Why twenty-one days, and what actually happens in them

The twenty-one day figure comes from a misreading. A plastic surgeon in the 1960s observed that patients took about twenty-one days to adjust to a changed appearance; a generation of self-help books turned that observation into a law about habit formation. It was never a finding about habits.

The actual research on automaticity suggests a much wider range — on average a couple of months, with enormous variation depending on the person and the behaviour. So if the promise is that three weeks makes something effortless, the promise is false.

Cycles have a fairly consistent arc, and knowing it removes most of the alarm when each stage arrives. Days one and two are easy and slightly self-congratulatory. Days three to five contain the first serious doubt. The second week is the flattest and the most important — nothing is new, nothing is dramatic, and the sessions are simply done.

Somewhere in the third week two things tend to happen at once: the practice stops requiring a decision, and something small moves externally. The second is rarely dramatic and is easy to dismiss — a conversation goes differently, an opportunity appears, you say something you would not have said in week one.

Review in writing, decide in writing, and take at least a day off before starting the next cycle. The gap matters: cycles run back to back without a break turn into an obligation, and obligations are abandoned rather than adjusted.

Then choose deliberately: same intention refined, or a new one. Most intentions deserve three or four cycles. The urge to start something completely different is usually the same avoidance that produced the doubt on day three, arriving in a more respectable form.

For anything with a structural gap — a career change, a qualification, a move — think in terms of four cycles rather than one. Roughly three months, with four reviews, one refinement each time. That is long enough for the automaticity research to become relevant and for external results to have had a fair chance to appear.

Keep the unit at three weeks even so. Long commitments made in one piece are hard to sustain; the same commitment made four times, each with a decision point at the end, is considerably easier to keep.

It is worth being explicit about what usually changes by the end of a first cycle, so that you are looking in the right place. Almost never: the outcome. Frequently: what you notice, what you say out loud, how quickly you recognise something is not right, and how long you wait before asking for something. Those are not consolation prizes — they are the mechanism by which outcomes eventually change, and they are visible in three weeks if you have been keeping any kind of record.