How long does it take to form a healthy eating habit?

About 10 weeks. That is how long it takes to form a healthy eating habit in the one study that measured it day by day, and the honest answer has a wide range on both sides. The 21 days you have seen on posters comes from nowhere in particular. This article is part of the guide to healthy eating habits that stick; here we look at the number, why food sits at the long end of it, and what to do with the weeks in the middle where nothing seems to happen.

Where the 21 days came from

The 21-day figure is usually traced to a 1960s book by a plastic surgeon who noticed that patients took about three weeks to get used to a new face. That is a story about adjusting, not about forming a habit, and it was never tested for eating. It survived because it is short enough to sound achievable and long enough to sound serious.

The trouble with a wrong number is not that it is wrong. It is that people who are doing everything right conclude, on day 22, that it is not working for them.

What the study found

In 2010 a research group asked 96 adults to pick one new daily habit, tied to a daily cue, and to report every day how automatic it felt. Most chose something to eat or drink, some chose an activity. The sense of automaticity rose quickly at first, then more slowly, and levelled off after 66 days on average. The range was 18 to 254 days (Lally, 2010).

Two details in that study matter more than the average. The habits that levelled off fastest were the simplest ones, a glass of water with lunch; the ones that took longest involved more effort. And missing a single day made no material difference to the curve. The advice later written for doctors says the same in plain words: one small action, the same context every day, and about 10 weeks before it runs on its own (Gardner, 2012).

Why food takes longer than flossing

A habit becomes automatic when the cue reliably produces the action. Flossing has one cue and one action. A meal has a dozen moving parts: what is in the house, who cooked, how late it is, what everyone else wants, how the day went. Every extra part is a place where the cue fails to produce the action, and every failure slows the curve.

The way round this is not more discipline. It is a smaller action with fewer parts. One taste of one plant, next to whatever you were eating anyway, has the same number of parts as the glass of water. That is the version that reaches the plateau in weeks rather than in a season.

The valley where people quit

Habit curves are steep at the start and flat later. In the first two weeks each repetition adds a lot; by week five each one adds a little, and from the inside that feels like nothing. You are still doing the thing, nothing visible has changed, and the obvious conclusion is that it has stopped working.

It has not. The flat part is what forming looks like when you are inside it. The people who get through it are usually not the most motivated; they are the ones with something to look at that goes up a little every day, a count or a line of marks, so the week has a result even when the mirror does not.

One missed day changes nothing

The fear of breaking the chain is bigger than the chain. In the data, a single missed day did not knock the habit back; it was the long gaps that did. So the useful rule is not "never miss", which nobody keeps, but "never miss twice", which anybody can.

That also says what a tracker is for. A counter that resets to zero on one missed day punishes exactly the event the data says is harmless, and the rational move after a reset is to stop. A counter that gives you a free pass on a missed day, and only asks you to get back at the next meal, is the one that gets you to week 10. The rule has its own article: never miss twice.

And how long to break one

Breaking an eating habit is not measured the same way, because you cannot count the days you did not do something you no longer notice. What the research suggests is that the old habit does not vanish; it stops being cued. The late snack goes when the cue goes, the bag in the cupboard, the screen on the sofa, and it comes back the week the cue does. So the honest answer is the same 10 weeks of building the replacement, plus whatever it takes to change the shelf.

What to count as a day

Count the smallest true thing. A day where you tasted one plant you do not usually eat is a day. Not a day where every meal was perfect, which is a standard nobody meets and which turns the count into a judgement.

Three things help the count survive the long middle:

• A fixed moment to do the counting, usually right after the last meal of the day.
• A record that only shows what you did, never what you missed.
• Something that accumulates across weeks, so a quiet week still adds to a total.

The feeling that eating well is simply hard for you, rather than slow for everyone, has its own explanation in how to stick to healthy eating.

How Project Food counts the days

Project Food is an app on your phone, for you alone or for everyone at your table. After the last meal of the day it asks what you tasted, and you tap the plants. A day with at least one plant logged counts towards the streak, a missed day uses a free pass before the streak ends, and every new plant becomes a card that stays. The stats page shows the days and the weeks side by side, so the flat stretch has something to look at. It never shows what you did not eat.

Worth remembering

• About 10 weeks on average, 18 to 254 days in the range. Not 21.
• Simple actions level off fastest. One taste is a simple action.
• The flat middle is the habit forming, not failing.
• One missed day changes nothing. Two in a row is the thing to avoid.
• Count the smallest true thing, every day, at the same moment.