Two documented rules frame professional revision. Stephen King's, from On Writing (2000): write with the door closed and rewrite with it open, and the second draft equals the first draft minus ten percent — the arithmetic of subtraction rather than addition. And Anne Lamott's, from Bird by Bird (1994): shitty first drafts are the raw material, not the verdict. Between those poles sits the working method most editors and revising novelists converge on: revision in passes, one concern per pass, each pass blind to the others' work. Trying to fix structure, scenes, and sentences simultaneously is the documented cause of the endlessly rewritten first chapter and the never-finished book.
Why one concern per pass?
Because attention is the scarce resource and every concern competes for it. A reader can hold a novel's structure in mind, or its chapter openings, or its verb column — not all three, and the attempt produces the half-fix: paragraphs polished at sentence level that a structural cut will delete next week. The pass system sequences the concerns from largest to smallest, so that nothing gets fine-polished before the big decisions have stabilized. It also converts an impossible task ("fix the book") into a list of finite ones ("this pass: every scene's exit"), which is the difference between revising and worrying.
What does the full sequence look like?
- The cooling pass. King's drawer: six weeks minimum away from the draft, occupied by different work. The draft returns strange, and strangeness is the reader's-eye the writer must borrow.
- The read-through. One sitting per day, no fixing, margin notes only — where you laughed, where attention dropped, what you forgot. This produces the diagnostic map every later pass uses.
- The structure pass. The card test at manuscript scale: state changes per scene, the midpoint, the cost ladder. Cut, merge, reorder. Expect the minus-ten-percent to be found here or nowhere.
- The scene pass. Each scene individually: want, obstacle, turn, exit. Scenes that survive get their entrances sharpened — late in, early out.
- The line pass. Sentence length, verbs, redundancy — the audits covered elsewhere on this site, run now that the sentences are permanent. Print helps; the page reads differently than the screen.
- The continuity pass. Names, timelines, eye colors, the gun from chapter two. A spreadsheet, one row per fact-class, catches what affection for the book forgives.
- The proof. Read aloud, once, slowly. Then stop. The documentable failure of good revisers is not stopping.
How do you run a single pass, mechanically?
With the tool narrowed to the concern. Structure pass: index cards or a spreadsheet, no prose editing. Scene pass: one chapter at a time, tracked changes off — fixing as you read corrupts the reading; margin-flag instead, then fix in a second sweep of the same chapter. Line pass: one device per sweep (sentence lengths, then verbs, then openings) because the eye trained to one fault finds it; trained to five, finds none. And every pass begins with a fresh cold read of the section it covers, because the pass's authority comes from reading like a reader, and readers have not memorized your intentions.
How many drafts does that make?
Fewer than it looks like and more than one: the passes compose into three or four real drafts for most published novels — the diagnostic surveys aggregated across author interviews put completed-and-revised manuscripts at three to five full versions, with the pass system compressing what undisciplined revisers spend across a dozen. The count matters less than the direction: each draft should be one rung smaller, tighter, and closer to its final shape, per the minus-ten rule. Drafts that grow — the research novel that accretes scenes each revision — are moving backward, and the pass audit will show it: if the structure pass adds scenes two drafts running, the problem is upstream of revision, in a premise that has not decided what it is.
What breaks the system, and how do you fix the breaks?
| Symptom | Cause | Repair |
|---|---|---|
| Chapter one rewritten endlessly | Line-passing an unstable opening | Freeze chapters 1–3 until the structure pass is done |
| Draft keeps growing | Addition mistaken for revision | Enforce the minus-ten; cut list before add list |
| Every pass finds new problems | Concerns bleeding across passes | Log new finds for their own pass; do not chase them now |
| Cannot see the draft anymore | Cooling too short | Back in the drawer; six weeks is a floor, not a ritual |
| Revision ends but feels unfinished | No stopping rule | The proof pass is the last one; schedule it like a deadline |
When is it actually done?
When a full cold read produces no notes that require prose changes — only the typos. Not when it feels perfect: the documented experience of every published novelist is that the book never feels finished, only released, and the feeling is identical at every level of career. The pass system's quiet gift is substituting a checklist for that feeling. A writer with seven completed passes holds a defensible manuscript; a writer with a feeling holds a mood. Send the checklist.
What changes between the first book and the later ones?
The passes stay; the writer's speed through them changes. Debut revisions spend their heaviest weeks in the scene pass — everything is the first time, every problem is new. Career novelists, per the interviews they give mid-career, run the structure pass faster (pattern recognition from a shelf of previous books), spend more relative time in the line pass (the taste got pickier), and treat the cooling drawer as non-negotiable precisely because they have felt what shipping an un-cooled draft costs. The sequence is also what makes multiple books survivable: a writer with a pass system can put a project down for a contract's gap year and re-enter it at a known phase, while a writer who revises by mood must restart each time from dread.
Frequently asked questions
How long should each pass take?
Proportionally: the structure pass costs a reading plus a card table — days; the scene pass costs the most, chapter by chapter; line and continuity run a week each on a standard novel. Budget the sequence in months for a first novel, and let the calendar, not anxiety, set the pace.
Should I revise while drafting?
Only yesterday's pages, lightly, to restart momentum — the documented Trollope method. Full revision mid-draft fixes prose a structural decision may cut, and the fixing becomes a sophisticated form of not finishing.
Does the minus-ten rule always apply?
As a direction, yes; as arithmetic, roughly. Some second drafts cut a quarter; a very few add. The rule's function is refusing the default — that revision means more scenes, more explanation — and the refusal is right nearly always.
What if readers give notes mid-sequence?
Log them, thank them, change nothing until the pass their notes belong to arrives. Feedback arriving out of sequence is the most common contaminant of disciplined revision, and triage is the whole defense.
For more context, read How Many Drafts Is Enough.
For more context, read sentence length pacing.
For more context, read Backstory Without Stalling the Scene.