How Long Does It Take To Form A Habit?

Three things follow, and none of them is a target.

Plan in months. The studies below report group averages of roughly two to five months, with individual estimates outside that range. If you are three weeks in and it still takes effort, you are not behind. You are at week three.

Stop treating the day count as a finish line. The studies measure when the curve stops rising, not when you are allowed to stop paying attention. Lally's own conclusion is that repetition in response to a cue was enough for many people, and that "although consistency in repetition is required, the degree of consistency is not yet known." These studies do not establish a universal minimum frequency.

Pick a view that survives month three. This is the one place where a tile grid earns its keep: it shows months at a time rather than the last few days, which is the timescale the evidence actually operates on. A run of 66 filled tiles looks like something. A streak counter reading 4 does not, even when the history behind it is identical.

The demo grid keeps January through June in view. It records repetitions across months, not a deadline for forming a habit.

If you want the mechanics of how the streak number itself is calculated, that is in how streaks are counted. If you are worried about what happens when you miss a day of your two to five months, the answer is measured and it is reassuring: what one missed day does to a habit.

Before the 66 there was 21, and 21 was never about habits at all.

The figure traces back to Psycho-Cybernetics, a 1960 book by the plastic surgeon Maxwell Maltz, who observed that his patients took around three weeks to adjust psychologically to a new face. Benjamin Gardner, Phillippa Lally and Jane Wardle summarised the lineage in a 2012 review written for general practitioners: the belief that habits take 21 days to form "appears to have originated from anecdotal evidence of patients who had received plastic surgery treatment and typically adjusted psychologically to their new appearance within 21 days."

We cite Maltz through that review rather than through the book. We have not read the 1960 edition, and the sentence people quote from it circulates in several different forms.

The 66 comes from a 2010 paper by Lally, van Jaarsveld, Potts and Wardle at University College London. Ninety-six volunteers, recruited by email circular through a university, each chose one eating, drinking or activity behaviour they did not already do and could attach to a daily cue, for example eating a piece of fruit with lunch or running for 15 minutes before dinner. They performed it daily for 84 days and filled in a habit questionnaire each day.

The numbers under the headline are worth having in full.

• 82 of the 96 provided enough data to analyse. Those 82 logged on a median of 47 of the 84 days.
• A curve was fitted to 62 of them. It was a good fit for 39, which the paper puts at 48% of the analysable sample.
• Among those 39, the median time to reach 95% of their plateau was 66 days, with quartiles at 39 and 102 days and a full range of 18 to 254 days.

So 66 days is a median drawn from 39 people, not from all 96. And because the study only ran for 84 days, any estimate beyond that point is the fitted curve projecting forward, not something anyone sat and watched happen. The 254 in particular is a projection.

The plain-language version, from the same research team in that 2012 review, is the fairest one-line summary of the study that exists: automaticity "plateaued on average around 66 days after the first daily performance, although there was considerable variation across participants and behaviours."

Worth keeping in view while you read the rest:

• The volunteers were self-selected university students in one city, paid a flat fee for taking part.
• Automaticity was measured by asking people daily how automatic the behaviour felt, not by observing them.
• The 66 belongs to the subgroup whose curves fitted well. The people whose data did not fit a curve had generally performed the behaviour less often.

You will see it written that exercise takes 91 days. That number is real and the claim around it is not.

Within the 39 good-fit participants, the behaviours split into eating (10 people, median 65 days), drinking (15 people, 59 days) and exercise (13 people, 91 days). A Kruskal-Wallis comparison of those three medians came out at p = 0.328, which is to say not significant. The height of the plateau did not differ either (p = 0.708). With subgroups of 10, 15 and 13, the study was not powered to find a difference, and the authors say the negative result should not be treated as definitive.

What they do argue, carefully, is that exercise took around one and a half times as long as the other two, and that this fits the idea that more complex behaviours take longer to become automatic. Their 2012 review states it as a plain claim: automaticity "peaked more quickly for simple actions (for example, drinking water) than for more elaborate routines (for example, doing 50 sit-ups)."

That is a descriptive difference plus an interpretation, not a significant test. It is still the best reason there is to make the first version of a new habit smaller than feels worthwhile, which is its own article: start smaller than feels worthwhile.

A 2024 systematic review by Singh, Murphy, Maher and Smith pulled together 20 studies covering 2,601 people. Its conclusion lands in the same neighbourhood and widens it.

• Only 4 of the 20 studies reported how long formation took at all.
• Those four gave medians of 59 to 66 days and means of 106 to 154 days, with an individual range of 4 to 335 days.
• Pooled across 12 studies, habit scores rose from before to after by a standardised mean difference of 0.69 (95% CI 0.49 to 0.88).
• 11 of the 20 studies were rated high risk of bias, and most had fewer than 100 participants.

The gap between those medians and those means is the shape of the thing. A median near two months with a mean near four means a long tail of people for whom it took much longer, and they are not outliers to be trimmed off.

The one preregistered randomised trial in this area points the same way, and it wants the same reading as the 66. Keller and colleagues enrolled 192 German adults, aged 18 to 77, and put them through 84 days of daily questionnaires. Among the people who successfully formed a habit, peak automaticity arrived after a median of 59 days.

That 59 belongs to 27 people: the ones whose fitted curve levelled off above the middle of the automaticity scale. The paper's own denominator is the 117 participants who followed the protocol, which makes it 23% of them, out of 192 who enrolled. The age range is doing some work as well. The mean age was 24.76 and 86.5% of the sample were women.

Same order of magnitude as Lally, different country, different behaviour, better design, and another median resting on a couple of dozen people.

• Lally, P., van Jaarsveld, C. H. M., Potts, H. W. W., & Wardle, J. (2010). How are habits formed: Modelling habit formation in the real world. European Journal of Social Psychology, 40(6), 998-1009. 10.1002/ejsp.674
• Gardner, B., Lally, P., & Wardle, J. (2012). Making health habitual: the psychology of 'habit-formation' and general practice. British Journal of General Practice, 62(605), 664-666. 10.3399/bjgp12X659466
• Singh, B., Murphy, A., Maher, C., & 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. 10.3390/healthcare12232488
• Keller, J., Kwasnicka, D., Klaiber, P., Sichert, L., Lally, P., & Fleig, L. (2021). Habit formation following routine-based versus time-based cue planning: A randomized controlled trial. British Journal of Health Psychology, 26(3), 807-824. 10.1111/bjhp.12504
• Maltz, M. (1960). Psycho-Cybernetics. Prentice Hall. Cited here only as the origin of the 21-day figure, by way of Gardner, Lally & Wardle (2012). We have not consulted the book directly.