1. Introduction: The Problem of the Last Page

1.1 The inverse asymmetry

Chapter 9 (The Opening Pages) began by comparing page 3 with page 203. Page 203 arrives loaded, underwritten by two hundred pages of accumulated investment; page 3 has no account, and must secure every claim it makes out of its own resources. That asymmetry — the threshold problem — is the reason openings need a chapter of their own.

This chapter begins from the same comparison, run in the other direction, and arrives at an asymmetry of the opposite shape.

Consider page 403 of a four-hundred-and-four-page novel. By the standards of Chapter 9 it is the best-funded page in the book. Everything is available to it: every character the reader has come to know, every object the book has put on a shelf, every phrase it has repeated, every question it has raised, every register it has taught the reader to hear. Nothing has to be established. Nothing has to be explained. The account is not merely well funded; it is at its maximum balance, and it will never be topped up again.

Which is the difficulty. Two constraints follow from that position, and they pull against each other.

The first constraint is that the account cannot be added to. An ending may spend anything the book has already deposited and nothing it has not. A new character on page 403 is not a new character; it is a withdrawal against an account that was never opened. This is the structural fact underneath the whole folk vocabulary of endings that “cheat” — a vocabulary readers reach for immediately and reliably, whatever they know about narrative theory. What they have detected is an unfunded transaction.

The second constraint is that there is no page 405. Every other unit in this textbook operates inside a story still running. A slack chapter can be redeemed by the one after it. A transition that loses the reader can be repaired two pages later, at the cost of a little goodwill. A scene whose purpose is invisible when it occurs can acquire one at the moment its plant pays off — Chapter 6 (The Chapter Break) calls this the relation of proposition and payoff, and the whole of Chapter 14 (The Subplot) depends on the reader’s willingness to wait for a thread to justify itself. The ending is the one unit for which waiting is not available. Whatever it does is what the book did.

Put the two constraints together and a third fact emerges that is more consequential than either:

The ending is read backwards. The reader’s last act of interpretation is not local to the last pages. It is an act of interpretation upon the whole, performed at the moment the whole first exists. Until the final page, a novel is an incomplete object about which the reader is still forming hypotheses; at the final page it becomes a completed object with a shape, and the shape is conferred from the end. This chapter calls it the retroactive lens, and it is the chapter’s first principle. It is also, as §3 will show, one of the most thoroughly theorised phenomena in twentieth-century criticism, under several different names.

The practical consequence is a disproportion that every working novelist has felt and few can explain. A handful of unearned pages can flatten two hundred pages of careful work. Not obscure them — flatten them, retrospectively, so that scenes which were doing real work while the reader was inside them come to seem like machinery for a destination that did not deserve them. The reverse holds equally, and is the entire reason for taking the unit this seriously: an ending that lands can make a slack middle read as patience, an obscure early scene read as design, and a novel that was merely good read as inevitable.

1.2 The claim in its strong form

Every earlier scene waits to find out what it was for. The ending is where they all find out at once.

The metaphor of accumulated trust runs through this textbook, and here it reaches its terminus. Chapter 9 argued that the opening pages are where trust is minted. This chapter argues that the ending is where it is spent — spent completely, with no reserve held back, because there is nothing left to hold it for.

The strong claim needs its limits stated.

• It does claim that a reader’s final judgement of a novel is disproportionately determined by its last pages, and that this judgement is applied retroactively to material the reader has already consumed and enjoyed.
• It does not claim that endings must resolve. Resolution is one option among several, and §2.3.2 and §6.4 spend some time separating it from the felt completion that readers actually respond to. A novel may leave every question open and close absolutely.
• It does not claim that endings must be positive, redemptive, consoling, or even satisfying in any ordinary sense. The Trial ends with a man knifed in a quarry and the observation that it is like the shame of it outliving him. Nobody finds this satisfying; almost everybody finds it complete.
• It does not claim that the ending is the part of a novel that matters most. It claims that the ending is the part with the least margin for error, which is a different and more modest proposition. A poor ending damages a great book. A great ending does not make a poor one.
• And it does not claim, though the craft literature routinely assumes it, that endings should be the last thing written. Many are the first. The evidence from working practice is mixed and the chapter takes no position on the composition order; what it insists on is that the ending must be revised last, because it is the only unit whose obligations cannot be known until the rest exists.

There is one further qualification, and because it is a serious one it gets its own section rather than a bullet. The strong claim assumes the reader wants the book to end well. A large and distinguished tradition of criticism holds that the desire for a well-made ending is itself the problem — that closure is the moment a novel stops being interesting and starts being tidy, and that the demand for earned resolution is a demand for the extinction of what made the reader read. That argument is D. A. Miller’s, and it is put at full strength in §3.9 before this chapter attempts to answer it.

1.3 Where this chapter sits in the series

Three placements matter, and each does work later in the chapter.

First: the formal inverse of Chapter 9. The two boundary chapters of this textbook describe the same privileged position under opposite conditions. Chapter 9 §3.2 states the symmetry from its side, noting that both openings and endings are what Rabinowitz calls privileged positions, and that Kermode and Herrnstein Smith describe the ending’s work as “the retrospective conferral of shape” while the opening’s work is “the prospective creation of the conditions under which shape can later be conferred”. This chapter completes that argument from the other end. The formulation the two chapters share can be stated as a single line:

The opening works with no accumulated context. The ending works with nothing but accumulated context.

Every difference between the two units follows from that. The opening’s characteristic failure is default — a contract left unopened, a reader given nothing to hold. The ending’s characteristic failure is unfunded expenditure — a claim made on capital the book never deposited. The opening’s discipline is establishment; the ending’s discipline is scarcity. And where Chapter 9’s zero-context test is an austere idealisation (readers do in fact arrive with a cover and a blurb), this chapter’s total-context assumption is close to literally true: a reader on the last page really does have the whole book behind them, and really does have nothing after it.

Second: the terminus of the load-bearing concepts. This textbook is organised so that a small number of ideas recur and accumulate, and nearly all of them fall due in this chapter:

Third: the last chapter. In the first edition this was the final unit of the book, and it laboured under the obligation it describes: a textbook whose final chapter is about endings and which ends badly has produced its own counter-example. In this edition Chapter 18 (The Structural Edit) closes the book, and this chapter closes Part III. §16 is written in that awareness, and the household narrative that runs through every chapter is concluded in §9.10 using nothing that has not already been planted — which is either the strongest available demonstration of the chapter’s thesis or a stunt, and the reader is entitled to decide which.