Lally on how long a habit takes
📚SOURCE
💎PERMANENT NOTES
🧵SYNTHESIS
Paraphrase, not quotation: the full paper, 🔬 How are habits formed - Modelling habit formation in the real world.
Ninety-six volunteers each chose one small daily behaviour to do in a fixed situation, such as a glass of water with breakfast or a walk before dinner, and reported for twelve weeks whether they had done it and how automatic it felt. The researchers fitted a curve to each person's automaticity scores: steep gains early, then a plateau.
For the people whose data the curve described well, the time to reach 95 percent of their own plateau had a median of 66 days and a range from 18 to 254. Harder behaviours, exercise for one, took longer than drinking water. And one missed opportunity did not measurably lower the curve.
So the popular number is real, and it is the least interesting thing in the paper. The finding is the spread: a factor of fourteen between the quickest person and the slowest.
✨KEY IDEAS
Automaticity rises quickly at first and then levels off; the curve is an asymptote, not a straight line.
Median 66 days, range 18 to 254.
The upper end of the range is an extrapolation: the study ran for 84 days, so 254 comes from extending a fitted curve.
Missing a single day made no material difference to habit formation.
More complex behaviours took longer to become automatic than simple ones.
📝READING NOTES
What they did
- Who: 96 volunteers, mostly students, recruited at a London university.
- What: each picked an eating, drinking or activity behaviour and tied it to a daily cue of their choosing.
- Measure: a daily self-report of whether they did it, and a short questionnaire on automaticity (does it happen without thinking, is it hard not to do).
- Analysis: an asymptotic curve fitted per person. Of the 96, 82 gave enough data, and the curve was a good fit for only 39 of them. The headline figures come from those 39.
What they found, in the order I care about
- The spread is enormous. 18 days to 254.
- A miss is noise. One skipped opportunity did not dent the curve.
- Consistency early matters most, because that is where the curve is steepest.
- The plateau differs by person. Some people's ceiling of automaticity was simply lower.
How much weight it will bear
Thirty-nine people. Self-reported feelings of automaticity. Behaviours the participants chose themselves, so nobody was trying to form a habit they disliked. Twelve weeks. Hale would call this a pilot. I am treating it as the best available estimate and not as a law, which is more than most uses of the number do.
Also: the paper measures one missed day. It says nothing about a week and a half in a row, which is what I actually did in February. I cannot claim it for that.
Against James
📖 James on habit says never allow an exception. This says the first exception costs almost nothing. They can both be right if James is describing what a lapse does to resolve and Lally what it does to automaticity. My rule is in between: one miss is noise, two are a decision.
What I changed
- Planning horizon for any new habit: about ten weeks, with a shrug.
- Each habit gets a 🧩 Minimum viable habit so that a bad day still counts as an opportunity taken.
- The fact itself is filed at 💡 The median time to form a habit in one study was 66 days.
What it produced
💎 One missed day does not break a habit and 💎 A habit takes months and not three weeks, both this morning.
Related notes
Reading record
- Weight of the evidence: suggestive
🗣️IN SHORT
- In : How are habits formed: Modelling habit formation in the real world (2010), the whole paper
- Lally and colleagues argue that : habit strength grows along a curve that flattens, and the time it takes varies widely between people and behaviours
- by stating that : in their sample the median time to near-maximum automaticity was 66 days with a range of 18 to 254, and that a single missed day did not materially change the trajectory