How Long Does It Take to Form a Habit? What the Research Shows

Nearly everyone has heard that it takes 21 days to form a habit. The number is tidy, it fits on a challenge calendar, and it is wrong. It did not come from a habit study. It came from a 1960 self-help book by a plastic surgeon who noticed that patients seemed to take about three weeks to get used to a changed face, and the observation drifted from anecdote to folk law over the next sixty years. Gardner, Lally and Wardle (2012) put it plainly in the British Journal of General Practice: the 21-day claim has no evidence base.

That matters more than a trivia correction, because the wrong number sets the wrong expectation. If you believe three weeks buys you automatic behaviour, then day 25 of still having to talk yourself into a workout feels like personal failure. It isn't. You're roughly a third of the way through the typical curve. Most people quit somewhere in that gap between what they were told and what the research actually found.

Below is what the prospective evidence says: the real medians, the enormous individual range, why exercise habits take longer than almost any other health behaviour, what speeds the process up, and where the measurements are still shaky. If you want the mirror image of this question, our post on how long it takes to break a habit covers the unlearning side, which runs on a different mechanism.

The Research: What Studies Show

Lally 2010: The Study Behind the 66-Day Number

Lally, van Jaarsveld, Potts and Wardle (2010) ran the study that most modern habit advice quotes, usually without the caveats. Ninety-six volunteers each picked one eating, drinking or activity behaviour and committed to doing it daily in the same context, for example "after breakfast" or "when I get home from work". They rated automaticity every day for 12 weeks using a short form of the Self-Report Habit Index. Eighty-two participants supplied enough data to model.

The key findings:

• The curve is asymptotic, not linear. Automaticity rose steeply at first, then flattened toward a plateau. Early repetitions bought far more automaticity than late ones.
• The median time to plateau was 66 days. That is the figure that escaped into popular culture, usually stripped of everything that follows.
• The individual range was 18 to 254 days. Two people running the same protocol could be four months apart. The average describes the group, not you.
• Complexity slowed things down. Simple behaviours like drinking a glass of water reached the plateau fastest. Multi-step behaviours, including exercise, took considerably longer.
• A single miss did not reset anything. Skipping one planned opportunity had no meaningful effect on the modelled curve. That finding alone should retire the "start over on day one" rule.

The limitation worth naming: only about half of participants produced a curve that fit the model cleanly enough to estimate a plateau. The 66-day figure describes those people. It is the best estimate available from that dataset, not a law.

Singh 2024: What 20 Studies and 2,601 People Show

The first systematic review to pool this literature arrived in late 2024. Singh and colleagues in Healthcare screened the prospective habit-formation studies and included 20, covering 2,601 participants with mean ages from 21.5 to 73.5. The behaviours ranged across physical activity (8 studies), diet (8), flossing (3), drinking water (2), vitamin taking, and reducing sedentary time.

Four studies reported a median or mean time to habit formation. The medians clustered at 59 to 66 days. The means ran higher, 106 to 154 days, which tells you the distribution has a long right tail: a minority of people take a very long time and drag the average up. Across individuals, the reported spread was 4 to 335 days.

The review also flagged which factors moved habit strength: frequency of repetition, time of day, the type of behaviour, whether the person chose the behaviour themselves, how they felt about doing it, behavioural regulation, and whether a preparatory habit was already in place. Morning practice and self-selected behaviours produced stronger habits. Those are levers you can actually pull.

Keller 2021: 192 Adults, a Median of 59 Days, and Cues That Tied

Keller and colleagues (2021) in British Journal of Health Psychology ran a randomised controlled trial with 192 adults aged 18 to 77, testing whether it's better to anchor a new behaviour to an existing routine ("after I brush my teeth") or to a clock time ("at 7am"). Participants rated automaticity daily using the Self-Report Behavioural Automaticity Index.

Two results are worth carrying away. First, among participants who successfully formed the habit, the median time to peak automaticity was 59 days, independently replicating the two-month ballpark. Second, routine-based and time-based cues both worked, and neither beat the other. That is a useful correction to the confident advice you see about habit stacking being categorically superior. What matters is that the cue is specific and that it actually fires every day, not which flavour of cue you picked.

Judah 2013: Cue Order Changes the Outcome

Judah, Gardner and Aunger (2013) in British Journal of Health Psychology followed adults building a flossing habit and found something small and very transferable. People who flossed after brushing formed stronger habits than people who flossed before, and at eight-month follow-up they still had stronger habits and flossed more often. Positive attitude toward the behaviour predicted habit formation independently of how many times it was repeated.

Two lessons for training. Attach the new behaviour after a cue that already runs on autopilot rather than before it, because the established behaviour reliably triggers the new one. And enjoyment is not decoration. How you feel about the session feeds directly into how fast it becomes automatic.

How Long Does It Take to Form an Exercise Habit?

Longer than a flossing habit, and the number that decides it is weekly frequency rather than elapsed days. Kaushal and Rhodes (2015) in Journal of Behavioral Medicine recruited 111 people who had just joined a gym and tracked them for 12 weeks with repeated surveys. Using linear mixed models, they identified a threshold: at least four exercise bouts per week, sustained for six weeks, was the minimum before habit strength climbed reliably. Below that frequency, intention kept doing all the work and automaticity never took over.

Their follow-up analysis found that habit and intention operate as parallel predictors during the transition. Early on you're running on intention, which is expensive and fragile. Later, habit carries a growing share of the load. The six-week, four-session threshold is roughly where the handover starts.

There's a reason exercise is slower than most behaviours in this literature. Taking a vitamin is one action. Training is a chain: decide, change clothes, travel or clear a space, warm up, work hard, cool down, shower. Each link can break, and the effort cost at the end of the chain is high, which is exactly the shape of behaviour that resists automation. This is also why the honest weekly minimum is higher than people expect, and why our guide on consistency over intensity argues for shorter sessions done more often rather than heroic ones done occasionally.

What Actually Speeds Habit Formation

Pull the determinants out of the studies above and a short list of levers emerges. None of them is a hack. All of them show up repeatedly in the data.

Repeat in the same context, every time

Context stability is the mechanism. A habit is a learned association between a situation and a response, so the situation has to stay put long enough for the association to form. Same place, same trigger, same slot in the day. Lally's protocol required exactly this, and it is the single design feature that separates habit-formation studies from ordinary adherence studies.

Hit a real weekly frequency

Four sessions a week for six weeks is the exercise threshold from Kaushal and Rhodes. If four is not realistic yet, shrink the session rather than the frequency. A 12-minute session four times a week builds automaticity faster than a 50-minute session twice a week, because automaticity counts repetitions, not minutes.

Consider training in the morning

Fournier and colleagues (2017) in Health Psychology gave 48 students a daily stretch to adopt for 90 days and randomised them to perform it either on waking, when cortisol is near its daily peak, or before bed, when it is near its low point. Extrapolating the automaticity curves, the morning group reached automaticity at around 106 days and the evening group at around 154 days, and circadian cortisol level at the moment of performance statistically mediated the difference. Singh's 2024 review reached the same practical conclusion from a different direction: morning practice was associated with stronger habits. Our guide to building a morning workout habit turns that into a schedule.

Put the new behaviour after an anchor, not before it

Judah's flossing result argues for sequence. Attach training to something that already happens without thought, and let the established behaviour pull the new one along. After you make coffee. After you close your laptop. After the school drop-off. The anchor needs to be reliable, not impressive.

Pick something you can stand doing

Affective judgement predicted habit formation independently of repetition count in both Judah 2013 and the Singh review. Behaviours people chose themselves produced stronger habits than behaviours assigned to them. If you hate running, running is a slower route to a training habit than something you mildly enjoy, even if running burns more calories per session. The research on gamified exercise points at the same lever from the reward side.

How Long Does It Take to Break a Habit?

People assume the two numbers are the same. They aren't, because the processes aren't symmetrical. Forming a habit builds a new cue-response link through repetition in a stable context, which is what every study on this page measured. Breaking one means weakening a link that already exists, and the evidence there points at changing the cue or installing a competing behaviour in the same slot rather than waiting out a countdown. There is no clean median for unlearning in the literature, which is exactly why calendar answers are less useful on that side. We cover the disruption evidence in detail in our post on how long it takes to break a habit.

Common Misconceptions

Misconception: "It takes 21 days"

No prospective study has ever produced that number. It came from a plastic surgeon's clinical impression about patients adjusting to a changed appearance, published in a 1960 self-help book, and Gardner, Lally and Wardle flagged the absence of evidence in 2012. Every dataset that actually tracked automaticity daily lands near two months for the median and far higher for the slower half of participants.

Misconception: "Miss one day and you start over"

Lally 2010 modelled this directly. A single missed opportunity did not meaningfully change the automaticity curve. What did slow progress was repeated missing, because automaticity accrues per repetition. The practical implication is the opposite of the streak-shaming version: one skipped session costs you one repetition, so the useful question is only whether you show up at the next scheduled slot. Our piece on the psychology of streaks covers why all-or-nothing framing backfires here.

Misconception: "Once it's a habit, motivation stops mattering"

Automaticity reduces the decision cost of starting. It does not make effort disappear, and the plateau in these studies is a plateau in automaticity ratings, not a switch that flips. Kaushal and Rhodes found habit and intention working as parallel predictors well past the six-week threshold. A mature training habit means you stop negotiating about whether to start. It does not mean the fourth set stops being hard.

Misconception: "The average applies to me"

The range in this literature is four to 335 days. That is not measurement noise, it is real between-person variation shaped by behaviour complexity, cue reliability, how much you like the activity, and how consistently life lets you repeat it. Use 66 days as a planning horizon, not as a deadline you can fail.

What the Research Suggests Going Forward

The evidence base here is better than it was a decade ago and still thinner than the confident advice built on top of it. Four caveats worth holding.

First, automaticity is measured by self-report. The Self-Report Habit Index and the Self-Report Behavioural Automaticity Index are well validated, but they ask people to rate how automatic something feels, which is not the same as an objective measure of the learned association. There is no blood test for a habit.

Second, plateau times are usually extrapolated from a fitted curve rather than observed directly. Fournier's 106 and 154 day estimates came from extrapolating past a 90-day study window. That is standard practice and it is also an assumption.

Third, the samples are small and narrow. Twenty studies and 2,601 people is a thin base for a claim repeated as often as this one, and the participants skew toward educated volunteers in high-income countries who agreed to log a behaviour daily for three months. Those are not typical conditions.

Fourth, most measured behaviours were simple. Physical activity appeared in 8 of the 20 studies in the Singh review, but much of the habit literature rests on flossing, water and vitamins. Training is a longer chain with a higher effort cost, so the exercise-specific numbers from Kaushal and Rhodes deserve more weight than the pooled median when you are planning a training routine.

The practical read for anyone starting: plan for two months before it feels automatic, aim for four short sessions a week rather than two long ones, anchor them to something that already happens reliably, put them early in the day if your schedule allows, and pick a format you can tolerate on a bad day. If you want that turned into an actual weekly structure, our roundup of the best fitness apps for consistency compares the tools built around this problem rather than around workout variety.

References

1. Lally P, van Jaarsveld CHM, Potts HWW, Wardle J. "How are habits formed: Modelling habit formation in the real world." Eur J Soc Psychol. 2010;40(6):998-1009. doi:10.1002/ejsp.674
2. Singh B, Murphy A, Maher C, Smith AE. "Time to Form a Habit: A Systematic Review and Meta-Analysis of Health Behaviour Habit Formation and Its Determinants." Healthcare (Basel). 2024;12(23):2488. doi:10.3390/healthcare12232488
3. Kaushal N, Rhodes RE. "Exercise habit formation in new gym members: a longitudinal study." J Behav Med. 2015;38(4):652-663. doi:10.1007/s10865-015-9640-7
4. Keller J, Kwasnicka D, Klaiber P, Sichert L, Lally P, Fleig L. "Habit formation following routine-based versus time-based cue planning: A randomized controlled trial." Br J Health Psychol. 2021;26(3):807-824. doi:10.1111/bjhp.12504
5. Fournier M, d'Arripe-Longueville F, Rovere C, et al. "Effects of circadian cortisol on the development of a health habit." Health Psychol. 2017;36(11):1059-1064. doi:10.1037/hea0000510
6. Judah G, Gardner B, Aunger R. "Forming a flossing habit: An exploratory study of the psychological determinants of habit formation." Br J Health Psychol. 2013;18(2):338-353. doi:10.1111/j.2044-8287.2012.02086.x
7. Gardner B, Lally P, Wardle J. "Making health habitual: the psychology of 'habit-formation' and general practice." Br J Gen Pract. 2012;62(605):664-666. doi:10.3399/bjgp12X659466