How long it takes, and what happens when you miss
90 min
Two hosts talk the lesson through. The voices are synthetic; the script was written from this lesson and checked against it, and asserts nothing the lesson does not.
- State what has been measured about how long habit formation takes, with its range, its scope, and the read level this course holds it at
- State what was measured about missing a single opportunity, and say why days are the wrong unit for any of this
- Trace the twenty-one-day claim to where it comes from, and say what its source actually said
This is the lesson the course was built to deliver, and the reason is one sentence in a three-page paper written for doctors. Almost nothing you will read about habits contains it.
Before that, the number you have certainly met.
What was measured, and what this course has actually read
A note about read level first, because it governs the rest of the lesson.
The study everybody quotes is Lally, van Jaarsveld, Potts and Wardle, published in the European Journal of Social Psychology in 2010.3 This course hasn't opened it. It sits behind a paywall that could not be got past, so every figure below is reported at one remove, from two open-access papers that have Phillippa Lally as an author and that describe her own study.12
That is a weaker position than quoting the paper and it is not a bad one: the reporting comes from the person who ran it. But it is not the same thing, and a course that asks you to check where claims come from should say where its own stand.
The design. Ninety-six volunteers each chose an eating, drinking or activity behaviour to do daily in the same context, and were followed for twelve weeks.12 Each day they recorded whether they had done it and rated how automatic it felt. Eighty-two provided enough data to analyse. Curves were fitted for sixty-two of them, of whom thirty-nine fitted well.2
Notice how the sample narrows. Ninety-six started, sixty-two could be modelled, thirty-nine fitted the curve. That's normal for this kind of study and it is the sort of attrition that disappears when a finding is retold.
The curve, and then the number
Automaticity didn't rise in a straight line. It rose quickly at first and then flattened, which the papers describe as "an asymptotic increase, with an initial acceleration that slowed to a plateau".1
That shape is more useful than the number, and it predicts something you can check on yourself. The first fortnight should feel like more progress than the sixth, not because you are losing momentum but because that's what the curve does.
The table below is this course's own rendering of that shape, and the two columns are not on the same footing.5 The middle column restates the measured curve. The right-hand column is what this course infers it would feel like from the inside, and nothing in the studies measured how any of it felt.
| Stage | What is happening (measured) | How it might feel (inferred) |
|---|---|---|
| The first repetitions | Largest gains in automaticity | Effortful, and noticeably easier each time |
| The middle | Gains continuing but smaller | Easier, and apparently stalling |
| Approaching the plateau | Little further change to measure | Not something you decide about any more |
Now the number. The plateau was reached after an average of 66 days, with individuals ranging from 18 days to a predicted 254.12
Sixty-six is a mean with an enormous spread around it. The fastest person in that sample got there in under three weeks. The slowest, on the model, would have taken more than eight months.
And complexity moved it. "Automaticity strength peaked more quickly for simple actions (for example, drinking water) than for more elaborate routines (for example, doing 50 sit-ups)."1
You have met 66 days somewhere. Before reading on: what do you think was attached to it in the paper that has fallen off by the time it reaches a reader, and why would that part in particular be the part that goes?
Show the answer
The range, and it goes for the same reason lesson 2's range went.
Eighteen to two hundred and fifty-four is not usable as advice. It doesn't tell a reader what to expect, it cannot be put on a slide, and it makes the finding sound less like a finding. Sixty-six does all three jobs.
There is a second thing that falls off and it is worse, because losing it changes what a reader does. Sixty-six days isn't sixty-six ticks on a calendar, and the next section is about why.
Days are the wrong unit
This is the error that survives even in careful retellings, and it is worth being exact about.
What builds a habit is the pairing: the cue occurs, the response happens, the association strengthens a little. The thing that accumulates is pairings. Elapsed time only tracks pairings if you never miss, and nobody does not miss.
The researchers say this outright in their guidance to other researchers. One of their criteria for a well-run habit study is that it "should not infer effects of repetition from measures of time".2 The reason, in this course's words rather than theirs, is that days are not repetitions: a day the behaviour did not happen is still a day, so a graph plotted against time counts it as progress.2
Worked: two people, seventy days
Both cases are constructed.5
Asha picks a cue she meets every day and does the behaviour on 64 of the next 70 days.
Ben picks the same behaviour and a cue he meets about four days in seven. He does it on 31 of the 70.
Both of them have "done seventy days". If either is counting on a calendar, they're at the same place. If you ask how many times the cue and the response have occurred together, one has had more than twice as much learning as the other.
So the honest thing Ben can say after ten weeks is not that it didn't work for him. It is that he has had about thirty-one repetitions, which the curve says is early, and that his cue is the thing to look at rather than his character. That is lesson 3's diagnosis arriving with a number attached.
And notice what this does to the 66. Even the study's own average is an average number of days in people who were asked to perform daily, which is the closest days ever get to repetitions. Outside a study, nobody's calendar is that clean.
What happens when you miss a day
The sentence that follows is the reason this lesson sits where it does in the course.
"Missing the occasional opportunity to perform the behaviour did not seriously impair the habit formation process: automaticity gains soon resumed after one missed performance."1
Read it twice. One missed day did not set the process back. Gains resumed. The curve carried on from about where it was.
Set that against what you have almost certainly been told, in an app, in a book, or by yourself: that a broken streak is a failure, that you are back at the beginning, that consistency means never missing.
None of that is in the research, and one part of it is contradicted by the one study this course has read on the question.1
So does this mean it does not matter whether you miss days? Somebody could take this lesson as permission to be irregular.
Show the answer
It would, and the finding does not say that. Two things keep it in bounds, and the second is the important one.
The first is what was actually measured. The claim is about the occasional missed opportunity and about one missed performance. Nothing here tells you missing four days in five is fine, and the research wouldn't expect it to be, because pairings are what build the association and missing means fewer pairings.
The second is that the two claims are about different things, and this is the part worth having. Frequency of repetition determines how fast the habit forms. A missed day is a lost repetition and it slows you down by roughly one repetition.
What the folk belief adds, and what is false, is that a missed day does something else as well: that it undoes accumulated progress, resets a counter, or breaks something. It does not. You're not back at the start. You are one repetition short of where you would have been.
Which is why this matters practically. The cost of a missed day is small and linear. The cost of believing you have to start again is that people stop, and stopping is the only thing here that actually costs you everything you have built.
Twenty-one days
The other number, and this is the course's worked provenance case.
Twenty-one days does not come from a study. As far as this course can establish, and it holds this at the level of a search summary because the book has not been opened,4 it comes from Psycho-Cybernetics, published in 1960 by Maxwell Maltz, a plastic surgeon who wrote a self-help book.
The sentence reported, in the preface, is that "It usually requires a minimum of about 21 days to effect any perceptible change in a mental image", and that after plastic surgery it takes about twenty-one days for the average patient to get used to a new face.
Three things about that, and each one is a step in the trace.
It's an observation, not a result. A clinician noticing a pattern in his patients, written in a preface.
It is about something else. Adjusting to a changed appearance is not forming a habit, and nothing in the sentence claims it is one.
And the qualifier is the part that vanished. "A minimum of about." Not twenty-one days, but at least around that, before there's any perceptible change. By the time it reaches you it has become a deadline rather than a floor, which is very nearly the opposite of what it said.
This is the same shape as the case Time Management traced in its last lesson, where a sentence offered as a commonplace observation in a satirical essay became a law. Different subject, and the same mechanism: a memorable sentence outliving its qualifier because the qualifier is what makes it hard to repeat.
Four things people believe about the timeline
"Twenty-one days." Traced above. A 1960 self-help book, an observation about surgical patients, and "a minimum of about" lost in transmission.
"Sixty-six days." Better, and still wrong as usually stated, and reported here at one remove from a paper this course has not opened. An average, on 96 self-selected volunteers doing self-chosen simple behaviours, with a range from 18 to a predicted 254, and simple actions plateauing sooner than elaborate ones.
"Miss a day and start again." Contradicted by the one study this course has read on it. Gains resumed.
"It gets easier every day at the same rate." The curve flattens instead. The stretch where it seems to have stalled is the stretch where you are closest to the plateau, which is an unfortunate design and it is worth expecting.
Practice
Take 20 minutes to set up, then about ten seconds a day for ten weeks.
This is modelled on a real instrument, the tick sheet in Box 1 of the British Journal of General Practice paper this course keeps quoting, which is a sheet a doctor can hand to a patient.1 The sheet's own text gives the ten weeks, and the reason this course reads into it is the one you now know: 66 days is about nine and a half weeks.5
- One behaviour, small and daily. The two examples the papers give are drinking water and doing 50 sit-ups, and the first of those is much closer to what the study's participants mostly picked.
- One cue, from lesson 3, tested against both bars: would you notice it tired, and do you meet it every day?
- A grid, seven columns and ten rows. A tick on every day you do it and nothing on days you do not. Don't leave the missed days blank and forgotten; leave them visibly empty.
- At the end of each week, one number: how automatic did it feel, from 1 for not at all to 10 for completely. That rating is the study's measure, in miniature.
At ten weeks you have two things to look at. Your curve, which is the ten weekly ratings, and your repetition count, which is the ticks. Write one sentence on each:
- Where were your largest gains? Early, if the research holds on you.
- How many repetitions did you actually get, out of seventy, and what happened to the rating in the week after each missed day?
If it is not automatic at ten weeks, you haven't failed. The range reaches 254 days. Write down where your rating got to and carry on if you want it.
Take 20 minutes.
Find one claim about how long a habit takes. An app's onboarding screen, a book, an article, a poster at work. Write it word for word, with where you found it.
Then four questions in writing.
- What number does it give, and is it presented as an average, a minimum, or a rule?
- Does it cite anything? If not, write "nothing", and you are done.
- If it cites something, is the number in what it cites? Not something like it. The number.
- What has been dropped? A range, a scope, a qualifier, a population. There is nearly always something, and naming it is the exercise.
This is the same trace Time Management lesson 8 asked for, on a different subject, and if you did it there you'll find this one faster. Lesson 7 asks you to do it once more on the whole subject.
Connections
Back. Lesson 3 gave you the cue and left open how long the association takes to build; this lesson answers it, and the answer is that time is the wrong question. Lesson 2's habit of putting a travelling number back beside its own sentence is the method for both numbers here. Lesson 1's definition is what makes the missed-day finding intelligible: the thing being built is an association, and an association is not a streak.
Forward. Lessons 5 and 6 leave habit behind and take up self-control, which is what happens on the occasions when there is no association to fall back on and a decision has to be made. Lesson 7 puts this lesson's two numbers beside lesson 2's and asks you to read the whole subject.
Go deeper
- Making health habitual: the psychology of 'habit-formation' and general practice (British Journal of General Practice, 2012). Read in full by this course, three pages, free. Every figure in the first half of this lesson is reported there, by one of the authors of the study it reports. Box 1 is the tick sheet the exercise above is modelled on, and it is worth seeing in its original form.
- How does habit form? Guidelines for tracking real-world habit formation (Cogent Psychology, 2022). Open access; this course has read it in substantial part. Written for researchers, and the reason to read it anyway is that it is a field saying plainly what its own methods cannot support. The criteria about time and repetition are in section 2.
Sources
- Benjamin Gardner, Phillippa Lally and Jane Wardle, "Making health habitual: the psychology of 'habit-formation' and general practice", British Journal of General Practice 62(605), 2012, pages 664 to 666. Read in full. Supports: the quoted description of the asymptotic increase and the average of 66 days; the quoted finding that missing the occasional opportunity did not seriously impair the process and that gains resumed after one missed performance; the quoted comparison of simple actions with more elaborate routines; and the ten-week tick sheet in Box 1.
- Benjamin Gardner, Amanda L. Rebar and Phillippa Lally, "How does habit form? Guidelines for tracking real-world habit formation", Cogent Psychology 9:1, 2022, article 2041277. Read in substantial part, open access. Supports: the 18-to-254 range; the numbers modelled and well-fitted; the quoted criterion that studies should not infer effects of repetition from measures of time; and the twelve-week follow-up and the 96 volunteers, which this source reports alongside source 1. The gloss about a missed day still being a day is this course's own and is unquoted in the body.
- Phillippa Lally, Cornelia van Jaarsveld, Henry Potts and Jane Wardle, "How are habits formed: Modelling habit formation in the real world", European Journal of Social Psychology 40(6), 2010, pages 998 to 1009. Paywalled, and this course has not opened it. It is the source of every figure in the first half of this lesson, and all of them are reported here at one remove through sources 1 and 2, both of which have Lally as an author. The body says so where the figures appear.
- Maxwell Maltz, Psycho-Cybernetics, 1960. Not read; search-summary level only, which the body
states where the claim appears. At that level: the reported sentence about "a minimum of about
21 days" to effect a perceptible change in a mental image, offered in the preface as an
observation about patients adjusting after plastic surgery rather than as a study of habits.
Nothing else in this course rests on it, and it is on the
unreadline ofresearch/SOURCES.md. - Asha and Ben are constructed, and so is every figure attached to them; the lesson says so where they appear. The three-stage table of the curve is this course's own rendering of the asymptotic shape sources 1 and 2 describe, its right-hand column is an inference about how that shape would feel rather than anything measured, and the table says both where it appears. The reading of Box 1's ten weeks as following from the 66 days is also this course's own: the sheet gives ten weeks, and the derivation is the lesson's.
Check your understanding
This lesson has a 6-question quiz. Pass it and the questions come back on a schedule in Review, so what you learned stays learned. Your progress is saved in your browser; no account needed.