Part 9 of 10 · Emotional eating, measured

Why eating less makes you eat more: the restrict-then-overeat loop in a million meals

You've been told the fix for overeating is more restraint. Eat less. Wait longer. Be stricter tomorrow. You've probably tried it, more than once.

Here is what 1.1 million logged meals say about that advice. Arriving at a meal famished doubles the chance it ends in overeating, compared with arriving at mild hunger: 13.2% against 6.4%. Waiting too long doesn't build discipline. It builds the next episode.

This isn't a theory about willpower. It's a pattern that shows up whenever meals start at the bottom of the hunger scale.

This article is part of a series built on 1.5 million meals logged in the I feel food app. → What 1.5 million logged meals show about emotional eating

Restricting food, then overeating: the other end of the U

Every meal in the app starts with a hunger reading from 0 (empty) to 9 (very full) and ends with a fullness reading on the same line. We took every meal with both readings and asked how the ending depended on the start.

The result is a U-shape. The top of the U, eating when already full, is the single worst place to start a meal. → Physical hunger vs emotional hunger This article is about the other end.

Famished (0.0–0.4)13.2%Strong hunger(0.5–1.4)12.0%Moderate hunger(1.5–2.4)7.3%Weak hunger (2.5–3.4)6.4%Neither hungry norfull (3.5–4.4)9.8%Already slightly full(4.5–5.4)18.1%Already comfortablyfull (5.5+)29.8%
I feel food tracker · 1,132,266 meals with valid hunger and fullness readings
Show the numbers
Hunger when the meal startedShare of mealsEnded uncomfortably fullEnded in the overeating zone
Famished (0.0–0.4)0.1%13.2%3.6%
Strong hunger1.8%12.0%4.1%
Moderate hunger18.6%7.3%1.8%
Weak hunger65.9%6.4%2.3%

From weak hunger to famished, the overeating rate doubles. Strong hunger, the band most people would call "properly hungry", is nearly as bad: 12.0%, and the highest rate of landing in the app's overeating zone of any band on the hungry side of the scale.

It isn't a willpower problem

By the time you're at a 0.4, the decision has already been made for you. A body that has been told no for too long doesn't negotiate at the table. It eats until the signal to stop finally arrives, and that signal is slow.

That's why the advice to be stricter fails on its own terms. The meal you overate at wasn't the failure. The skipped one before it was, because it set the start point. Willpower at the plate is being asked to undo a decision that was made three hours earlier.

The compounding version

Restriction also has a second cost, and it runs through guilt.

When the strict day breaks, as the table says it tends to, the meal leaves guilt behind. And meals that begin in guilt start without any physical hunger 27% of the time, well above the 19% baseline. Guilt is a feeling that sends people to food. → The guilt loop

So the loop closes: restrict, arrive famished, overeat, feel guilty, eat from the guilt, restrict harder. Each step is a reasonable response to the one before. The loop is the problem, not any single meal in it.

RestrictArrive famishedOvereatGuilt13.2% vs 6.4% at mild hunger27% of guilt mealsstart without hunger
Show the numbers
Link in the loopFigure
Famished meals that ended uncomfortably full13.2%
Weak-hunger meals that ended uncomfortably full6.4%
Meals that begin in guilt and start without hunger27%
I feel food tracker · 1,132,266 meals with valid hunger and fullness readings

What this article is not claiming

This is worth saying plainly, because this reader has been promised a lot.

We have no weight data. None. The app doesn't ask for it and never has, so nothing here says anything about weight in either direction. We have no diagnoses and no clinicians involved. These are not medical findings. They're a description of what happened in meals that real people logged, before and after, and that's all they are.

What they describe is simple: starting a meal very hungry is associated with ending it uncomfortably full, twice as often as starting mildly hungry.

The alternative: eat at mild hunger on purpose

Two thirds of all logged meals, 65.9%, already start at weak hunger. That's where most people eat most of the time, and it's the safest place on the scale. The goal isn't a new regime. It's to protect the start point you already use.

  1. Don't let yourself get to 1. If you notice you're at a 2 and heading down, that's the moment to eat, not an hour later. The data says the hour costs you.
  2. Treat skipped meals as a cost, not a saving. A missed lunch doesn't disappear. It arrives at dinner as a 0.8 on the scale, and dinner goes the way the table says.
  3. After a meal you regret, eat the next one normally. Not less. Normal. The instinct to compensate is exactly what sets up the next episode. → Why it happens at night
  4. Read the number before you eat. 0 to 9. The whole method is knowing where you're starting from.

There is nothing here about eating less. If anything, the data argues for eating sooner. The decision that determines a meal's outcome is made before the first bite, and "wait longer" moves that decision in the wrong direction.

Tomorrow, if you catch yourself skipping a meal to make up for one, notice what the table says about the meal after it.

Next in the series → What changes in 14 days, and what doesn’t

How this was measured. Figures come from 1,132,266 meals logged in I feel food between 2021 and September 2026 with valid hunger and fullness readings. "Uncomfortably full" is a fullness reading of 7.5 or above; "overeating zone" is 8.5 or above. The comparisons are observational, with no control group, and describe association rather than cause. Full methodology →