Insulin delivered to the brain makes people eat less, not more#
The popular account has insulin as the hormone that fattens you and drives you back to the fridge: carbohydrate raises insulin, insulin locks fat away and leaves you hungry, so keeping insulin low is how you eat less. The direct experiments point the other way. When healthy women received 160 IU of insulin as a nasal spray — a route that delivers it to the brain without meaningfully raising blood insulin — their intake of chocolate chip cookies after a meal fell to 76.11 kcal against 112.74 kcal on placebo (P = 0.049), and they rated those cookies as markedly less appealing (56.48 vs 72.88, P = 0.025)1.
So more insulin acting at the brain produced less eating, and specifically less eating of the most tempting food. Insulin's rise after a meal is part of the signal to stop, not the signal to start. That said, the everyday experience the popular story is trying to explain — a big fast-carbohydrate breakfast, then ravenous by eleven — is real, has been measured, and does involve insulin. It just doesn't work the way the slogan says. What follows is the difference between those two claims, because it decides whether "keep insulin low" is useful advice.
One useful detail from the same paper: the effect appeared only after a meal. In a second experiment with fasted participants, intranasal insulin did nothing to snack intake or palatability. Brain insulin is a postprandial satiety signal — it tells your brain that energy has arrived, which is exactly the job a naive reading assigns to leptin.
Where the rebound-hunger feeling actually comes from#
The strongest evidence for the folk model is a small, careful, and genuinely persuasive experiment. Twelve adolescent boys with obesity ate breakfast and lunch on three separate occasions, with the meals matched for energy, macronutrients, fiber and palatability, and differing only in glycemic index. Then they ate freely.
Voluntary intake after the high-glycemic-index meal was 53% greater than after the medium-GI meal and 81% greater than after the low-GI meal. The high-GI condition produced higher serum insulin, lower postabsorptive glucose and free fatty acids, and — the detail worth pausing on — elevated plasma epinephrine. The area under each meal's glycemic response curve accounted for 53% of the within-subject variance in how much was subsequently eaten2.
That epinephrine reading is what makes the "crash" more than a figure of speech. When glucose falls faster and further than expected, the body mounts a counter-regulatory response — adrenaline among it — and that response is not a calm state in which to be near food. The sequence is a fast, large glucose rise, an insulin response scaled to it, an overshoot on the way down, and a stress signal that arrives with hunger attached.
The rebound isn't insulin being present. It's insulin being matched to a rise so steep that the fall overshoots — and the adrenaline that answers the overshoot.
Hold the study's size in view: twelve boys, three test days, obesity-specific, and a high-GI condition built to be extreme. It demonstrates a mechanism cleanly rather than describing an average adult's Tuesday. The broader question of which carbohydrates behave this way, and how much glycemic index predicts anything useful, is worked through in are there really 'good' and 'bad' carbs.
Scale it up to whole diets and the prediction inverts#
The carbohydrate-insulin model makes a testable prediction: cut carbohydrate, cut insulin, and people should spontaneously eat less. Somebody ran that experiment under about the tightest conditions available.
Twenty adults lived as inpatients and were randomized to two weeks of a plant-based, low-fat diet (10.3% fat, 75.2% carbohydrate, glycemic load 85 g per 1,000 kcal) and two weeks of an animal-based ketogenic diet (75.8% fat, 10.0% carbohydrate, glycemic load 6), in alternating order, eating as much as they wanted. The low-fat, high-carbohydrate diet — the one producing far more insulin — resulted in 689 ± 73 kcal per day less energy intake over the two weeks, and 544 ± 68 kcal per day less over the final week, both P < 0.0001. The authors' conclusion is unambiguous: the predictions of the carbohydrate-insulin model were inconsistent with their observations3.
| What was manipulated | Effect on insulin | Effect on eating |
|---|---|---|
| 160 IU insulin to the brain after a meal1 | Raised centrally, not systemically | Cookie intake 112.7 → 76.1 kcal |
| High- vs low-GI meals, energy and macros matched2 | Higher after high-GI | Later intake 81% greater after high-GI |
| Ketogenic vs high-carb whole diets, 2 weeks each3 | Far lower on ketogenic | Ate 689 kcal/day more on ketogenic |
Three rows, and the middle one appears to contradict the other two. It doesn't, and working out why is the most useful thing on this page.
What those results actually disagree about#
Start by being precise about the comparison each one made, because they are not the same comparison wearing different clothes.
Ludwig held energy, macronutrients, fiber and palatability constant and moved one variable — how fast the carbohydrate hit the bloodstream — then measured eating a few hours later. Hall moved entire diets that differed in fat, carbohydrate, fiber, energy density, degree of processing and food form all at once, and measured eating across two weeks. Ludwig's question is within a diet: does a steep glucose curve provoke a rebound? Hall's question is between diets: does the lower-insulin diet win on spontaneous intake?
Both answers can be true, and probably are. A fast rise-and-fall can drive extra eating in the hours after a specific meal, while insulin level is nonetheless a poor predictor of how much you eat across a whole diet — because at the diet level, other properties of the food are doing far more work. Nothing in Hall's result denies the crash; nothing in Ludwig's result rescues the claim that lowering insulin lowers intake.
Worth naming who is arguing, in both directions. Ludwig is among the most prominent proponents of the carbohydrate-insulin model, and his 1999 experiment is one of its foundations. Hall has been that model's most prominent critic and designed the 2021 trial specifically to test it. Neither fact alters a measurement, and each team published results a fair reader can check — but knowing the priors is part of reading the literature properly, and the standard should apply to both sides or neither.
One limitation cuts against Hall's arm too: two weeks is short, participants were housed in a metabolic ward, and one withdrew with hypoglycemia on the low-carbohydrate diet. What the trial licenses is the narrow claim it makes — that the low-insulin diet did not produce lower intake — rather than a verdict on ketogenic eating generally. Which carbohydrate level suits you is a separate and much more individual question.
What this means between meals#
The practical translation is smaller than either camp's rhetoric and more usable than both.
Insulin is not a dial you should be trying to keep down. It rises after eating because that is its job, it participates in telling your brain the meal happened, and pushing it lower did not make people eat less when somebody actually checked. Meanwhile the thing you can act on is the shape of the curve, not its existence: a meal that arrives fast and alone produces the steepest rise and the sharpest fall, and a meal with protein, fat and fiber alongside the carbohydrate does not. That is the same lever as protein's effect on fullness, arriving from the glucose side.
Two boundaries are worth keeping straight. Insulin sensitivity is a different variable from insulin level, and it is the one that short sleep genuinely damages — walked through in how one bad night hurts insulin sensitivity, with the same meal landing harder at night for reasons covered in why the same meal hits harder at night. And none of this changes the arithmetic: hunger determines how hard the deficit is to hold, not whether the deficit is what moves your weight. Insulin shapes the experience of dieting. It does not repeal the accounting — and the wider set of hormones people hope will explain their appetite gets audited in the cluster's pillar.
If you get ravenous ninety minutes after breakfast, the fix suggested by all three studies above is the same: change what the carbohydrate arrives with, not whether insulin shows up.
FAQ#
Does insulin make you hungry?#
Not at the brain, where the decision is made. Insulin delivered intranasally after a meal reduced women's intake of chocolate chip cookies from 112.74 to 76.11 kcal and lowered how appealing they found them1. The same study found no effect in fasted participants, so insulin acts as a satiety signal specifically after eating. What people experience as insulin-driven hunger is better described as the aftermath of a steep glucose rise and fall.
Why do I get hungry two hours after a big carb-heavy breakfast?#
Because of the descent, not the peak. In energy- and macronutrient-matched meals differing only in glycemic index, the high-GI condition produced lower post-absorptive glucose and elevated plasma epinephrine, and was followed by 81% more voluntary eating than the low-GI condition2. A glucose fall that overshoots triggers a counter-regulatory stress response, and hunger rides along with it. Adding protein, fat or fiber to the same carbohydrate flattens the curve.
Does keeping insulin low help you eat less?#
When this was tested directly, it did the opposite. Adults living as inpatients ate 689 ± 73 kcal a day more on an animal-based ketogenic diet than on a plant-based low-fat diet with a glycemic load fourteen times higher, with no difference in reported hunger or meal enjoyment3. Low-carbohydrate eating suits plenty of people for plenty of reasons, but a reduced insulin response is not the mechanism by which it would reduce intake.
Sources#
- Hallschmid M, Higgs S, Thienel M, Ott V, Lehnert H. Postprandial administration of intranasal insulin intensifies satiety and reduces intake of palatable snacks in women. Diabetes. 2012;61(4):782-789.
- Ludwig DS, Majzoub JA, Al-Zahrani A, Dallal GE, Blanco I, Roberts SB. High glycemic index foods, overeating, and obesity. Pediatrics. 1999;103(3):E26.
- Hall KD, Guo J, Courville AB, et al. Effect of a plant-based, low-fat diet versus an animal-based, ketogenic diet on ad libitum energy intake. Nat Med. 2021;27(2):344-353.
- Garaulet M, Lopez-Minguez J, Dashti HS, et al. Interplay of dinner timing and MTNR1B type 2 diabetes risk variant on glucose tolerance and insulin secretion: a randomized crossover trial. Diabetes Care. 2022;45(3):512-519.
- Sparsø T, Bonnefond A, Andersson E, et al. G-allele of intronic rs10830963 in MTNR1B confers increased risk of impaired fasting glycemia and type 2 diabetes through an impaired glucose-stimulated insulin release: studies involving 19,605 Europeans. Diabetes. 2009;58(6):1450-1456.
