The ten tests are for revision, not composition; applied while drafting they make the writer stop before every scene to ask whose head it belongs in, which is a reliable way to write nobody’s. Run them on grants, not scenes, since the unit under test is a viewpoint character across the whole manuscript. Run them in order and stop when one fails badly: tests 1 to 3 (Necessity, Domain and Investment) ask whether a grant should exist, 4 to 7 concern the shape of the allocation, and 8 to 10 concern execution and consequence. Record a sentence, not a verdict (“Dov’s vantage serves sympathy and contrast, and appears once, at 78 per cent” is useful; “fails test 7” is not), and treat every failure as a question, since each test is failed by at least one novel this chapter admires.
Test 1 — Necessity
Could report, inference, reassignment, document or ellipsis carry this vantage’s scenes, and what would be lost?
Failure. An access grant to convert.
Exception. The Demeter’s log in Dracula, whose facts the novel also prints as a newspaper report.
Test 2 — Domain
Which domains does the grant serve, with a passage as evidence for each?
Failure. Access only: under-employed. None: remove.
Exception. Addie’s section in As I Lay Dying, which serves the book rather than its own pages.
Test 3 — Investment
Before the first grant, can the reader say what this character wants?
Failure. A cold start; the grant reads as an appointment.
Exception. “Long, Long Time” in The Last of Us, which introduces Frank and gives him the episode.
Test 4 — Distinguishability
Swap two hundred words between two vantages, changing only the names: could a reader tell?
Failure. Two cameras, one mind.
Exception. Dracula, whose similar journals corroborate one another.
Test 5 — Page-share
What share of the book does each vantage hold, in how many sections, from when, with what longest gap?
Failure. Equal shares across five or more; a vantage re-entered as a stranger.
Exception. Wide Sargasso Sea, whose Antoinette is re-entered as a stranger on purpose.
Test 6 — Declaration
On what page does the reader learn the book has more than one vantage, and is it before her first deep investment?
Failure. An undeclared architecture; resistance rather than confusion.
Exception. Me Before You, whose first chapter outside Lou comes about a fifth of the way in.
Test 7 — Late grant
Does the payoff cover formal surprise, attention withdrawal and cold start?
Failure. An unearned late arrival.
Exception. The Sound and the Fury, which turns to an outside narration for its last section.
Test 8 — Knowledge boundary
Does the vantage know only what it could know?
Failure. Scattered: an omniscient slip. Clustered at plot points: a vantage that knows too much.
Exception. The Great Gatsby, whose Nick narrates scenes he never saw.
Test 9 — Scene integrity
Inside a continuous scene, does the narration stay with one mind, and if not, has the book established the right to move?
Failure. Head-hopping: a right exercised but never established.
Exception. Anna Karenina’s races, told among Vronsky, Anna and Karenin.
Test 10 — Character compression
For each named character with no viewpoint: how many scenes, whose explanation of her motives, what repeated tag?
Failure. A character-system that argues something the writer did not intend.
Exception. Lolita, whose single grant compresses Dolores Haze and makes the compression its indictment.
Three refinements sharpen the tests. Test 1 (Necessity) asks about loss, not information, since report can always convey a fact; what it cannot do is dramatise a decision as it is taken, render a misapprehension from inside, or make a silence feel like a choice rather than an absence, and a loss of one of those passes the test. Test 10 (Character compression) is the only one that looks away from the pages under revision, and it asks three things of each outsider: whether the only character whose motives are never available is the one the plot needs to be wrong, whether a relative appears in many scenes only to make a viewpoint character feel something, and whether a whole group is present in the book without a single interior.
Reading the results together
Failures cluster. Tests 1 (Necessity) and 2 (Domain) together mean the grant is a delivery mechanism: convert it to report, inference or document and reassign its scenes. Test 2 (Domain) alone means an under-employed vantage: give it a want and a blind spot, or bound it. Test 3 (Investment) alone means a cold start, and the remedy is presence before the grant, not more interior inside it. Test 4 (Distinguishability) alone means a redundant vantage: repair it by what each mind notices, appraises, cannot know and calls things, and if that fails, merge the two characters, which is nearly always an improvement. Test 6 (Declaration) means the first change should move to the first chapter break. Test 8 (Knowledge boundary) clustered means replacing knowledge with inference, and the scenes usually improve, because inference is dramatic and knowledge is not. Test 9 (Scene integrity) in a limited architecture means deciding: repair to the convention, or commit to a mobile narrator from page one, not both. Test 10 (Character compression) is not a repair but a decision about what argument the distribution makes.
A page-share audit that shows a beloved secondary vantage holding four per cent of the book prompts the instinct to expand it, and the instinct is usually wrong. The useful question is whether those pages are the right four per cent, and whether they should be one bounded section rather than six thin ones.
See also: the Reference protocol, with each test’s procedure and a table of combined failures.