The 66-Day StudyWhat it actually says – and how it turned into yet another rule

You have probably met the number by now. In a podcast, under a video, in a book about self-improvement: it takes 66 days for a new habit to stick. Not 21, as people used to say – 66. It sounds more precise, more serious, more scientific.

The number is even correct. It comes from a real study at University College London, published in 2010. And yet it is almost always passed on wrongly – not in the digits, but in what it is taken to mean.

Because 66 is the least interesting number in the entire study.

Where the 21 days came from

A word on the prehistory, because it shows how such rules come about in the first place. The famous 21-day rule goes back to Maxwell Maltz, a plastic surgeon who wrote “Psycho-Cybernetics” in 1960. He had noticed that after an operation his patients needed roughly three weeks to get used to their altered face in the mirror. He wrote that it takes a minimum of about 21 days for an old self-image to dissolve and a new one to set.

That was an observation of patients, not an experiment. It concerned the image of one's own face, not habits. And the word “minimum” got lost somewhere on the way into the self-help literature. A lower bound for something else entirely became a deadline for everything.

What was actually studied in 2010

The study the 66 comes from was run by Phillippa Lally and her team at University College London. The design is pleasingly unspectacular.

Ninety-six volunteers each picked one single everyday behaviour to carry out daily for twelve weeks – and to carry out in the same context every time. For example: a piece of fruit with lunch. A bottle of water with lunch. Fifteen minutes of running before dinner. Every day they logged on to a study website, recorded whether they had done it, and answered a short questionnaire about how automatic the whole thing now felt.

From these daily self-ratings a curve could be drawn for each person: automaticity rises quickly at first, then more and more slowly, until it approaches a plateau. The study's question was when that plateau is reached in practice.

The median was 66 days. That is the number that went around the world.

The three findings that matter

1. 01The range is the real newsIt ran from 18 to 254 days. For some people the new behaviour had settled after two and a half weeks; for others the model implied almost a year. A median is not a deadline – it is the point where one half was faster and the other half slower. Anyone who writes 66 in their calendar stands a good chance of concluding on day 67 that it simply does not work for them.
2. 02Missing a day costs almost nothingThis is the finding that genuinely helps people, and it is the one least often quoted. The researchers analysed 140 occasions on which someone skipped a day after several successful ones. Automaticity dropped by 0.29 points on a scale whose maximum was 42 – and was made up again the following day. A single omission had no measurable long-term cost. The sentence “I missed one day, the streak is ruined” is not merely harsh. It is wrong.
3. 03Context was a condition, not a detailIt is worth rereading the instruction: daily, in the same context. Not “drink more water”, but “a bottle of water with lunch”. The behaviour was coupled to a fixed point in the day. So the study never measured how long good intentions take. It measured how long a coupling takes.

What the study does not say

This is where it gets uncomfortable, but without this section the article would be dishonest – and would do exactly what I accused the self-help literature of above.

• The median comes from a subgroup. Of the 96 who signed up, 82 provided enough data. A curve could be fitted at all for 62, and for 39 the model was a good fit. The figure of 66 comes from those 39 people. That is careful work, openly reported in the paper – it is simply not the basis for a law of nature.
• Nobody waited out the 254 days. The study ran for 84 days. The upper end of the range is extrapolated from the curve, not observed. That is a normal procedure, but it ought to be mentioned.
• It was about fruit and water. What was studied were simple, clearly bounded single actions. Not: becoming more patient. Not: no longer making yourself small. Not: responding differently to criticism. Transferring the number to patterns like those is not something the study supports.
• What was measured was a feeling. How automatic something feels was rated by the participants themselves. For this question that is a sensible approach – but it is self-report, not a measurement in the brain.
• They were students. Predominantly postgraduates, average age 27, at a single university in London.

One nice detail in passing: split by type of behaviour, the medians were 65 days for eating, 59 for drinking and 91 for exercise. The difference was not statistically meaningful – but if you have ever suspected that exercise beds in more stubbornly than a glass of water, the data do not contradict you.

Why a number is so tempting

You might ask why both numbers persist so stubbornly – first 21, now 66. I think it is because they contain a promise the thing itself does not make: that change has an end date.

A deadline turns something open-ended into something foreseeable. You only have to hold out, and then it is over. That is precisely why the range of 18 to 254 days is so unpopular: it gives you nothing to steer towards.

But it gives you something else, which I consider more valuable. It strips day 67 of its meaning. If something still does not run by itself after nine weeks, that is not a finding about you. It is the normal range.

And the patterns that are already there?

So far this has been about how a new behaviour comes about. But most people who arrive with an intention to change have a different concern: something already runs automatically for them – just not the thing they want.

This study offers no number for that, and I am not going to invent one here. Two things can nevertheless be carried across.

First, the time scale. If even a bottle of water with lunch takes around two months to bed in, that says something about everything larger. A pattern that has been running for twenty years does not give way in a fortnight. That is not bad news – it is simply a different expectation.

Second, the trigger, which is the more practically important point. The new behaviour in the study had a visible starting point that the participants had chosen themselves. With ingrained patterns that is exactly the difference: the trigger is there, but you do not know it. You notice the outcome, not the beginning. Why we observe ourselves so poorly in this respect I have described elsewhere.

That is why the first step with patterns is rarely discipline. It is attention: finding out what actually happens beforehand. And for that another person is useful – not because they know more, but because they are not standing inside the habit.

Sixty-six days is a good number if you take it for what it is: an indication that change takes longer than we tell ourselves – and forgives more than we fear.

This article reflects the state of research on habit formation and is intended as everyday orientation. The coaching methods described support general wellbeing and holistic balance. They do not constitute a medical diagnosis, promise no cure, and do not replace medical or therapeutic treatment.