How many days to break a habit? Not 21

It does not take 21 days. The best-known measurement, by Lally and colleagues in 2010, found a median of 66 days for a simple daily habit to become automatic, with individuals ranging from 18 to 254 days. No study produced the 21-day figure.

If day 21 came and went and it was still hard, that fits the data. The real numbers are also more useful than a short deadline would have been.

What the 66-day study measured

Lally and colleagues, publishing in the European Journal of Social Psychology in 2010, gave 96 volunteers one simple daily thing each: eat a piece of fruit, go for a walk, drink a glass of water. They tracked how automatic it felt over time and measured how long each person took to reach the point where that curve flattened out.

The median was 66 days. Individual times ran from 18 days to 254, a gap of nearly eight months between the fastest and slowest person, on tasks that took under a minute a day.

Why the range matters more than the median

A median tells you where the middle was. The range tells you what normal covers.

Half the volunteers took longer than 66 days. Some took nearly four times that, doing the same trivial task under the same conditions. A tight cluster would have justified a deadline. A spread this wide means no single day predicts anything for any one person.

The same authors note that automaticity arrives faster for a simple action, drinking a glass of water, than for an elaborate one, fifty sit-ups. Any timeline measured against 66 days is being measured against people drinking water.

The takeaway is the spread. Taking longer than expected is the expected result, and it reflects the range, not the person.

If you passed 21 days and it was still hard

That is inside the measured range.

The damage the 21-day figure does is to supply a date. When the date passes and nothing feels different, the only variable left in the sum is the person, and a bad week turns into a verdict on character. That verdict comes from a wrong number, not from the evidence.

At day 66, half the people in Lally's study had still not reached the plateau, on the least demanding habits anybody has been asked to form.

What happens to the process when you miss a day

Less than most streak counters assume. Gardner, Lally and Wardle wrote the research up in the British Journal of General Practice in 2012, and the paper is free to read: "Missing the occasional opportunity to perform the behaviour did not seriously impair the habit formation process: automaticity gains soon resumed after one missed performance."

One missed day did not send anybody back to the start. The gains resumed where they left off. Every counter that resets to zero after a slip is built on the opposite assumption, that the slip undid the progress. The best habit-formation data available says it did not.

The limit is real. "Did not seriously impair" is not "does not matter". One missed glass of water is not a relapse into something with an urge behind it. The study is about building a simple routine, not stopping a compulsive behavior, and nobody has run the equivalent test on that. It is evidence that a missed day does not end the process. It is not permission.

What the study does not say about quitting porn

Nearly everything, and its limits should travel with the number.

Ninety-six people is a small study. They were adding a simple routine, not stopping something they wanted and could reach in two seconds. The paper is called "How are habits formed", and it is about starting one. Automaticity was self-reported: people were asked how automatic it felt. Nothing in it touches porn, compulsive behavior, or anything with an urge attached.

So 66 days is not a prescription, and it is not a timeline for this. A claim that science says quitting porn takes 66 days, or 90, or any other number, is quoting a study about eating fruit.

Why the plan runs to ninety days, then

Ninety is a generous place for a checkpoint. Lally et al. never produced the number 90, and the study does not endorse it as a target.

The plan in the app runs to 90 days and gives you a real date. Ninety sits past the 66-day median with room to spare. Day 90 is followed by day 91. The plan frames the first run and does not end anything.

A date helps because a run with a visible end is something a person can start on a Tuesday. "Never again, for the rest of my life" is not.

What to do with a range this wide

Nothing that follows from 18 to 254 involves picking a better deadline.

Count the days you have instead of counting down to one you do not. A number that goes up survives being wrong about the timeline. A countdown to a graduation has to be right about it.

Take the timeline out of the argument. The 21-day myth is expensive because it keeps the question of whether this is working open every night, and that argument is easiest to lose at 1am. Removing the one-tap path removes most of it. iOS has a content filter you can turn on yourself in about five minutes, free, and the steps are in how to block porn on an iPhone.

Keep marking the days, including the ones that go wrong. If a missed day does not derail the process, a counter that wipes to zero tells a story the habit research does not support, at the moment it is hardest to argue with. A calendar that holds the whole year is still there at day 254 to show what happened. Most attempts come apart much earlier, and that is covered in why you keep relapsing on day two or three. If an attempt came apart later, after a month, there is a short answer in is it normal to relapse after 30 days.

None of this shortens the timeline. It stops the timeline being the thing you argue with yourself about.