Reading a claim about your habits
90 min
Two hosts talk the lesson through. The voices are synthetic; the script was written from this lesson and checked against it, and asserts nothing the lesson does not.
- Sort a piece of habit advice into measurable, unmeasurable as stated, or false, and say what would settle it if it is the first
- Trace a repeated figure in this subject to its source, or establish that what travels is not what the source says
- Decide, for one behaviour of your own, whether to build a habit, to keep deciding each time, or to stop carrying it, with what each costs
This is the last lesson, and it's the one that stays useful after the rest of the course has dated.
Everything before it was about habits. This one is about the claims you'll meet about them, and you will meet a great many, because this is a subject with far more written about it than measured in it.
Three habits of reading, and then the decision the course has been deferring since lesson 1.
Habit one: sort the claim
Almost every piece of advice about habits is one of three things.
Measurable. It makes a claim a study could check. Take "stacking a new behaviour onto an existing routine makes it stick faster than putting it at a fixed time". Work it. Is there something to count? Yes: how automatic the behaviour feels, which is how this literature measures sticking, rated over some weeks. Are there two conditions to compare? Yes: stacked against fixed-time. Could one person run it? Yes, on themselves, badly, and a lab could run it properly. So it is measurable, and the useful question is not whether it sounds right but whether anybody has run it. Nothing this course read has.
Unmeasurable as stated. It uses a term nobody has defined, so no study could settle it either way. Take "become the kind of person who does this". Work it the same way. What is to be counted? A kind of person. By what criterion? None is given, and the sentence supplies no way of telling somebody who has become that kind of person from somebody who merely does the behaviour every day. Without that, there is no result the advice could fail. That isn't the same as useless, and it may be good advice; it is simply not the kind of sentence evidence bears on until somebody says what the words mean.
False. It makes a checkable claim that has been checked and didn't hold. Take "miss a day and you start again". Work it. What would make it true? That automaticity ratings after a missed day resume from lower than where they were. What was found? That gains soon resumed after one missed performance. So the claim is not merely unsupported; it is contradicted by the one study this course has read on the point, which is the strongest thing you can say about a piece of advice and the rarest.1
And a fourth, which is the commonest of all: two claims stuck together. Most habit advice is a recommendation with a reason attached, and the two are separately checkable. "Do it first thing in the morning, because that is when your willpower is highest" is one claim about when to act and one about why. The second is the resource model, which lesson 5 is about; the first would survive somebody's morning being reliably quiet whatever willpower turns out to be. A sorter who marks the whole sentence false because its reason failed has thrown away a claim nobody has tested, and habit three is what to do about that.
This sort isn't new here and the course says so. It is Digital Literacy lesson 8's move and Using AI Effectively lesson 11's, and Time Management lesson 8 applied it to a different subject. What is new is this subject's own failure mode, which is the next section.
Habit two: trace the number, and expect the same failure
Lesson 2 traced one number and lesson 4 traced another. Put all three of this subject's travelling numbers side by side and something becomes obvious that none of them shows alone.1
| The claim | What the source says | What fell off |
|---|---|---|
| 21 days to form a habit | "[A] minimum of about 21 days" to change a mental image, in a 1960 self-help book, about patients adjusting after surgery | "a minimum of about", and that it was not about habits |
| 66 days to form a habit | An average, with individuals from 18 days to a predicted 254, on 96 volunteers doing self-chosen eating, drinking or activity behaviours | The range, the sample, and that simple actions plateaued sooner |
| 43% of what you do is habit | "Between a third and a half", 43% in one study and 35% in another, on students, by a stated criterion | The range, the criterion, and the second study |
None of these was invented.1 That's the point of putting them together. Every one is a real thing somebody measured or observed, reported honestly in its source, and every one arrives at you with the part that made it meaningful stripped away.
That's a harder failure to catch than invention, and it is why lesson 2 came so early. A course that taught you only to spot fabrications would leave you defenceless against all three of these.
Before reading on: of the three, which do you think is the most misleading as it travels, and why?
Show the answer
There is a case for each, and the one this course would argue for is the 66.3
Twenty-one days is the most wrong and the least dangerous, because it has no study behind it at all, and once you know that you can discard it whole.
Forty-three percent is misleading and doesn't change what you do. Believing a third when it is between a third and a half costs you nothing in practice.
Sixty-six is the one that changes behaviour. It sounds like a finding, because it is one. It sounds precise, because it's a real average. And a reader who takes it as a rule will conclude at week ten that they've failed, when the study they're relying on fitted curves running out to eight months. A number that is true and stripped of its range is worse than a number that is simply wrong, because there's nothing about it to distrust.
Habit three: separate the advice from its reason
This one is particular to this subject, because half of it has just changed underneath.
Lesson 5 showed that the resource model of self-control lost most of the evidence that supported it.1 An enormous amount of practical advice was built on that model: decide important things in the morning, do not change two things at once, do not go shopping hungry, save your discipline for what matters.
The reason has gone. The advice hasn't necessarily gone with it.
"Don't start two habits at once" was argued from a willpower budget. It can also be argued from something this course did establish: what builds a habit is repeated pairings of cue and response, and two behaviours started at once split whatever attention you have for noticing either cue between them, so each gets fewer usable repetitions. Same advice, different support, and the second one survives lesson 5.
So a piece of advice and its stated reason are two separate things to check, and finding the reason unsupported does not settle the advice.3 What it does tell you is that nobody has checked the advice either, and that you're on your own with it, which is what the exercises are for.
The decision this course has been deferring
Lesson 1 said there was a question the course wouldn't answer, and that lesson 7 would come back to it. Here it is.
Should you want your life run by habit?
The case for is most of this course. A habit is behaviour that survives your worst weeks. It costs nothing to run, it doesn't depend on your wanting it that day, and it frees attention for the things that need deciding. A life with more of them spends less of itself on getting started.
The case against is in the definition. A habit is by construction something you do without deciding to. The mechanism is indifferent to whether the thing is worth doing, it built your phone habit exactly as it would build your reading habit, and a person can automate their way into a life they never chose, one cue at a time. There is a smaller version of the same worry too: a life with very few decisions left in it may be a life somebody stopped examining.
This course has no standing to settle that, and it is worth saying why rather than just saying it. Nothing in the evidence this course read bears on the question. The studies measure how habits form, not whether a person should form them, and a course that answered it would be handing you a preference dressed as a finding.
What it offers instead is the same in either direction. Know which of your behaviours are habitual and which aren't. Know what a cue is, and how long an association takes. And decide, rather than default, which is the one thing both answers require.
This course has spent seven lessons being careful about what the evidence supports. So apply that to the course. Which of the things you have been told here should you hold most loosely, and which most firmly?
Show the answer
Worth doing, and the ordering is not the obvious one.3
Hold most firmly: the definition, and the distinction between automaticity and frequency. A definition is a convention rather than a finding, so it cannot be wrong in the way a figure can, and this one is what makes the rest of the literature intelligible. Nothing that happens to the numbers will touch it.
Hold firmly: the cue. Two field experiments with a mechanism, a clean disruption, and a result that inverts what people expect. It is the best-evidenced practical thing in the course.
Hold with its scope attached: the timeline. The 66 days and the missed-day finding are the strongest practical results here and they come from a study this course has not opened, reported at one remove through two papers by its own author. That is a real limitation and lesson 4 says so where the figures appear.
Hold most loosely: what lesson 6 proposed. The process account is the newest thing in the course and the least established. Its authors call their paper a blueprint for future research and say where evidence is lacking, and this course read its abstract and two passages. It is in the course because leaving a hole where the resource model used to be would have been worse, not because it is settled.
And hold this lesson somewhere in the middle. The three-number table is the course's own assembly, and the argument that a true number stripped of its range is worse than a wrong one is reasoning rather than a finding.
The books you will be handed
You will meet this subject mostly through four very widely read books: Duhigg's, Clear's, Fogg's, and Baumeister and Tierney's. A course that never named them would be pretending you were not going to.
This course hasn't opened any of them.4 That is on the unread line of its research file and
enforced by the build, and it means this section is not an assessment of them. No claim anywhere in
these seven lessons rests on any of the four.
What to do when somebody hands you one is the three habits above. Sort each claim. Trace each figure, and expect the failure in the table rather than invention. And check whether the advice survives its reason, which matters most for the fourth of them: at the same search-summary level as everything else here, it is the trade statement of the position lesson 5 is about, so a reader has to ask what of it survives lesson 5.
One of them is worth naming separately, on the same search-summary basis as everything else this course knows about them. If a later reader opens one properly, the one to open is Fogg's, because of the first three it is the one written by a researcher about his own research programme, which means its claims can be checked against papers with his name on them. The fourth is a researcher's book too, and it is the trade statement of the position lesson 5 is about, which is a different reason to open it and a worse one to start with. That is a note about where a reading would go furthest rather than a recommendation.4
Three things people believe about this literature
"Somebody must have checked." Three of this subject's best-known numbers reached you with their conditions removed, and the checking took twenty minutes each. The sources that handed them to you had not done it.
"The science says." There is less science here than the phrase implies and more than there is in most of the genre. The habit half has a measurement paradigm and researchers who publish their own methodological doubts. The self-control half had a famous model that did not survive testing. Saying which half a claim comes from is most of reading this subject well.
"An old study is out of date." Sometimes, and the distinction is the one Time Management lesson 8 drew. Figures describing a time and a place go stale. A finding about how association works is a claim about a mechanism, and 2002 being a long time ago is not in itself an argument against it.
Practice
Take 20 minutes, and stop at twenty whether or not you've finished.
Find one claim about habits with a number in it. An app's onboarding, a book, a poster, an article. Write it down word for word, with where you found it.
Then follow it back, writing down each step:
- What does this source cite? If nothing, write "nothing", and you are finished.
- Open what it cites. Is the number there? Not something like it. The number.
- If it is there, what is attached to it in the source that is not attached to it in the claim? A range, a criterion, a sample, a qualifier. This is the question the table above exists to prompt, and it is the one most traces in this subject end on.
Write your finding in one sentence. Five endings are all results: the number is there with its conditions; the number is there stripped of them; the trail ends in nothing; the trail leads somewhere you cannot get to, which is what happened to this course with the Lally paper; or twenty minutes ran out.
Take 30 minutes. This is the last thing the course asks, and it is the point of all of it.
Take one behaviour you have been meaning to take up and have not. Something real, that you have carried for months or years.
Three endings, and all three are legitimate. Write whichever you choose in four lines.
One: build the habit. What is the cue, tested against both bars from lesson 3? What is the smallest version of the behaviour? What will you record, and for how long? And what comes out of your week to make room, because something must.
Two: keep deciding each time. Not everything should be automatic, and lesson 1's second worked case was somebody whose monthly accounts are more reliable than most people's habits and are not habitual at all. What makes this one worth deciding about each time? What will you put in place to make the decision arrive, since it will not arrive by itself? And what does it cost you to keep deciding?
Three: stop carrying it. Write down that you are not going to do it, what that costs, and what would bring it back. This is a real answer and the genre almost never offers it. Carrying an undecided intention has a price that is paid every week and buys nothing.
Then one line, whichever you chose: who has to be told?
If you cannot choose, that is a result too. Write down what you would need to know in order to choose, and whether it is something you could find out in an afternoon.
Connections
Back. This lesson uses the whole course. Lesson 1's definition is what the checkpoint says to hold most firmly; lesson 2's and lesson 4's traces are two of the three rows in the table; lesson 3's cue is the best-evidenced practical thing here; lessons 5 and 6 are why the third habit exists. Lesson 1 named the value question and this lesson faces it. From earlier Core courses: the sort is Digital Literacy lesson 8's and Using AI Effectively lesson 11's, adapted again, and Time Management lesson 8 is the nearest relative of this lesson in the whole institute.
Out. Focus and Deep Work and Memory sit later in this same term of the Core and are not written yet; when they exist they take up what this course stops short of. What survives from here, once every figure in the course has dated, is the definition, the sort, and the trace.
Go deeper
- Making health habitual (British Journal of General Practice, 2012). Read in full by this course, three pages, free. If you keep one thing from this course's reading list, keep this one: it is the whole habit half, written for practitioners, by people who ran the studies.
- Guidelines for tracking real-world habit formation (Cogent Psychology, 2022, open access). Read in substantial part by this course. Lesson 4 sends you here for the timeline; come back to it for its criteria, the second of which is a field writing down plainly what its own measures cannot support. That is a rarer document than it ought to be.
- A Multilab Preregistered Replication of the Ego-Depletion Effect (Perspectives on Psychological Science, 2016). Abstract and limitations read by this course. Lesson 5's exercise sends you to the limitations, and they are worth the trip on their own.
Sources
- This lesson introduces no new evidence, and the marker above appears wherever it leans on an earlier one. Every measured claim here belongs to an earlier lesson and is cited there: lesson 2 for the diary studies and the 43 percent, lesson 3 for the popcorn experiments, lesson 4 for the timeline, the missed day and the twenty-one days, and lessons 5 and 6 for the self-control literature.
- The three-row table is this course's own assembly. No source read for this course presents those three claims together; each row's content is cited in the lesson it comes from, and the body says where the assembly is the course's own work.
- The three habits of reading, the fourth kind of claim, the argument that a true number stripped of its range is worse than a wrong one, and the three-way decision in the second exercise are the course's own reasoning, marked as such where they appear, and the checkpoint says where to put each of them on a scale of confidence. The decision's form is taken from Time Management lesson 6, which is this institute's own earlier handling of a value question.
- Charles Duhigg, The Power of Habit; James Clear, Atomic Habits; B. J. Fogg, Tiny Habits;
Roy Baumeister and John Tierney, Willpower. Titles and dates from
research/SOURCES.md, which holds them at search-summary level. None of the four was opened for this course, and all four are on theunread:line ofresearch/SOURCES.md, which exempts this lesson so that it can name them. They are cited here for nothing. The judgement that Fogg's would be the one to open first rests on the same search-summary level as everything this course knows about the four, and the body says so.
Check your understanding
This lesson has a 6-question quiz. Pass it and the questions come back on a schedule in Review, so what you learned stays learned. Your progress is saved in your browser; no account needed.