6. Earned and Unearned

6.0 What this section covers

“Earned” is the central term in craft discussion of endings and the least analysed. It is used as though its meaning were obvious, and it is not: writers apply it confidently to endings they like and cannot say what test they have run. This section takes the term apart.

It proceeds in seven steps. §6.1 states what the earned/unearned distinction actually claims, which turns out to be two claims rather than one. §6.2 separates the claim from the thing it is most often confused with, predictability, and argues that the separation is the whole art. §6.3 grounds both in Aristotle and in the plant economy this textbook has been building since Chapter 6. §6.4 supplies a taxonomy of sixteen closural strategies. §6.5 shows how to use a taxonomy of strategies in revision, which is not by classification. §6.6 addresses the proportion problem of climax and aftermath. §6.7 historicises the whole apparatus and warns against universalising the present.

6.1 What “earned” claims

The working distinction is the standard one:

Earned — every element the ending relies on was planted or built earlier in the book, so that the resolution feels, in hindsight, like a culmination the reader could have anticipated even if they did not predict it.

Unearned — the ending depends on something introduced at the last moment: new information, a sudden capability, an outside rescue, a fact withheld not for effect but for convenience.

This is correct and incomplete. Unpacked, it contains two distinct claims, which have different evidence, different failure modes and different remedies, and which are conflated at some cost.

Claim 1 is about preparation, and it is a fact about the text. Every element the ending uses as load-bearing must exist in the book before the ending. This is checkable, mechanically, by anyone with a copy: it is the plant audit (§7.1, test 2). Its failure mode is the unearned rescue (§8.1). Its remedy is to plant the missing element earlier, which is nearly always a small edit — a sentence in chapter four — rather than a structural repair.

Claim 2 is about retrospective inevitability, and it is a fact about the reader. The ending must produce, in hindsight, the sense that this was what the book was going towards. This is not checkable. It is a felt property, it can fail in books where every element is dutifully planted, and its remedy is not planting but pressure — the sense that the material was building towards something rather than merely accumulating.

The distinction matters because the two failures look identical from inside a draft and require opposite responses.

The second failure is far more common in competent drafts and far less often correctly named. A writer who has been told to plant things will plant them; the resulting ending will be defensible, traceable, and inert. Claim 1 is a necessary condition and not a sufficient one, and a chapter that taught only the plant audit would produce endings that pass an audit and land on nothing.

The full statement the chapter uses from here on:

An ending is earned when everything it relies on was already in the book, and when what it does with that material answers a pressure the book has been building. The first is a matter of deposit; the second of interest accrued on it.

A note on why this is the answer to Miller. §3.9 left the chapter with a concession: every ending is a loss, and the question is what the reader gets in exchange. Claim 2 is the answer. An ending that merely closes the narratable has taken something and given nothing; an ending that closes the narratable by answering a pressure the narratable itself generated has converted instability into shape. The exchange is real, and it is what “earned” names. Miller’s thin endings are thin because they satisfy Claim 1 and ignore Claim 2 — they dispose of the disequilibrium administratively. His diagnosis is right about the endings he examines, and the remedy is not to abandon closure but to stop treating it as disposal.

6.2 Inevitability is not predictability

A damaging misunderstanding in this area, because it produces two opposite errors, is the belief that an earned ending is one the reader can see coming. It is not, and the confusion produces two opposite errors: writers who fear predictability and therefore withhold preparation, and writers who fear the accusation of cheating and therefore telegraph.

The distinction:

Predictability is a property of the reader’s state before the ending: they can say what will happen.

Inevitability is a property of the reader’s state after it: they cannot imagine it having happened otherwise.

These are independent. All four combinations exist:

The top-left quadrant deserves emphasis, because contemporary craft discussion tends to treat predictability as a defect tout court. It is not. Greek tragedy is entirely predictable to its original audience; the pleasure is in watching the known outcome arrive by the particular route this dramatist has chosen. So is the marriage plot: nobody reading Pride and Prejudice is uncertain that Elizabeth will marry Darcy, and the novel is not damaged by it. Surprise is one source of satisfaction at an ending and not the principal one.

What produces inevitability without predictability is a specific and teachable structure, and it needs naming because most drafts get it backwards:

Plant the materials in plain sight; withhold the combination. The reader should be able, afterwards, to point at every component and say they saw it. What they could not do is assemble them, because the assembling required a piece of judgement — a decision by a character, a shift of vantage, a recognition — that only the ending supplies.

This is Aristotle’s anagnorisis doing structural work. The recognition is not new information arriving; it is existing information reorganised. And it is why the strongest twist endings are also the most heavily planted: the elements were all present, in the open, doing other apparent jobs.

Two practical consequences.

(1) Plants should not look like plants. An element inserted so that the reader will notice it has been half-spent already. Genette’s term for the unemphatic advance notice is the amorce — a seed that passes as ordinary and becomes significant only retrospectively (Chapter 10 §6.2 uses the same term for the prologue’s fragments). The good pen going into the coat pocket in the household narrative’s prologue is an amorce: at the time it is a small oddity about a man’s habits.

(2) The test for a plant is not “did the reader notice?” but “can the reader find it?” These are different questions with different answers, and the second is the one the earned/unearned distinction actually depends on. A reader who reaches the last page and cannot locate the preparation feels cheated. A reader who located it in advance feels clever. A reader who could not have located it in advance but can locate it now is having the intended experience.

6.3 Aristotle’s necessity or probability, and the plant economy

The ancient form of the earned test is the requirement in the Poetics that the incidents of a plot follow one another by necessity or probability. Aristotle is precise about the alternative he is rejecting: a sequence in which events follow one another merely after rather than because. The distinction is the same one Chapter 2 (Scene and Sequel) makes with the causal chain — the test of whether each beat “had to happen next” — and applying it at the scale of the whole book gives the earned test its logical form.

Two refinements from the Poetics are worth carrying into revision.

“Or probability” is doing a great deal of work. Aristotle does not require that endings be necessary; probable is enough. This is a lower bar than modern craft discussion usually assumes, and it licenses a good deal. An ending need not be the only possible outcome; it needs to be one that the established characters, in the established world, would plausibly produce. Chapter 14’s Levin-and-Anna problem is the standard case: nothing necessitates Levin’s arrival at faith, and the novel’s claim is only that a man like that, in that situation, might.

The plot’s own structure is the source. Aristotle’s stipulation that the unravelling arise from the plot itself rather than from a contrivance is the ancestor of the whole modern prohibition on the deus ex machina, and it contains a discrimination that the modern version has lost. What Aristotle objects to is not implausibility but externality: an outcome produced by something outside the causal system the play has built. A wildly improbable event generated by the plot is acceptable; a probable one imported from outside is not. This is why coincidence at an ending is so much more damaging than coincidence at a beginning — a point worth teaching as a rule, since it explains a large asymmetry that puzzles students:

Coincidence may start a story and may not end one. A chance meeting on page one is a premise; the same meeting on page four hundred is a contrivance. The reason is causal position: an opening coincidence generates the causal chain, and a closing one substitutes for it.

The plant economy, formalised. This book has built the machinery for this in the chapters before this one and it can now be stated as a single system.

The whole system has one purpose: to make the fourth stage possible without the reader detecting the first three as preparation. And it produces the chapter’s most testable single instruction:

Nothing new and load-bearing arrives at the ending. Everything the ending uses was deposited, maintained, and instalment-paid; what the ending contributes is not material but combination.

That instruction has a limit, and the limit should be stated because students over-apply the rule. It prohibits new load-bearing elements, not new elements. An ending may perfectly well contain a room the reader has not seen, a minor character who has not appeared, a sentence about a bus going past. What it may not do is rest on them. The test is substitution: if the element were removed, would the ending still work? If yes, it is furniture and is welcome. If no, it must have been planted.

6.4 A taxonomy of sixteen closural strategies

The strategies below are adapted from Torgovnick’s typology of closural relations (§3.5) and Herrnstein Smith’s account of closural devices (§3.2), with additions from practice. They are not mutually exclusive — most strong endings braid two or three — and they are not ranked. Each is a shape the ending can take, with characteristic purchases and characteristic costs.

Four clusters behave alike in revision.

• Strategies 1, 2 and 5 are return strategies. They close by revisiting. They are cheap (the material is paid for), reliable, and require nothing new. They are also the strategies most likely to be available in a draft that already has an ending problem, and the first place to look.
• Strategies 6, 7, 8 and 14 are displacement strategies. They close by moving — in time, voice or register — away from the narrative. They buy perspective and cost immediacy, and every one of them opens the question Chapter 10 asks of prologues: does this material require its exteriority?
• Strategies 9, 10, 11 and 12 are terminal strategies. They are decisions about the last page rather than the last chapter, and they can be substituted for one another almost freely without touching the structure. This makes them the cheapest thing to experiment with and the least likely to fix a structural problem.
• Strategies 3, 4, 13, 15 and 16 are disruptive. They close by refusing the expected relation between ending and book. They have the largest effects, the highest failure rate, and — this is the significant point — they are almost never available as a repair. A book either was always going to end this way or cannot be made to.

6.5 How to use the taxonomy: substitution testing

The taxonomy’s purpose is not classification. Knowing that a draft ends with a proleptic close is worth nothing by itself. Its purpose is substitution testing, as with Chapter 9 §6.5’s opening gambits and Chapter 10 §6.4’s prologue types.

The procedure:

1. Name the strategy the draft currently uses. Most drafts use one without having chosen it, and the commonest unchosen default in contemporary literary fiction, as far as this book can judge without a count, is 9 (the terminal image) and in commercial fiction 8 (the proleptic close).
2. Write the same ending in two other strategies, chosen from a different cluster. Not sketches — actual prose, at least a page each.
3. Do not judge until all three exist. The finished draft always feels correct; that is a property of finish, not of fit.
4. Ask what each version makes the book about. This is the diagnostic question, and it is more informative than asking which is best. An ending in strategy 6 makes a book about consequences; the same material in strategy 12 makes it about a moment; in strategy 3 about a world that continues regardless.

The exercise is set formally at §12.1, exercise 2. Its reliability is high enough to be worth a claim: in a workshop setting, by this book’s estimate rather than any measurement, roughly a third of endings that “aren’t working” are the right material in the wrong strategy, and the substitution locates the problem faster than any amount of line-editing.

A specific and common finding is worth flagging. Drafts that feel too tidy are frequently strategy 6 or 8 material that wants to be 2 or 9 — a novel that has written a finale when it needed a parallel scene. And drafts that feel abrupt are frequently strategy 12 that wants to be 6: the writer, correctly wary of the endless aftermath, has stopped at the climax and left four domains unserved.

6.6 Climax and aftermath: the proportion problem

The most consequential structural decision at an ending is not which strategy but how much aftermath. Both errors are severe and they are opposite (§8.2, §8.3), which means the correction for one produces the other, and drafts commonly oscillate.

The framing that resolves the oscillation is Aristotle’s (§3.3): the aftermath is not an appendage to the structure but the lusis, the second of two movements. The question is therefore not how much can I get away with? but what does the unravelling have to do?

What only the aftermath can do:

What the aftermath must not do:

The operational rule the chapter offers on proportion:

The aftermath ends at the last decision, not at the last consequence. Consequences are infinite; decisions are countable. Find the final decision any character takes — however small, and it may be as small as putting on a coat — and end within a page of it.

This rule is Chapter 2’s (Scene and Sequel) sequel structure applied to the whole book (§2.3.1): reaction, dilemma, decision. A book’s aftermath that stops at reaction feels indulgent; one that reaches a decision feels complete; one that continues past the decision into further consequence feels like a failure to notice that it had finished.

How much do the traditions permit? The following is descriptive and approximate, derived from the works discussed in this chapter. It is offered as a corrective to the assumption that “short aftermath” is a rule rather than a period preference.

Two things follow. First, the range is enormous, which means no absolute prescription can be right. Second, the contemporary literary norm is at the extreme low end of the historical distribution — which is worth knowing before concluding that a draft’s forty-page final movement is self-indulgent. It may be Victorian.

6.7 Historical drift, and a caution against universalising the present

§3.7 laid out the history. This section states its consequence for the advice in this chapter, in the same terms Chapter 9 §6.7 and Chapter 10 use, because it is this book’s standing intellectual obligation.

Almost every prescription in §§7 and 8 is calibrated to Anglo-American literary and commercial practice since roughly 1960. That is a real tradition with real standards, and a writer working inside it needs to know them. It is not narrative.

Specifically:

The properly critical position is the same double one this book has maintained throughout: learn the current conventions well enough to break them deliberately, and read historically well enough to know they are conventions. A pedagogy that can only say “a modern editor would cut this” has explained the market and not the book — and at the ending, more than anywhere else, the market’s preferences have hardened into an aesthetic without anyone deciding that they should.

One final observation, which §16 returns to. The contemporary preference for the unresolved close is frequently defended on the grounds that life does not resolve. This is true and is not an argument, for the reason Kermode gives (§3.1): novels are not like life in any other respect either, and the demand for shape is not a demand for accuracy. A novel that refuses to conclude because life does not conclude has confused mimesis with form. The good reasons for an open ending are formal and thematic — that this book’s material admits no settlement, that the refusal is the argument — and they are available. The mimetic reason is not one of them.