Why you can’t stop eating processed foods—it’s the design, not your willpower.
What the evidence says
- Whole foods carry the properties that track with satiety: across 38 foods tested at matched energy, satiety correlated positively with serving weight, water, fiber and protein content, and negatively with fat — the paper reports correlations, not prediction, and serving weight, water, fibre and protein all describe a food being heavier per calorie, so these may not be four separate mechanisms[1]
- Palatability itself predicts lower satiety: across those same 38 foods, palatability ratings correlated negatively with satiety — the more a food was liked, the less it filled at matched energy[1]
- People can consume substantially more calories without registering it as more food: participants ate 508 kcal/day more on the ultra-processed diet while reporting no difference in the meals’ pleasantness (p = 0.13) or familiarity (p = 0.57), and no significant difference in hunger, fullness, or satisfaction on visual analogue scales[2]
- The appetite hormones moved in the direction that fits: on the unprocessed diet, appetite-suppressing PYY rose (28.9 → 34.3 pg/ml, p = 0.047) and hunger-signalling active ghrelin fell (61.4 → 48.3 pg/ml, p = 0.01), while the ultra-processed diet shifted neither[2]
- Ultra-processed diet caused 508 kcal/day excess intake vs matched unprocessed diet — excess came entirely from carbohydrate (+280 kcal/day) and fat (+230 kcal/day), not protein[2]
- Unprocessed diet reduced fasting insulin, C-peptide, and insulin resistance (HOMA-IR) relative to baseline within 2 weeks; ultra-processed diet did not (corrected 2026-09-10: the bullet added ”metabolic benefits from unprocessed food beyond calorie effects”, which the paper does not say; intake and weight both fell on that diet)[2]
- Difficulty stopping is not simple weakness — it has a measurable neural signature: higher food-addiction (YFAS) scores went with greater anticipatory activation in anterior cingulate (r = .74), medial orbitofrontal cortex (r = .58) and amygdala (r = .55), while the high-scoring group showed reduced lateral orbitofrontal activation during actual consumption (r = −.66) — heightened anticipation paired with weakened inhibitory control, the pattern of substance dependence[3]
- That finding is 39 young women, cross-sectional, and it measures the response rather than its cause — it does not establish that the foods were designed to produce it[3]
- ”Vanishing caloric density” is investigative reporting, not a measured finding: the proposal that foods engineered to dissolve quickly in the mouth keep satiety signals from registering the calories comes from Moss’s reporting, and this library holds the book’s existence rather than its text[4]
- REMOVED 2026-08-23: the correlation coefficients — water 0.64, fibre 0.46, protein 0.37, fat −0.43, palatability −0.64 — and the comparison “as strong as water content’s positive one.” `csi01` is held as abstract only and states directions without magnitudes. The directions stand; the numbers had no source here. Restore only against the full text.