← Longevity & Supplement Guides
How fast you eat is a portion size
Every nutrition argument of the past decade has been about what is in the food. The label obliges: grams of sugar, grams of fat, a percentage of something you are meant to want less of. Nobody prints the variable a Dutch laboratory has just spent four weeks demonstrating, because it is not in the food at all. It is in how long the food makes you sit there.
They screened volunteers by watching them eat a carrot
Before the trial began, forty-one healthy young adults had their natural eating rate measured with a carrot. Chew, swallow, divide grams by minutes. Anyone slower than 5 g/min or faster than 35 g/min was excluded.
It is a faintly comic image — a screening room with a bowl of carrots — and it tells you precisely what the researchers thought they were studying. Not appetite. Not willpower. Not whether anyone knew which foods were virtuous. Mechanics.
Two ultra-processed diets, one difference
The trial, led by Ciarán Forde and Marlou Lasschuijt and published in the American Journal of Clinical Nutrition in 2026, was a single-blind, block-randomised crossover: two fourteen-day diets, a two-week washout between them, everything served ad libitum. Twenty-one of the forty-one participants were men, the mean age was 27, and the mean BMI was 23.4.
Both diets were built almost entirely from ultra-processed foods. They were matched for palatability, portion size served, total energy served, non-beverage energy density and meal variety. One difference survived that matching: texture. One menu was assembled from foods known to be eaten slowly, the other from foods that go down fast.
Daily energy intake was 369 kcal lower on the slow-texture diet (95% CI: 221, 517). Across the fortnight that is more than five thousand calories, from a diet nobody would call healthier.
The 2026 crossover trial
Forde et al., Am J Clin Nutr 2026 — n = 41, two 14-day ad libitum ultra-processed diets
Palatability, portion size served, total energy served, non-beverage energy density, meal variety. Liking and familiarity ratings came out the same.
Food texture only — one menu chosen to be eaten slowly, one to be eaten fast. Measured eating rate differed by 43%.
Daily intake 369 kcal lower on the slow menu (95% CI 221–517), a 14% difference, sustained across all 14 days. Body fat mass 0.43 kg lower.
Body weight, on either diet. And appetite ratings — hunger, fullness, desire to eat — before and after meals.
How they knew people actually ate slower
Weekday meals were video-recorded on a concealed webcam and an action camera, then annotated frame by frame by trained coders using ELAN behavioural-coding software. Eating rate was grams consumed divided by the minutes between first bite and last swallow. Weekend meals eaten at home were estimated from the timestamps on appetite questionnaires participants filled in before and after eating, and all packaging and leftovers were returned and weighed.
Numbers as reported in the trial. Nothing here is a bar: the four panels are categories, not quantities on a shared scale.
Nobody was told to slow down
That is the part of the design that matters most, and it is easy to read past. Participants were never instructed to chew more, put the fork down between bites, or count to twenty. They ate the way they normally eat. The food did the pacing.
A 43% difference in eating rate produced a 14% difference in daily calories — in line with the rough relationship the authors cite from earlier work, where a 20% change in rate moves intake by about 10%. It held for the full fortnight. Nobody adapted, and nobody made the calories back elsewhere in the day.
And they did not notice. Ratings of hunger, fullness and desire to eat, collected before and after every meal, did not differ meaningfully between the two diets. People ate 369 fewer calories a day and reported feeling the same about it. That is the most interesting sentence in the paper and also the least comfortable, because it means the sensation you are trusting to tell you how much you ate is not doing that job with any precision.
Why “ultra-processed” keeps under-explaining itself
In 2019, Kevin Hall's group ran twenty inpatients at the NIH through two weeks of ultra-processed and two weeks of unprocessed food, matched for presented calories, energy density, macronutrients, sugar, sodium and fibre. People ate 508 kcal a day more on the ultra-processed arm and gained about 0.9 kg, having lost about the same on the other. It is the most-cited experiment in the field and it changed how governments talk about food.
What it did not do was name a mechanism. Forde and colleagues went looking for one in 2020, pooling 327 foods from five studies and multiplying energy density by eating rate to get a single number: calories per minute. It climbs steeply with processing. In Hall's own trial, the ultra-processed arm was eaten at 48 kcal/min against 31 for the unprocessed one.
Energy intake rate
How many calories a food lets you take on board per minute
Forde 2020, 327 foods pooled from five studies Hall 2019 inpatient trial
Bar widths are proportional to the published kcal/min values, on one shared axis. The two blue bars come from a different study than the three above them and are shown for comparison, not pooled with them.
Then the argument got sharper. In 2022, Teo and Forde gave fifty healthy-weight adults four ad libitum lunches: soft and hard versions of both minimally processed and ultra-processed meals. The hard meals were eaten more slowly and produced roughly 21% less food by weight and 26% fewer calories — in both processing categories. Hardness cut intake whether the meal came from a factory or a field.
Which reframes the whole quarrel. The 2026 authors say it themselves, and it is a striking admission from inside a paper about ultra-processed food: treating NOVA group 4 as one category is not coherent, because the sensory and nutritional range inside that bin is enormous. Our own guide to ultra-processed food holds up, but this is the amendment — a good part of what the classification detects may be softness wearing a lab coat.
Eating rate is a portion lever, not a satiety lever
Here is the limit that keeps this useful instead of merely exciting. In 2023, Wallace and colleagues gave twenty-four people fixed-portion breakfasts — harder or softer, with or without instructions to chew slowly. Combining texture and instruction slowed eating the most, exactly as predicted. But there was no difference in post-meal satiety, and no difference in what anyone ate for the rest of the day.
Read the two findings together and the mechanism comes into focus. Eating rate changes how much of the meal in front of you you get through, when the amount is up to you. It does not appear to buy a longer-lasting fullness or a smaller dinner. Your mouth is metering, roughly, and a fast bite gets counted carelessly.
That distinction matters, because the version of this advice everyone repeats — eat slowly and you will feel full sooner — is the version the fixed-portion trial looked for and did not find.
What these trials cannot tell you
Whether it works for everyone. In 2014, Shah and colleagues fed thirty-five normal-weight and thirty-five overweight or obese adults the same lunch, fast and slow. The normal-weight group ate less when slow (804 vs 893 kcal, P = 0.04). In the overweight and obese group the gap did not reach significance (667 vs 725, P = 0.18). One meal, small groups — but a caution against assuming the effect transfers most strongly to the people most often told to slow down.
Whether it changes anything you can weigh. Forty-one lean young adults, fourteen days. Body weight did not move on either diet; body fat mass was 0.43 kg lower on the slow arm. The authors attribute the flat weight to closely matched energy density and a short trial, and say plainly that whether this approach sustains a change in body weight is not yet answered.
Who funded it. The trial sits inside a Dutch public–private research programme on food texture and eating rate, and the senior authors declare extensive food-industry relationships. All of it is disclosed. It does not undo a crossover design with video-coded eating behaviour, but a texture consortium concluding that texture is the lever is a result to read with your eyes open.
Which way the arrow points in the population data. Pooling twenty-three studies in 2015, Ohkuma and colleagues found self-described fast eaters carried 1.78 kg/m² more BMI and about twice the odds of obesity. Cross-sectional and self-reported: eating quickly may partly be a consequence of a larger appetite rather than its cause. It is the same problem that haunts every number the history of the calorie left us with.
Buy the resistance instead of remembering the rule
The practical lesson is not mindfulness. Mindful eating is an instruction you have to reissue at every single meal, and the trials that lean on instruction show the weaker, more fragile effect. Texture is one decision, made at the shop, that then works unsupervised for a fortnight while you think about other things.
So the useful move is to make sure something on the plate fights back:
- Whole fruit rather than the smoothie version of it.
- Nuts, and vegetables raw or cooked only far enough to still crunch.
- Coarse, dense bread instead of soft.
- Meat you have to cut, rather than something already minced.
- Beans and lentils — slow to eat, and carrying the second benefit covered in our fibre and gut microbiome guide.
And one measurement worth doing exactly once: time a meal, first bite to last swallow. Most people have never looked. The screening range in the trial — 5 to 35 g/min — is wide enough that knowing what is normal tells you nothing about you. You are not chasing a target. You are finding out whether dinner takes eight minutes or twenty-five, because if it takes eight, texture has far more room to work than resolve does. It belongs in the same category as the other cheap, unglamorous readings in building a longevity protocol: not impressive, just informative.
The bottom line
Eating rate is real, it held for two weeks, and it works without your attention — which is exactly why it is worth arranging in advance rather than remembering in the moment. It shrinks the meal in front of you; it does not appear to shrink tomorrow's appetite. It has been shown in lean young adults over a fortnight and not yet in anybody's long-term body weight, and the trial that showed it best was funded by people with an interest in the answer.
But the result stands, and it is a strange one. The same ultra-processed diet, re-textured, was worth 369 calories a day. The food category never changed. The chewing did.
