Goal Setting

How Long Does It Take to Form a Habit? What Studies Show

By iDel Published

The best-known measurement says a new daily habit took a median of 66 days to reach its plateau of automaticity, and that individuals ranged from 18 to 254 days. That comes from a 2010 study by Phillippa Lally and colleagues at University College London. The popular “21 days” figure has no such study behind it. Later research agrees with the shape of the answer: a 2024 systematic review puts most health habits at two to five months, a large field study found gym habits took a median of about four to seven months against a couple of weeks for hospital handwashing, and individual times varied widely.

Quick answer

QuestionShort answerSource
Typical time to automaticity (one daily health behaviour)Median 66 daysLally et al. 2010
Range between individuals18 to 254 daysLally et al. 2010
Systematic review of health-behaviour studiesMedians of 59 and 66 days in the studies that timed it; 2 to 5 months for most behaviours; individuals 4 to 335 daysSingh et al. 2024
Gym attendance (field data)Median of about 4 to 7 monthsBuyalskaya et al. 2023
Handwashing in a hospital (field data)A couple of weeks (9 to 10 shifts)Buyalskaya et al. 2023
Effect of missing one dayNo material effect on the processLally et al. 2010
Origin of “21 days”Usually traced to a 1960 self-image book by a plastic surgeonMaltz, Psycho-Cybernetics

What did the 2010 study measure?

Lally, van Jaarsveld, Potts and Wardle published “How are habits formed: Modelling habit formation in the real world” in the European Journal of Social Psychology (volume 40, pages 998 to 1009). The abstract describes the design:

  • 96 volunteers each chose one eating, drinking or activity behaviour to carry out daily, in the same context, for 12 weeks. An example context in the paper is “after breakfast”.
  • Each day they completed the Self-Report Habit Index (SRHI), a standard questionnaire on habit strength, and recorded whether they had carried out the behaviour.
  • 82 of the 96 supplied enough data for analysis. Automaticity was calculated from a subset of the SRHI items; the British Psychological Society’s research digest gives “I do it automatically” and “I do it without thinking” as example items.
  • The researchers fitted a curve to each person’s automaticity scores across the 84 days. The model fitted for 62 people, and for 39 of them the fit was good. People who performed the behaviour more consistently had a better model fit.

Automaticity was measured by self-report, which the authors treat as a limitation: as the BPS digest summarises it, the study depended entirely on participants’ ability to report how automatic their own behaviour felt.

How long did habits take in that study?

The abstract reports that the time to reach 95 percent of each person’s own asymptote of automaticity ranged from 18 to 254 days, which it calls “considerable variation in how long it takes people to reach their limit of automaticity.” The 66-day figure is not in the abstract. The Singh et al. review quotes it from the paper as a median (“the median time to reach 95% automaticity was 66 days (range: 18 to 254 days)”), while the BPS digest and a 2012 article by Benjamin Gardner, Lally and Wardle describe it as an average.

Three details change how the number should be read.

  1. The study ran for 84 days. Because the range reaches 254 days, the longest figures extrapolate beyond what participants were observed doing. The BPS digest calls the top of the range “a predicted 254 days”, and Ben Gardner’s UCL blog post says the slowest participant “did not get there in the 84 days, but was forecast to do so”.
  2. 66 days is the time to a plateau, not the time to perfection. The 95 percent point is where each person’s own curve is nearly flat. Automaticity rose fastest early on and then levelled off; the review summarises this as early repetitions producing larger gains than later ones, until a point where more repetition adds nothing. The plateau itself varied: Gardner notes that for some people habit strength peaked below the halfway point of the 42-point scale and for others at the very top.
  3. Only part of the sample could be modelled. The curve fitted for 62 of 82 analysed participants and fitted well for 39. The 66-day median describes people whose habits formed in a way the model could capture, which excludes the people for whom it did not.

What happens if a day is missed?

The abstract states that “missing one opportunity to perform the behaviour did not materially affect the habit formation process.” Gardner, Lally and Wardle later summarised it as automaticity gains soon resuming after one missed performance. The BPS digest adds the other half: there was “some evidence” that too many missed repetitions, even spread out over time, reduced the maximum automaticity eventually reached, although the digest also notes that the amount of data made clear conclusions about consistency hard to draw. The abstract also reports that more consistent performance went with a better model fit.

A single lapse therefore did not reset anything in this study. The evidence points to frequent gaps, not one missed day, as the pattern to avoid.

Where did “21 days” come from?

The claim is usually traced to Maxwell Maltz, a plastic surgeon, and his 1960 book Psycho-Cybernetics. In the preface (pages xiii to xiv of the 1960 edition, as quoted in Gardner’s UCL post), Maltz wrote:

It usually requires a minimum of about 21 days to effect any perceptible change in a mental image. Following plastic surgery it takes about 21 days for the average patient to get used to his new face. When an arm or leg is amputated the “phantom limb” persists for about 21 days. People must live in a new house for about three weeks before it begins to “seem like home”. These, and many other commonly observed phenomena tend to show that it requires a minimum of about 21 days for an old mental image to dissolve and a new one to jell.

The 2015 Perigee edition prints the same sentences under the heading “Reserve Judgment for 21 Days”, where Maltz asks readers to keep practising the book’s self-image techniques for a minimum of 21 days before judging them.

Read in context, the passage concerns how long a person’s mental image of themselves takes to change, and its evidence is a list of anecdotes: surgery patients, amputees, people in a new house. It names no study, and it says “a minimum”, which later retellings dropped. The book does tie the number to habits elsewhere. Maltz writes that “our self-image and our habits tend to go together” (page 108 of the 1960 edition, as quoted by Gardner), and a practice exercise in the 2015 edition asks readers to decide that “for the next 21 days you are going to form a new habit” by putting on the other shoe first and tying their laces a different way. Even there, 21 days is a practice period he prescribes, not a measured time for a habit to form.

Gardner calls the 21-day figure “unfounded as a basis for habit formation” and offers two possible routes to the myth: the distinction between habituation (getting used to something) and habit formation being “lost in translation”, or readers reasoning that if self-image takes 21 days to change and habits follow self-image, habits must take 21 days too. The Singh review notes that the timeline, widely repeated in popular culture, “has been questioned”, and concludes that its results, together with Lally’s, “clearly refute the popular notion that habits can be formed in approximately 21 days.”

Putting the two side by side:

Maltz (1960)Lally et al. (2010)
What the number describesChange in a mental image or self-imageTime for a repeated behaviour’s automaticity to plateau
EvidenceAnecdotes (surgery, amputation, a new house)96 volunteers, daily reports for 84 days
Result”A minimum of about 21 days”Median 66 days, range 18 to 254
Variation between peopleNot addressedCentral finding

What does later research show?

The 2024 systematic review

Ben Singh, Andrew Murphy, Carol Maher and Ashleigh Smith (University of South Australia) published “Time to Form a Habit: A Systematic Review and Meta-Analysis of Health Behaviour Habit Formation and Its Determinants” in Healthcare on 9 December 2024. They searched six databases for experimental studies using habit or automaticity questionnaires and included 20 studies with 2,601 participants (study mean ages from 21.5 to 73.5 years). Eleven studies had a high risk of bias.

  • Time: Only four studies reported a time to habit formation. Two gave medians: 59 days (Keller and colleagues, 2021) and 66 days (Lally 2010). The others gave means: 91 days in a 2008 study by Lally and colleagues on healthy eating and self-weighing, in which participants estimated the time themselves, and 106 days for morning and 154 days for evening stretching (Fournier and colleagues). Individual times ranged from 4 to 335 days.
  • Strength: Across 12 studies, habit scores rose from before to after the interventions (standardised mean difference 0.69, 95% confidence interval 0.49 to 0.88). The studies differed a lot from one another (I-squared of 87 percent). Flossing showed the largest effect (1.11); the single study of sedentary behaviour showed no significant effect.
  • The authors’ conclusion: habits can start forming within about two months, and the time required varies significantly between individuals; in their implications section the authors put the typical time for most health behaviours at two to five months. They name the small number of studies, with flossing and diet most represented, as a limitation.

The 59-day median comes from a 2021 trial by Keller and colleagues in which, according to the review, only 23 percent of participants reached the predetermined habit threshold. The “about two months” summary therefore describes people who did form the habit, and most participants in that trial did not.

The 2021 randomised trial of cue planning

Keller, Kwasnicka, Klaiber, Sichert, Lally and Fleig (British Journal of Health Psychology, 2021) randomly assigned 192 adults aged 18 to 77 to link an everyday nutrition behaviour of their choice either to a daily routine (such as after breakfast) or to a set clock time. Over 84 days of daily questionnaires, participants who successfully formed habits took a median of 59 days to reach peak automaticity. Both kinds of cue increased automaticity and plan enactment, with no difference found between them; the authors call the two “similarly effective”. Repeatedly carrying out the plan in response to the cue was a key predictor of automaticity.

The 2023 machine-learning study of gym and handwashing habits

Anastasia Buyalskaya, Hung Ho, Katherine Milkman, Xiaomin Li, Angela Duckworth and Colin Camerer (PNAS, April 2023) used objective records rather than questionnaires: gym attendance (over 12 million observations) and RFID-tracked hospital handwashing (over 40 million), with 30,110 gym members and 3,124 healthcare workers in the final analyses. Their Predicting Context Sensitivity method finds which context variables best predict each person’s behaviour, then dates habit formation to the point where that predictability reaches 95 percent of its plateau, the threshold Lally’s study used. Contrary to a “magic number” of days, they report that “it typically takes months to form the habit of going to the gym but weeks to develop the habit of handwashing in the hospital.” In the full paper the median was about 4 to 7 months (122 to 226 days) for gym attendance and a couple of weeks (9 to 10 shifts, or about 220 handwashing opportunities) for handwashing.

They also found that more predictable gymgoers were less responsive to an intervention designed to increase attendance, consistent with earlier experiments showing that habits are insensitive to reward devaluation; they found no such insensitivity in their other tests, including the handwashing data. As one possible explanation for the gap in timing, the authors note that going to the gym is a less frequent and more complex behaviour than handwashing. For planning, the point is that the timeline depends on the behaviour.

What helps a habit form?

Most of these findings concern how strong a habit became rather than how quickly it formed, and several rest on a single study.

FactorWhat the research foundPaper
Consistent contextRepeating the behaviour in the same context is the study design that produced automaticity; people who performed it more consistently had better model fitsLally et al. 2010
Repetition of the planRepeated enactment of the planned behaviour in response to its cue was a key predictor of automaticityKeller et al. 2021
Choice of cueA routine cue (such as after breakfast) and a time cue (such as 9:00 a.m.) were similarly effectiveKeller et al. 2021
SimplicitySimpler, repetitive behaviours with clear cues and immediate rewards (flossing, drinking water) tended to show larger gains in habit strength than complex ones (healthy eating, physical activity)Singh et al. 2024
Type of behaviourDrinking-water habits were stronger than exercise and fruit-and-vegetable habits. In the 2010 study automaticity peaked sooner for simple actions such as drinking water than for routines such as 50 sit-ups, though the review notes that the time to automaticity did not differ significantly between exercising, drinking water and eating fruitSingh et al. 2024, reporting Lally 2010 and van der Weiden et al.; Gardner, Lally and Wardle 2012
Enjoyment and planningEnjoyment, planning the time and context, and a preparatory habit (such as a daily exercise-clothes routine) mediated physical-activity habit formation in one studySingh et al. 2024, reporting Kaushal et al.
Self-chosen habitsHabits chosen by individuals themselves were stronger, in one studySingh et al. 2024, reporting van der Weiden et al.
Morning practiceStretching habits were stronger when practised in the morning than in the evening (mean 106 versus 154 days to form), in one studySingh et al. 2024, reporting Fournier et al.

Two cautions apply. The review’s authors note that differences in effect size between behaviours may reflect how effective each intervention was rather than how readily each behaviour becomes a habit. And rewards play two roles: the review lists immediate rewards among the features of behaviours that are easier to automate, while in the 2023 gym data more habitual (more predictable) gymgoers responded less to an intervention designed to increase attendance, which the authors read as consistent with reduced sensitivity to reward.

How long does it take to break a habit?

Every study above timed how long a new behaviour took to become automatic. None of them timed how long an established habit takes to fade, so they give no figure for breaking one. The Singh review’s background describes lasting behaviour change as both establishing positive habits and breaking negative ones, but its timing evidence covers only formation. What the formation research does describe is the mechanism: Gardner’s UCL post explains a habit as a mental link between a context (getting home in the evening) and a response (eating a snack) that strengthens each time the response follows the cue. In these studies that link built up over weeks to months, with wide differences between people, so a single day count for breaking a habit has no support in this evidence.

A practical plan built from the evidence

These steps apply the findings above to goal setting. They are guidance drawn from the studies, not a protocol that any of them tested.

1. Pick a behaviour small enough to repeat daily

The evidence favours starting small: in the Singh review, simpler, repetitive behaviours with clear cues showed larger gains in habit strength. A goal such as “get fit” is an outcome; a daily action such as “put on running shoes and walk for ten minutes” is a behaviour. The distinction is covered in process goals versus outcome goals. The review’s background also lists modest, attainable and time-bound goals among the strategies believed to help; setting SMART goals covers how to write one.

2. Attach it to one fixed cue

A routine cue and a clock-time cue were similarly effective in the Keller trial, so the choice can follow whichever is more stable in a given week. A shift worker with changing hours has a steadier cue in “after my first coffee” than in “7:00 a.m.”

3. Plan for a timeline measured in months

The table converts the studied day counts into weeks, so that a review date can be set in advance.

Reference pointDaysWeeks (rounded)What it represents
Fast end of Lally 2010182.6Quickest participant to reach 95% of own plateau
Median in Keller 2021598.4Among those who formed the habit
Median in Lally 2010669.4Middle participant among those whose curves could be modelled
End of observation, Lally and Keller8412Length of both trials
Gym median, Buyalskaya 2023122 to 22617.4 to 32.3About 4 to 7 months for gym attendance
Slow end of Lally 201025436.3Forecast from the fitted curve, beyond the 84 observed days

Writing for family doctors in 2012, Gardner, Lally and Wardle suggested telling patients to expect a habit built on daily repetition to take around 10 weeks. A first checkpoint at 12 weeks also matches the trials’ 84-day observation window. For a harder behaviour such as regular exercise, the gym data argue for a horizon of several months.

4. Count misses as a pattern

One missed day did not materially affect habit formation in the 2010 study. Tracking the share of days completed, rather than an unbroken streak, follows from that: a single miss had little effect, while there was some evidence that many missed days lowered the plateau. A weekly look at the share completed fits naturally into the Sunday weekly review.

5. Add enjoyment and a preparatory step

In one study the review summarises, enjoyment of the behaviour and a preparatory habit, such as a daily exercise-clothes routine, were significant mediators of physical-activity habit formation; in another, habits people chose themselves were stronger. Choosing a version of the behaviour the person enjoys, and picks for themselves, applies both findings.

6. Expect the first weeks to feel the hardest

Automaticity rose fastest in the early repetitions and then flattened toward a plateau, which for the median participant in the 2010 study came at about 66 days, or 9.4 weeks. A behaviour that still takes deliberate effort at week three is therefore well within the normal range.

What the evidence cannot say

  • Most studies rely on self-reported automaticity; the 2023 PNAS study uses objective records but infers habit from predictability, and its authors note that mechanisms other than habit could produce that predictability.
  • The behaviours that were timed were everyday health actions such as eating fruit, drinking water, stretching, going to the gym and handwashing. Complex or emotionally loaded changes were not timed.
  • The Singh review names the small number of studies as a limitation, and its own results show a high risk of bias in most of them and wide differences between their findings.
  • A median describes the middle of a distribution. In Lally’s data the middle was 66 days and the fastest participant needed 18.

Sources