Companions
Challenge
Blog
←Back to Blog
Science

How Long Does It Take to Form a Habit? Where 21 Days Came From, and What the Studies Found Instead

ILTY TeamILTY Team•November 21, 2026•15 min read

In crisis? Call or text 988 in the US, 9-8-8 in Canada, or find your country's line at findahelpline.com.

Day 22 of the 6:30 walk. You read that habits take 21 days, so by now your feet should find the shoes on their own. They don't. This morning you stood in the hallway with one shoe on, negotiating with yourself just like on day 3.

Nothing is wrong with you. The number was. So how long does it take to form a habit? The best studies found medians of about two months (59 days in one, 66 in another) among a subset of participants, individual times from 4 to 335 days, and plenty of people who still hadn't formed the habit when the study ended.

It also depends on the habit. In records from more than 30,000 gym users, the gym habit took months; in records from 3,124 hospital staff, handwashing took a couple of weeks (Buyalskaya and colleagues, 2023). The 21 days was never a research finding.

Where "21 days to form a habit" really came from

Maxwell Maltz was a plastic surgeon. The 21 days comes from the preface of his 1960 book Psycho-Cybernetics, under the subhead "Reserve judgment for 21 days." It was advice on how long to keep practicing the book's self-image exercises before judging whether they worked.

His evidence was what he had observed: surgery patients took about 21 days to get used to a new face, a phantom limb persisted for about 21 days after an amputation, and a new house took about three weeks to feel like home. So, he concluded, it takes "a minimum of about 21 days" for an old mental image to dissolve and a new one to set.

Three things got lost in the retelling:

  • It's a floor, not a deadline. "A minimum of about" is not "it takes."
  • It's about self-image, not behavior. Maltz meant adjusting to a change, not an action running on autopilot.
  • Nothing was tested. Scientific American noted in 2024 that "No formal experiment was conducted to verify this."

Gardner and Meisel, writing on a UCL behavioral science blog in 2012, traced the figure to that preface and called its basis anecdotal. The same year, Gardner, Lally and Wardle called it a myth in the British Journal of General Practice and advised doctors to tell patients to expect about 10 weeks. Who first turned a self-image floor into "21 days to form a habit," we couldn't verify, so we won't name anyone.

What is the 21/90 rule, and what is the 3-3-3 rule for habits?

Neither has a research basis.

The 21/90 rule says 21 days builds a habit and 90 days makes it a permanent lifestyle. It has no author and no study: a PubMed search in October 2026 for "21/90 rule" and "21-90 rule" returned no habit-formation papers. It reads like Maltz's number with a second stage bolted on. Some pages name a creator; we couldn't verify any of them. Both numbers fail as fixed rules, because study medians sit at 59 to 66 days and individual times run from 4 to 335.

The 3-3-3 rule for habits has no single definition: depending on the page, it's stages of 3 days, 3 weeks and 3 months, three small actions a day, or a diet or workout split. The best-known 3-3-3 rule is an unrelated anxiety grounding exercise (name three things you see, three you hear, move three parts of your body). PubMed has no habit-formation research on any habit version, and the stage version repeats the 21-day error: a fixed calendar for something that varies by person and behavior.

What Lally's 66 days does and doesn't mean

The 66 comes from one 2010 study, and it describes fewer people than most retellings suggest.

Lally, van Jaarsveld, Potts and Wardle recruited 96 volunteers. Each picked one eating, drinking or activity behavior, such as a piece of fruit with lunch, to do daily in the same context, and for 84 days recorded whether they did it and how automatic it felt. Then the numbers narrowed:

  • 96 people were recruited.
  • 82 provided enough data to analyze.
  • 62 had data a habit curve could be fitted to.
  • 39 of those showed a good fit.

The 66 days is the median of those 39, not an average across the study and not a population fact. Individual times ran from 18 to 254 days, but the study ended at day 84, so the 254 is projected from a curve, not observed. The BPS Research Digest adds that about half of the participants still weren't automatic enough to count as a habit after 84 days.

"Formed" here means a curve flattening, not a date. Researchers score how automatic a behavior feels, and Lally's team used items from Verplanken and Orbell's 12-item Self-Report Habit Index (2003). Automaticity climbs fastest early, then levels off, and a habit counted as formed when a person reached 95% of their own plateau. Ten people interviewed for a 2011 study described the early phase as "cognitively effortful," easing as automaticity rose (Lally, Wardle and Gardner, 2011). If day 22 still takes effort, you're where they were.

The limits: self-reported automaticity, a small analyzed group, an extrapolated tail. Like several later studies, Lally's had people log the behavior daily; our post on goal tracking covers ways to keep that kind of record.

How long does it take to form a habit, according to newer studies?

Weeks for some behaviors, months for others, depending on the behavior and on what the researchers counted as "formed."

StudyBehavior and sampleMeasured byTime to habit
Armitage (2005)Gym attendance; 94 new membersAttendance recordsStable within 5 weeks
Lally et al. (2010)One eating, drinking or activity behavior; 96 recruited, 82 analyzed, 39 good fitsSelf-rated automaticityMedian 66 days; 18 to 254 (upper end projected)
Kaushal and Rhodes (2015)Exercise; 111 new gym membersSelf-reported habitAt least 4 sessions a week for 6 weeks
Fournier et al. (2017)One stretch, on waking or at bedtime; 48 studentsSelf-rated automaticityAbout 106 days (morning), 154 (bedtime), projected
Keller et al. (2021)Nutrition behavior, most often water; 192 adults, 27 of 117 modeled formed a habitSelf-rated automaticityMedian 59 days among the 27 who formed one; 4 to 335 (modeled)
Buyalskaya et al. (2023)Gym visits, 30,110 users; hospital handwashing, 3,124 staffPredictability from contextGym: median 122 to 226 days; handwashing: about 9 to 10 shifts
Singh et al. (2024)Review: 20 studies, 2,601 peopleMixedMedians 59 to 66 days; individual range 4 to 335

In Keller and colleagues' 2021 randomized trial, adults aged 18 to 77 tied a nutrition behavior, most often drinking water, to a daily routine or a set time and reported on it daily for 84 days. Only 23% of those whose curves could be modeled formed a habit in that time; for them, the median was 59 days. The sample skewed young and female.

The gym gave three answers. Armitage's attendance records showed stable habits within five weeks, and Kaushal and Rhodes's survey pointed to a minimum of four sessions a week for six weeks. Buyalskaya's machine-learning analysis found a median of four to seven months across more than 12 million observations, against a couple of weeks for hospital handwashing. Buyalskaya's team framed that as evidence against "the popular belief in a 'magic number' of days to develop a habit."

In our reading, the gap between five weeks and seven months is mostly about measurement. Armitage looked for a stable attendance pattern; Buyalskaya's team looked for the point where context made the next visit highly predictable. The samples differ too: Armitage followed people just joining a gym, while Buyalskaya's gym data left out members with under a year of records.

Fournier's 2017 morning-versus-bedtime result rests on 48 students, one stretch and projected times: a hint, not proof that mornings win.

Singh and colleagues' 2024 systematic review concluded the results "clearly refute the popular notion that habits can be formed in approximately 21 days." Its headline numbers need a footnote: the medians are Keller's and Lally's, the means of 106 to 154 days are Fournier's two groups, and the 4-to-335 range is Keller's. Eleven of its 20 studies were rated high risk of bias.

Gardner, Rebar and Lally (2022) found only five existing studies that met their 14 criteria for tracking real-world habit formation. The honest summary: 21 days is too short for most people, no single number is right, and the evidence for any precise figure is thin.

Why do some habits take longer than others?

Because behaviors, cues and people vary, and a few factors move the timeline.

Simpler behaviors tend to get there sooner. In Lally's data, simpler behaviors like drinking water tended to become automatic sooner than more elaborate ones like 50 sit-ups (Gardner, Lally and Wardle, 2012). Among new gym members, low complexity predicted habit growth (Kaushal and Rhodes, 2015).

Repetition does most of the work. In Lally's study, performing the behavior more consistently went with a better-fitting habit curve. In Keller's trial, "Repeated plan enactment was a key predictor for automaticity." Lally put it simply to Scientific American: "repetition is the key to making it work."

A stable cue matters; a rigid clock may not. In Keller's trial, routine cues and set-time cues worked about equally (medians of 60 and 59 days among those who formed a habit). In a field experiment with 2,508 Google employees, Beshears and colleagues (2021) found that paying people only for gym visits inside a planned daily two-hour window produced fewer visits than paying for visits at any time, during the four-week incentive and, marginally, in the four weeks after. They read that as rigid timing building weaker habits. One employer and a short follow-up, so our takeaway is modest: keep the cue stable and let the exact time flex.

Liking it helps. In Keller's trial, the people who went on to form habits had rated their behavior as more intrinsically rewarding at the start, 4.84 against 4.38 on a 1-to-6 scale. Kaushal and Rhodes found enjoyment predicted habit growth, and the 2024 review found self-chosen habits came out stronger.

Willpower doesn't look like the variable. Van der Weiden and colleagues (2020) analysed data from 146 people over 90 days, expecting self-control to matter, and found that "self-control capacity did not seem to affect the habit formation process."

The method itself (cues, if-then plans, habit stacking) is in our guide on how to build good habits. These studies all looked at building a behavior; if yours is one you're trying to stop, our page on quitting a bad habit when willpower keeps losing is the better next step.

Does missing a day break a habit?

No. In Lally's study, "Missing one opportunity to perform the behaviour did not materially affect the habit formation process." The BPS Research Digest noted that William James wrote in 1890 that a behavior had to be repeated without omission to become a habit, and that these results suggest he overestimated the cost of one miss.

Many misses are a different story. The same digest reports that "too many missed repeats of the behaviour, even if spread out over time, had a cumulative effect, reducing the maximum automaticity level that was ultimately reached." The damage isn't a reset to zero. It's a lower ceiling.

Gaps showed up in objective data too: in Buyalskaya's gym records, time since the last visit was the best predictor of attendance, and for 76% of gymgoers a longer gap meant a lower chance of going.

Weekends and holidays are the classic gap. In Lally's 2011 interviews, "Weekends and vacations temporarily disrupted performance due to absence of associated cues, but habits were reinstated on return to work." A formed habit isn't permanent; take away its cue and it can stall.

The morning after a miss is its own subject, covered in our post on how to be consistent. The timeline version: one miss costs almost nothing, a pattern of misses costs you height on the curve, and day 22 is early. Mr. Relentless would put it this way: you didn't fail at day 21, you were sold the wrong finish line. Put the other shoe on.

And if you're not sure this habit deserves two months of your mornings, start further back, with our goal-setting guide.

Frequently asked questions

Is it 21 or 66 days to form a habit? Neither, as a rule. The 21 comes from Maltz's 1960 preface about adjusting to a new self-image and was never tested. The 66 is the median for the 39 of 82 analyzed participants in Lally's 2010 study whose data fitted well, and individual times ran from 18 to 254 days.

Do people with ADHD struggle to form habits? We didn't find a study that measured habit-formation timelines in people with ADHD, so we won't give you a number. Our post on how to be consistent covers what changes with ADHD and what tends to help.

How do you know when a habit has formed? Score it the way researchers do. The four-item Self-Report Behavioural Automaticity Index asks whether a behavior is something you do automatically, do without having to consciously remember, do without thinking, and start doing before you realize you're doing it (Gardner, Abraham, Lally and de Bruijn, 2012). Our suggestion: rate yourself every week or two, and when the ratings stop climbing, you're near your own plateau, which is how the studies define "formed."

Sources and further reading

If you're on day 22 and still negotiating in the hallway, tell Mr. Relentless the one behavior you're building and the cue you're tying it to. ILTY remembers what you tell it, each conversation ends with two to four concrete action items, and the next time you talk, he asks how they went, so a missed morning comes up in the next conversation instead of going unmentioned. It isn't therapy, it isn't a crisis service, and it can't do the repetitions for you. Start the 7-day trial.

#how long does it take to form a habit#21 days to form a habit#21/90 rule#habit formation#lally 2010#habit automaticity

Share this article

ILTY Team

ILTY Team

AI Mental Health Companion

Building an AI companion that actually helps with your mental health.

Get mental health insights in your inbox

No fluff, no toxic positivity — just what actually helps.