Between 18 and 254 days
The median is 66 days, measured across the 39 of 82 participants (48 percent) whose progress followed the model. The top of the range, 254 days, lies beyond the last measured day (day 84) and comes from the model curve.
For eating habits the median was 65 days (middle half: 35 to 106 days, 10 people). Of these, the upper quartile from 106 days lies beyond the last measured day (day 84), so it comes from the model curve. The difference between the three behavior types was not statistically significant.
No days logged yet.
Enter how many days ago you started and your position appears here.
Missed a day? A single skipped opportunity did not materially affect habit formation in the study. For longer gaps the authors point to an earlier study (Armitage 2005), where only a full week off counted as a lapse, and that lowered the odds of carrying on.
What your result means
The range above is not an evasion, it is the finding. In the real-life study behind all of these numbers (Lally and colleagues), 96 volunteers picked one small eating, drinking or activity behavior and were asked to do it daily in the same context, right after breakfast for example. 82 of them provided enough data. For 39 of those (48 percent), felt automaticity followed the expected curve cleanly: steep gains first, then a plateau. How long it took to reach that plateau varied enormously, from 18 days to 254 days.
Most articles leave out two things. First, the denominator: for more than half of the participants the model did not fit well enough; their curves followed it only loosely, stayed flat, or reached no plausible plateau. Second, the measurement boundary: observation ran for 84 days. Any figure beyond the last measured day, including the top of the range, comes from the mathematically extended curve. The calculator flags such values right next to the number.
The 21-day myth and the half-truth of the 66 days
The 21-day rule is commonly traced to a 1960 self-help book about self-image and adjustment, not to any study of habit building. That origin story offers no evidence in the modern sense; the number just got copied often enough.
The 66 days are better, but incomplete. That is the median of the 39 of 82 people the model fit well. And a median means about half of that group needed at least that long. Anyone promising you autopilot after two good months is citing a study that says no such thing.
How the calculation works
The calculator invents nothing. It shows the published distributions: the overall range with its median, and the subgroups by behavior type (eating: median 65 days across 10 people, drinking: median 59 days across 15 people, movement: median 91 days across 13 people). The difference between the types was not statistically significant, so we treat it as a hint and never as a rule.
The unit is days, not repetitions, and that is the study's own unit rather than our shortcut: because behavioral data had gaps, the authors explicitly modeled automaticity against the day of the study, not against the number of performed repetitions. So the calculator asks how many days ago you started and compares day with day. Your progress appears as a share of the group median, because no study on earth knows your personal plateau.
Transparency also means naming a discrepancy inside the source itself: the paper states 39 people with a good model fit, yet its three subgroups add up to 10 + 15 + 13 = 38. The text does not explain the difference; a plausible candidate is the study's fourth category (4 participants chose another behavior, such as meditation), which the subgroup sentences do not enumerate. We report the figures exactly as the source states them and compute nothing from the sum.
Two examples
Example 1: movement, daily, started 30 days ago. Day 30 of 91 toward the movement-group median, which is 33 percent. 61 days remain to the median and 88 to the upper quartile. Both landmarks, the movement median of 91 days and the upper quartile from 118 days, lie beyond the last measured day (day 84), so they come from the extended model curve.
Example 2: drinking, 3 days a week, started 12 days ago. Here the calculator converts nothing. The study's participants were asked to practice daily; a three-day rhythm was never measured. Comparing your own day count with the medians is therefore labeled explicitly as an assumption, not a study result.
Three honest notes
First: one missed day is not a disaster. Automaticity kept climbing at a comparable rate after a single gap. Second: the fixed trigger is what counts. Every behavior in the study was tied to a stable context; without an anchor you are measuring something else. Third: for longer gaps the authors point to an earlier study (Armitage 2005), where a full week off made carrying on less likely. At that point a smaller restart beats a motivational quote.
Frequently asked questions
How long does it really take to form a habit?
In the real-life study behind all of these numbers (Lally et al. 2010), participants needed between 18 and 254 days for a new behavior to reach its automaticity plateau. The median was 66 days, across the 39 of 82 participants whose progress followed the model. The study does not offer one number that fits everyone, and that is the finding.
Is the 21-day rule true?
No. The 21 days are commonly traced to a 1960 self-help book, not to any habit study. And the median of the real-life data from Lally et al., 66 days across 39 of 82 participants, sits far above the rule; even the fastest participant, at 18 days, only just beat it.
Where do the 66 days you read everywhere come from?
From the same study, minus its qualifiers. 66 days is the median of the 39 of 82 participants (48 percent) whose progress followed the automaticity model well. For more than half, the model did not fit well enough. Quoting the number without the denominator is quoting half the result.
Is exercise different from food habits?
The medians differed in the study (65 days for eating, 59 for drinking, 91 for movement habits), but the difference was not statistically significant. With groups of 10, 15 and 13 people, that is not a rule you can build on.
I missed a day. Is it all ruined now?
No. Single gaps are what the study itself tested: after one missed opportunity, automaticity kept rising the next day almost as before. For longer gaps the authors point to an earlier study (Armitage 2005), in which a full week off made carrying on less likely.
Why does the calculator count days rather than repetitions?
Because that is how the study measured it. Some days had missing data, so the authors modeled automaticity against the day of the study, not against performed repetitions. And for anyone practicing less than daily: the study never measured that case, so the comparison with the medians is labeled as an assumption.
What counts as a habit here?
In the study: a behavior that feels automatic and no longer runs on willpower, measured with a self-report scale over 84 days. The threshold was the point where 95 percent of a person's own maximum was reached.
Sources
All sources retrieved September 9, 2026. Lally P, van Jaarsveld CHM, Potts HWW, Wardle J: How are habits formed: Modelling habit formation in the real world. European Journal of Social Psychology 40 (2010), 998 to 1009, doi 10.1002/ejsp.674. Every raw number on this page (range, median, subgroups, denominator, observation window) comes from this paper; the number-by-number mapping with verbatim quotes lives in our verification record.
Published September 9, 2026 · Last updated September 10, 2026 · Wild Garden Editorial Team · Last verified September 10, 2026 · Calculation tested against an independent reference implementation (8,421 test cases, 0 mismatches) · 41 of 41 anchor values confirmed · 16 of 16 source citations confirmed against the primary document; all runs in the verification record. Our editorial standards · Spotted an error? redaktion@wildgarden.io
Self-report data from one volunteer study, no guarantee for any individual case. Values beyond the 84-day observation window are model estimates.
More from the garden
You can find every calculator under Tools. A good companion: the Wheel of Life shows you in 2 minutes which area of your life a new habit would help most. If the new routine touches sleep, the sleep calculator with caffeine check tells you when to go to bed. And for eating routines, the calorie calculator supplies the numbers.
