How are habits formed: modelling habit formation in the real world
The study behind the 66-day figure - and the study that refutes how it is used. Ninety-six people repeated a daily behaviour for 12 weeks; the median time to reach automaticity was 66 days, but the range ran from 18 to 254 days, so there is no universal number of days to build a habit.
Everyone quotes ‘66 days’. The paper’s actual finding is that there is no single number, and that is far more useful.
Ninety-six volunteers each picked an eating, drinking or activity behaviour to perform daily in a consistent context - ‘after breakfast’ - for 12 weeks. Every day they rated automaticity and recorded whether they had done it. The researchers then fitted an asymptotic curve to each individual’s scores.
The median time to reach automaticity was 66 days. The range was 18 to 254 days.
That spread is the finding. The slowest habit-former took roughly fourteen times longer than the fastest, and the median was drawn from about half the participants - those whose data fitted the model.
There is a further caveat the authors state and almost nobody repeats: the instrument measured automaticity - whether the behaviour felt effortless - rather than habit in the technical sense of persisting when the reward is removed.
Two things follow for anyone starting a journal. Consistent context matters more than willpower: the behaviour was tied to an existing daily anchor. And if it does not feel automatic after a couple of months, you are not failing - you are inside the range the study actually reports.
The paper also found that missing a single day did not meaningfully damage the trajectory.
It does not show that habits take 66 days. That median came from the subset whose data fitted the model - roughly half the participants - and the range was 18 to 254 days, meaning the slowest took about fourteen times longer than the fastest. It measured self-reported automaticity, not habit in the strict behavioural sense, and studied simple eating, drinking and activity behaviours rather than complex ones.
Articles citing this study
Summarized by Tiago Martins · Last updated: · How we write these
