What Is a Calorie? What the Number on a Food Label Actually Measures

A food calorie is one kilocalorie (kcal), 1,000 of the small calories that physics uses. That is the heat that warms 1 kg of water by 1 degree Celsius, and it equals 4.184 kJ. The number on a pack isn’t a heat reading. It estimates the energy your body can take from one serving. UK and EU law has makers calculate it from the grams of protein, carbohydrate and fat inside.

What is a calorie, exactly?

Calories measure energy and nothing else. They don’t say how filling a food is, and they don’t grade it.

The small calorie of physics is the heat that warms 1 g of water by 1 degree Celsius, and NIST sets it at 4.184 joules. Food uses a unit 1,000 times larger, written Calorie with a capital C or kcal. It warms 1 kg of water by 1 degree. Do the sum and 100 kcal takes a liter of water from freezing to boiling, near enough.

Officially, energy is counted in joules. FAO gives the link between the two as 1 kcal = 4.184 kJ. Australian and New Zealand packs must show kilojoules (FSANZ). The conversion takes one multiplication, laid out in calories to joules and kilojoules.

Take the cup of milk worked out below. It holds about 149 kcal. In the physicist’s unit that is 149,000 small calories, which is why nobody prints them that way. Check the capital letter before you convert: a thousandfold slip is easy to make.

Foods differ in calories per gram. Water has none. Carbohydrate and protein have about 4, and fat, as the next section shows, has about 9, more than twice as much.

How does a food label turn grams into calories?

A label figure is often arithmetic, not a measurement. A maker takes the grams of protein, carbohydrate and fat in a serving and multiplies each by a fixed factor. UK and EU law requires that route, and the US rule allows it alongside others, such as food-specific factors or bomb calorimetry data. The factors carry the name of W.O. Atwater, who built the system with colleagues at a USDA station in Connecticut at the end of the nineteenth century. FAO describes it as resting on the heat each nutrient gives off when burned. That heat is then adjusted for what the gut fails to absorb and for the urea lost in urine.

One factor per nutrient is a simplification, and the FAO says so. A refinement from 1955 gave each food its own factors, because protein in rice releases about 20% more heat than protein in potatoes. That is why USDA records can list two energy figures for one food.

Annex XIV of the UK version of EU food-information law supplies the factors below.

Component	kcal per gram	kJ per gram
Protein	4	17
Carbohydrate (except polyols)	4	17
Fat	9	37
Alcohol	7	29
Fiber	2	8
Polyols (sugar alcohols)	2.4	10

The two columns are separate legal factors, not conversions of each other: 9 kcal would convert to 37.7 kJ, yet the annex gives 37. The same annex gives erythritol, a sugar alcohol, 0 kcal per gram. Fiber is the line that changes by region. The US rule in 21 CFR 101.9 lets a maker use 4-4-9. Carbohydrate is counted less non-digestible carbohydrate and sugar alcohols, and the soluble kind gets 2 kcal per gram. The rule also allows factors specific to a food. Protein, carbohydrate and fat per gram compares the systems on eight foods.

Here is the sum on a real record. I took whole milk in USDA FoodData Central, per 100 g: 3.27 g protein, 4.63 g carbohydrate and 3.2 g fat.

• Protein: 3.27 × 4 = 13.08 kcal
• Carbohydrate: 4.63 × 4 = 18.52 kcal
• Fat: 3.2 × 9 = 28.8 kcal
• Total: 60.4 kcal

The record carries two energy lines for milk, 61 kcal on general factors and 60 on food-specific ones. A cup is 249 g, which the record lists at 149 kcal (60 × 2.49), and a US label would print 150. Milk is an easy case, since it’s mostly water plus the three macronutrients. My sum lands within 1% of both record figures.

Fiber makes the arithmetic less tidy. I invented a bar to show it: 4 g protein, 25 g total carbohydrate (3 g of it fiber) and 6 g fat per serving. A maker who uses plain 4-4-9 gets 16 + 100 + 54 = 170 kcal. A maker who counts the fiber at 2 kcal per gram gets 16 + 88 + 6 + 54 = 164. Both round to the nearest 10, so one prints 170 and the other 160. Same bar, two different labels.

Labels count the energy in food. To see the energy you use, put your age, height, weight and activity into the calorie calculator. It runs the 2023 NASEM equations and never suggests an intake below the guideline floors.

Do you absorb every calorie the label lists?

Not always. Label factors estimate metabolizable energy, which is what is left once the losses are taken off. The ceiling is the heat released when a food burns completely to carbon dioxide and water inside a bomb calorimeter. You never get all of that, and FAO notes that how much reaches you depends on the food, its matrix and how it was prepared. Its own example is wheat: whole-wheat flour and heavily milled flour give different available energy from the same weight.

Nuts show the gap best. Take the walnut record, per 100 g: 14.6 g protein, 10.9 g carbohydrate (5.2 g of it fiber) and 69.7 g fat. I worked out the energy three ways.

Method on the same walnut record	kcal per 100 g
General factors, 4-4-9	730
UK rule, fiber at 2 kcal per gram	719
USDA specific factors	679

The 4-4-9 line is 58.4 + 43.6 + 627.3, which adds to 729.3, and the record lists 730. The UK line takes the 5.2 g of fiber out of the carbohydrate and counts it at 2: 58.4 + 22.8 + 627.3 + 10.4. A 30 g handful comes to 219 kcal on the first line and 204 on the last. A US label would round those to 220 and 200, and both routes are allowed.

Feeding trials measured it in people. In a controlled feeding trial, 18 healthy adults ate the same diet for two 3-week periods. One period added 42 g of walnuts a day. The team ran food, urine and feces through a bomb calorimeter. Energy eaten minus energy lost gave the usable part. A 28 g serving came out at 146 kcal, against 185 calculated with the Atwater factors, so the walnuts delivered 21% less.

A trial of almonds gave 18 adults 0, 42 or 84 g a day over 18-day periods. It found 129 kcal per 28 g, against 168 to 170 by the Atwater factors, which the authors call a 32% overestimate. Put the other way round, the almonds delivered about a quarter less than predicted.

Structure seems to matter. A follow-up found that chopped almonds gave 5.04 kcal per gram and whole natural ones 4.42, both below prediction. Almond butter, at 6.53, came out close to it, with no significant difference. The authors suggest that whole almonds break into fewer, larger pieces that keep their fat from being released. They offer that as a possible reason, not a proven one.

Processing changes the cost of digestion as well. In a crossover study, 17 adults each ate two cheese sandwiches of equal energy. One had multigrain bread and cheddar, the other white bread and processed cheese. Over the next five to six hours, energy use came to 19.9% of the meal’s energy after the first and 10.7% after the second. That is about 46% lower.

The trials have limits. The nut studies tested one nut at a time, at 42 g a day or more, in healthy adults on a controlled diet. The sandwich study compared two meals, once, in 17 people. I find the sandwich result the weakest of the lot, since it rests on one pair of meals. None of it tells you what your own gut does with a handful of something else. It gives you no discount to apply to your plate.

How far off can a printed number be?

By more than the tidy figure suggests. Rounding is the small part: a US label states calories to the nearest 5 up to 50, and to the nearest 10 above that. Under 5, it may print zero, as 21 CFR 101.9 says. A serving of 4 kcal can read 0. Ten servings add up to 40 kcal.

The margin is bigger. The same rule treats a food as misbranded when the calories in a composite of samples run more than 20% above the declared value. The limit only looks upward. A pack labeled 200 kcal can hold up to 240 before it breaks the rule, and one holding 170 doesn’t.

Testers have measured the gap, too. In a pilot study, researchers bought reduced-energy foods and measured them in a bomb calorimeter. The 29 restaurant items averaged 18% above their stated energy, and the 10 supermarket frozen meals 8%. Neither gap reached statistical significance, because the spread was wide. Some restaurant items held double, and free side dishes took the entrees they came with to an average of 245% of the stated figure. Even so, most items sat inside the 20% allowance. The sample was small and covered one kind of food, so it shows the direction of the miss and not the size of yours.

I put those averages onto a day. A diary of 1,800 kcal in which every item ran 8% over is really 1,944 kcal, which is 144 more. At 18% over it is 2,124. In that test the averages leaned over, not under, and a lean in one direction adds up instead of canceling.

What should you do with an estimate?

Use it for comparison, and judge your intake over weeks. One meal can sit 20% off its label. A week of meals averages out the random part, and a few weigh-ins each week show what your total did. Compare like with like, too. Switching between a label, a database record and a restaurant’s website adds noise you can’t tell from change. Why calorie trackers disagree shows how far apart one food lands across records.

If you eat walnuts or almonds daily, the trials point to a label that overstates what you absorb. A logged handful runs a little high. Leave it alone. If you subtract a guessed 20%, you’ve added a second error to the first.

Weigh when it counts. A label per serving inherits the maker’s idea of a serving, while grams are yours. How to weigh food covers the steps, including raw against cooked. How to read a nutrition label shows where each figure sits on a US, UK or Australian pack.

Try this today. Pick one packaged food you eat most days. Weigh the serving, read the protein, carbohydrate and fat grams from the label, and multiply by 4, 4 and 9. If the sum lands within rounding distance of the printed number, the label is consistent. A bigger gap points to fiber, sugar alcohols, food-specific factors or rounding. Rounding and fiber moved the bar. Then log the grams you weighed, not the serving on the front.

A pack tells you the energy in the food. The other half is what you need in a day. How many calories per day you need explains how that is estimated and how wide its error runs.

The unit is exact: 1 kcal is 4.184 kJ. The number printed beside it is an estimate. Weigh what you can, and let a few weeks of your own scale readings outvote any single label.