Ask the internet and you'll hear 21 days. Ask the research and you'll get a different answer, one that's honestly more encouraging.
In 1960, plastic surgeon Maxwell Maltz observed that his patients took "a minimum of about 21 days" to get used to their new faces. His book Psycho-Cybernetics sold 30 million copies, the "minimum of about" got sanded off, and a myth was born [5]. No habit study was ever involved.
The best real-world study we have comes from University College London, where health psychologist Dr Phillippa Lally and colleagues tracked 96 people forming everyday habits: drinking water with lunch, running before dinner. Published in the European Journal of Social Psychology in 2010, the study found behaviours took 66 days on average to become automatic, with a huge range: 18 days for the simplest habits to a projected 254 for the hardest [1].
Sixty-six days sounds worse than 21, until you realise it means you're not failing. You're just not finished yet.
Lally's participants rated each day how automatic the behaviour felt, and the ratings traced a curve rather than a line. It rises quickly at first, when each repetition adds a lot, then flattens as the behaviour approaches its ceiling of automaticity [1]. The practical consequences:
The 18-to-254 range isn't noise; it tracks the behaviour. Drinking a glass of water at lunch has no barrier, an immediate cue and an immediate small reward, and it automated in weeks. Fifty sit-ups after morning coffee has effort, discomfort and a delayed payoff, and it was projected at many months [1]. The lesson isn't to avoid hard habits. It's that the version of a hard habit you start with should be an easy one: five sit-ups, not fifty. You are building the cue-to-action link, and the link doesn't care how many reps were attached. Once the link is automatic, raising the reps is a small change to a running system rather than a new habit from scratch.
Buried in Lally's study is the finding that matters most: "missing one opportunity to perform the behaviour did not materially affect the habit formation process" [1]. Habits aren't glass. One rough morning doesn't reset the clock. Quitting after one rough morning does.
Researchers measure habit strength with a short questionnaire developed by Bas Verplanken and Sheina Orbell, the Self-Report Habit Index [6]. You don't need the form. Its questions, in plain English, are a decent self-test:
Mostly yes: it's a habit, whatever day it is. Mostly no: keep going, and check that the cue is the same every time, because that is usually the missing piece.
Same cue, every time. Habits attach to context, what researchers call cue-consistent repetition [1]. An action performed after the same trigger daily, like an alarm going off, automates far faster than one squeezed in "whenever". As Wendy Wood's research at Duke University showed, around 43% of our daily actions already run this way: triggered by context, not decisions [2].
Start smaller than feels impressive. Stanford behaviour scientist BJ Fogg built his Tiny Habits method on this: a behaviour small enough to be easy gets repeated, and repetition, not ambition, is what wires the loop [3].
Never miss twice. The one-day grace is real (see Lally above); the two-day slide is how habits actually die. As James Clear puts it in Atomic Habits: "You do not rise to the level of your goals. You fall to the level of your systems." [4]
No. The figure comes from a 1960 book by a plastic surgeon describing how long patients took to adjust to a new face. The best real-world study, from UCL in 2010, found an average of 66 days with a range from 18 to 254 depending on the behaviour and the person.
No. The UCL study found that missing a single opportunity made no measurable difference to how the habit formed. What matters is not letting one miss become two, and then a pattern.
When you would feel odd not doing it, when you start it without deciding to, and when it happens in the same place at the same cue without much thought. Researchers measure this with a short questionnaire called the Self-Report Habit Index; those are its main questions in plain English.
Complexity and reward. Drinking a glass of water at lunch became automatic in a few weeks in the UCL study; fifty sit-ups after coffee was projected at months. Simpler, more immediately rewarding behaviours automate faster, which is why starting smaller than feels impressive works.