How long does it take to form a habit?

Where “21 days” came from

It came from a book about cosmetic surgery. In Psycho-Cybernetics (1960), the plastic surgeon Maxwell Maltz noted that his patients seemed to need about 21 days to get used to a new face, and that an amputee took about 21 days to stop feeling a phantom limb. He was describing how long people took to accept a change in their own body.

That observation was not a study, it was not about habits, and Maltz described it as a minimum rather than an average. It became a rule through repetition alone.

What the research actually found

In 2010, Phillippa Lally and colleagues at University College London published the study that replaced it. 96 volunteers each chose one eating, drinking, or activity behaviour and performed it daily in the same context, such as “after breakfast”, for 12 weeks. Each day they recorded whether they had done it and how automatic it felt.

Plotting automaticity over time produced a curve for each person. The time to reach 95% of that curve's plateau had a median of 66 days and a range of18 to 254 days. A systematic review published in 2024 covering 2,601 people found means running even higher, from 106 to 154 days.

So the headline number is real, and the spread around it is the more useful finding. A habit that takes you four months is not a failed habit. It is inside the normal range.

What changes your number

• How hard the action is. Simple, low-effort behaviours became automatic much faster than demanding ones. Drinking a glass of water was the quickest kind of behaviour in the data; exercise was the slowest.
• How often you repeat it. Automaticity is driven by repetitions, not by elapsed time. A habit practised three days a week needs more calendar days to accumulate the same repetitions as a daily one.
• How stable the cue is. The study asked participants to attach the behaviour to a consistent moment in the day. A cue that happens at the same point every day does most of the work of remembering for you.

How this estimate is worked out

In full, so you can disagree with it. The page starts at the study's median of 66 days and multiplies, then holds the result inside the study's own 18 to 254 day range. These multipliers are our reading of what the researchers observed, not published coefficients: there is no equation in the paper that takes a habit and returns a number.

• What you are building. Something you drink or eat ×0.75, a small daily action ×1, a thinking practice ×1.2, exercise or training ×1.5. Drinking behaviours reached automaticity fastest in the data and exercise slowest.
• Effort each time. Small ×0.85, moderate ×1, demanding ×1.3. Effort is the strongest single thing separating the fast curves from the slow ones.
• How often. Every day ×1, most days ×1.2, a few days ×1.55. Automaticity is built by repetitions, so a part-time cadence needs more calendar days to reach the same count.
• The spread. Half your estimate to double it, clipped to 18 and 254. It is a deliberately wide band, because the real distribution was deliberately wide.

Which is the honest shape of the answer: a direction with an error bar, not a date. Any page that gives you one confident number for this is telling you something the evidence does not support.

What happens when you miss a day

This is the part of the study that almost never gets quoted, and it is the most useful thing in it. The researchers specifically tested whether a missed day damaged the curve:

“Missing one opportunity to perform the behaviour did not materially affect the habit formation process.”Lally, van Jaarsveld, Potts & Wardle (2010), European Journal of Social Psychology

They also found no relationship between when a miss happened and the change in habit strength. Missing an occasion early was no worse than missing one later.

Which makes the standard habit-tracker punishment, a streak that resets to zero after one bad day, a measure of something the research says does not matter. The thing that builds the habit is your overall rate of repetition over weeks. One gap does not undo the weeks behind it.

How to use the number

Use it to set your expectations, not your deadline. If your estimate says 80 days, the useful conclusion is “this will take a season, not a fortnight”, so plan for a stretch where it still feels like effort and do not read that effort as evidence of failure.

Then stop watching the number. The only input you control is whether you do the thing today, and then again tomorrow, and again after the day you miss.