How Long Does It Take to Form a Habit?
There is no single number. The most cited study, Lally and colleagues in the European Journal of Social Psychology in 2010, reported a median of 66 days to reach near-automatic performance, with a range of 18 to 254 days. It was a median, not an average, and it describes 39 of the 96 people who started. Plan in months, and stop counting days.
That is the honest answer, and the rest of this page is why the popular version of it is wrong in three specific ways that matter.
The short answer, and why “66 days” is repeated wrong
Search this question and nearly every result on the first page says the same thing: researchers at University College London found it takes an average of 66 days to form a habit, with a range of 18 to 254 days. Scientific American says it. James Clear says it. Psychology Today says it. Fitness and app blogs say it.
The number is real and the study is real. The framing around it has drifted, and three errors have hardened into consensus:
- It was a median, not an average. Those are different statistics and the difference is not pedantic here, as the section on the 2024 evidence shows.
- It is not the time to “form a habit.” It is the modeled time to reach 95 percent of an automaticity curve, based on people rating how automatic a behavior felt to them each day.
- It describes fewer than half the participants. This is the big one, and almost nobody mentions it.
None of this makes the study bad. It is a careful piece of work and its authors were considerably more cautious than the pages quoting it. The problem is entirely downstream.
What Lally and colleagues actually did
The study is Phillippa Lally, Cornelia van Jaarsveld, Henry Potts and Jane Wardle, “How are habits formed: Modelling habit formation in the real world,” published in the European Journal of Social Psychology, volume 40, issue 6, 2010, pages 998 to 1009.
Ninety-six volunteers, mostly postgraduate students, each chose one eating, drinking or activity behavior to perform daily in the same context. Something like drinking a glass of water with lunch, or walking for fifteen minutes before dinner. They did this for twelve weeks, and each day they recorded whether they had done it and completed a short questionnaire measuring how automatic the behavior felt.
The researchers then fitted a curve to each person’s automaticity scores over the 84 days, and read off how long it took each curve to reach 95 percent of its ceiling. Across the participants whose curves fitted the model well, the median was 66 days, and individual times ran from 18 to 254 days. That upper figure is itself a projection, since the study only ran for 84 days.
So far this is roughly what the popular version says, apart from median versus average. The next part is not.
The part almost nobody quotes: the 66 days describes 39 people
Ninety-six people enrolled. Fourteen dropped out. That leaves 82 with usable data, and here is what happened when the researchers tried to fit an automaticity curve to each of them:
- For 12 participants, the software could not fit the curve at all.
- For 8 participants, the fitted line was flat. Their sense of automaticity did not rise over twelve weeks of daily repetition.
- Others were excluded for poor statistical fit, for never reaching habit-level automaticity, or for projections so extreme they were not credible.
- That left 39 participants, about 48 percent of those with adequate data, for whom the curve was a good fit.
The 66-day median and the 18-to-254-day range describe those 39 people. They are the ones the model worked for.
This is not a scandal and it does not invalidate anything. It is ordinary behavioral research, honestly reported by its authors, who also noted that they were measuring the automaticity of performing a behavior rather than habit as a construct, and that the study was not powered to compare sub-groups. But it changes what you should take from the number. It is not “habits take 66 days.” It is closer to “among people whose automaticity rose in a clean, curve-shaped way, the middle case reached near-ceiling in about 66 days, and for a substantial number of people it did not go that way at all.”
If you have been treating 66 as a deadline you are behind on, that is the correction.
What a 2024 review of the wider evidence found
The Lally study is one study from 2010 with 96 participants. There is now a broader picture.
Ben Singh, Andrew Murphy, Carol Maher and Ashleigh Smith at the University of South Australia published “Time to Form a Habit: A Systematic Review and Meta-Analysis of Health Behaviour Habit Formation and Its Determinants” in Healthcare in 2024. It pulls together the studies that have measured this for health behaviors: physical activity, drinking water, flossing, diet changes, taking vitamins, reducing sitting.
Among the studies reporting a central figure, medians ran from 59 to 66 days and means ran from 106 to 154 days, with individual times spanning 4 to 335 days.
Look at the gap between those medians and those means. That is what the median-versus-average distinction is actually about. A handful of people who take the better part of a year drag the average far above the typical experience. The average is not the case you should plan around. The median is closer, and even the median hides an enormous spread.
The practical reading: for most everyday health-adjacent habits, expect somewhere in the range of two to five months before it stops feeling like a decision, and accept that your own number could sit well outside that without meaning anything is wrong.
Why the type of habit changes the answer more than you do
The single largest factor is not discipline or personality. It is what the habit is.
A behavior that takes ten seconds, always happens in the same place, and is bolted to something you already do every day will consolidate quickly. Drinking a glass of water when you sit down to lunch is that kind of behavior: the cue is fixed, the response is trivial, and you were going to sit down to lunch anyway.
A behavior that takes forty minutes, requires changing clothes, depends on weather and daylight, and has no fixed slot in your day will take far longer, if it consolidates at all. Exercise is the standard example and it consistently sits at the slow end of the evidence.
This is a design lever, not just an observation. The parts of a habit that make it fast to form are the parts you control: how small the action is, how reliable the cue is, and how consistently the context repeats. That is the practical argument behind building a habit around a cue and a floor rather than around motivation, and it is why two people can attempt what sounds like the same habit and get completely different timelines.
The finding that matters most: one missed day did not derail it
Buried in the same 2010 paper is the result that should be quoted far more than the 66 days.
Missing a single opportunity did not materially affect the habit formation process. One skipped day did not measurably change the trajectory toward automaticity.
If you have ever restarted a habit from zero because you broke a streak, sit with that for a moment. The most cited study in this field, the one everyone mines for a number, found that the missed day was not the event people treat it as. The automaticity kept building.
What does damage a habit is not the miss. It is the response to the miss: the conclusion that the attempt is contaminated, the decision to wait until Monday or the first of the month, and the fresh start that never comes. One day off is a rounding error. Three weeks off, taken because one day felt like failure, is not.
This is why the useful thing to decide in advance is not a target date but a recovery rule: what specifically you will do at the next opportunity after a miss, chosen while you are calm rather than while you are disappointed.
What to watch instead of the calendar
Counting days is a poor instrument, partly because the number you are counting toward is not real for you specifically, and partly because the count itself becomes something to fail at. Three better signals:
Does the cue still fire? The first thing to go is not the will to do the habit. It is the moment that reminds you. If the reminder is still landing and you are choosing not to act, that is a different problem than the reminder having quietly disappeared, and the four-part loop is worth walking through to tell them apart.
How fast do you come back? After a miss, did the next opportunity happen or did it slide? Return speed predicts survival far better than any streak length, and unlike a streak it does not reset to zero and punish you for one bad day.
How much deciding is involved? This is the closest thing to a real measure of automaticity you have access to. Early on, the habit costs a small negotiation every time. Later, you notice you have already started. Watching that negotiation shrink tells you more than any day count, and a tracking method built around what actually happened rather than an unbroken chain will show it to you.
Frequently asked questions
Is it 21 days or 66 days?
Neither is a rule. The 21-day figure has no research behind it at all and traces to a 1960 self-help book, which is covered in where the 21-day habit myth came from. The 66-day figure is real but is a median from one 2010 study, drawn from the 39 participants whose data fitted the model, with a range of 18 to 254 days. A broader 2024 review found medians of 59 to 66 days and means of 106 to 154 days. The honest expectation is months, with wide variation.
Why do some habits take so much longer than others?
Complexity and context. A short action with a fixed daily cue consolidates quickly. A long action that needs preparation, equipment, or a slot that moves around takes much longer. In the published evidence, simple consumption behaviors like drinking water sit at the fast end and exercise sits at the slow end. The habit’s design matters more than the person’s willpower.
Does missing a day reset the clock?
No. Lally and colleagues found that missing one opportunity did not materially affect the habit formation process. What causes real damage is the pattern that often follows a miss: deciding the attempt is spoiled and waiting for a clean restart. Do the smallest version at the next opportunity and carry on.
How do I know when a habit has actually formed?
There is no moment when it switches over. The realistic marker is that the amount of deliberation drops toward nothing: you find you have started before you have finished deciding, and skipping it would take more effort than doing it. That shift is gradual and it can partly reverse if the context changes, which is why a habit you held for a year can still disappear after a move or a new job.
Should I do a 30-day or 90-day challenge?
A fixed-length challenge is a reasonable way to start something and a poor way to keep it. It supplies structure and a decision you only have to make once, which is genuinely useful early on. The failure mode is what happens on the last day: the challenge ends and nothing replaces it. If you run one, decide before day one what the ongoing version looks like, and make that version much smaller than the challenge.