Sixty-six days, not twenty-one: what habit science actually says

You have probably heard that it takes 21 days to form a habit. It is a tidy number. It fits on a fridge magnet, it sells a lot of 21-day challenges, and it is almost certainly wrong. The real picture is slower, messier and, oddly, far kinder than the folklore. So here is what the research actually found, where the myth came from, and why missing a day next Tuesday does not mean you have to start again.

Where the 21 days came from

The 21-day rule traces back to a plastic surgeon. In 1960, Maxwell Maltz published Psycho-Cybernetics, a self-help book that went on to sell tens of millions of copies. Maltz had noticed that his patients seemed to take about three weeks to get used to a new face after surgery, and that people who had a limb amputated tended to feel the "phantom limb" for roughly the same stretch. From this he drew a careful, hedged inference. His actual words were that "it usually requires a minimum of about 21 days for an old mental image to dissolve and a new one to jell."

Read that again, because two words did most of the damage on the way to the fridge magnet. Minimum. About. Maltz was describing how long it takes someone to adjust to a changed body, and he offered three weeks as a rough floor, not a deadline. He was not talking about habits at all. Over the following decades the self-help industry quietly deleted the hedges, swapped "mental image" for "habit", and turned a surgeon's field note into a supposed law of behaviour. Colin Camerer, a behavioural economist who has since studied habits at scale, puts it bluntly: you may have heard it takes about 21 days to form a habit, but that estimate "was not based on any science." It is a good story about how a caveat gets lost in translation.

What the Lally study actually found

The number worth knowing comes from a 2010 study led by Phillippa Lally, then at University College London. Her team recruited 96 volunteers, mostly postgraduate students with a mean age of 27, and asked each to pick one new healthy behaviour and repeat it every day for 12 weeks in response to a daily cue. The behaviours were things like eating a piece of fruit with lunch, drinking a glass of water, or doing 50 sit-ups. Each day, participants logged whether they had done it and rated how automatic it felt, using a questionnaire called the Self-Report Habit Index.

The headline figure that everyone quotes is 66 days. More precisely, among the participants whose data fit the model well, the median time to reach 95 per cent of their maximum automaticity was 66 days. That is the honest, boring, useful answer: for a simple daily behaviour, expect it to take around two months, not three weeks, before it starts to feel like it runs itself.

It is worth being straight about the size of this study, because Rise would rather tell you the truth than a rounder number. Of the 96 who started, 82 provided enough data to analyse, and the clean 66-day figure comes from just 39 of them whose automaticity followed the expected curve. The study measured automaticity, which is how automatic something feels, rather than habit in a stricter laboratory sense. The authors say as much themselves. It is a small, single study, and it is the best real-world evidence we have.

The range is the real story

If you take one thing from the research, make it this: the average is almost useless on its own. In Lally's data the time to reach that automaticity plateau ranged from 18 days at the fast end to 254 days at the slow end. Some people got there in under three weeks. Others were still climbing after eight months.

There is an important caveat buried in that 254. The study only ran for 84 days, so the longest figures were not observed directly. They were projected from each person's curve using the fitted model. In plain terms, for the slowest participants the researchers were estimating where the line would eventually flatten, not watching it happen. So treat 254 as an educated extrapolation rather than a stopwatch reading. It still makes the point: the spread between people is enormous.

Behaviour type mattered too, in the direction you would expect. Among the participants whose curves fit well, those who chose an exercise behaviour reached their plateau in a median of 91 days, against 59 for drinking and 65 for eating, so exercise took roughly one and a half times as long. Lally is careful to note this difference was not statistically significant, with a Kruskal-Wallis comparison giving a p value of 0.328, and the study was not designed to test it, so hold it loosely. Still, it fits common sense and later work. Drinking a glass of water after breakfast is a small, bounded action. Fifty sit-ups is a project.

Bigger studies since then point the same way. A 2023 analysis in PNAS by Buyalskaya and colleagues applied machine learning to two large datasets, more than 12 million gym check-ins and more than 40 million hospital handwashing observations, and found no single magic number: as the authors put it, "it typically takes months to form the habit of going to the gym but weeks to develop the habit of handwashing in the hospital." A 2024 systematic review by Singh and colleagues, pooling 20 studies and 2,601 participants, reached a similar verdict, with median times clustering around 59 to 66 days but individual cases ranging all the way from 4 to 335 days. The consistent finding across all of it is not a tidier number. It is that the number depends on you and on what you are trying to do.

The kind part: one missed day is not a relapse

Here is the finding that should take the pressure off. In Lally's study, missing a single opportunity to perform the behaviour did not meaningfully dent the process. One skipped day lowered automaticity by an average of less than a third of a point on the scale, and the score recovered the next time the behaviour was done. Just as usefully, it did not matter whether the miss happened early on, when the habit was fragile, or later, when it was more established. The researchers found no relationship between the timing of a missed day and the size of the effect: a missed Tuesday early in the process was no more damaging than a missed Tuesday in week ten.

There is a limit to this generosity, and it is worth naming. The reassuring result was about single, occasional misses, and it rested on a fairly small number of such moments, so the authors themselves called for more work on it. Other research they cite suggests that missing a whole week's worth of repetitions is a different matter and does set you back. So the honest summary is not "skip whenever you like". It is closer to this: one missed day is noise, and treating it as failure is the actual risk. The Surrey researchers who have carried this work forward make exactly this point, that the danger is a single slip making you feel you have blown it, so that you abandon the whole thing over one sugary breakfast.

What this means if you are trying to build one

The practical advice that falls out of this research is refreshingly dull, which is usually a good sign. The 2012 guidance from Gardner, Lally and Wardle, written for GPs, comes down to a few things. Pick one specific action. Attach it to a cue you already meet every single day, like "after I make my morning coffee", because a stable, repeated context is what does the quiet work of turning a decision into a reflex. Keep the action small enough that you will actually do it on a bad day. Repeat it in the same place at the same point in your routine, rather than chasing variety, because consistency of context matters more than novelty. And expect it to take around ten weeks, not three.

Then, crucially, plan for the miss before it happens. You will forget a day. Everyone does. The people who succeed are not the ones with perfect records. They are the ones who treat a gap as a single data point and simply do the thing again the next day.

A short close

The truth is less catchy than 21 days. It is a range, not a rule, and it is longer than you were promised. But it is also gentler than the folklore, because it says the thing most of us most need to hear: you are not behind, and you have not failed. You are somewhere on a curve that bends slowly upwards, and one bad day does not reset it.

We built Rise.Habits around that curve rather than around streaks. It tracks how consistent you have been over weeks rather than punishing you for a single blank square, because that is what the evidence actually supports. If that is a more honest way to think about the slow work of change, you might find it a calm place to keep track.

References

Buyalskaya, A., Ho, H., Milkman, K.L., Li, X., Duckworth, A.L. and Camerer, C. (2023) 'What can machine learning teach us about habit formation? Evidence from exercise and hygiene', Proceedings of the National Academy of Sciences, 120(17), e2216115120. doi: 10.1073/pnas.2216115120.

Gardner, B., Lally, P. and Wardle, J. (2012) 'Making health habitual: the psychology of "habit-formation" and general practice', British Journal of General Practice, 62(605), pp. 664-666. doi: 10.3399/bjgp12X659466.

Gardner, B., Rebar, A.L., de Wit, S. and Lally, P. (2024) 'What is habit and how can it be used to change real-world behaviour? Narrowing the theory-reality gap', Social and Personality Psychology Compass, 18(6), e12975. doi: 10.1111/spc3.12975.

Lally, P., van Jaarsveld, C.H.M., Potts, H.W.W. and Wardle, J. (2010) 'How are habits formed: Modelling habit formation in the real world', European Journal of Social Psychology, 40(6), pp. 998-1009. doi: 10.1002/ejsp.674.

Maltz, M. (1960) Psycho-Cybernetics. Englewood Cliffs, NJ: Prentice-Hall.

Singh, B., Murphy, A., Maher, C. and Smith, A.E. (2024) 'Time to form a habit: a systematic review and meta-analysis of health behaviour habit formation and its determinants', Healthcare, 12(23), 2488. doi: 10.3390/healthcare12232488.

Wood, W. and Neal, D.T. (2016) 'Healthy through habit: interventions for initiating and maintaining health behavior change', Behavioral Science & Policy, 2(1), pp. 71-83. doi: 10.1353/bsp.2016.0008.