One Story Moment
BlogTry the web app
← Blog

66 days is not a rule for your journal

The 66-day habit figure is the median from one 2010 study of eating, drinking, and exercise. It did not measure journaling, and it was not a law.

"It takes 66 days to form a habit" is one of those sentences that escaped a paper and lost the paper on the way out. The study is real. The sentence, as people repeat it, is not what the study found. It also has nothing direct to say about a two-minute journal.

What the study measured

Phillippa Lally, Cornelia H. M. van Jaarsveld, Henry W. W. Potts, and Jane Wardle published "How are habits formed: Modelling habit formation in the real world" in the European Journal of Social Psychology in 2010 (first published online July 16, 2009). The DOI is 10.1002/ejsp.674.

Ninety-six volunteers each chose an eating, drinking, or activity behaviour to do every day in the same context for 12 weeks. The paper's example of a context is "after breakfast." Each day they filled in a subset of the Self-Report Habit Index, a questionnaire about how automatic the behaviour felt, and they recorded whether they had done it.

This was self-reported automaticity. It was not a lab test of memory, and it was not a test of writing.

Eighty-two participants provided enough data to analyse. The researchers fitted a curve to each person's automaticity scores over 84 days. The model fitted 62 people. For 39 of those, the fit was good enough to use. Ten of that group of 39 were doing an eating behaviour, 15 a drinking behaviour, and 13 an exercise behaviour.

The participants were mostly postgraduate students. The paper reports a mean age of 27, with a range of 21 to 45.

Where 66 days comes from

For those 39 people, the median time to reach 95% of their own automaticity plateau was 66 days. The range was 18 to 254 days. Table 2 in the paper lists that median as 66, with the middle half of the group between 39 and 102 days.

The discussion also calls 66 days "the average modelled time to plateau in this sample." The results section and the table call it the median. Either way, the figure describes that subsample of 39, not the original 96, and not people in general.

The abstract states the range and the conclusion the authors were willing to put up front: there is considerable variation, it can take a very long time, and missing one opportunity to perform the behaviour did not materially affect the process. With repetition in a consistent context, automaticity increased along a curve that flattened.

A median of 66 days with a range of 18 to 254 days is a description of scatter. It is a poor candidate for a rule.

What the study did not study

Journaling was not one of the behaviours. The habits on offer were things like eating fruit, drinking water, or exercising, tied to a cue such as breakfast. A nightly note about a moment from the day is a different behaviour, with a different cue, and this paper did not measure it.

The paper also did not find that everyone plateaus. Many participants never produced a clean curve. Fourteen did not enter data beyond day 60 and were treated as dropouts. Treating "66 days" as the time it takes a person to become a journaler borrows a number from a group the study already had to narrow, then applies it to an activity the study never assigned.

Subgroup times for eating, drinking, and exercise were not significantly different in the paper's test, which is another way of saying the study was not precise enough to hand you a personal timeline. It was precise enough to show that the timelines differed a lot.

Two minutes is a design choice

Matthew Dicks, describing Homework for Life, tells people to spend about five minutes. His 2015 essay says "five minutes." His current site still says "Spend 5 minutes each day." One Story Moment is built to be done in about two. Neither number comes from Lally and colleagues.

What the habit paper does support, and only inside its own limits, is narrower: a behaviour repeated in the same context became more automatic for many of the people whose data fitted the model, on a curve that rose and then leveled, and a single missed day was not, in their data, the thing that broke the process.

A journal you open at the same time each night is the kind of consistent context they were interested in. That is an analogy, not a result. We are not claiming that two minutes of story-capture will reach an automaticity plateau in 66 days, or in 18, or in 254. We have not run that study. Nobody in this paper did either.

If you want the practice Dicks describes, the useful instruction is the one he published: ask for the moment, write a few sentences, do it again tomorrow. The calendar in the app is there so the days you kept are visible. It is not a score, and it is not a 66-day contract.

Sources