REPETITION · DOSE OVER INTENSITY
On two minutes daily beating twenty minutes weekly — and on the number everybody quotes wrongly.
Somebody cuts across a field because it is quicker. Then somebody else, because there is the faintest suggestion of a line. Within a season there is a path, and it was never decided on by anyone.
Nine essays have offered things to do. None of them will do anything unless the doing becomes ordinary, and this essay is about the only mechanism by which that happens.
It is the least interesting essay here and it is the one the other nine depend on.
One is too much at once. The other is too little too rarely.
Two weeks of committed daily work, abandoned in week three, remembered as a failure of discipline. It was a dose error: the amount was set at what felt serious rather than at what could survive a bad week.
Twenty minutes once a week, which feels substantial and installs nothing. Frequency does the work here and duration mostly does not.
I would take two minutes daily over twenty minutes weekly, and this is a design choice rather than something the research establishes. Nobody has run that comparison. What the habit literature supports is that repetition in a consistent context produces automaticity; it does not rank dosing schedules against each other.
The reasoning behind the choice: what is being built is a cue-response link, links are built by repetitions rather than by minutes, and a small daily dose yields more repetitions and survives a bad week. That is an argument, and it should be read as one.
People resist this because the small version feels unserious. That feeling is the main obstacle to the whole collection working, which is worth saying plainly.
The Sufi orders built their practice around dhikr — remembrance — a short phrase repeated, daily, for years.
The technical insight underneath it is about dose rather than about content: something very small, done with high frequency over a long horizon, changes what is available to a person in a way that an occasional intense session does not.
The orders were not making a psychological claim and would not recognize this description of what they were doing. I am taking the structure and leaving the content, which is what this collection does throughout and is worth saying each time.
Habit formation research gives the mechanism: repetition in a consistent context produces automaticity, and automaticity is what removes the need to decide.4
The number everybody quotes is wrong. Lally and colleagues found a median of 66 days to reach asymptotic automaticity — with a range from 18 to 254 days in a small sample, and considerable variation by behaviour.5 The twenty-one-day figure in circulation comes from a 1960 plastic surgery book about patients adjusting to their appearance and has no evidential basis whatever.
What the research supports: consistency of context matters more than duration of session; missing a single day does not meaningfully impair formation, which is worth knowing because people abandon after one miss; and the curve is asymptotic rather than linear, so early gains are steep and later ones are not.
The sample sizes in this literature are small and it has not been the subject of large replication efforts. The direction is well supported; the precise numbers are not.
The heroic material has nothing, and the gap is structural. Homer has no training montage, no practice, no acquisition — excellence is a property a figure has, demonstrated rather than developed.
That absence explains something about the register generally. A world with no account of how anyone became capable is a world where capability is a fact about birth and favour, and the tradition’s politics follow from that more than from anything it says explicitly.
The Confucian material is the opposite extreme and the Confucian collection covers it; its whole account of character is repetition.
The objection is that this essay is training the thing several earlier essays warned about.
The case against
If the exhale becomes automatic at every trigger, it will fire at triggers that were carrying information — which is the fifth essay’s objection, installed permanently. Automaticity removes the decision, and the decision was sometimes doing work.
There is also a version where the practice becomes the identity. Somebody two years into a daily regulation practice has a substantial investment in it working, which is a poor position from which to evaluate whether it does.
The guard is a scheduled review: once a quarter, ask what this is still for. A practice that cannot survive that question was not worth installing.
Choose one thing from the previous nine essays. One. People who choose three install none, which is the Mussar collection’s finding arriving here.
Make it smaller than feels serious. Two minutes. If you cannot imagine failing to do it on your worst day, the dose is right.
Fix the context rather than the duration. Same time, same place, same trigger. Consistency of context is what the evidence points at.
And treat a missed day as a missed day. The research is clear that one miss does not impair formation; the thing that does is treating the miss as evidence and stopping.
Two weeks is not enough for this station and saying so is more useful than pretending.
Nothing in this collection works without this, and this is the least interesting thing in it. Small, frequent, same context, over a longer horizon than anybody wants.
The twenty-one days everybody cites comes from a 1960 book about cosmetic surgery patients. The real answer is a median of about two months with enormous variation, from a small study, which is the honest version.
And the review is part of the practice rather than a concession to it. An automatic response is one that stopped being examined, and some of these were worth examining.
Choose one thing from the previous nine essays. Make it smaller than feels serious — if you cannot imagine failing to do it on your worst day, the dose is right.
Fix the context, not the duration. Treat a missed day as a missed day. And schedule the quarterly question: what is this still for?
Research is cited by author and publication, with effect sizes and replication status where they matter. Where a claim rests on a single study, the essay says so.