What the trials measured, what they excluded, and where the effect actually comes from.
3 branches open from this page.
Short enough to survive an interrupted evening, structured enough to audit.
Continue 02ChapterA cue that depends on memory is not a cue. It is another decision.
Continue 03ChapterA mark on a page outperforms an app notification for the behaviour it is meant to protect.
ContinueA part changes the scale of the argument. The story is split so each movement can be read in one sitting and still stand on its own.
A part is a movement, not a chapter. It changes the scale of the question so the next level can be smaller and more exact. The three parts of this investigation move from the problem, to the evidence, to the default.
A result that disappears when adherence is controlled is a result about recruitment, not treatment.
Each part ends with a question the next part has to answer. If the question is weak, the part is rewritten.