Negative Calorie Foods: One Small Study Put Celery 86% of the Way There
No. But celery gets closer than we expected, and closer than we said the first time we published this.
Somebody actually measured it, in people, and the number is worth your attention.
86% of the way there
Fifteen healthy young women, an overnight fast, and a ventilated hood over the head that works out calories burned from the air they breathe. Resting metabolic rate first (the calories a body burns lying still), then 3.5 oz (100 g) of celery, about a cup chopped — 16 calories — eaten in ten minutes. For the next three hours, energy use was measured for 20 minutes in every 30 [1].
Digesting that celery cost 13.76 calories. Against the 16 it delivered, that is 86% [1].
Eighty-six percent. The stalk very nearly pays for itself. Anyone who told you celery is basically free was not far off, and we would rather say that plainly than pretend the folklore had nothing in it.
Then read the last two words of the arithmetic: nearly free. Sixteen minus 13.76 leaves about 2.2 calories in your favour, on the wrong side of the ledger for the claim. The researchers’ own conclusion is that eating celery “did not induce a negative energy balance” [1].
Close is the interesting finding here. How close, this study cannot say. It is a one-page conference abstract on fifteen people, and worked out from its own figures, the range the true average probably sits in runs from about 8 to 19 calories: roughly half of what the celery delivered, up to about a fifth more than all of it [1]. So it cannot tell nearly free from break-even. What it does rule out is celery costing much more than it gives you.
Why celery does so much better than a steak
This is the part we got wrong first time round, so here it is properly.
Measured by macronutrient, digestion is cheap. Protein is the priciest at 20 to 30% of the energy it carries. Carbohydrate runs 5 to 10%. Fat is almost free, 0 to 3% [2].
So how does celery reach 86%, three times the best macronutrient figure? Not by being small. If digestion cost a fixed share of the calories, a small food would have a small cost: even if all 16 of celery’s calories were protein, the priciest to process, they should cost about five [2]. The measured cost was nearly fourteen [1]. In the published version of their paper, the lizard researchers below call that 86% “extremely large” next to other human studies, which found 8 to 12% [3]. So something that does not grow with the calories fills the gap, and the abstract does not say what. What follows is the desk’s reasoning, not a finding: it could be the work of moving 3.5 oz of water and fibre through the gut, or it could be the measurement itself, because the abstract describes no comparison session without celery, so anything else that nudged energy use up over those three hours would have counted as digestion. Until someone tests it, the gap is unexplained.
Which is exactly why quoting the macronutrient ranges as a ceiling — as we did — is the wrong move. Those describe what happens to nutrients you absorb. A stalk of water and string is a different question.
Is popcorn a negative calorie food?
No, and it is not close in the way celery is close.
Popcorn is a whole grain heated until it explodes: it is mostly starch, and starch is a carbohydrate you absorb efficiently at a digestion cost of 5 to 10% [2]. Air-popped popcorn is genuinely low in calories per cup because it is mostly air — but the calories that are there, you keep.
Low-density and negative-calorie keep getting swapped for each other, and they are not the same idea.
Is any food actually zero, or below?
Not one that has been shown to be, once you count the right thing. The best documented case in a human laboratory is the celery above, which averaged 86% [1].
There are two other tests, and the first is stranger than you would guess. Researchers fed nine bearded dragons a celery meal weighing 5% of their body weight, tracked their metabolism for five days, and burned samples of the celery, the droppings and the urate (the paste reptiles pass instead of urine) to measure the energy in each. Digesting the meal cost 33% of its energy, the droppings carried off 29% and the urate 14%, leaving the lizards a net gain of 24% [3]. The researchers add that the true gain is probably higher, because not all of the droppings came from the celery.
Lizards are not people. These ate a meal that was 5% of their body weight, where the women in the celery study ate 0.17% of theirs, and at a lower body temperature they took days rather than hours to digest it [3]. But followed all the way to what the body keeps, celery still came out ahead. On paper, the same researchers work out that the celery study’s average would slip just below zero, by less than one calorie, if you also subtracted the fibre that passes straight through and the cost of chewing; they build that on the 86% they call extremely large [3]. A calculation is not a measurement.
The second test said yes, and it made the papers. In June 2016 the Daily Mail ran “Eating celery really DOES burn more calories than it contains” over a test for Channel 4’s Food Unwrapped: one man, twelve hours in a calorie chamber, celery worth 53 calories, and 72 calories burned while he ate it (112 when he drank it as a smoothie). By the paper’s own account, the chamber counted every calorie he burned, and that includes the ones a body burns just keeping itself alive, spent with or without celery. In the 2019 version of their paper, the lizard researchers put the clash between that result and the celery study down to exactly this: one counted resting metabolism and the other subtracted it [3].
The myth is older than the measurement
So where did “more calories than it contains” come from? Not from the lab. The celery study’s own abstract calls the idea “a renowned health myth” [1]. The researchers were testing it, not starting it.
It was in print at least twenty years before the measurement. A 1992 diet book was titled Foods That Cause You to Lose Weight: The Negative Calorie Effect, and The Negative Calorie Diet followed in 2015, subtitled “Lose Up to 10 Pounds in 10 Days.” Who first used the phrase, we could not trace. What we can place is its most quotable modern form: the 2016 headline above, which said it outright.
Then the format takes over. “Ten negative calorie foods” is a list you can screenshot. “Diet-induced thermogenesis was 86% of energy intake in fifteen women” is not. The list travels; the measurement stays home.
When we searched the phrase on 26 July 2026, Google’s own AI summary assembled its answer from a hospital site, Healthline, a fitness app’s blog, the Tufts nutrition letter, a research-summary tool — and four YouTube videos, one of them a daytime-television segment. The Tufts newsletter files it under “Myth of the Month.” It was sharing the citation list with daytime TV.
What to actually do
Eat the celery. On that study’s average it nets you about two calories a cup, chopped, and less than one a medium stalk, which is as close to free as food gets, and the water and fibre do real work on fullness. The same goes for most of the list — cucumber, lettuce, zucchini. They are excellent, and they are excellent for the boring reason.
Just do not build a plan on the idea that they run a deficit for you. No food has been shown to. The best-measured case in the literature kept an average of 14% of what it gave, and even if the true average sat at the far end of that study’s range, a cup of celery would come out about three and a half calories behind. That is not a diet.
Nearly free is a genuinely good deal. It is just not the deal that was advertised.
This page was corrected on 16 September 2026 after an independent editorial review. We said the negative-calorie claim began when celery’s measured 86% was rounded up to “more than.” That had the history backwards: the celery study’s own abstract calls the idea “a renowned health myth,” and a diet book was subtitled The Negative Calorie Effect in 1992, twenty years before the measurement. The section and the provenance chain now tell it in that order, and say we could not trace the phrase’s first use. We also wrote that celery is “14% short of free, not negative” as if 86% were settled. It comes from a one-page conference abstract on fifteen women, and worked out from the abstract’s own standard error, its measure of how precisely fifteen people pin down an average, the range the true average probably sits in runs from about half of celery’s calories to about a fifth more than all of them. The body, the evidence panel, the headline and the search description now say the study cannot tell nearly free from break-even.
Also changed: our explanation for the high percentage, that celery’s calories are few, did not hold, because a cost that kept pace with the calories would have stayed small; the page now calls the gap unexplained and labels our reasoning about it. A 2016 one-man chamber test reported as proof that celery burns more than it contains is now described, with why it does not count: its total included the calories he would have burned at rest anyway. The lizard study we said we could not read opened on the first try on 15 September 2026, and its figures, funding and caveats are now printed. The funding row now gives each source’s own statement; “two net calories a stalk” was the figure for 100 g, about a cup chopped, so a medium stalk nets less than one; and the Google AI summary we describe is now dated to 26 July 2026, when we saw it. The 26 July correction stands. The rating is unchanged: Unsupported.
An earlier version of this article, published the same day, argued that no food can cost more than about 30% of its energy to digest, citing the thermic effect of protein, carbohydrate and fat. That reasoning was wrong. Those ranges describe absorbed macronutrients, not whole foods, and celery is mostly water and indigestible fibre. Clegg and Cooper measured celery's diet-induced thermogenesis at 86% of its energy — nearly three times any macronutrient figure. The verdict is unchanged, because 86% is not 100% and the researchers concluded there was no negative energy balance. But the claim is much closer to true than we originally said, and the article has been rewritten around the measurement instead of around our arithmetic.
