How long it really takes to build a habit
You are on day 25 and it still takes a decision every time. Nothing is wrong with you. The number you were given came from somewhere else entirely.
Where "21 days" came from
In 1960 a plastic surgeon named Maxwell Maltz published a book about self-image. In it he mentioned something he had noticed in his practice: patients seemed to need about three weeks to stop seeing a stranger in the mirror after surgery, and people who lost a limb tended to feel the phantom of it for roughly the same period. He wrote that it takes a minimum of about 21 days for an old mental image to dissolve and a new one to set.
The book sold in the millions. Over the following decades the hedges fell away one by one. "A minimum of about 21 days" became "21 days". An observation about how long it takes to get used to your own face became a rule about brushing your teeth, running, and flossing. Nobody checked, because the number was convenient: three weeks is short enough to be motivating and long enough to sound earned.
It is worth being precise about what is wrong here. Maltz was not sloppy and he did not lie. He described adaptation to a changed body and said so. What happened afterwards was that a specific claim about one thing got repeated until it became a general claim about something else.
What the research actually found
The study people should be quoting instead was run at University College London and published in 2010. Ninety-six participants each picked one new eating, drinking or activity behaviour and committed to doing it in response to a daily cue β after breakfast, when they got home from work. Every day for twelve weeks they recorded whether they had done it and how automatic it felt, and the researchers fitted a curve to each person to find where automaticity levelled off.
The median was 66 days. That is the number the headlines took, and taking it is a mistake, because the spread underneath it is the actual finding: individual participants reached their plateau anywhere between 18 and 254 days. Same study, same design, same twelve weeks of daily logging. One person was done in under three weeks and another was still climbing after eight months.
One more detail deserves mention, because it changes how you should read your own progress. The curves were shaped like a plateau, not a switch. Automaticity climbed steeply at first and then flattened, which means the difference between day 40 and day 60 is much smaller than the difference between day 1 and day 20. Most of what you get, you get early.
What decides where you land in that range
Three things move you along the 18-to-254 axis, and none of them is willpower.
- How small the behaviour is. Drinking a glass of water after breakfast became automatic fastest in the UCL data. Fifty sit-ups did not. A behaviour that costs almost nothing to perform gets repeated on the days you feel terrible, and those are the days that do the work.
- How fixed the cue is. "After I pour my morning coffee" fires at the same moment every day and needs no decision. "Sometime in the evening" is not a cue at all, and a behaviour without a cue is a behaviour you will remember at bedtime.
- How often the opportunity comes round. This one is arithmetic rather than psychology, and it is the one people forget.
That last point is worth doing on paper. Automaticity is built by repetitions, and repetitions arrive at whatever rate your schedule allows. If a behaviour needs roughly 66 of them, the calendar decides how long that takes.
| How often you do it | Calendar time |
|---|---|
| Every day | Just over two months |
| Three times a week | Around five months |
| Once a week | Well over a year |
This is the strongest practical argument for daily over weekly when you have a choice, and it explains a common frustration: the gym routine that still feels like effort in March was only ever performed thirty times. It is not lagging. It is exactly where thirty repetitions puts you.
Missing one day costs less than you think
The same researchers looked at what happened when someone skipped. A single missed day had no measurable effect on the curve. People who missed one and carried on ended up in the same place as people who did not miss at all.
What did damage the curve was a run of missed days. The mechanism is not mystical: repetitions build automaticity, and a week off is a week of no repetitions. But it also means the popular idea that one lapse resets your progress to zero is wrong, and believing it is expensive β it turns a Tuesday you were tired into a reason to abandon the whole thing.
Treat one miss as what it is: one repetition you did not get, out of dozens. The thing to protect is not the perfect record. It is the return the next day.
Stop counting days. Watch this instead.
Counting days measures the calendar, not you. Two questions tell you far more about whether a routine has taken hold.
The first: does starting still require a decision? A habit that has formed does not feel like discipline, it feels like a thing you were already doing before you noticed. When you catch yourself halfway through without having debated it, that is the plateau arriving.
The second: what does the last month look like? A completion rate over a rolling thirty days survives a bad week, which is precisely why it is more honest than a streak. Twenty-five out of thirty is a routine that lives in your life. Six out of thirty is a routine that needs redesigning, and knowing that in week five is much cheaper than discovering it in month six.
Questions people ask
Is 66 days the target I should aim for?
No. It is the median of one study of 96 people, which means half of them took longer. The useful reading is the range β 18 to 254 days β and the fact that where you land depends mostly on how small the behaviour is and how reliably the cue fires.
Does missing a day reset the clock?
No. In the UCL data a single missed day had no measurable effect on how quickly the behaviour became automatic. A run of missed days does, because automaticity is built by repetition and a week off contains none.
Why does it feel harder around week three?
Because the early novelty has worn off and the results have not arrived yet. That gap is the plateau of latent potential, and it is the single most common point at which a working routine gets abandoned.
Where these numbers come from
The formation figures here are from Lally, van Jaarsveld, Potts and Wardle, "How are habits formed: Modelling habit formation in the real world", published in the European Journal of Social Psychology in 2010.
It is worth stating the limits plainly, since this article exists to argue against a number that was repeated without them. This was one study of 96 people over twelve weeks, tracking simple eating, drinking and activity behaviours. The 66-day median is a median, not a law, and nothing in it establishes how long a complex behaviour takes.
Give it enough repetitions to work
Pick something small enough that a bad day cannot stop it, attach it to a moment that already happens, and let the reminder carry the cue while it is still new.