The mechanism has four parts: a map of the draft made once, an order in which to read it, a list that records each repair and what it disturbs, and a rule for stopping.
The book map
The book map has one row for each scene and a column for each fact the other chapters’ instruments need. Its first five columns follow the five dimensions readers track; the rest record the novel’s own structure. A map of a three-hundred-page draft takes a day to build, from the draft rather than from memory, and its value lies in being built from the draft: the writer’s memory holds the book she meant, and the map has to hold the book she wrote.
| Column | What it records | What is read off it |
|---|---|---|
| Position | Where the scene falls, as a percentage of the text | Every instrument that places events in the book |
| Story time | When it happens, by date or by order | The order chart and strand table (Non-linear and Multi-timeline Architecture, Chapter 8) |
| Place | Where it happens | The count of change at each gap (The Transition, Chapter 5) |
| Viewpoint | Whose mind, if any | The grant and the admission test (The Viewpoint Shift, Chapter 13) |
| Goal and decision | What the viewpoint character wants, and what she decides at the end | The option list (The Middle, Chapter 11) and the criterion log (Arc and Theme, Chapter 12) |
| Thread | Which chain of cause it belongs to | The convergence test (The Subplot, Chapter 14) |
| Change | The Purposes of a Scene’s two-sentence statement (Chapter 1), compressed | The scene’s purpose (The Purposes of a Scene, Chapter 1) and the value-shift log (The Middle, Chapter 11) |
| Learns | What the reader learns, and which characters learn it | The disclosure map (The Reveal, Chapter 15) |
| Deposits | What it plants, and what it pays | The plant audit (The Ending, Chapter 16) |
| Promise | Any convention invoked, kept or declined | The promise inventory (Series and Genre, Chapter 17) |
A row that cannot fill its change column is the first finding of the edit, before any test is run. A map also gives the writer what no single chapter can, which is a view of the book at a size she can hold: three hundred pages become about ninety rows.
One map, many readings
The instruments of the earlier chapters are views of the same map. The order chart sets the position column against story time; the strand table groups rows by thread; the option list reads the decision column at three positions; the criterion log reads it for one character from end to end; the disclosure map reads the learns column; the plant audit reads the deposits column backwards from the ending; the promise inventory reads the promise column forwards from the opening. A writer who builds the map once can run every test in the book without rereading the draft, and when she changes the book she changes the map, so that every instrument stays current. This is the practical reason for a single map: the instruments of separate chapters, made separately, disagree with one another as soon as one repair has been made, and the seventeen contradictions found when the household novel’s cases were first set side by side, over where the ledger was first seen, who found it and what each sibling knew, began in that way.
The order of dependence, and where it bends
The ladder is read from the top for repair and from the bottom for diagnosis. A writer revising works rung by rung downwards, finishing the decisions on each rung before the next, because a decision on a rung can make work below it wrong. A writer diagnosing starts where the reader felt the problem and looks upward. The rung-up rule states the second direction: before repairing a failure, check the rung above it for a cause, and treat a failure that appears in several places on one rung, or that recurs after it has been repaired, as caused above it.
The order bends in one way. A writer working on a lower rung sometimes finds that a decision above is wrong: a scene written to fill a gap shows what the book is about, or a cut that was meant to tighten a chapter makes a thread unnecessary. This is upward discovery, and it is common, since writers learn what their books are by writing them. The order does not forbid it; it says what to do with it. When discovery runs up the ladder, the writer stops work on the lower rung, makes the higher decision, and comes down again, because the lower work was done against the old decision and will need checking against the new one.
The edit list
The edit list records every note and every failure the tests find, one row each: the note and its source, where it was felt, the rung of the symptom, the rung of the cause, the repair, the tests the repair reopens, and its status. The status is one of three: mended, kept and defended in a sentence, or passed to the line edit. Reopened tests are the tests on lower rungs that a repair can make wrong for the units it touches: moving the arc’s decision reopens the option list at the midpoint, the plant audit, the chapter break nearest the decision and the scenes on either side. The list is what keeps the order honest. A repair whose reopened tests are not rerun is the cascade of Failure Patterns.
The economy of the repair
A scene that does several jobs with one set of material is economical, as The Purposes of a Scene (Chapter 1) argues, and a repair is economical in the same way when it answers several notes with one change. Repairs differ in economy by rung. A repair on a high rung often answers notes felt on several lower rungs at once, since their cause was shared, and a repair on a low rung answers one note and reopens little. The economy of the repair is the number of notes a repair answers set against the number of tests it reopens and the words it adds. Between two repairs that mend the same failure, prefer the one that answers more notes; between two that answer the same notes, prefer the one that reopens fewer tests; and between two that reopen the same number of tests, prefer the one that moves or cuts over the one that adds.
When to stop
The stopping rule has three conditions. A pass down the first three rungs changes nothing below them. Every failure that remains on the edit list is kept and defended in a sentence, or passed to the line edit, which The Purposes of a Scene (Chapter 1) already asks of a scene that fails its tests: its author should be able to say why. And a cold read, by a reader who has not read an earlier draft or by the writer after an interval long enough to have forgotten what she meant, produces notes that do not cluster above rung 5. The third condition is the one most often skipped, and the one that tests the first two, since a writer who has revised a book knows it too well to see what it now does.
See also: the full treatment, section 5, with the procedure for building the map and keeping the edit list.