How long it actually takes to form a new habit — a median, not an average, over a wide range.
What the evidence says
- Median time to reach 95% of automaticity asymptote: 66 days — a median, not a mean; the wide range below is why “average” overstates how typical it is[1]
- And it is a median over 39 people. 96 volunteers enrolled, 82 had adequate data, and the asymptotic curve was a good fit for 39 of those 82 (48%) — the 66 days is computed on that subset, the participants whose habit formation actually followed the modelled shape. Interquartile range 39:102 days. The sample was 30 men and 66 women, predominantly postgraduate students[1]
- The interquartile range is 39 to 102 days — Table 2, ”Time to reach 95% of asymptote (days): median 66, quartiles (Q1:Q3) 39:102, minimum 18, maximum 254″[1]
- Full range: 18–254 days; the paper does not show that behaviour type explains the spread[1]
- Automaticity follows asymptotic curve: early repetitions yield larger gains, diminishing returns over time[1]
- Missing a single day does not materially affect habit formation — the paper’s own word is ”materially”: ”Missing one opportunity to perform the behaviour did not materially affect the habit formation process”, which is an observation rather than a reported significance test[1]
- By behaviour type (eating N = 10, drinking N = 15, exercise N = 13), median times to asymptote were 65, 59 and 91 days, not a significant difference (p = 0.328). The authors call the exercise group’s ”one and a half times longer” notable, and say the study ”was not powered for sub-group analyses”[1]
- These data do not support the popular “21 days to form a habit”: the median was 66 days, over a range of 18 to 254. The paper itself does not discuss the 21-day figure[1]