The ten tests are for a finished draft. Applied while drafting they produce paralysis at a scale from which manuscripts rarely recover; applied in revision they are structural triage. Run them on every thread, the main plot included, since several produce information only by comparison, and run them in order: tests 1 (Engine) and 2 (Kernel) ask whether the thread is a chain at all, 3 to 5 what it does, 6 to 8 how much of the book it holds and when, and 9 and 10 whether it should exist and in what rank. Record a sentence, not a verdict (“converges only at the party in chapter 31, and nothing before chapter 27 prepares it” is usable; “fails test 5” is not). Run the protocol twice, some weeks apart: the first run reports the writer’s intentions and the second the text.
Test 1 — Engine
Can I state the thread’s goal, conflict and consequence in three sentences without mentioning the protagonist’s goal?
Failure. Presence, not plot; or main-plot machinery spread across other people.
Exception. Platon Karataev in War and Peace, a term the novel needs rather than a thread.
Test 2 — Kernel
At each of its four or five chief moments, could the thread have gone another way?
Failure. One kernel and a long tail of satellites.
Exception. The intercalary chapters of The Grapes of Wrath, general chapters between the Joads’ story, which have no kernels and carry the novel’s claim.
Test 3 — Tax
What does it supply that the main plot cannot by its nature?
Failure. Only relief, or nothing.
Exception. Middlemarch, which has no single primary investment to be taxed.
Test 4 — Analogy
What does it say about the main plot, and what would be lost if it were not said?
Failure. Duplication, a decorative rhyme, or an analogy no reader sees.
Exception. The inserted tale of the curious impertinent in Don Quixote, whose relevance the novel’s second part admits was questioned.
Test 5 — Convergence
At what point would deleting it break the main plot’s logic, not merely thin it?
Failure. Parallel play, or a meeting mistaken for convergence.
Exception. Anna Karenina, whose Levin thread never converges in event.
Test 6 — Proportion
Are its pages proportional to its functions, and to its weight in the reader’s experience?
Failure. Padding, or an under-served co-plot.
Exception. The Ilyusha thread in The Brothers Karamazov, a small share that takes the ending.
Test 7 — Page-share
What share of the book does each thread hold, and does the distribution match the book I think I wrote?
Failure. A different book from the one the writer believes in.
Exception. Vanity Fair, whose pages follow Becky and whose subtitle admits it.
Test 8 — Dormancy
What is the longest gap between appearances, and what does resumption cost?
Failure. A re-establishing scene where a crossing should have been.
Exception. The Corrections, whose long sections leave whole threads dormant for a hundred pages.
Test 9 — Removal
Cut it on the outline: what is lost, in which domain, and does the loss exceed the pages recovered?
Failure. Nothing nameable; or a loss one scene could supply.
Exception. Koznyshev and Varenka in Anna Karenina, two chapters whose removal would cost only a rhyme.
Test 10 — Promotion
Would the book be better if this thread were the main plot?
Failure. The wrong book has been written.
Exception. Great Expectations, which builds the transfer of primacy into its design.
Reading the results together
Failures cluster, and the clusters point to remedies. Test 1 (Engine) alone usually means presence, not plot, and the remedy is cheap: supply a goal and a reversal using scenes that already exist. Tests 1 (Engine) and 2 (Kernel) together mean texture wearing a subplot’s page-share; reclassify it as motif or runner and cut it to the size of what it is. Tests 3 (Tax) and 4 (Analogy) together mean a thread that neither does nor says anything; cut it, and check whether its one useful relation can be carried by a scene elsewhere. Test 4 (Analogy) alone is usually a braiding failure disguised as a thematic one, repaired by bringing the compared scenes nearer. Tests 5 (Convergence) and 8 (Dormancy) together mean a late convergence that has not been paid for: braid the thread into the quarter before it, or move the convergence earlier. Tests 6 (Proportion) and 7 (Page-share) point opposite ways according to the mismatch, towards cutting where share is high and weight low, and towards expanding where share is low and weight high, the remedy writers in structural revision least often consider.
Test 10 (Promotion) comes last because it can overturn every other answer. When it returns an uncomfortable result, writers tend to defend the existing hierarchy, since the main plot is the more finished object and the one the book was planned around. A thread that keeps growing across drafts is evidence of where the book’s energy lies, and the two honest responses are to promote it or to recognise a double plot and stop ranking.
See also: the Reference protocol, with each test’s procedure and a table of combined failures.