7.0 How to use the protocol
The ten tests below are for revision, not composition. Applied during drafting they induce paralysis, and they induce it in a particular way: the value-shift test asks a question that can only be answered by a finished scene, so a writer who asks it in the third paragraph will stop writing in the fourth. Applied to a completed draft, chapter by chapter, they are the most efficient triage instrument in this chapter.
Two procedural notes.
- Record the answer, not the verdict. The output of a test is a sentence describing what the scene does, not a pass or a fail. A protocol run that produces ten ticks has taught the writer nothing; a run that produces ten sentences has produced a description of the scene, which is what revision needs.
- A failed test is a question, not a death sentence. Every one of these tests is failed by at least one scene discussed in this chapter, most of them by scenes that are among the finest in the European novel. The test’s function is to make a choice visible so that it can be confirmed as a choice.
7.1 The ten tests
Test 1 — The want test
State, in one sentence, what the viewpoint character wants in this scene, right now.
Interpretation. A missing want is the most common single fault in draft scenes and the one that most reliably explains a vague sense of flatness. But note the legitimate exceptions: scenes of pure reception, in which something is done to a character who wanted nothing (the arrival of bad news); and scenes in which the absence of a want is characterising, as with Camus’s Meursault (Chapter 9, §9.4). In both cases the writer should be able to say which exception applies.
Test 2 — The resistance test
Complete the sentence: “___ wants ___, and ___ is in the way.”
Interpretation. An unresisted want produces activity rather than scene, and the resulting passage tends to be short, pleasant and forgettable. The most frequent draft error is nominal resistance: an obstacle exists but is overcome without cost, so the scene has the shape of conflict and none of its consequences. The follow-up question is therefore: what did getting it cost? If the answer is nothing, the resistance was ornamental. Failure pattern §8.8 treats this at length, together with the important counter-case in which a want is granted immediately and the granting is itself the disaster.
Test 3 — The value-shift test
Name what is true at the scene’s start and what is no longer true at its close.
Procedure. Write two sentences in parallel, in the same grammatical form.
Start: Nadia believes the ledger stopped when her father did. End: Nadia knows someone continued it, and has not asked who.
Interpretation. Name the value being moved — trust, safety, knowledge, hope, standing, belonging — and its direction. A scene that moves a value the book has not established as charged has produced a change without consequence; McKee’s insistence on charged values (§3.7) is this point. And note the class of scenes for which the test is inapplicable rather than failed: normal-establishing scenes (§4.4), whose function is to make a later shift legible. For these, run the test on the pair.
Test 4 — The extraction test
Remove the scene entirely. What would the reader now fail to learn, feel or anticipate — and what later scene becomes unearned?
Interpretation. “Nothing” means the scene is a satellite that has been given kernel resources (§3.4). The available remedies, in order of preference: compress into a paragraph of summary; fuse with a neighbouring scene that does have a job; relocate the one load-bearing element into a scene that already exists; cut. Note that cutting is last. Most purposeless scenes contain one sentence that matters, and the efficient revision extracts the sentence rather than mourning the scene.
Test 5 — The domain-count test
Which of the five domains does this scene actively serve, as opposed to merely occupying?
Interpretation. A scene marked with only one letter is thin unless it is short, is a deliberate rest, or is establishing a normal. Two is the working standard. Three is efficient. Four or five in a scene of ordinary length is a warning rather than an achievement (§5.4). The most useful output of this test is usually not the count but the identification of the latent domain — the second function the scene could serve without additional length (§5.3).
Test 6 — The redundancy test
Has the reader already received this value shift?
Interpretation. Redundancy is the failure pattern most invisible from inside the draft, because each scene was written on a different day and felt new at the time. It is also the fault that most reliably produces the reader complaint that a book “sags in the middle”: the middle of a novel is where redundant scenes accumulate, since it is the region with the least structural pressure. The remedy is almost never to improve the second scene. It is to ask which of the two is better placed and delete the other.
Test 7 — The summary-substitution test
Rewrite the scene as three sentences of summary. What is lost?
Interpretation. Three outcomes, tabulated at §6.4. The crucial discrimination is between lost and merely absent: a scene’s atmosphere is always absent from its summary, and that is not by itself an argument for the scene. The question is whether the book needs that atmosphere at that point. This is the test that most often produces the answer a writer does not want, and it is the reason the protocol is placed in the revision sequence rather than the drafting one.
Test 8 — The entry-and-exit test
Does the scene start as late as it can and end as early as it can?
Interpretation. Late entry and early exit are the cheapest revisions available, because they need no new writing. A scene that opens before the want is in operation spends its first page on arrival; a scene that continues past its last movement muffles its own shift, usually by explaining it. The exception is the slow entry that builds the normal the scene will break: the first chapter of Suzanne Collins’s The Hunger Games (2008) spends most of its length on the morning of the reaping before the draw. Where the opening pages do that work, run test 3 on the pair.
Test 9 — The pressure test
Does the information in this scene arrive under pressure, or is it being delivered?
Interpretation. This is the diagnostic for the expository scene in disguise (§8.3). Information delivered under pressure is retained, because it is bound to a transaction — the situation-model account (§3.5) supplies the mechanism: facts attached to a person in a situation have somewhere to be stored. Information delivered as briefing is not retained, which means the info-dump does not even achieve its own stated purpose. The remedy is rarely to remove the information; it is to make someone reluctant to give it.
Test 10 — The register test
Is the story currently due for an action beat or a reflective one, and which is this?
Interpretation. Four consecutive action scenes produce incident without consequence: things keep happening and nothing lands, because the reader has not been given the space in which landing occurs. Four consecutive reflective scenes produce brooding without motion. The register test catches a failure that is invisible at scene level and severe at book level, and it is the point at which this chapter hands over to Chapter 2 (Scene and Sequel), which treats the rhythm as its subject.
7.2 The reverse outline: the protocol at manuscript scale
The tests above examine one scene. The instrument that examines a book is the reverse outline, and it is worth separating because it is a different operation with a different output.
Procedure.
- Go through the finished draft and write one line per scene, in order, in a single document.
- Write each line in the language of consequence, not of event. Not “Elena and Dov talk in the doorway” but “Elena gets the key and not the reason, and stops believing her brother is protecting their mother.”
- Add, for each line, the scene’s length and its domain letters.
- Read the resulting document straight through, as a text in its own right.
What the reverse outline exposes that scene-level testing cannot:
| Symptom in the outline | Diagnosis |
|---|---|
| A line that will only come out event-shaped | That scene has not earned a consequence-shaped summary; it is a satellite |
| Two lines that say substantially the same thing | Redundancy (§8.2), now visible where it was not before |
| A run of six lines with no R or no P | A domain has gone dark for a stretch of the book |
| A stretch where every line begins “X learns…” | The middle has become an information-delivery sequence |
| Lines getting longer as the book proceeds | Scenes are accreting functions rather than being chosen |
| The last line of Act One arriving at 40 per cent | A proportion problem invisible from inside |
7.3 Reading the results
| Failed tests | Likely diagnosis | First move |
|---|---|---|
| 1 and 2 | No engine. The passage is an occasion, not a scene | Find the want, or convert to summary (§6.4) |
| 3 alone, with 1 and 2 passing | The scene has an engine that idles: a want is pursued and the situation is unchanged | Make the pursuit cost something, or let the character succeed and discover the cost |
| 4 alone | A satellite at kernel length | Compress, fuse, relocate the load-bearing sentence, then cut |
| 5 alone (one domain) | Under-served real estate | Run the multiplicity pass (§5.3) |
| 6 alone | Redundancy | Choose the better-placed of the two scenes; delete the other |
| 7 alone | Mode error: summary inflated into scene | Restore the summary; keep at most a fifty-word scenic instance |
| 8 alone | Entry and exit slack | Cut to the first operative moment and the last moving one |
| 9 alone | Exposition in disguise | Give the information to someone reluctant to part with it |
| 10 alone | Register mismatch | Move the scene, or convert it to the other register |
| 3, 4 and 6 together | The scene is a fossil of discovery drafting | Cut without ceremony; extract any usable sentence |
| 1, 2 and 3 pass; 5 gives four or five | Overstuffed (§5.4) | Simplify deliberately; the scene needs less, not more |
A warning about test 3. When a scene fails the value-shift test, the instinctive remedy is to add an event — a phone call, an arrival, a revelation. This is almost always the wrong move, because it treats a structural absence as a content shortage. The stronger remedy is to find the shift that is already latent in the scene and remove whatever prevents the reader from registering it. Most inert scenes contain a value shift the writer has muffled by continuing past it (test 8) or by having a character announce it (§4.5, “statement”).