The short answer: a median of 66 days, with real people landing anywhere between 18 and 254 days. The popular answer — 21 days — is not a research finding at all. It comes from a 1960 book by a plastic surgeon.
Where the 21-day myth came from
Maxwell Maltz was a cosmetic surgeon who noticed that patients seemed to need about three weeks to get used to a new face, and that amputees reported roughly three weeks of phantom-limb sensation before adjusting. He wrote this down in Psycho-Cybernetics as a rough clinical observation: it takes a minimum of about 21 days for a mental image to change.
That is an observation about adjusting to a change in self-image, made by one clinician, about surgical patients. It was never a study of habit formation. Over sixty years of repetition sanded off the qualifier — the “minimum of about” disappeared and the number hardened into a rule.
What the research actually found
The most cited real measurement comes from a study by Phillippa Lally and colleagues at University College London, published in the European Journal of Social Psychology. Participants each picked a simple daily habit — a glass of water with lunch, a piece of fruit, a short run — and logged every day for twelve weeks, rating how automatic the behaviour felt.
Plotting automaticity over time produced a curve that rises steeply and then flattens. The point where it plateaus is what “formed” means in practice: the behaviour no longer requires a decision. Three findings matter for you:
- The median was 66 days. Not the average of a tight cluster — the middle of a very wide spread.
- The range was roughly 18 to 254 days. Some habits locked in faster than the 21-day myth promises. Others took the better part of a year.
- Missing a single day did not measurably hurt. One lapse did not reset the curve or meaningfully slow it down.
Why the range matters more than the median
Sixty-six days is a fine trivia answer and a bad planning number. The useful finding is the spread, because the spread is mostly explained by what you chose to build.
Complexity drives the curve. Drinking a glass of water after lunch automates quickly: it is one action, it takes seconds, and the cue is unmissable. Fifty sit-ups before breakfast automates slowly, because every repetition still costs effort and effort never becomes invisible. In the study, the easy habits clustered near the fast end and the effortful ones stretched out toward the long tail.
Cue reliability drives it too. A habit anchored to something that happens every single day at the same time gets a clean repetition every day. A habit anchored to something irregular — “when I get home” on a schedule that varies by three hours — gets fewer real repetitions per calendar week, so the calendar count runs ahead of the actual practice count.
What to do with the number
Stop counting down to a finish line
The most common failure mode is treating day 21 or day 66 as graduation — the point where you stop paying attention. There is no day where a habit becomes permanent and maintenance-free. The curve flattens; it does not lock. If you built a gym habit over four months and then travel for three weeks, you come back with a weaker habit than you left with.
Count repetitions, not days
Automaticity is built by repetition in a consistent context, so a three-times-a-week habit needs more than twice the calendar time of a daily one to reach the same number of reps. If you want a habit to form quickly, daily beats intense. A ten-minute daily walk will automate long before a punishing Saturday hike.
Size the habit for the timeline you can survive
If a realistic estimate is two to eight months of deliberate effort, the version of the habit you pick has to be one you would still do in month three on a bad week. This is the actual argument for starting absurdly small — not that small habits are magic, but that the formation window is long enough that anything ambitious gets abandoned before it automates. See how to build a habit that sticks for how to size it.
Plan the miss in advance
The research is reassuring here and most people get it backwards: one missed day did not derail habit formation. What derails it is the response to the miss — the abandoned tracker, the “I've already ruined it” reasoning, the second and third miss that follow. Decide your recovery rule while things are going well, not in the moment. More on that in getting back on track after breaking a streak.
The honest summary
Nobody can tell you how long your habit will take, because the answer depends on how complex the behaviour is, how reliable your cue is, and how many real repetitions you get per week. What the evidence does support: expect months rather than weeks, choose something small enough to survive that long, get a repetition most days rather than a heroic one occasionally, and treat a single miss as noise.
The part almost nobody solves alone is honest measurement across that many weeks. Self-reported tracking drifts — days get marked done from memory, and the record slowly stops describing reality. That is the specific problem Amber was built for: each check-in is a photo, verified by a friend or by AI, so a 66-day record means 66 days that actually happened.