Mark ở vũ trụ khácVIMark học AI
Mark ở vũ trụ khác · Another Universe Common Myths · Habits · Tracing the source

Common Myths #5: It takes 21 days to form a habit

2166

One sentence in a 1960 preface, sixty years of self-help, and a study that finally measured it in 2010.

The claim"Do it 21 days in a row and it becomes a habit"
OriginMaxwell Maltz, Psycho-Cybernetics, 1960
Measuredmedian 66 days, from 18 to 254 days
VerdictMostly wrong, with a true core

Source: Lally, van Jaarsveld, Potts and Wardle (2010), European Journal of Social Psychology 40(6), abstract.

Re-read in 1 minute

Essence

Born in a plastic surgery clinicMaltz wrote about patients getting used to a new face. He never measured habits.
The word "minimum" fell off"A minimum of about 21 days" was retold as "21 days and you're done".
Measured: median 66 daysAnywhere from 18 to 254 days, depending on the person and the behaviour (Lally, 2010).
Fast first, slow laterThe first repetitions help the most, then the curve flattens.
66 isn't magic eitherDon't swap "21 days" for "66 days". Plan for 2–3 months, and a missed day is fine.

Sources: Maltz, Psycho-Cybernetics (1960), preface, as quoted by the UCL Health Behaviour Research blog (2012, secondary source); Lally et al. (2010), abstract.

Part 1

The question

If 21 days is enough, why does day 22 still take willpower? Because the number was never measured for habits. When someone finally measured it, the answer was three times longer and varied enormously from person to person.

1960An observationDr Maltz: patients need "a minimum of about 21 days" to adjust to a new self-image.
→
DecadesA sloganSelf-help shortens it to "21 days to form a habit".
→
2010A measurementLally follows 96 people for 84 days: median 66 days.

Sources: UCL Health Behaviour Research blog (2012, secondary source for Maltz's words); James Clear, "How long does it actually take to form a new habit?" (secondary source on how the quote was shortened); Lally et al. (2010).

Part 2

Where the myth came from

The sentence was distorted in two steps: it lost the word "minimum", and "self-image" turned into "habit".

Three retellings of one sentence. In 1960 Maltz wrote: it usually requires a minimum of about 21 days to effect any perceptible change in a mental image. The first retelling drops the word minimum. The third swaps mental image for habit: it takes 21 days to form a habit. 1960 · Maltz writes "it usually requires a minimum of about 21 days to effect any perceptible change in a mental image" observing patients Retelling 1 "it takes a minimum of about 21 days to change your self-image" "minimum" is gone Today self-image "It takes 21 days to form a habit" a different topic The original is about ADJUSTING to a change, and says "minimum".Today's version is about HABITS, and says "enough".
Nobody made up the 21. Maltz really wrote it, but he was talking about something else and was far more careful than today's slogan.Sources: the original sentence in the Psycho-Cybernetics preface (1960), quoted by the UCL Health Behaviour Research blog, "Busting the 21 days habit formation myth" (2012), secondary source; how the quote was shortened: James Clear, jamesclear.com/new-habit (secondary source).
1960 · Maxwell Maltz, Psycho-CyberneticsA plastic surgeon. He saw patients take about 21 days to get used to a new face, amputees feel a "phantom limb" for about as long, and people need about three weeks in a new house before it feels like home.
The following decades · Self-help books and coursesThe line was retold countless times, getting shorter, until it became the "21-day challenge".
2009–2010 · Phillippa Lally and colleagues, UCLThe first day-by-day measurement of habit formation in real life. Median 66 days.
2012 · UCL's behavioural science blogTraces Maltz's sentence and calls its use for habits "unfounded".
2021–2024 · Later studiesA randomised trial, millions of gym visits, a systematic review: all find weeks to months, with no single number.

Sources: UCL Health Behaviour Research blog (2012); UCL press release (8/2009); Keller et al. (2021); Buyalskaya et al. (PNAS, 2023); Singh et al. (Healthcare, 2024).

Round and memorable

Three weeks. Short enough that anyone dares to start, long enough to sound serious.

A doctor's name on it

A real doctor, a real book. Few people open the preface to read the original sentence.

Sometimes it's true

Some people, with simple behaviours, really do find it easier after three weeks. They tell the story, and it spreads.
Post 1

Where the myth came from

Maltz 1960a new face"minimum" lost
Post 2

What the evidence says

96 people66 days18–254a missed day
Post 3

Work lesson

12-week planfixed contextMark's score
Part 3

What the evidence says

The original study: 96 people, 84 days

Lally et al., 2010EJSP 40(6), 998–1009
The data funnel in Lally 2010. 96 people started. 82 provided enough data. The curve model could be fitted for 62. 39 showed a good fit, and the 66-day figure comes from this group. started96 people enough data82 model fitted62 good fit39 ← the 66 days come from this group each picked one eating, drinking or activity behaviour, daily, same context
The kind of behaviour: drinking water after breakfast, eating fruit with lunch, going for a run before dinner. Every day they rated how automatic it felt on a short questionnaire.Sources: Lally et al. (2010), abstract (96 people, 84 days, same context); the 82, 62 and 39 steps come from a reproduced full text and a commentary in Frontiers in Psychology (2020), secondary sources. The example behaviours are Mark's illustrations of the study's three groups: eating, drinking, activity.

Fast first, slow later: an asymptotic curve

asymptotic curveLally 2010
Automaticity rises fast in the first days, then slows and levels off. Day 21 is still on the climb. Median in Lally's study: 66 days to reach 95 percent of the maximum. that person's maximum automaticity day 21still climbing day 66median to 95% of peak 0306090120 days automaticity
There's no "switch-on day". Each repetition adds a bit of automaticity, and each addition is smaller than the last. So day 21 usually feels easier than day 1, but it isn't the finish line.Source: Lally et al. (2010), abstract (asymptotic curve, 95% mark, median 66 days). The curve's shape is Mark's illustration, not drawn from individual data.

No single number: from 18 to 254 days

individual variationLally 2010
A day axis from 0 to 260. Fastest person: 18 days, close to the myth's 21. Median: 66 days. The study ended on day 84; every figure after that is an extrapolation of the curve. Slowest person: 254 days. after day 84: extrapolated from the curve 18fastest 21: the myth 66median 84study ends 254slowest scale: days to reach 95% of maximum automaticity, the 39 people with a good fit
The myth's true core: the fastest person needed 18 days, so "about three weeks" is real for a few people. But the slowest needed more than ten times as long, and 254 is an estimate beyond the observation period.Sources: Lally et al. (2010), abstract (18–254 days, median 66, 84-day study); on the extrapolation: The Behavioral Scientist, "What Lally et al. (2010) actually found" (secondary source).

Later studies: weeks to months

later evidence2021–2024
Compared on one day axis. Lally 2010: median 66, range 18 to 254. Keller 2021, 192 people: median 59 days. Singh 2024 review of 20 studies: medians 59 to 66 days, means 106 to 154 days, individuals from 4 to 335 days. The 21-day mark sits at the far left. 050100150200250300daysLally 2010 · 39 peoplemedian 66 · range 18–254Keller 2021 · 192 peoplemedian 59Singh 2024 · medians59–66Singh 2024 · means106–154Singh 2024 · individuals4–33521: the myth
Means are longer than medians because a few very slow people pull the number up. For a personal plan, the median (the person in the middle) is the more sensible mark.Sources: Lally et al. (2010); Keller et al., British Journal of Health Psychology (2021), abstract (192 people, median 59 days among those who formed a habit); Singh, Murphy, Maher and Smith, Healthcare 12(23):2488 (2024), abstract (20 studies, 2,601 people, 4 studies reporting timing).

Different behaviours, different times

no magic numberPNAS 2023; J Behav Med 2015
Hospital handwashing, repeated many times a day, became a habit within weeks. Going to the gym took months. For new gym members, the minimum threshold was about 4 sessions a week for 6 weeks. handwashing (health workers) weeks going to the gym months over 40 million handwashing events · over 12 million gym visits (PNAS 2023) gym, new members minimum threshold: about 4 sessions/week × 6 weeks (111 people, 2015)
Habits are counted in repetitions in the same context, not in calendar days. Handwashing repeats dozens of times a shift; the gym happens a few times a week and is far more complex.Sources: Buyalskaya, Ho, Milkman, Li, Duckworth and Camerer, PNAS (2023), from the abstract and the Caltech press release (the "weeks" and "months" summary is a secondary source); Kaushal and Rhodes, Journal of Behavioral Medicine (2015), abstract.

A missed day is fine

missing one opportunityLally 2010
A row of dots, one per day, along a rising automaticity curve. One hollow dot is a missed day. The curve barely changes and keeps rising. one missed daythe curve keeps rising Missing once "did not materially affect" the process. Dropping out often is different.
Don't count streaks. Count the share of days you did it each week. A missed day is just one hollow dot, not a return to the starting line.Source: Lally et al. (2010), abstract: "Missing one opportunity to perform the behaviour did not materially affect the habit formation process." The figure is Mark's illustration.
Part 4

Fact check: Solid, Caution, Wrong

10 things people say about habits, including variants of "21 days". Tap each line to see why.

SOLIDRepeating a behaviour in the same context makes it gradually automatic

That is how Lally (2010) designed and measured it: same behaviour, same context, every day. Keller (2021) found that tying it to an existing routine or to a fixed time worked equally well; what matters is repeating it at the planned moment.

SOLIDThe first repetitions help the most

Lally describes automaticity rising along an asymptotic curve: steep at first, slower later. The "it's getting easier" feeling after a few weeks is real.

SOLIDA missed day doesn't ruin the process

Lally (2010) abstract: missing one opportunity "did not materially affect" it. But that is evidence about a single miss, not about long breaks or frequent dropping out.

CAUTION"Actually, it's 66 days"

66 is the median of the 39 people whose data fitted well, not a figure for everyone. The range was 18–254 days. Turning "66 days" into a new slogan repeats exactly the mistake of "21 days".

CAUTION"For simple things, 21 days is enough"

Possibly, for some people: the fastest person in Lally needed 18 days, and hospital handwashing became habitual within weeks (PNAS 2023). But no study shows 21 days as a general mark.

CAUTION"Go to the gym 4 times a week for 6 weeks and it's a habit"

Kaushal and Rhodes (2015) called this the minimum threshold among 111 new gym members in a 12-week observational study. A floor, not a promise.

CAUTIONThe 66-day figure is firmly established

Small sample, self-ratings, everyone chose a different behaviour, and no exact replication yet. Later studies (Keller 2021: median 59; Singh 2024: medians 59–66) find numbers of the same size, so the general direction is fairly solid; the exact figure is not.

WRONG"Dr Maltz proved that 21 days forms a habit"

Maltz wrote about adjusting to a self-image, based on watching patients, and wrote "a minimum of about 21 days". There is no habit experiment here. (Original sentence quoted via the UCL blog 2012, secondary source.)

WRONG"After 21 days the habit runs itself, no more effort"

At the median, day 21 is still on the rising part of the curve. Many people quit right here, thinking it "isn't for them".

WRONG"Miss one day and you have to start over"

The opposite of Lally (2010). Streak counting makes a single miss look like failure, while the curve barely changes.

Sources: Lally et al. (2010); Keller et al. (2021); Kaushal and Rhodes (2015); Buyalskaya et al. (2023); Singh et al. (2024); UCL Health Behaviour Research blog (2012).

Part 5

Verdict on the myth

Scores are Mark's assessment based on the sources in Part 8, not figures from a study.

How wrong
4/5
How persistent
5/5
How harmful
2/5
The true core
  • Regular repetition makes things easier
  • The first few weeks improve the most
  • For a few people and behaviours, three weeks really is enough
What's wrong
  • The number was never measured for habits
  • It promises one mark for everyone and everything
  • It makes people give up on day 22

A gentle myth: nobody is badly hurt, but it sets the wrong expectation, so many people quit just as the habit is about to form.

More: why harm is only 2/5

"21 days" still encourages people to start and repeat, and that is the true part. The harm is in the expectation: when day 22 is still hard, people conclude "I lack willpower" or "this isn't for me", when they are simply in the middle of the curve.

The second harm comes from streak counting: miss a day and the "21-day challenge" makes you start over, so many people stop altogether. The research says a single miss barely matters.

Part 6

Work lesson: plan for 2–3 months, tie it to a fixed context

For a new habit or skill at work: keeping a work log, learning a new tool for a few minutes a day, preparing before each meeting.

A 12-week plan. Weeks 1 to 2: pick one small behaviour and tie it to a fixed context. Weeks 3 to 8: repeat, count the share of days done, and if you miss a day carry on the next. Weeks 9 to 12: ask whether you do it without thinking, and only then raise the bar. weeks 1–2 weeks 3–8 weeks 9–12 pick and tie1 small behaviour1 fixed context repeat, count the share"this week: 5 of 7 days"missed a day: carry on tomorrow, don't double up check, then raise the bar"do I do it without thinking?"not yet: keep the level a few more weeks day 21: not done, don't quit The 12-week mark (84 days) is exactly the observation period in Lally's study.
Why 2–3 months: the medians in the studies are 59–66 days, about 9–10 weeks. A 12-week plan leaves room for people slower than the middle, and is still short enough to stay motivated.Sources: Lally et al. (2010) (66 days, 84-day observation); Keller et al. (2021) and Singh et al. (2024) (59–66 days). The 12-week schedule is Mark's suggestion.
Tie it to something already fixed"After I make my morning coffee, I open my plan file for 5 minutes." Same place, same time, every day.
Small enough for your worst day5 minutes with the new tool, 3 lines of log. Raise the bar later, once it runs itself.
Missed a day: carry on the nextDon't restart the count, don't double up. Avoid missing two days in a row.
For a team: don't promise "21 days"When rolling out a new process, say up front "about a quarter to get used to it". Fewer people lose heart in week four.
Missing one opportunity to perform the behaviour did not materially affect the habit formation process.Lally, van Jaarsveld, Potts and Wardle, abstract, European Journal of Social Psychology (2010).
Part 7

Read next

Phillippa Lally et al., "How are habits formed" (2010)The original paper, short, with individual curves. Read it to see how different the curves are.
Wendy Wood, Good Habits, Bad Habits (2019)A psychologist who has studied habits for decades, writing for general readers: context matters more than willpower.
Singh et al., "Time to Form a Habit" (Healthcare, 2024)The latest review, free to read, pooling 20 studies.
Common Myths #06: "10,000 hours makes you an expert"Another round number about practice, and what the original research actually says.
Part 8

Sources

Primary sources first. Where only a secondary source was available, it is noted right next to the number.

  1. M. Maltz, Psycho-Cybernetics, Prentice-Hall, 1960, preface. Original sentence quoted via a secondary source (item 3), not checked against the printed book.
  2. Lally, van Jaarsveld, Potts and Wardle, "How are habits formed: Modelling habit formation in the real world", European Journal of Social Psychology 40(6), 998–1009, 2010. doi:10.1002/ejsp.674. onlinelibrary.wiley.com · UCL press release 8/2009: ucl.ac.uk
  3. UCL Health Behaviour Research, "Busting the 21 days habit formation myth", blog, 29 June 2012. blogs.ucl.ac.uk (secondary source for Maltz's sentence)
  4. Keller, Kwasnicka, Klaiber, Sichert, Lally and Fleig, "Habit formation following routine-based versus time-based cue planning: A randomized controlled trial", British Journal of Health Psychology, 2021. doi:10.1111/bjhp.12504.
  5. Kaushal and Rhodes, "Exercise habit formation in new gym members: a longitudinal study", Journal of Behavioral Medicine, 2015.
  6. Buyalskaya, Ho, Milkman, Li, Duckworth and Camerer, "What can machine learning teach us about habit formation? Evidence from exercise and hygiene", PNAS, 2023. PMC10151500
  7. Singh, Murphy, Maher and Smith, "Time to Form a Habit: A Systematic Review and Meta-Analysis of Health Behaviour Habit Formation and Its Determinants", Healthcare 12(23), 2488, 2024. PMC11641623
  8. Wood, Quinn and Kashy, "Habits in everyday life: Thought, emotion, and action", Journal of Personality and Social Psychology 83(6), 1281–1297, 2002 (the 43% in the game).
  9. J. Clear, "How Long Does it Actually Take to Form a New Habit?", jamesclear.com (secondary source on how the quote was shortened).
  10. Secondary sources for sample details (82, 62, 39 people; 27/31/34 choosing eating, drinking, activity): a reproduced full text of Lally (2010), and a commentary in Frontiers in Psychology (2020), doi:10.3389/fpsyg.2020.578108.