Everything in this chapter is an instrument for noticing. None of it is a rule, and the chapter contains its own evidence for that claim.
Middlemarch ends almost every chapter softly and would fail the variety test in most of its Books. Moby-Dick violates the proportion principle so comprehensively that a manuscript following its practice would be returned unread, and is one of the great segmented works in the language. The Count of Monte Cristo commits cliffhanger fatigue on an industrial scale. One Hundred Years of Solitude has no architecture for a reader to hold on to and is architecturally perfect. Don Quixote contains the most flagrant false cliffhanger ever printed. Tristram Shandy is a four-hundred-page demonstration that the whole apparatus is a convention. A model that rules these out has not found a defect in them; it has found its own edges.
But the deeper argument of this chapter is not the usual one about exceptions. It is historical, and it is stated plainly here at the end.
The chapter is roughly two thousand years old, and it has meant several different things. It began as a retrieval device, an aid to finding a passage in a long text — a piece of information technology, no more expressive than an index. For most of its life the divisions were made by editors, scribes and scholars rather than by authors, and they arrived with headings whose purpose was to tell the reader in advance what was coming. Only in the last three centuries, and decisively in the last two, has it become what a contemporary novelist assumes it to be: an instrument for shaping a reader’s experience of time, a place to leave someone hanging, a unit of tempo, an offered exit with a price.
Even that recent meaning is not one thing. The chapter of the monthly part-issue and the chapter of the paperback thriller share a name and almost nothing else. The first was designed to survive thirty days in the reader’s ordinary life; the second is designed to be crossed without the reader noticing they had the option to stop. Both are called chapters. Both are printed the same way. They are different tools.
The practical consequence for a writer is exact:
A novelist who treats the current form of the chapter as natural has mistaken a technology for a fact — and will therefore use it at only one of its settings.
The whole repertoire is available. Titled chapters, summarising headings, mottoes, Parts with names, chapters of a paragraph, chapters of nine thousand words, chapters that are lists, chapters that are dialogue, blocks of chapters designed together as a number, and the refusal of chapters altogether. Every one of these has been used well by somebody, and every one of them is presently unfashionable enough to be nearly unused. A writer who knows the history has more instruments than a writer who knows the convention.
What the model in §§4–8 does reliably is convert a vague dissatisfaction into a specific question. This chapter doesn’t land becomes: there is no proposition here, or the payoff resolves what should have been held, or this is the fourth expensive exit in a row and the reader has stopped believing me, or the real boundary was two pages ago. Specific questions can be answered. Vague ones can only be worried at.
And beneath all of it is a fact about reading that no taxonomy reaches. A chapter break is the point at which a book stops being something the reader is inside and becomes an object in their hands that they may set down. Everything a novelist does at that boundary is an attempt to make the setting-down matter — to ensure that the interval, however long, is time in which the book continues to exist.
Which suggests one last test, to be applied after the other ten:
What is this book doing while it is closed?
If the answer is nothing, the ending was free and should have been. If the answer is waiting, in the reader’s head, with something unfinished in it, the price was set deliberately, and the writer knows what they are doing.