7.0 How to use the protocol
The ten tests below are for revision, not composition. Applied while drafting they produce paralysis and, worse, produce chapters shaped like the tests. Applied to a finished draft they are the most efficient structural triage available in this textbook, because they inspect a level — the boundary and the sequence — that no amount of sentence work reaches.
Four procedural notes.
- Run them in order. Tests 1 and 2 are structural, and their failures invalidate the rest: there is no point tuning the permission price of a chapter that has no proposition to be priced.
- Tests 1–6 are applied to one chapter. Tests 7–10 are applied to the manuscript. This is the protocol’s main internal division, and it is the reason writers who revise chapter by chapter never find their pacing problems. Half of what is wrong with chapter breaks is only visible in the aggregate.
- Record the answer, not the verdict. The output of a test is a sentence describing what the break does. “Fails test 3” is useless; “stopping here is free, and this is the fourth free stop in a row” is a revision plan.
- A failed test is a question. Every one of these tests is failed by a novel discussed in this chapter, usually deliberately. The test’s job is to make a decision visible so it can be confirmed as one.
7.1 The ten tests
Test 1 — The proposition test
Can I state, in one sentence, the question or tension this chapter raised at or near its start?
Write it down as a reader would phrase it, not as the author knows it. If the sentence requires two clauses joined by “and also”, the chapter has more than one proposition and you should identify which the ending addresses. If the sentence contains “somehow”, “a sense that”, or “the feeling that”, there is no proposition — there is atmosphere, which is a different thing and cannot be paid off.
The eighteenth-century form of this test is quicker and more brutal, and is Exercise 1 in §12.1: write the chapter’s summarising heading. In which ___. A chapter that resists the formula is a chapter without a subject.
Test 2 — The payoff test
Does the ending do something identifiable with that proposition — resolve, escalate, hold open, re-interpret, or exchange it — rather than simply stopping?
Name the operation from the list. If none of the five verbs applies, the proposition has been abandoned, which is the fault at §6.3 and, by this chapter’s estimate, a common structural defect in competent manuscripts.
A useful refinement: read the chapter’s first page and its last paragraph consecutively, skipping everything between. A well-shaped chapter survives this treatment — the two ends are audibly related. A chapter that does not survive it is either two chapters (test 9) or has drifted.
Test 3 — The permission test
Does this ending make stopping cost the reader something, or let it be free — and is that the right price at this point in the book?
Answer in two parts, and record both. Part one is descriptive: free / moderate / expensive. Part two is justificatory: because the last two chapters were expensive and the reader needs release, or because the next chapter changes viewpoint and needs the reader to want to come back.
If part two cannot be completed, the price was not chosen. That is not necessarily wrong, but it is not yet a decision.
Test 4 — The stopped-reader test
If a reader closed the book here and did not open it again for three days, what would they be carrying?
An informative test, and the hardest to run without self-deception, for the reason given at §1.1: the writer never stops. The procedure that makes it work is mechanical. Read to the end of the chapter. Close the file. Do something else for ten minutes. Then, without looking, write down: what happened, what is unresolved, what you want next, and how you feel.
Three failure signatures:
| What you wrote | Diagnosis |
|---|---|
| Nothing you want next | The exit is free — confirm that was intended (test 3) |
| A want, but attached to a character you cannot name | The chapter’s tension is generic; it belongs to the plot, not to anyone |
| A summary of events and no feeling at all | The ending is informational; consider an echo or recasting close (§4.4, §4.6) |
Test 5 — The placement test (substitution)
If I moved this break one beat earlier, and then one beat later, could I say why the current position is right?
Run the substitution procedure at §6.5 in full. This is a slow test — twenty minutes per chapter — and should be reserved for the six or eight breaks that matter most: Part boundaries, viewpoint handoffs, and the ends of movements.
The test’s characteristic result is worth anticipating so that it is not mistaken for failure: very often all three versions work. That is information, not a null result. A boundary with three viable positions is a boundary the writer can place for sequence reasons — variety, proportion, the needs of the next chapter — rather than for local ones.
Test 6 — The proportion test
Does this chapter’s length match its actual weight, or has it been padded or compressed towards a habitual target?
The operational form: state the chapter’s weight in the book’s economy — major turning point / substantial movement / connective / brief — and then compare its length ranking with its weight ranking. Systematic disagreement is the finding.
Two specific patterns to look for. A connective chapter of full length is padding (§8.3) and is nearly always the result of a writer who felt a chapter had to be a certain size. A major turning point of below-average length is compression, and is more often a good sign than a bad one: turning points frequently want to be short, because contraction reads as impact.
Test 7 — The variety test
Checked against the five chapters on either side, is this ending type and this length a deliberate choice or a default?
The first of the manuscript-level tests. Produce the four-column table from §5.6 for eleven consecutive chapters and read the ending-type column aloud. If it says cliffhanger, cliffhanger, cliffhanger, pivot, cliffhanger, the book has one gear. If it says soft, soft, soft, soft, soft, the book has no gear.
Note that the target is not maximum variety. A deliberate run of soft landings through a novel’s quiet middle third is a design; the test asks whether it is one.
Test 8 — The re-entry test
What does the next chapter have to spend on orientation because of where I cut?
The point at which this chapter’s decisions hand over to Chapter 5 (The Transition)’s. Placement determines the orientation bill: a mid-scene cut leaves the reader fully oriented and the next chapter can open in the middle of a sentence; a cut at a decision pre-pays the next chapter’s where and why; a time-jump or a viewpoint handoff leaves the next chapter owing three dimensions at once.
The diagnostic is a comparison. Count the words the next chapter spends before its own business begins. If the number is above about sixty, ask whether a different placement would have paid the bill in advance. A costly opening is often a symptom of a badly placed break rather than a badly written transition, and writers who revise the transition instead of the break do the same work repeatedly.
Test 9 — The container test
Is this one chapter, or two chapters that have not been separated?
Look for the internal seam: a point at which the chapter’s proposition changes, a location shifts substantially, or a section break already exists. Then ask whether that seam is better than the current ending. The signature of a chapter that is really two is that test 2 fails while both halves would pass it separately.
The inverse form of the test is equally useful: are these two chapters, or one chapter that has been cut in half? Its signature is a chapter whose ending is a shrug and whose successor opens by re-establishing everything the reader already had (§8.9).
Test 10 — The architecture test
Read only the chapter headings, the Part titles and the first and last sentence of every chapter. Is the shape of the book legible?
The whole-manuscript instrument, and the one most likely to find something a writer did not know was there. Compile the document; it takes an hour for a novel and is worth it once, late.
Four questions to ask of the compiled document:
- Does the sequence of last sentences have a shape — does it tighten, quicken, darken?
- Do the Part boundaries fall where the book’s movements actually change?
- Are there stretches of five or six chapters whose last sentences are interchangeable?
- Could a reader who saw only this document describe what kind of book it is?
7.2 Reading the results
| Failed tests | Likely diagnosis | First move |
|---|---|---|
| 1 and 2 | The chapter has no subject; it is a stretch of narrative that happens to be bounded | Do not fix the ending. Find the proposition or merge the chapter with its neighbour |
| 2 alone | Abandoned proposition — the chapter drifted from its own opening | Either return to the opening question in the last page, or move the break to where the first proposition actually completes |
| 3 alone | The price was not chosen | Decide it. Usually the answer is dictated by the two chapters on either side, not by this one |
| 4 alone | The ending is informational rather than experiential | Change the last paragraph, not the chapter. An echo (§4.4) or recasting (§4.6) close is the usual remedy |
| 5 | The cut is in a defensible position but not the best one | Move it. This costs nothing; no prose is rewritten |
| 6 and 3 | The chapter has been sized and ended by habit | Suspect a whole-manuscript pattern; go to test 7 |
| 7 | The book has one gear | Convert every third close to a different type; the sequence, not the instance, is the object |
| 8 | Orientation costs are being paid on the wrong side of the break | Move the break to a decision or an announced destination (Ch 5 §5.3) |
| 9 | Two chapters in one container, or one split into two | Split at the internal seam, or merge and re-cut |
| 10 | The book’s architecture is invisible to a reader | Consider Parts; consider titles; consider whether the movements are where you think they are |
A warning about test 7. Writers who discover that their endings are monotonous tend to respond by adding cliffhangers, because the cliffhanger is the type that feels like an improvement. This is nearly always the wrong direction. A manuscript with too many expensive exits and one with too few both fail the same test, and the commoner condition in ambitious drafts, by this chapter’s estimate, is too many. The corrective move is usually to install soft landings, which is counter-intuitive, unglamorous, and works.