The core vocabulary

Wilbur Atwater in the 1890s measured these losses and published the Atwater general factors that are still used in food labeling today:

• Carbohydrate: 4 kcal/g
• Protein: 4 kcal/g
• Fat: 9 kcal/g
• Alcohol: 7 kcal/g
• Fiber (not absorbed): typically 2 kcal/g for soluble fiber, 0 for insoluble

These are round numbers averaged across foods; Atwater specific factors adjust for the actual digestibility of individual foods (corn protein is less digestible than egg protein, for example). Nutrition labels use the general factors unless a food differs substantially.

Macronutrients

The three macronutrients are carbohydrate, protein, and fat. Alcohol is sometimes treated as a fourth. Each provides both energy and, in the case of protein and fat, essential building blocks the body cannot synthesize from other sources.

• Carbohydrates include sugars (mono- and disaccharides), starches (polysaccharides that humans can digest), and fiber (polysaccharides humans cannot digest). Fiber still matters nutritionally even though it is not absorbed — it shapes the gut environment, slows glucose absorption, and binds bile acids.
• Proteins are chains of amino acids. Nine amino acids are essential for adult humans (cannot be synthesized and must come from the diet): histidine, isoleucine, leucine, lysine, methionine, phenylalanine, threonine, tryptophan, valine. Protein quality is judged by how closely a food's amino acid profile matches human requirements and by digestibility.
• Fats are triglycerides composed of a glycerol backbone and three fatty acids. Fatty acids are categorized by saturation (number of double bonds) and by chain length. Two fatty acids are essential: linoleic acid (omega-6) and alpha-linolenic acid (omega-3).

Micronutrients

Micronutrients are vitamins and minerals — required in small amounts but essential for enzymatic and structural functions. Thirteen vitamins and approximately sixteen minerals are established as essential for humans. The field of micronutrition was built on deficiency-disease mapping: scurvy to vitamin C, beriberi to thiamine, pellagra to niacin, rickets to vitamin D, xerophthalmia to vitamin A. Each of these diseases was endemic before the responsible nutrient was identified.

Reference intakes

The US and Canada use the Dietary Reference Intakes (DRIs), which replaced the older RDA system in 1997. DRIs include:

• EAR (Estimated Average Requirement) — intake meeting the needs of 50% of healthy individuals
• RDA (Recommended Dietary Allowance) — intake meeting the needs of 97–98% of healthy individuals; typically EAR + 2 standard deviations
• AI (Adequate Intake) — used when data are insufficient to calculate an EAR
• UL (Tolerable Upper Intake Level) — highest intake likely to pose no risk of adverse effects

The distinction between EAR and RDA matters. Dietary intake assessments compare populations against EAR, not RDA, because the question is "what fraction of the population has inadequate intake," not "does this individual meet the high-confidence recommendation."

The measurement problem

Nutrition science is harder than many adjacent fields for measurable, structural reasons. The student who does not internalize these limits ends up confused when two respectable studies reach opposite conclusions.

Problem 1 — Dietary recall is noisy

Most population-level nutrition data comes from dietary recall instruments: 24-hour recalls, food-frequency questionnaires (FFQs), diet diaries. All of these depend on people remembering and reporting what they ate. The noise is large. Underreporting of total energy intake in FFQs is routinely 20–30%, and the underreporting is non-random: people underreport foods they think of as unhealthy.

Implication: Observational nutrition studies that rely on FFQs and find small effect sizes are essentially uninterpretable. A 10% relative-risk difference in a study where the exposure is measured with 30% error is statistically fragile.

Problem 2 — Controlled feeding is expensive and short

The alternative to recall is controlled feeding: the study provides all the food, so intake is known. Controlled-feeding studies are the gold standard for metabolic endpoints, but they are expensive, tightly controlled-feeding trials rarely exceed a few weeks, and they cannot answer questions about long-term outcomes like cancer or cardiovascular events. The tradeoff between precision (controlled feeding) and duration (observational cohort) is the central methodological problem of the field.

Problem 3 — Foods are not nutrients

Nutrients interact. Calcium absorption depends on vitamin D and is inhibited by oxalates. Iron absorption depends on heme form and vitamin C and is inhibited by phytates and tannins. A study that measures "iron intake" without measuring the food matrix in which it was delivered is not actually measuring bioavailable iron. The "nutrient" frame is useful for biochemistry and deficiency diseases; it breaks down for complex chronic-disease questions where food-matrix effects may dominate.

Worked example — reading a nutrition label

A standard US nutrition label for a cup of 2% milk reads:

Serving size: 1 cup (240 mL)
Calories: 120
Total fat: 5 g (saturated 3 g)
Total carbohydrate: 12 g (sugars 12 g)
Protein: 8 g
Calcium: 300 mg (23% DV)
Vitamin D: 2.5 mcg (15% DV)

Using Atwater general factors to check the calorie count:

• Fat: 5 g x 9 kcal/g = 45 kcal
• Carbohydrate: 12 g x 4 kcal/g = 48 kcal
• Protein: 8 g x 4 kcal/g = 32 kcal
• Total: 125 kcal

The label says 120. The 5-kcal discrepancy is within the rounding tolerance FDA permits. The exercise is useful because it demonstrates that the label's macronutrient values are the ground truth and the calorie count is derived from them — not the other way around. A product whose macronutrients don't match its calorie count is almost certainly mislabeled.

Worked example — comparing two plant-based protein sources

Lentils and hemp seeds are both plant proteins but have very different profiles.

                    Lentils (100g cooked)    Hemp seeds (100g)
Protein                9 g                      31 g
Essential AAs
  - lysine             0.65 g                   1.28 g
  - methionine         0.08 g                   0.93 g
PDCAAS                 0.52                     0.66

Lentils are lysine-rich but methionine-limited (the limiting amino acid in most legumes). Hemp seeds are high in both. A meal of lentils with a small amount of rice complements the lysine limitation of rice, producing a complete protein match even though neither food is complete on its own. This is the classic amino acid complementation strategy used in traditional diets around the world — rice and beans, pita and hummus, corn and beans.

Assessment protocol for a nutrient-intake question

When a user asks whether an intake level is adequate, the department follows a protocol:

1. Identify the population. Age, sex, pregnancy/lactation status, activity level. DRIs vary across these.
2. Look up the EAR, RDA, AI, and UL. Use the current DRI tables from the US National Academies.
3. Compare against EAR for inadequacy assessment, not RDA. This is a common error. The RDA is a target; the EAR is the threshold below which probability of inadequacy is at least 50%.
4. Check the UL. Supplement-heavy diets can approach the UL for fat-soluble vitamins and several minerals.
5. Consider bioavailability and matrix effects. Iron from a steak is not iron from a spinach leaf. Calcium from yogurt is not calcium from kale.
6. Report uncertainty honestly. Dietary recall noise, nutrient-database limits, and individual variability all contribute.

Routing heuristics

• Questions about calorie accounting or macronutrient breakdown → atwater.
• Questions about controlled-trial design or study interpretation → lind.
• Questions about long-term cardiovascular or dietary-pattern evidence → ancel-keys.
• Questions about food policy or industry influence on research → marion-nestle.
• Questions about child nutrition and feeding relationships → satter.
• Questions about popular nutrition writing or food-system framing → pollan.
• Questions about claims that sound unusually confident or unusually alarming → route to contested-claims-in-nutrition skill first, then to an appropriate specialist for critique.

Common failure modes

• Treating the RDA as a threshold for individual adequacy. The RDA is designed so that individuals at the RDA are almost certainly adequate; it is not a threshold below which individuals are inadequate.
• Reporting calorie counts without noting the Atwater-factor assumptions. Label calories are not measured; they are computed from macronutrients using rounded factors.
• Aggregating nutrients from wildly different food matrices as if "10 g protein" means the same thing everywhere.
• Confusing observational correlation with causal effect in a literature where controlled feeding is short and dietary recall is noisy.

Further reading

• Atwater and Bryant, 1900, The Chemistry of Foods and Nutrition (foundational US nutrition text)
• Lind, 1753, A Treatise of the Scurvy (the first controlled trial)
• Institute of Medicine DRI volumes, 1997–2011 (the reference-intake framework)
• National Academies Press, Dietary Reference Intakes: The Essential Guide to Nutrient Requirements