eating back the calories you burn, not a thermostat quietly turning down. Non-exercise activity, the other usual suspect, did not fall either.
What separates the two camps: whether you're dieting#
Here the disagreement resolves into something you can act on, and it resolves the same way in two independent places.
A doubly-labeled-water study of 584 older US adults aged 50 to 74 asked which model fit, then split the sample by energy balance. Across the whole group the activity-expenditure association was essentially a straight line (P < 0.0001 for linearity; P = 0.920 for curvature), running from 2,354 ± 351 kcal/day in the least active decile to 2,693 ± 480 in the most active — additive. Among participants in negative energy balance, total expenditure was flat across deciles: 2,428 ± 285 kcal/day at the bottom, 2,372 ± 560 at the top. The conclusion: additive in positive energy balance, constrained in negative4.
The review from the other camp independently reports the same moderator — compensation is "reduced with resistance training and amplified when aerobic exercise is paired with diet restriction"1. Two teams with opposite priors, converging on the condition that switches the model.
Read against your own situation, it is unwelcome news. Well-fed and adding training, the calories mostly count. Running a deficit and adding training to accelerate it is exactly the condition under which the body reclaims the most — so dieting harder and training harder is the one combination the evidence says compensates worst. The deficit still works; it just doesn't stack the way the spreadsheet suggests, which is one more reason to build it from intake rather than workouts.
The ceiling is real, and it's nowhere near you#
One piece of evidence for constraint comes from a design with none of the correlational problems: watching what happens when humans push expenditure as hard as biology permits.
Researchers compiled total expenditure and basal metabolic rate across endurance events lasting from half a day to over 250 days, adding new measurements from six athletes (five men, one woman, aged 32 to 73) running a transcontinental race — 4,957 km in 20 weeks, roughly a marathon a day, six days a week. Sustained metabolic scope fell curvilinearly with event duration and plateaued below 3× basal metabolic rate; in the race itself it started at 3.76× BMR in week 1 and fell to 2.81× by the final week. Adding overfeeding studies, the authors put the alimentary limit — the ceiling set by how much energy the gut can absorb — at about 2.5× BMR, beyond which expenditure must come out of body stores5.
So a hard constraint exists, and its cause is digestive rather than mysterious. It also sits at roughly two and a half times basal metabolism — for most adults, north of 4,000 calories a day sustained indefinitely. Both camps agree on this; the argument was never about the Tour de France.
What to take from an unsettled model#
- Assume you keep a fraction, not the whole — about 30 percent of the additive prediction, in the pooled intervention data.
- Expect the leak to be appetite first. The best controlled trial found compensation entirely in intake, not in a slowed metabolism or reduced daily movement — a leak you can see, unlike one you can't.
- Don't stack aggressive training on an aggressive deficit. Both camps agree negative energy balance is when compensation is strongest.
- Ignore the ceiling. Real at around 2.5× basal metabolic rate, irrelevant to anyone not running across a continent.
The constrained model is neither the settled law it is quoted as nor the overreach its critics were right to interrogate. What survives both is narrower and more useful: energy is reallocated rather than simply added, the reallocation is bigger when you are eating less, and the largest leak sits between the workout and the fridge, where you can see it — which is why the argument in can you outrun a bad diet holds whichever model eventually wins.
FAQ#
What is the constrained energy expenditure model?#
It proposes that the body holds total daily energy expenditure within a limited range, so more physical activity causes reductions elsewhere — basal metabolism, immune and reproductive function, other movement — rather than adding to the day's total. Its rival, the additive model, says exercise calories stack on top of baseline. Pooled human interventions land between: total expenditure rose by roughly 30 percent of the additive prediction.
Is there an upper limit to how many calories a human can burn?#
Yes, and it has been measured. Across endurance events from half a day to over 250 days, sustained metabolic scope declines with duration and plateaus below 3× basal metabolic rate, with an absorption-based ceiling near 2.5× BMR. Athletes running a marathon a day across a continent fell from 3.76× BMR in week one to 2.81× by week twenty — and for almost everyone that limit is far above anything they will approach.
Why does exercise seem to compensate more when you're dieting?#
Because energy balance appears to be the moderator that separates the two models. In 584 older adults, the activity-expenditure relationship was additive in people in positive energy balance but flat — constrained — in those in negative energy balance. A review from the opposing research camp independently reports that compensation is amplified when aerobic exercise is combined with dietary restriction. Adding hard training to an aggressive deficit is the condition where the least of it survives.
Sources#
- Pontzer H, Trexler ET. The evidence for constrained total energy expenditure in humans and other animals. Curr Biol. 2026;36(4):1013-1025.e4.
- Gonzalez JT, Batterham AM, Atkinson G, Thompson D. Perspective: Is the response of human energy expenditure to increased physical activity additive or constrained? Adv Nutr. 2023;14(3):406-419.
- Martin CK, Johnson WD, Myers CA, et al. Effect of different doses of supervised exercise on food intake, metabolism, and non-exercise physical activity: the E-MECHANIC randomized controlled trial. Am J Clin Nutr. 2019;110(3):583-592.
- Willis EA, Creasy SA, Saint-Maurice PF, et al. Physical activity and total daily energy expenditure in older US adults: constrained versus additive models. Med Sci Sports Exerc. 2022;54(1):98-105.
- Thurber C, Dugas LR, Ocobock C, Carlson B, Speakman JR, Pontzer H. Extreme events reveal an alimentary limit on sustained maximal human energy expenditure. Sci Adv. 2019;5(6):eaaw0341.
- Pontzer H, Durazo-Arvizu R, Dugas LR, et al. Constrained total energy expenditure and metabolic adaptation to physical activity in adult humans. Curr Biol. 2016;26(3):410-417.
