On this page
If you have ever downloaded a habit app, you have met the number 21. It is on the onboarding screen, the milestone badge, the motivational notification on day 20. It is also wrong, and the place it came from has nothing to do with habits.
The number that replaced it — 66 days — is better sourced, and it comes with a warning from the people who produced it that almost nobody passes on.
Where 21 days came from
Researchers at UCL went looking for the source of the 21-day claim and traced it to the preface of a 1960 book: Psycho-Cybernetics, by the plastic surgeon Maxwell Maltz. Their write-up quotes him directly:
It usually requires a minimum of about 21 days to effect any perceptible change in a mental image.
Read that again, because the mismatch is the whole story. Maltz is talking about a mental image — how long it took his patients to stop being startled by their own face in the mirror after surgery. The UCL post suggests the figure may have grown out of those anecdotes, and that "habituation" got confused with "habit formation" somewhere along the way.
So the rule on your onboarding screen is a surgeon's observation about adjusting to a new nose, repeated until it sounded like science.
What the study actually found
Phillippa Lally and colleagues at UCL ran the study the 66-day figure comes from, published in 2010. Participants each picked a health-promoting diet or activity behaviour and reported every day how automatic it felt. They were tracked for 84 days.
Automaticity rose quickly at first and then levelled off — the familiar shape of a learning curve. The average time to reach that plateau was 66 days.
But the average is the least interesting number in the study. From the same write-up:
for one person it took just 18 days, and another did not get there in the 84 days
That second participant was projected to need as long as 254 days.
| Days to automaticity | |
|---|---|
| Fastest participant | 18 |
| Average | 66 |
| Slowest, projected | 254 |
A fourteen-fold spread between the fastest and the slowest. Any single number drawn from that distribution tells you about the group, not about you.
The asterisk the researchers added
This is the part that gets dropped when the 66-day figure is passed around. The UCL authors wrote:
it's unwise to attempt to assign a number to this process
and that the duration "is likely to differ depending on who you are and what you are trying to do."
The people who produced the number are telling you not to treat it as a target. Which is, of course, exactly what the apps did with it — 21 became 66 and the badge stayed.
What this means for your streak
A streak counts consecutive days. That is a property of the counter.
Automaticity — whether the behaviour still requires a decision — is a property of you, and the study above says it arrives on a timeline that varies by a factor of fourteen between people. These two things are not the same measurement, and the first is a poor proxy for the second.
The practical consequence is the failure mode everyone who has used a habit app knows. You run 40 days, miss one because you were ill, and the counter goes to zero. Nothing about your behaviour has reverted 40 days. But the number says it has, and the number is what you have been looking at every morning, so a lot of people stop entirely — not because the habit collapsed, but because the scoreboard did.
Day 41 after a missed day is still day 41 of doing the thing. The counter is the only part that disagrees.
What to do with this
Count totals, not just consecutive days. "I have done this 54 times since March" survives a bad week. "Current streak: 0" does not.
Expect two months, and do not read anything into day 21. If a behaviour has not gone automatic by three weeks, you are not failing; you are at the start of the range the research describes.
Match your expectation to the difficulty. The simple behaviours in the study settled far sooner than the demanding ones. A glass of water with breakfast and a 5k are not on the same clock, and treating them as though they are mostly produces an abandoned 5k.
Judge a month by its ratio. Twenty-two days out of thirty is a good month by any honest standard. Under streak logic, if the miss falls on day 15, it reads as two short streaks and a failure.
A note on the app I build
Disclosed interest: Streakly is mine, and I am going to be straight about where it sits in this.
It is a streak tracker. It has milestone badges at 7, 21 and 66 days — and having just spent a post explaining that 21 came from a 1960s surgery anecdote and that the researchers behind 66 warn against treating it as a number, I am not going to pretend those badges are science. They are encouragement with round numbers on them, and the 21 is there because people expect it.
What the app does do is show the other measurements alongside the streak: a twelve-week heatmap, total completions, best week, and per-habit completion rates. Those are the numbers that survive a missed day, and they are the ones worth looking at when the streak counter resets and you are deciding whether to carry on.
If I were giving you one instruction from this post, it would be: on the day your streak breaks, open the heatmap instead.
Frequently asked questions
How long does it actually take to form a habit?
In the best-known study, an average of 66 days to reach peak automaticity — but the range was enormous. One participant got there in 18 days; another had not after the full 84 days of tracking and was projected to need as long as 254. There is no single number that applies to you.
Where did the 21-day rule come from?
From the preface of Maxwell Maltz's 1960 book Psycho-Cybernetics, which says it usually takes a minimum of about 21 days to effect any perceptible change in a mental image. He was describing people adjusting to a changed self-image, not behaviours becoming automatic.
Is 66 days the real number?
It is a better number than 21, and it is still only an average. The UCL researchers who published it wrote that it is unwise to attempt to assign a number to this process, because the duration differs depending on who you are and what you are trying to do.
Does breaking a streak ruin the habit?
The research does not support treating one missed day as a reset — what it shows is that the timeline varies hugely between people and behaviours. A streak counter measures consecutive days, which is a property of the counter, not of how automatic the behaviour has become.
Why do simple habits form faster than hard ones?
In the study, participants chose their own health-promoting diet or activity behaviour, and the simpler ones settled in far sooner than the demanding ones. Drinking a glass of water with breakfast asks less of you than a run, so it needs fewer repetitions before it stops requiring a decision.