How long does it take to build a habit?

The short answer

Phillippa Lally and her team at University College London did something that sounds obvious and almost nobody had done: follow real people building a real habit for twelve weeks. Ninety-six volunteers each picked a behaviour (drinking water with lunch, walking before dinner, eating a piece of fruit) and logged it daily alongside a measure of how automatic it felt.

The result, published in 2010 in the European Journal of Social Psychology, was a median of 66 days before the behaviour reached its automaticity plateau. And, more importantly, an individual range of 18 to 254 days.

Fifteen years later, a systematic review published in Healthcare by Ben Singh and colleagues pooled 20 studies and 2,601 participants on health habit formation. The medians it found run from 59 to 66 days, closely matching Lally, but with even wider individual variability: 4 to 335 days.

Where the 21 days came from

The figure traces back to Maxwell Maltz, a plastic surgeon who published Psycho-Cybernetics in 1960. Maltz reported an observation from his practice: patients needed a minimum of about three weeks to stop being startled by their new face in the mirror, and people who had lost a limb took a similar stretch to stop sensing it.

Three things worth noticing. It was a clinical impression rather than a study. It described adapting to a change in body image rather than building habits. And it said "minimum", which is precisely the word that got lost. The book sold tens of millions of copies, the self-help industry turned a floor into a deadline, and sixty years on it's still on the cover of every three-week challenge.

Why the range is so wide

The useful question is a different one: what puts you nearer 18 days or nearer 254. There are four known factors, and you control three of them.

1. How hard the behaviour is

By far the biggest factor. In Lally's study, drinking a glass of water with lunch became automatic far faster than doing fifty sit-ups. A behaviour that costs physical or mental effort never fires by itself the way a trivial one does. Which is why the advice to start absurdly small isn't false modesty: it's what shortens the timeline.

2. Frequency

Automaticity is built out of repetitions, not calendar days. A daily habit banks seven repetitions a week; a three-times-a-week habit banks three. If you're going to do something three times a week, expect the timeline in weeks to stretch proportionally.

3. How stable the context is

A habit automates when a stable cue in your environment triggers the behaviour. Same time, same place, after the same thing. If you do it at a different moment every day, you aren't repeating one habit ninety times. You're starting ninety different habits.

4. Whether you chose it

The 2024 review found that self-selected habits produce greater habit strength than assigned ones, and that practices done in the morning tend to consolidate better. That makes sense: mornings have a more stable context and fewer surprises negotiating with you.

How automaticity rises

The shape of the curve matters more than the number

The most quoted finding from Lally is the 66. The most useful one is different: automaticity follows an asymptotic curve. It rises fast at first, then flattens towards a plateau.

Two practical consequences. First, early repetitions pay more than late ones: the jump from day 3 to day 10 is enormous, the one from day 60 to day 70 is barely perceptible. Second, the "this isn't improving any more" feeling that shows up around the second month isn't a failure signal, it's the shape of the curve.

And there's a third finding, the most reassuring of all: missing a single opportunity had no appreciable effect on the process. One lost day doesn't send you back to the start. What interrupts formation is stopping the repetitions for a long stretch, and that almost never begins with the lost day. It begins with what you tell yourself afterwards. We unpack that in the piece on streaks and when they turn on you.

Realistic timelines by habit type

With the factors above you can make a decent estimate. Treat it as an expectations guide rather than a measurement, and mainly as a way to stop judging the process too early.

What to do with this

Knowing it's 66 and not 21 only helps if it changes how you go about it. Four things change:

1. Set your review at 60 days, not 21. Almost all quitting happens before the habit has had a chance to automate. Judge yourself at three weeks and you'll judge yourself at the worst possible moment: effort still high, automaticity not yet arrived.
2. Lower the difficulty to shorten the timeline. It's the only lever that acts directly on the number. A smaller habit reaches the plateau sooner, and from the plateau you can always raise the bar.
3. Measure automaticity, not days. The right question isn't "how many days am I on?" but "how much did deciding cost me today?". When the answer has been "I didn't decide, I just did it" for a couple of weeks, you're there.
4. Get external support for exactly that stretch. The gap between motivation running out at two weeks and automaticity arriving in the second month is where nearly everyone is lost. It's six weeks of no-man's land, and it's precisely the stretch where having someone watching changes the outcome.

Frequently asked questions

---

Honest summary: plan for two months, not three weeks, and don't judge yourself before day 60. The exact number matters far less than its practical consequence, which is this: the moment almost everyone quits falls before the point where the habit would have become easy.