How long does it take to form a habit? Not 21 days

In the study that most answers to this question go back to, a new daily behaviour took a median of 66 days to become as automatic as it was going to get, and the estimates ran from 18 to 254 days depending on the person. Phillippa Lally and colleagues at University College London ran it and published it in 2010. The 21-day figure is older. It’s usually traced to a 1960 self-help book by a plastic surgeon, who was describing how long his patients took to get used to a new face.

So if you’re 3 weeks into something and it still takes effort, nothing has gone wrong. In that study, at least 3 in 4 of the people whose curves fitted well took 39 days or more.

Where does the 21-day rule come from?

It isn’t a finding. Maxwell Maltz was a plastic surgeon, and in 1960 he published Psycho-Cybernetics, a book about self-image subtitled A New Way to Get More Living Out of Life, with a preface called “The Secret of Using This Book To Change Your Life”. In that preface he writes:

It usually requires a minimum of about 21 days to effect any perceptible change in a mental image. Following plastic surgery it takes about 21 days for the average patient to get used to his new face.

He adds amputees, whose “phantom limb” he says persists for about 21 days, and people who need about 3 weeks in a new house before it feels like home. Then he draws the conclusion: these “and many other commonly observed phenomena tend to show that it requires a minimum of about 21 days for an old mental image to dissolve and a new one to jell.” Ben Gardner, writing on the blog of the UCL department that ran the habit study, traced the popular number back to this passage.

Read in full, it’s a different claim from the one people repeat. Maltz was reporting what he’d observed, in his practice and in everyday life, with no sample and nothing measured. He was describing adjustment: the time it takes for something that happened to you, a new face or a new address, to stop feeling strange. Forming a habit runs the other way. Something you chose to do starts to happen without a decision. And he called 21 days a minimum, and an approximate one. Somewhere in the retelling, “minimum of about” fell off, and a floor for one thing became a deadline for another.

To be fair to him, he did connect the two. Later in the book he writes: “Our self-image and our habits tend to go together. Change one and you will automatically change the other.” It’s easy to see how 21 days for one became 21 days for both. The book measures neither.

What the UCL study actually measured

When Lally’s team set out, they wrote that they knew of no study that had followed habit formation within individuals, in everyday life. So they built one.

They advertised to university students by email. 101 people came to a first meeting and 96 took part, mostly postgraduates, with an average age of 27. Each chose one healthy eating, drinking or exercise behaviour they didn’t already do, and tied it to something that happened once a day. The paper’s examples are “eating a piece of fruit with lunch”, “drinking a bottle of water with lunch” and “running for 15 minutes before dinner”. The instruction was to do it every day for 84 days.

Each day, participants logged on to the study website, said whether they’d done it the day before, and rated how automatic it felt. The rating used 7 of the 12 items of the Self-Report Habit Index, a questionnaire that asks how far you agree with statements like “I do automatically” and “I do without thinking”, for a score from 0 to 42.

The team then fitted a curve to each person’s ratings: a shape that rises fast at first and flattens toward a plateau. A curve like that never quite arrives, so they counted the habit as formed on the day it reached 95% of its plateau.

The numbers narrow as you go:

• 96 people started.
• 14 entered no data after day 60, leaving 82 with enough to analyse.
• The curve could be fitted for 62 of those, and fitted well for 39.

The 66 days is the median of those 39. The paper notes that people whose curves fitted poorly had typically done their behaviour less often.

That’s worth holding onto when you see the number quoted. It’s one study, 39 usable curves, from students who picked a healthy behaviour they wanted and were paid £30 to take part. Automaticity is self-rated, and the authors point out that the questionnaire had never been answered daily like this before. They call their good-fit sample “relatively small”. None of that makes the result wrong. It makes it a careful first measurement.

What does 66 days actually mean?

66 days, one square each. The 4 dashed squares are days 67 to 70, past the median.

66 is the middle of the 39: half reached their plateau sooner, half later. The middle half fell between 39 and 102 days. The quickest took 18. The slowest, 254.

That last figure needs a footnote. The study lasted 84 days. Any time past day 84, including the 102 at the top of the middle half and the 254 at the end of the range, is where that person’s fitted curve says they would have reached 95% of their plateau if recording had carried on. Nobody was observed on day 254. The paper is open about this. It calls 66 days the “modelled time to plateau”, and it notes that some curves were still climbing when the 84 days ran out. So 254 isn’t wrong, but it’s a projection from someone who was still getting there. Read it the way the abstract does: as “highlighting that it can take a very long time”.

Does the kind of habit matter? It seems to, though this is the thinnest part of the data. The exercise group’s median was 91 days, against 65 for eating and 59 for drinking, with 10 to 15 people in each group. The authors note the exercisers “took one and a half times longer to reach their asymptote than the other two groups”, and read it as support for the idea that more complex behaviours take longer to become automatic. They also say the difference wasn’t statistically significant and the study wasn’t large enough to test it properly. A bottle of water with lunch and a 15-minute run before dinner are probably not on the same clock. This study can’t tell you how far apart the clocks are.

Does missing a day break a habit?

The paper tested this directly. It cites William James, who argued in 1890 that forming a habit needs uninterrupted performance, and points out that if a single miss were catastrophic, it would be very hard to help anyone build a healthy habit.

The team defined a missed opportunity as a day the behaviour wasn’t done, straight after 3 days when it was. They found 140 across 55 people. On the day after a miss, the automaticity score was on average 0.29 points lower, on a scale that runs to 42. Where people did the behaviour again the next day, their score had still risen across the gap, by 0.55 points, which the authors call comparable to the 0.79-point rise over 3 days with no miss.

The abstract puts it in one sentence: “Missing one opportunity to perform the behaviour did not materially affect the habit formation process.” It didn’t seem to matter whether the miss came early or late either, though the authors say there were relatively few cases to test that on.

The finding has limits. It’s about one missed day, not a missed week. The paper contrasts its result with an earlier study of exercise habits in which a lapse meant missing a whole week, and concludes that “missing one opportunity does not preclude habit formation, but missing a week’s worth of opportunities reduces the likelihood of future performance and hinders habit acquisition.” And consistency still counted overall: people who did their behaviour more consistently had curves that fitted the habit pattern better.

What to do with this

None of this calls for a new system. A few small adjustments follow from it.

Tie it to a moment, not a time. Lally’s volunteers attached their behaviour to a situation that came round once a day, like “after breakfast”, rather than to a clock time. The paper gives the reason: a situation can prompt you, while a time has to be watched for. Written down as “after X, I’ll do Y”, that’s an if-then plan, which has research of its own behind it.

Expect a dull middle. On the curve the study fitted, the early days add the most automaticity and each one after adds a little less. A few weeks in, it can feel as if nothing is changing, when the change has only got smaller. The paper’s closing paragraph is honest about the cost: new habits “will require self-control to be maintained for a significant period” before they run on their own. If the thing is effortful, like that run before dinner, plan for the longer end, or shrink it until it’s easy enough to do on a bad day.

Count the rate, not the run. One missed day cost the volunteers almost nothing. A missed week, on the evidence the paper cites, is a different matter. So the number worth keeping is how often you did it, 5 of the last 7, say, rather than how many days in a row. An unbroken count turns one ordinary miss into a reset, which is the opposite of what the data showed. There’s more on that in why streaks make you quit.

Maltz’s preface asks its readers to “reserve critical judgment for at least three weeks”. For a new habit, hold off longer than that. Pick the moment it hangs on, and if you want a date to take stock, put it at day 66, not day 21. Even then, you’ll only be at the middle.

Sources

• Gardner, B. (2012). Busting the 21 days habit formation myth. Health Chatter, UCL Research Department of Behavioural Science and Health blog. https://blogs.ucl.ac.uk/bsh/2012/06/29/busting-the-21-days-habit-formation-myth/
• Lally, P., van Jaarsveld, C. H. M., Potts, H. W. W., & Wardle, J. (2010). How are habits formed: Modelling habit formation in the real world. European Journal of Social Psychology, 40(6), 998–1009. https://doi.org/10.1002/ejsp.674
• Maltz, M. (1960). Psycho-Cybernetics: A New Way to Get More Living Out of Life. Prentice-Hall. Quoted from the Wilshire Book Company printing (1965), published by arrangement with Prentice-Hall.