How long a habit takes to form: 66 days, not 21
In the best-known study, 96 volunteers repeated a new daily behaviour for 12 weeks; among the 39 whose data fitted the habit curve well, the median time to reach 95% of their automaticity plateau was 66 days, with individual values from 18 to 254 days. A 2024 review of 20 studies with 2,601 participants found only 4 that timed the process, with medians of 59 to 66 days, means of 106 to 154 days and individual values from 4 to 335 days. The popular claim of 21 days has no support in any of these trials.
Where the 66 days come from
The number most people have heard comes from a 2010 study by Phillippa Lally and colleagues at University College London, in the European Journal of Social Psychology. Each of 96 volunteers chose one eating, drinking or activity behaviour, such as eating a piece of fruit with lunch or running for 15 minutes before dinner, and did it once a day in the same context, for example after breakfast, for 12 weeks. Every day they recorded whether they had done it and rated how automatic it felt on the Self-Report Habit Index.
Of the 96, 82 provided enough data to analyse. The researchers fitted a curve to each person's automaticity scores across the 84 days, and the model fitted for 62 people, 39 of them well. The curve had the same shape in almost everyone: fast early gains, a slowing, then a plateau. The famous figures come from those 39, not from the whole sample: their median time to reach 95% of their own plateau was 66 days, and the individual values ran from 18 to 254 days. That is a lowest and a highest value across 39 people, not a confidence interval, and since the study ran 84 days, anything above that is projected rather than observed.
Two further findings matter more than the headline. Consistent repetition gave a better fit to the curve, and missing one opportunity did not materially affect the process. Complexity mattered less clearly than it is usually reported. Median times were 65 days for eating, 59 for drinking and 91 for exercise, so the exercise group took about 1.5 times as long, but the comparison was not significant (p = 0.33) and the study was not powered for it. The 2012 guidance paper drawn from this work puts it as a tendency: automaticity peaked more quickly for simple actions such as drinking water than for elaborate ones like 50 sit-ups.
| Study | Behaviour | Time to habit | Note |
|---|---|---|---|
| Lally et al. 2010 | Eating, drinking or activity, once a day | Median 66 days (18 to 254) | 39 good-fit curves of the 82 analysed, self-report |
| Lally et al. 2008 | Healthy eating and self-weighing | Mean 91 days (SD 55) | Participants' own estimate, reported in the 2024 review |
| Keller et al. 2021 | Healthy eating | Median 59 days (4 to 335) | Only 23% reached the habit threshold |
| Fournier et al. 2017 | Stretching in the morning | Mean 106 days | Same stretch, different time of day |
| Fournier et al. 2017 | Stretching in the evening | Mean 154 days | About 7 weeks longer than the morning group |
| Buyalskaya et al. 2023 | Going to the gym | Months | 60,277 members, over 12 million check-ins |
| Buyalskaya et al. 2023 | Hand washing at work | Weeks | 5,246 hospital workers, over 40 million opportunities |
| Gardner et al. 2012 | Any simple daily action | About 10 weeks | Advice paper for GPs based on the 2010 data |
What the newer studies add
In 2024, a team at the University of South Australia published a systematic review and meta-analysis in Healthcare. They searched 6 databases and included 20 experimental studies with 2,601 participants, mean ages 21.5 to 73.5 years. The behaviours were physical activity (8 studies), healthy eating (8), flossing (3), drinking water (2), microwaving a dishcloth (2), taking a vitamin (1) and reducing sitting time (1), over interventions lasting 3 weeks to 2 years.
Only 4 of the 20 studies reported how long habit formation took. Besides Lally's 66 days, a 2021 study of healthy eating found a median of 59 days and a range of 4 to 335, but only 23% of participants reached the authors' threshold for a habit. A 2008 study of healthy eating and self-weighing gave a mean of 91 days (SD 55), and a 2017 stretching study a mean of 106 days in the morning and 154 in the evening. Across the 12 studies that could be pooled, habit scores rose by a standardised mean difference of 0.69 (95% confidence interval 0.49 to 0.88), a moderate to large change; the effect was largest for flossing (1.11) and smaller for diet (0.57).
The review also names what predicts a stronger habit: frequent repetition, morning rather than evening practice, choosing the behaviour yourself, enjoying it, planning when and where it happens, and a preparatory habit such as laying out exercise clothes. On the 21-day figure, which the authors trace to the 1960 book Psycho-Cybernetics by Maxwell Maltz, they are blunt: the results clearly refute it.
Why the gym takes months and hand washing takes weeks
The studies above rely on people rating their own automaticity. A 2023 PNAS paper took a different route. Researchers including Katy Milkman, Angela Duckworth and Colin Camerer applied machine learning to 2 sets of objective records: more than 12 million check-ins by 60,277 members of a large gym chain, and more than 40 million hand-washing opportunities among 5,246 healthcare workers at 30 hospitals. The model learned which context variables best predicted each person's behaviour, and treated the point where it became highly predictable as habit formation.
The answer was not one number. Hand washing at work became habitual within weeks; going to the gym typically took months. The authors say this directly contradicts the belief in a magic number of days. It fits the earlier signal that complex actions take longer: a review in the American Journal of Lifestyle Medicine notes that in Lally's data physical activity took 1.5 times as long to become automatic as eating or drinking, while saying in the same sentence that the difference was not significant. One more result: gym members whose attendance was highly predictable responded less to an intervention designed to increase visits, as you would expect once a behaviour no longer depends on rewards.
How to shorten the road
A 2012 paper in the British Journal of General Practice, by Benjamin Gardner, Phillippa Lally and Jane Wardle for family doctors, turns the research into advice you can use yourself. Decide on a goal. Choose one simple action towards it that you can do every day. Plan a time and a place, tied to something that already happens: after brushing your teeth, or when the kettle boils. Do it there every time, expect around 10 weeks before it runs without much thought, and only then add the next action.
The studies sharpen this. Pick a cue that occurs once a day, because the 66-day figure came from once-daily behaviours in a fixed context. Choose the behaviour yourself; assigned habits formed less reliably in the 2024 review. Put it in the morning if you can, which was linked to stronger habits and, in the stretching study, a mean of 106 rather than 154 days. Start small, with 1 glass of water or a 10-minute walk, since simpler actions appear to become automatic sooner. Keep a record: a tick on a calendar gives the daily feedback the study participants had, and the 30-Day Habit Challenge covers the first month, about half the road for a typical daily habit. If the habit you want is more movement, our guide on daily steps gives sensible targets.
A missed day is a data point, not a verdict. In Lally's study, one skipped opportunity did not materially affect the process; it was people who missed many days who fitted the curve poorly. If you miss 2 days, do the smallest version on day 3 at the planned time and place, so the cue keeps its meaning.
Limits of the evidence and when to ask for advice
The 66-day figure deserves its fame less than its precision suggests. It is a median from the 39 people, out of 82 with usable data, whose curves fitted the model well, in a study that ended on day 84, so any value above that is a projection. Every self-report study measures how automatic an action feels, not whether it is done, and the 2024 review judged 11 of its 20 studies at high risk of bias, with only 4 reporting a time to habit. The PNAS study is large and objective but covers 2 specific behaviours in 2 specific settings. What the studies agree on is the shape: fast early gains, a long tail, wide differences between people, and months rather than 3 weeks for anything demanding.
Habits are a tool, and the behaviour you attach them to must be safe for you. If you have diabetes and take insulin or a sulfonylurea, a new daily exercise or fasting habit can change your blood sugar, so agree the plan with your care team first. The same applies to a new exercise habit if you have heart disease or have not exercised for years, and to any restrictive eating habit if you have a history of an eating disorder, where daily self-weighing can do harm. If you are under 18, pregnant, or on medication such as blood pressure tablets, check with your doctor before starting a strict daily routine. None of this argues against habits; it argues for choosing the right one.
Six steps for a new daily habit
- Decide on one goal you want to reach, and write it down
- Choose a single simple action you can do every day that moves you towards it
- Plan when and where you will do it, tied to a cue that happens once a day
- Do it in that context every time; the earliest repetitions bring the largest gains in automaticity
- Tick it off daily; one missed day does not reset the clock
- Expect around 10 weeks before it feels automatic, then add the next action
Frequently asked
Is it true that a habit takes 21 days?
No. The 21-day figure traces back to a 1960 self-help book, and the 2024 review of 20 studies with 2,601 participants states that its results clearly refute it. In the 4 of those studies that timed the process, medians ran from 59 to 66 days and means from 106 to 154 days.
Does missing a day ruin a habit?
Not according to the data. In the 2010 study, 140 single missed days across 55 participants lowered the automaticity score by an average of 0.29 points the next day, yet once the behaviour resumed the score was 0.55 points higher than before the gap. Long runs of missed days did weaken the fit to the habit curve.
Why do some habits take longer than others?
Complexity and context. In the 2010 study the median was 59 days for a drinking behaviour and 91 days for an exercise behaviour, about 1.5 times longer, although that difference was not statistically significant, and the 2023 PNAS analysis found hand washing took weeks while going to the gym took months.
Is a 30-day challenge long enough?
It is a good start and covers roughly the first 30 of a typical 66 days, when gains in automaticity are fastest. Plan to carry on for around 10 weeks in total, the timeframe the 2012 GP paper gives, and longer for exercise.
Does the time of day matter?
The evidence points to mornings. In the 2024 review, morning practice was linked to stronger habits, and in the stretching study the morning group averaged 106 days to a habit against 154 days for the evening group.
How do I know when something has become a habit?
Research uses the Self-Report Habit Index, which asks whether you do the action without thinking, would find it hard not to do, and start before you realise. When most of those statements feel true and the action survives a busy week, it has become a habit.
Sources
- Lally et al. 2010, European Journal of Social Psychology: How are habits formed. Modelling habit formation in the real world · 96 volunteers, 12 weeks, 82 analysed, 62 fitted the model, 39 a good fit; among those 39 the median time to 95% of asymptote was 66 days, lowest 18 and highest 254; behaviour type differences not significant (p = 0.33); one missed day did not materially affect the process. Publisher page answers 403 to automated requests; figures read from the published article text
- Gardner, Lally and Wardle 2012, British Journal of General Practice: Making health habitual, the psychology of habit-formation and general practice · Advice paper for GPs: around 10 weeks, a once-daily cue, simple actions such as drinking water form faster than 50 sit-ups, one missed performance does not impair the process
- Singh et al. 2024, Healthcare: Time to form a habit, a systematic review and meta-analysis of health behaviour habit formation and its determinants · 20 studies, 2,601 participants, mean ages 21.5 to 73.5; medians 59 to 66 days, means 106 to 154 days, range 4 to 335; SMD 0.69 (95% CI 0.49 to 0.88); 11 studies at high risk of bias; refutes the 21-day claim
- Buyalskaya et al. 2023, PNAS: What can machine learning teach us about habit formation? Evidence from exercise and hygiene · Over 12 million gym check-ins (60,277 members) and over 40 million hand-washing opportunities (5,246 workers, 30 hospitals); months for the gym, weeks for hand washing. A 2023 correction revised the modelled gym figure to 68 to 78 days, still months. Free full text of the journal article
- Arlinghaus and Johnston 2019, American Journal of Lifestyle Medicine: The importance of creating habits and routine · Review of routines in lifestyle medicine; reports that in the 2010 data physical activity took 1.5 times longer than eating or drinking to become automatic, and states in the same sentence that the difference was not significant
Read on
Health
How many steps a day: what the studies support
Life
Morning light and the body clock: how much you need
Health
Strength training after 40: how much is enoughTools that fit
This guide is for general information. It does not replace medical advice or an examination.