The 4-4-9 Rule — Where It Breaks Down

The 4-4-9 calorie rule — protein provides 4 kcal/g, carbohydrates 4 kcal/g, fat 9 kcal/g — is the universal shorthand that every food label, nutrition app, and dietary guideline uses to convert macronutrient weights into energy estimates, but it is a deliberate simplification that Wilbur Atwater himself acknowledged was an average over many food types, not a physical constant, and modern research has identified at least six categories where the simplified factors produce systematic errors of 10–30% that compound meaningfully over a full day’s calorie count. The 4-4-9 rule collapses the actual metabolic energy yield of food into three broad buckets, ignoring the thermic effect of food, the variable digestibility of different protein sources, the non-caloric contribution of some carbohydrate fractions, and the distinct metabolic handling of medium-chain triglycerides versus long-chain fatty acids.

The original Atwater method: general vs specific factors

Wilbur Olin Atwater was an American agricultural chemist who conducted the first systematic calorimetric measurements of food energy in humans at Wesleyan University between 1896 and 1907. His method was rigorous for its era: he measured heat produced by burning food in a bomb calorimeter, measured heat produced by humans metabolising the same foods in a respiration calorimeter, and calculated the difference as energy lost in stool, urine, and digestive gases.¹

Crucially, Atwater developed two sets of factors: general factors (4/4/9) for use when food composition details are unknown, and specific factors for individual food groups, published in multiple tables between 1896 and 1902. The specific Atwater factors for wheat protein (3.91 kcal/g), wheat starch (3.87 kcal/g), and wheat fat (8.37 kcal/g) differ materially from the general 4/4/9 defaults. For egg protein, the specific factor is 4.36 kcal/g. For beef protein, 4.27 kcal/g. These are not rounding differences — they represent meaningful variation in the metabolically available energy from proteins of different origin.

Modern food databases and nutrition apps almost universally use the general factors rather than the food-specific ones, partly for simplicity and partly because accurate application of specific factors requires knowing the food’s exact botanical or zoological origin — information not always available for processed or mixed foods. The USDA’s FoodData Central does use modified Atwater factors for some food categories, but most consumer-facing apps that pull from USDA data do not surface this distinction to the user.

The World Health Organization and the Food and Agriculture Organization reviewed the Atwater factors in a 2002 expert consultation and recommended transitioning toward the FAO/WHO/UNU energy conversion factors, which differ from the general Atwater values for several macronutrients — but this transition has been uneven across national food regulatory agencies and essentially absent in consumer apps.¹

The fundamental problem is this: 4-4-9 was designed as a population-level approximation for an era before high-fiber foods, sugar alcohols, and MCT-enriched products became common in consumer diets. Applying it to a 2026 diet that includes keto snack bars with erythritol, protein shakes with hydrolysed collagen, and bulletproof coffee with coconut oil introduces systematic errors that Atwater never intended the general factors to handle.

Dietary fiber: the biggest systematic error

The general Atwater factors assign 4 kcal/g to all carbohydrates, including dietary fiber. This is the largest systematic error in mainstream calorie counting, and it disproportionately affects people eating high-fiber diets — exactly the population most likely to be tracking calories for health reasons.²

The error occurs because fiber is not a uniform entity — and the distinction between net carbs and total carbs matters here. Insoluble fiber (cellulose, lignin) passes through the gastrointestinal tract essentially unmodified. Human digestive enzymes cannot cleave the beta-glycosidic bonds in cellulose; it exits the body unchanged and contributes essentially 0 kcal to energy balance. Soluble fermentable fiber (inulin, pectin, beta-glucan, fructooligosaccharides) is fermented by colonic bacteria into short-chain fatty acids (SCFAs) — primarily acetate, propionate, and butyrate — which are absorbed through the colonic wall and contribute approximately 2 kcal/g of fiber fermented.

The upshot: the correct energy factor for dietary fiber is not 4 kcal/g. For insoluble fiber, it is 0 kcal/g. For soluble fermentable fiber, it is approximately 2 kcal/g. Blended as a weighted average across a typical mixed-fiber diet, the effective value is approximately 1.5–2 kcal/g.²

The US FDA adopted a corrected fiber energy factor of 2 kcal/g in its 2016 updated nutrition labeling regulations. This is reflected in how the FDA instructs manufacturers to calculate calories on Nutrition Facts panels — fiber grams are multiplied by 2, not 4, before being included in the total calorie count. However, many nutrition tracking apps that calculate calories from raw macronutrient inputs (rather than pulling from label data) still apply 4 kcal/g to total carbohydrates including fiber, then subtract fiber from “net carbs” without adjusting the calorie calculation.

The practical error magnitude: a 100g serving of black beans contains approximately 16g dietary fiber. At the correct 2 kcal/g factor, fiber contributes 32 kcal. At the incorrect 4 kcal/g factor, it contributes 64 kcal — a 32 kcal overestimate per serving. For someone eating 30g fiber per day (recommended intake is 25–38g), the daily overestimate from the fiber error alone is 30–60 kcal. Across a year, this represents 10,950–21,900 kcal of phantom calories — a meaningfully incorrect baseline for anyone trying to maintain a precise deficit.

Sugar alcohols: variable absorption, variable calories

Sugar alcohols (polyols) are used as low-calorie sweeteners in products marketed to people following low-carbohydrate or diabetic diets. The category includes erythritol, xylitol, maltitol, sorbitol, mannitol, isomalt, and lactitol — and their metabolic fates are so different that treating them as a single category is a serious analytical error.³

Erythritol is absorbed almost entirely in the small intestine (up to 90% of ingested dose) and excreted unchanged in urine, bypassing colonic fermentation. Net energy yield: approximately 0.2 kcal/g. It produces no glycemic response and contributes negligible calories regardless of dose.

Xylitol is partially absorbed (approximately 50%) in the small intestine and partially fermented in the colon. Net energy yield: approximately 2.4 kcal/g.

Maltitol is 75–85% absorbed in the small intestine with a glycemic index of approximately 35 (compared with 100 for glucose). Net energy yield: approximately 2.1 kcal/g — and it produces a meaningful, though blunted, insulin and blood glucose response.

Sorbitol is approximately 30–50% absorbed; the remainder reaches the colon where fermentation causes the gas and bloating familiar to anyone who has eaten too many “sugar-free” candies. Net energy yield: approximately 2.6 kcal/g.

The FDA has specific energy conversion factors for each polyol, and these are the values used in nutrition labeling regulations. The problem is that apps tracking “net carbs” (total carbs minus fiber minus sugar alcohols) often zero out all polyols from the calorie calculation — severely underestimating the calorie contribution of maltitol-rich products by up to 50%. Apps that count all sugar alcohols at the full 4 kcal/g overstate the calorie contribution of erythritol by a factor of 20.

The correct approach is polyol-specific energy factors applied per ingredient, which requires knowing which specific polyol is in the product — information available on the ingredient list but not always on the nutrition label. For practical tracking, erythritol can be excluded from the calorie count; maltitol and xylitol should be counted at approximately 2–2.5 kcal/g.

Alcohol: the missing macronutrient

Ethanol is the only commonly consumed food substance that has no slot in the 4-4-9 framework. It is metabolically significant — ethanol provides 7.1 kcal/g by bomb calorimetry and has priority metabolism that displaces both fat oxidation and glucose metabolism in the liver — yet nutrition labels in most countries do not list it as a macronutrient, and many tracking apps handle it inconsistently.⁴

The bomb calorimetry value of 7.1 kcal/g overstates metabolically available alcohol energy by approximately 20–25%. Ethanol metabolism is obligatory through the acetaldehyde-acetate pathway, which has a high thermogenic cost — approximately 20% of ingested alcohol energy is dissipated as heat from the metabolic process itself. Charles Lieber’s careful human calorimetry studies (American Journal of Clinical Nutrition, 1991) established that metabolically available energy from ethanol is approximately 5.6 kcal/g, not 7.1 kcal/g.⁴

Most apps that track alcohol use 7 kcal/g (some use the rounded 7.1 kcal/g from bomb calorimetry). The practical consequence: a person consuming 3 standard drinks per day (approximately 42g ethanol) has their alcohol calories overstated by approximately 63 kcal (42g × 1.5 kcal/g error). This is modest individually but directionally wrong — and it doesn’t account for the more significant metabolic effect of alcohol, which is the suppression of fat oxidation for 2–3 hours after drinking. This is why alcohol calories are not equivalent to food calories in terms of their effect on fat storage: the same calorie intake that would not cause fat gain in the absence of alcohol may promote fat storage when alcohol occupies the hepatic oxidation pathway.

Medium-chain triglycerides and novel proteins

The 9 kcal/g fat factor is correct for long-chain fatty acids (C14–C22, which constitute the majority of fat in most diets) but measurably incorrect for medium-chain triglycerides (MCTs, C8–C12), which are handled differently by the body from the moment of absorption.⁵

Long-chain fatty acids are absorbed into the lymphatic system, packaged into chylomicrons, and delivered to peripheral tissues (primarily adipose and muscle) before reaching the liver. MCTs, by contrast, are absorbed directly into the portal vein and travel to the liver, where they are rapidly oxidised rather than esterified for storage. The thermic effect of MCT is higher than for long-chain fats, and a larger proportion is oxidised rather than stored. Seaton et al. 1986 (American Journal of Clinical Nutrition) measured net energy yield of MCT at approximately 8.3 kcal/g versus 9 kcal/g for long-chain fats — a 7.7% overestimate for MCT-containing products like coconut oil and MCT oil supplements.⁵

For protein, the source-dependence of digestibility creates systematic errors in the opposite direction from how apps calculate them. Protein from whole plant foods (legumes, whole grains) with intact cell walls has a digestibility-corrected amino acid score (DIAAS) of 0.6–0.8, meaning 20–40% of the protein grams are not absorbed as amino acids. Yet all protein is counted at 4 kcal/g regardless of bioavailability. A person eating 30g of plant protein from intact lentils absorbs perhaps 20–24g of that amino acid content; the remaining 6–10g passes into the colon as undigested peptides and is fermented at a lower energy yield than 4 kcal/g. The practical implication for calorie counting is a modest overcounting of calories from whole-food plant protein sources.

What a corrected calorie count looks like

Applying updated energy factors — FDA’s 2016 fiber factor (2 kcal/g), polyol-specific factors, corrected alcohol factor (5.6 kcal/g), and MCT adjustment — to a typical high-fiber, high-polyol “keto” meal produces a meaningfully different calorie figure than the 4-4-9 default.^2,3

Example meal: keto protein bowl

• 200g ground turkey (protein: 52g, fat: 20g, no carbs): Standard calc = 388 kcal. Corrected: ~385 kcal (minimal change for animal protein).
• 50g erythritol-sweetened dark chocolate (total carbs: 28g, of which 22g erythritol, 6g net digestible): Standard calc at 4 kcal/g = 112 kcal for carbs. Corrected: 6g digestible carbs × 4 kcal + 22g erythritol × 0.2 kcal = 28 kcal. Difference: −84 kcal.
• 15g MCT oil: Standard calc at 9 kcal/g = 135 kcal. Corrected at 8.3 kcal/g = 125 kcal. Difference: −10 kcal.
• 30g insoluble fiber supplement: Standard calc at 4 kcal/g = 120 kcal. Corrected at 0 kcal/g = 0 kcal. Difference: −120 kcal.

Total standard calc: 755 kcal. Corrected: 538 kcal. Difference: −217 kcal (29%).

For people eating standard mixed Western diets without sugar alcohols or high-dose fiber supplements, the 4-4-9 error is closer to 3–5% — not clinically significant. For people following high-fiber, sugar-alcohol-rich diets, the error can exceed 20% and creates a systematic calorie overestimate that inflates the apparent deficit and obscures why weight loss is proceeding faster than the numbers predict — or why the app’s calorie targets feel too easy to hit.

References

1. Atwater WO, Benedict FG. Experiments on the Metabolism of Matter and Energy in the Human Body, 1900–1902. USDA Office of Experiment Stations Bulletin No. 136. Washington DC, 1903. [Referenced in: Food and Agriculture Organization of the United Nations. Food Energy: Methods of Analysis and Conversion Factors. FAO Food and Nutrition Paper 77. Rome: FAO, 2003.]

2. Livesey G. “Energy Values of Unavailable Carbohydrate and Diets: An Inquiry and Analysis.” American Journal of Clinical Nutrition 51, no. 4 (1990): 617–637.

3. Grabitske HA, Slavin JL. “Gastrointestinal Effects of Low-Digestible Carbohydrates.” Critical Reviews in Food Science and Nutrition 49, no. 4 (2009): 327–360.

4. Lieber CS. “Perspectives: Do Alcohol Calories Count?” American Journal of Clinical Nutrition 54, no. 6 (1991): 976–982.

5. Seaton TB, Welle SL, Warenko MK, Campbell RG. “Thermic Effect of Medium-Chain and Long-Chain Triglycerides in Man.” American Journal of Clinical Nutrition 44, no. 5 (1986): 630–634.

6. Colombo PE, Patterson E, Elinder LS, Ejlerskov K. “Removing Meat and Increasing Plant-Based Foods in the Diet: A Systematic Review.” Nutrients 15, no. 12 (2023): 2680. [DIAAS context for plant protein digestibility.]