The most filling foods, calorie for calorie, are the ones that weigh the most per calorie: foods full of water and fibre, with some protein and little fat. In the best-known test, boiled potatoes kept people fuller than any of 38 common foods, about three times more than white bread at the same calories1.
That test is from 1995 and ran for two hours. Below is how it worked, what later trials added, and where the ranking stops being useful.
The satiety index: how it was measured
Satiety is how full and satisfied you feel after eating, and how long until the next urge to eat. A satiety index puts a number on it for single foods.
In 1995, researchers in Sydney served 38 foods from six groups: fruits, bakery products, snack foods, carbohydrate-rich foods, protein-rich foods and breakfast cereals1. Every portion held the same energy: 1,000 kJ, or 240 kcal. So a portion of potatoes was much heavier than a portion of croissant.
Groups of 11 to 13 volunteers ate each food. They rated their fullness every 15 minutes for two hours. Then they could eat freely from a spread of foods and drinks, and what they ate was recorded1.
Each food's score was its fullness over two hours divided by the fullness for white bread, times 100. So white bread scores 100, and a food scoring 200 kept people twice as full.
| What | Detail |
|---|---|
| Foods tested | 38, in six groups |
| Portion | 240 kcal (1,000 kJ) each |
| People per food | 11 to 13 volunteers |
| Fullness ratings | every 15 minutes for 2 hours |
| Reference | white bread = 100 |
In plain terms: equal calories, different foods, and a simple question every quarter hour: how full are you?
What the ranking showed
The spread was large. Boiled potatoes scored highest at 323, and the croissant lowest at 47, a sevenfold range1. Most foods, 76% of them, scored at or above white bread.
Satiety index at 240 kcal, white bread = 100
The published abstract names only these three scores. Longer lists of the index circulate online; we have not been able to check them against the paper, so we do not repeat them here.
The ratings matched behaviour, loosely. Higher fullness ratings went with less eating from the spread that followed1. The link was modest (a correlation of −0.37), but it points the same way as the ratings.
What the filling foods had in common
The authors checked which features of a food went with a high score1. A correlation, written r, runs from −1 to 1. Zero means no link; closer to 1 or −1 means a tighter link.
How food features correlated with the satiety index · r
- Heavier portions scored higher (r = 0.66). At 240 kcal, a heavier portion means fewer calories per gram.
- More water (0.64), more fibre (0.46) and [more protein](/articles/does-protein-keep-you-full/) (0.37) went with higher scores.
- More fat went with lower scores (−0.43).
- Tastier foods, rated by the eaters, scored lower (−0.64).
These are links across 38 foods, not tests of one feature at a time. A food rich in water is also usually low in fat, so the features overlap.
In plain terms: the foods that filled people up gave them the most food for the calories.
What later trials added
The idea behind this is energy density: calories per gram of food. Water and fibre add weight without adding energy; fat adds a lot of energy in little weight.
A first course that is light per gram. In a 2004 trial, 42 women ate lunch in a lab once a week for seven weeks2. Before a pasta main course they ate one of six salads, or no salad. The salads varied in calories per gram (0.33, 0.67 or 1.33 kcal/g, changed through dressing and cheese) and in size (150 or 300 g).
Compared with no first course, the light salads cut total lunch intake by 7% (small portion) and 12% (large)2. The rich salads raised it by 8% and 17%.
Change in total lunch intake vs no first course · %
The pooled picture is mixed. A 2017 meta-analysis (a study that pools many trials) of 39 studies found that, overall, energy density did not change intake at the next meal3. Split by design, it did: when two first courses had the same calories, the lighter-per-gram one was followed by less eating. When they had the same weight, the lighter one was followed by more.
Fruit. A 2019 systematic review found that in single-meal trials, whole fresh fruit tended to lower energy intake, especially eaten before a meal or in place of denser foods4. The author graded that certainty as moderate.
In plain terms: more food for the same calories tends to fill people up, but it is the calories per gram that matter, not the food's name.
What this means in a real diet
These nutrition trials describe foods, not rules. The meals that led to less eating in these studies were heavy on water and fibre and light on fat, and they came first or replaced denser food.
That fits the energy-balance picture in calories in, calories out. Fibre shows up in both the index and the fruit review; see how much fiber per day for the reference values. If you track food, calories per gram is simple to read from any label: divide calories by grams.
What the studies found
| Study | Who was studied | Key result | Year |
|---|---|---|---|
| Holt et al., *Eur J Clin Nutr*1 | Groups of 11–13 volunteers per food; 38 foods at 240 kcal (abstract) | Boiled potatoes 323, white bread 100, croissant 47; weight, water, fibre and protein went with fullness | 1995 |
| Rolls, Roe & Meengs, *J Am Diet Assoc*2 | 42 women; seven weekly lab lunches (abstract) | Light first-course salad: lunch intake −7% to −12%; rich salad: +8% to +17% | 2004 |
| Rouhani et al., *Eat Behav*3 | 39 trials pooled (abstract) | No overall effect; lighter first course at equal calories was followed by less eating | 2017 |
| Guyenet, *Front Nutr*4 | Randomised trials and cohorts on whole fruit (abstract) | Fruit tended to lower meal intake; moderate certainty | 2019 |
| Lasschuijt et al., *J Nutr*5 | 69 healthy adults, average age 25; five lab lunches (abstract) | Slow, light sandwich lunch: about 570 kcal eaten vs about 1,143 for fast, dense | 2025 |
Limits worth knowing
- Two hours only. The satiety index measured fullness for two hours and the next snack, not a day, a week or body weight1.
- Small groups. Each food was tested in 11 to 13 people. The salad study had 42 women; the 2025 trial 69 young adults.
- Abstract only for every paper here. The 1995 paper is behind a paywall. We quote only the three scores its abstract gives.
- Links, not causes. The features that went with fullness were correlations across foods, and they overlap.
- Lab meals. Portions, timing and food choice were set by researchers, which is not how people usually eat.
Latest research
The newest trial we opened is from 20255. Sixty-nine healthy adults, average age 25, came to five lab lunches of sandwiches they could eat freely. The sandwiches were either light (1.9 kcal/g) or dense (3.8 kcal/g), and made to be eaten either slowly or quickly.
The slow, light lunch led to an average intake of about 570 kcal5. The fast, dense lunch led to about 1,143 kcal, twice as much. The two features interacted, so their combined effect was not a simple sum of each one alone. It supports the satiety index's central pattern, calories per gram, with a modern controlled design, though still over single meals.
The short version
- Calorie for calorie, the most filling foods in tests are heavy, watery and fibre-rich, with some protein and little fat.
- In the 1995 index, boiled potatoes scored about three times white bread; the croissant under half.
- Starting a meal with something light per gram cut total intake in lab trials.
- The evidence is short-term: hours, not months.
Educational overview only — not medical advice.
