Everything in this chapter is an instrument for noticing. None of it is a rule.
The evidence is in the chapter itself. Anna Karenina fails the convergence test. Middlemarch’s second plot occupies a third of the book and converges causally on one afternoon. Bleak House runs on coincidences no contemporary editor would permit and is the finest convergence structure in English. Moby-Dick drops a thread it introduced with enormous weight, and interrupts its own narrative for hundreds of pages to classify whales. Les Misérables contains a hundred pages about a battle at which nobody in the novel is present. Vanity Fair’s subordinate thread eats the book. Every one of these is a failure by the protocol at §7, and the protocol is not thereby shown to be wrong — only to be an instrument calibrated for a particular kind of book.
That kind of book is real, and most manuscripts are it. A novel with one primary investment, ninety thousand words, and a secondary thread that has grown without supervision is a common object, and the attention tax will diagnose it accurately every time. A writer who runs the ten tests on such a book will find the answer, and the answer will usually be that the thread has no chain, or supplies nothing, or converges by encounter. The model works. That is not in question.
What is in question is the inference writers draw from its working — which is that a leaner book is a better one, and that the discipline of pruning is the discipline of the novel.
It is not. It is the discipline of a form under length constraint, in a market that has been shortening books for a century, taught by a pedagogy that developed alongside that market and mistook its conditions for principles. The tests in §7 have no vocabulary for a book that wants to hold four lives at once and rank none of them. Applied to Middlemarch, the protocol asks which of the four plots is the main one, and the novel’s answer — that the question is the mistake, that the light must be concentrated on the web and not on any strand — is unavailable to a model built on hierarchy.
A form that can hold four plots at once is not a form that has failed to prune. It is a form doing something a single plot cannot do: representing a society as a set of relations rather than a protagonist as a set of obstacles. Levin’s analogy and Empson’s juxtaposition are not consolation prizes awarded to novels that could not manage a convergence. They are the mechanisms by which a book says something about how lives stand to one another — which is a subject, and arguably the novel’s largest one.
Two closing propositions, offered in the order in which a writer should hold them.
First: run the protocol. Most subplot problems are the ordinary problems, and the ordinary problems have ordinary solutions. A thread with no goal, no reversal and no convergence is not a multi-plot novel in embryo; it is a thread with no goal, no reversal and no convergence, and the fact that Tolstoy did something superficially similar is not a defence. The counterexamples in §8 are not permissions. Each of them succeeds by a mechanism that the chapter has tried to name, and none of them succeeds by accident.
Second: know what the protocol cannot see. It cannot see position, and a subplot placed last can assert what a subplot given twice the pages cannot (§9.7). It cannot see the surplus Empson describes, which is produced between threads and belongs to neither. It cannot see a writer’s conviction that the world contains more than one story, which is a conviction the greatest novels in the tradition were written to act on.
Which suggests the last and least mechanical question, to be asked after the other ten:
Does this book contain more life because this thread is in it, or only more pages?
The tests exist to make that question askable. They do not answer it, and a writer who wants them to has misunderstood what an instrument is for.