8. Failure Patterns

Each pattern below is presented with its symptom, its mechanism (what actually goes wrong, as distinct from what the surface looks like), its persistence (why competent writers keep producing it), its remedy, and — importantly — the exception: a published work that commits the fault and succeeds.

The exceptions are the intellectual heart of the section. A pedagogy of prohibitions produces writers who can avoid faults and not writers who can make choices. In every case the pattern fails because of a mechanism, not because of a surface feature, and where the mechanism is absent the surface feature is harmless.

8.1 The Arbitrary Cut

Symptom. The chapter ends because it reached a length. There is no proposition-and-payoff logic behind this particular placement, and moving the break two pages either way would make no perceptible difference.

Mechanism. The break has been treated as formatting rather than as a decision (§1.1). The reader receives an offered exit at a point where nothing has completed and nothing is pending, and the exit therefore carries no information. The damage is cumulative rather than local: one arbitrary cut is invisible, and thirty produce a book with no felt architecture, which readers describe as “meandering” and cannot locate.

Persistence. The chapter break is the only structural mark a writer makes that has no immediate effect on their own experience of the text. It is also the only one that can be made by pressing return. Nothing in the drafting process forces the decision to be conscious.

Remedy. Test 1 and Test 5. Write the eighteenth-century heading; if it cannot be written, the chapter has no subject. Then run the substitution test — very often the real boundary is two pages back.

The exception. Tristram Shandy (1759–67). Sterne’s chapters are of wildly arbitrary length and placement, and the arbitrariness is the argument: he devotes a passage in the fourth volume to the theory of chapter-making, announces that he has torn a chapter out and lets the pagination skip, and leaves two chapters of the ninth volume blank to be supplied later. The cuts are arbitrary in a novel about the impossibility of ordering a life into units, and every one of them is therefore doing the work an arbitrary cut normally fails to do. The distinction is the one this whole section rests on: Sterne’s cuts are arbitrary and legible as arbitrary. A draft’s arbitrary cuts are merely arbitrary, and the reader assumes they were meant.

8.2 Cliffhanger Fatigue

Symptom. Every chapter, or nearly every chapter, ends on unresolved danger.

Mechanism. The price of an exit is relative (§3.7). Where every exit costs the maximum, the reader recalibrates, and the device stops carrying information. Worse, the reader learns a procedural lesson — that this book always ends chapters this way — and begins to read the ending as a formal feature rather than as a fact about the situation. The result is the exact opposite of what was intended: the reader stops registering danger.

Persistence. It is the failure mode of good advice. Writers are told that chapters should end compellingly; the cliffhanger is the most legible way to be compelling; and the fault is invisible from inside any individual chapter, since each one is doing what it was told to do.

Remedy. Test 7, applied to eleven consecutive chapters. The corrective is not subtler cliffhangers; it is soft landings, installed deliberately at the end of movements, so that the cliffhangers regain their contrast.

The exception. Dumas’s The Count of Monte Cristo (1844–46), whose chapters end on unresolved turns with a regularity no modern editor would permit, and which is one of the most compulsively readable novels ever written. The reason is a delivery format: as a feuilleton, the novel was consumed one instalment a day, and a device deployed once every twenty-four hours does not fatigue. The fault is not in the frequency of the endings but in the ratio of endings to reading time, and the ratio was, in 1844, entirely different. A modern novelist who reproduces Dumas’s density is reproducing a solution to a problem they do not have. §9.3 works the case in full.

8.3 The Padded Chapter

Symptom. The chapter continues past its own ending. Material has been added — a further conversation, a second location, a stretch of reflection — that does not serve the chapter’s proposition.

Mechanism. Two overlapping failures. Structurally, the writer is filling a container to a habitual size (§5.1). Cognitively, the break has been placed after the reader’s own event boundary (§3.6): the reader closed the event at the point where the chapter’s question was answered, and everything since has been read as residue. This is the mechanism behind the universal manuscript-report complaint that a chapter “sags at the end”, and it explains why the sag is usually not where the writer looks for it.

Persistence. Padding is written, not left in. It is generated by the belief that a chapter must reach a length, which is generated by looking at the previous chapters, which were the same length for the same reason.

Remedy. Test 6, and then the crudest available operation: delete the last page and read what remains. If nothing is lost, the chapter ended a page ago. Repeat.

The exception. Moby-Dick (1851). By any proportion test, Melville’s cetological chapters are padding of the most flagrant kind: a chapter classifying whales by book format, a chapter on the whiteness of the whale, a chapter on the whale as a painted subject, none of which advances the hunt. They are also the reason the novel is what it is. The exception’s mechanism is this: the material is not serving the chapter’s proposition because Melville has made the digression itself the book’s proposition, and the segmentation announces it — each excursus is given its own chapter and its own title, so that the reader is never in doubt that the digression is deliberate architecture. Padding pretends to be story. Melville’s chapters do not pretend to be anything.

8.4 The Anticlimactic Soft Landing

Symptom. A chapter that should hold its tension open — mid-crisis, mid-decision, mid-danger — resolves instead, releasing pressure the book needed.

Mechanism. The payoff operation chosen is resolve where it should have been hold open or escalate. The specific damage is that a novel’s tension is cumulative, and each premature discharge resets the accumulation. A book with four consecutive premature resolutions will feel, at its midpoint, as though nothing is at stake, even though a great deal has happened.

Persistence. Writers finish things. The impulse to tidy a chapter before leaving it is a good impulse in most contexts, and it is very hard to see as a fault when the tidying is well written.

Remedy. Test 3, with the second part completed. Then the substitution test: move the break earlier, to the moment before the resolution, and read what the chapter now offers.

The exception. Tolstoy’s Anna Karenina (1878) and Eliot’s Middlemarch (1871–72) both resolve at the ends of chapters where a suspense-driven novel would hold, and do it constantly. The reason it works is that neither novel’s engine is suspense. Their propositions are moral and social, and moral propositions do not decay when local situations are settled: a chapter can resolve an argument completely and leave entirely intact the question of what kind of person Dorothea is becoming. Premature resolution is only premature relative to the kind of tension a book runs on, and a novel whose tension is characterological can afford to close its events.

8.5 The Buried Pivot

Symptom. A reveal or a turn is delivered in the middle of a chapter, and the chapter continues afterwards.

Mechanism. Rabinowitz’s rule of privileged position (Chapter 9 §3.2) confers emphasis on terminal material for free. A turn placed mid-chapter forfeits that emphasis and must generate its own, which costs words and usually fails. There is a second and subtler cost: the material after the reveal is read at reduced attention, because the reader is still processing the turn, so the chapter’s last pages are effectively wasted.

Persistence. Reveals are written where they occur in the drafting, and drafting follows chronology. Nothing prompts the writer to ask whether the reveal could be moved to the end, and moving it usually requires re-ordering rather than rewriting, which feels like more work than it is.

Remedy. For every chapter containing a turn, ask: could this be the last thing in the chapter? If not, ask what the following pages are for. If the answer is “the reaction”, consider that the reaction may be the next chapter’s opening (placement strategy 2, §6.4).

The exception. The Good Soldier (1915). Ford’s narrator buries his revelations constantly — deaths, betrayals and the novel’s central adultery are dropped mid-paragraph, mid-section, in subordinate clauses, and then abandoned. The burial is the characterisation: Dowell cannot organise his material because he cannot face it, and a novel whose subject is a man’s failure to know what happened to him could not have its turns properly placed. The exception’s condition is that the mis-placement must be attributable to the narration rather than to the author. Where a reader can see that the narrator buried it, the burial is a technique; where they can only see that the book buried it, it is a fault.

8.6 The Uniform Chapter

Symptom. Every chapter is within a few hundred words of the same length, throughout.

Mechanism. §5.1. Uniform length asserts uniform tempo. Since almost no novel has a uniform tempo, the form contradicts the content quietly and continuously, and the book’s climax arrives at the same measured pace as its exposition. The reader does not diagnose this; they report that the book “doesn’t build”.

Persistence. Uniformity is the path of least resistance and looks like professionalism. It is also produced by writing software that shows a word count, by contracts that specify a length, and by the entirely reasonable desire to make chapters feel like a set.

Remedy. The audit at §5.6. Then the cheapest structural revision in fiction: split three chapters and merge two, without altering a sentence.

The exception. Calvino’s Invisible Cities (1972; trans. William Weaver, 1974) is built from units of almost identical brevity, arranged in a rigid numerical scheme, and the uniformity is the form: a catalogue’s entries must be commensurable or it is not a catalogue. Closer to the novels used elsewhere in this chapter, Steinbeck’s The Grapes of Wrath (1939) alternates narrative chapters with short intercalary ones in a pattern so regular that the reader learns to expect it, and the regularity produces the novel’s liturgical rhythm. In both cases the uniformity is perceptible as a scheme, which is what distinguishes a form from a habit. The test is whether a reader could describe the pattern. If they could not, it is not a pattern; it is a default.

8.7 The Chapter That Is Really Two

Symptom. A chapter whose first half and second half have different subjects, different locations, or different propositions, joined by a section break the writer treated as adequate.

Mechanism. Test 2 fails at the level of the chapter and passes at the level of each half. The reader experiences the second half as an appendix to the first, because the chapter’s opening question has already been settled and a new one has begun without the signal that ordinarily marks a new unit. The offered exit at the end is attached to material the chapter did not open with.

Persistence. It results from drafting forward without stopping, which is a good drafting habit. It also results from an unwillingness to write short chapters, since splitting produces two units of half the accustomed length.

Remedy. Test 9. If the seam exists, ask whether it is a better break than the current one. It usually is, because the seam is where the material itself divides.

The exception. The Victorian multi-plot novel does this constantly and by design. A Middlemarch chapter may move from Dorothea to Lydgate to a piece of narratorial generalisation; a War and Peace chapter may carry two households. The reason it works is that the novel’s unit of design is not the chapter but the Book or Part, and the chapter is a subdivision within it rather than a self-contained movement — which is the model the “one chapter, one movement” convention displaced. Where a novel has a functioning larger architecture, chapters are allowed to be untidy; where it has none, the chapter is the only shape the reader has.

8.8 The False Cliffhanger

Symptom. A chapter ends on apparent danger which is dispelled in the first paragraph of the next chapter — the noise was the cat, the figure was a friend, the fall was a stumble.

Mechanism. A breach of trust with a measurable cost. The reader invested attention in a threat that the book knew was not a threat. The first time this produces mild irritation; the second time it produces a policy, and the reader stops responding to chapter endings altogether. This is the mechanism by which a novel can contain a genuine cliffhanger in chapter 30 that fails, because chapters 6 and 14 spent the credit.

Persistence. It is the easiest way to produce the appearance of a strong ending without altering the story, and it therefore appears in drafts at the points where the story has nothing to offer.

Remedy. Test 4, run across the boundary: read the ending and the following page together, and ask whether the reader’s three days of anxiety would have been repaid. If not, the ending is a fraud and should be replaced with a type that does not promise (§4.4, §4.5).

The exception. Don Quixote, Part I, Chapters 8 and 9 (1605; trans. Edith Grossman, 2003, or John Rutherford, 2000). The chapter breaks off in the middle of a duel, swords raised — and the next chapter opens with the narrator explaining that the manuscript he was working from stopped there, and describing how he found the continuation in an Arabic manuscript in the Alcaná of Toledo. It is the most flagrant false cliffhanger in the history of the novel, executed four hundred years ago, and it succeeds because the deflation is the joke and the joke is the book’s method. The exception’s condition is that the trivial resolution must be more interesting than the danger would have been. Cervantes replaces a duel with a metafictional crisis about the reliability of the entire narrative. A draft that replaces a prowler with a cat has not made that trade.

8.9 The Recap Opening

Symptom. The chapter after the break opens by re-establishing what the reader already had: where they are, who is present, what just happened, and how the character feels about it.

Mechanism. This is a chapter-break fault presenting as a transition fault, which is why it is so often misdiagnosed. Chapter 5 (The Transition) §8.6 treats the recap bridge as an orientation problem. The version that belongs here is different: the recap is a symptom that the break was placed where nothing changed. If the next chapter has to re-establish the situation, the situation did not end; the writer cut in the middle of a continuing unit and then paid, on the far side, for a discontinuity they had invented.

Persistence. The writer, reading their own manuscript, feels the white space as a real gap (because they wrote the two chapters on different days) and orients accordingly.

Remedy. Test 8. Count the words before the new chapter’s business begins. Then try deleting the break entirely and reading the two chapters as one. If the join is invisible, the break was doing nothing.

The exception. Serial publication, where the gap is real. A weekly instalment appearing seven days after the last one is entitled to remind its reader who is speaking, and Victorian serialists did so routinely; the same licence applies to a contemporary web serial, a newsletter novel and a long-running series (§10.4). The distinguishing question is whether the reader has actually been away. Recap is a fault in a codex and a courtesy in an instalment, and the same paragraph can be either depending on how the book is delivered.

8.10 The Sleep Ending

Symptom. The chapter ends with the viewpoint character falling asleep, going to bed, or otherwise ceasing to be conscious.

Mechanism. Herrnstein Smith’s closural allusion (§3.5) firing without anything behind it. Sleep, night and darkness are among the strongest terminal signals available in the language, and the reader receives a powerful sense of ending attached to no completion whatever. The chapter therefore feels finished while having resolved nothing — and, because the signal is so strong, the exit is maximally cheap at the moment the writer has stopped thinking about it. A secondary problem is repetition: a novel that runs on daily chronology will do this at every chapter unless the writer intervenes.

Persistence. It is where the day ends, and drafting follows chronology. It also feels like a natural break, and it is one — for the character. The reader is not tired.

Remedy. Move the break earlier, to the last moment at which the character wanted something. Sleep is almost never the chapter’s real boundary; it is the point past the boundary at which the writer stopped typing. If the sleep must stay, give the last thought a job: the last thought before sleep can do work that the sleep itself cannot.

The exception. In Search of Lost Time (1913–27; trans. C. K. Scott Moncrieff, revised by Terence Kilmartin and D. J. Enright; or Lydia Davis and others, 2002). Proust’s novel opens with a man who has long gone to bed early, and its first great movement is a study of the boundary between waking and sleep — the confusion of rooms, the collapse of chronology, the body’s memory of positions. Sleep works as a boundary here because sleep is the subject, and the segmentation of consciousness is what the novel is about. The general principle: a closural allusion is a fault when it is borrowed and a technique when it is thematic. The same holds for the other members of the family — chapters ending on departures, on nightfall, on a door closing, on a train pulling out. Each is free closure, and free closure is only worth having when the chapter has earned it.

8.11 The Epigraph Doing the Chapter’s Work

Symptom. A chapter is headed by a quotation that states its theme, and the chapter then demonstrates the theme.

Mechanism. The epigraph is paratext (§6.6) and is read first. Where it states the chapter’s meaning, the chapter’s own work is pre-empted: the reader arrives already knowing what to conclude and reads for confirmation rather than for discovery. This is the primacy effect used against the writer’s own material. There is a second cost — the epigraph is usually by a better writer, and the comparison is invited on the page.

Persistence. Epigraphs are enjoyable to collect, they make a manuscript look serious, and they solve, apparently for free, the writer’s anxiety that the chapter’s point may not be visible.

Remedy. The deletion test at §6.6. And a sharper diagnostic: if the epigraph could serve as the chapter’s title, it is doing the chapter’s work. If it is in tension with the chapter, it is doing its own.

The exception. Middlemarch. Eliot’s mottoes, many of them written by herself in an assumed idiom, are among the most successful chapter epigraphs in English because they are counterpoint rather than caption. A motto proposes a general truth and the chapter tests it, so that the epigraph and the narrative are in argument. Herbert’s Dune (1965) offers the other legitimate model: its epigraphs come from within the fictional world and deliver information — including foreknowledge of the outcome — that no narrator could supply without intrusion. The distinction is between an epigraph that tells the reader what the chapter means and one that gives the chapter something to disagree with.

8.12 The Undifferentiated Sequence

Symptom. A long novel with forty or sixty chapters, no Parts, no titles, no larger architecture — or the inverse: a novel whose Part boundaries fall where nothing changes.

Mechanism. §6.6. Readers use the larger divisions to build a model of the book’s shape, and to locate themselves within it. A long novel with no larger divisions gives them nothing to hold except a running count, and the effect is a loss of proportion: the reader cannot tell whether they are in the second act or the third, and therefore cannot feel the approach of the end. Kermode’s closural expectation has nothing to attach to.

The inverse fault is worse, because it is misinformation. A Part boundary is a promise that something has changed; where nothing has, the reader adjusts their expectations and is not met.

Persistence. Parts are added last, if at all, and by then the writer knows the book too well to see it as a shape. Many drafts acquire Parts by being divided into three at roughly even intervals, which produces boundaries that fall wherever the word count says.

Remedy. Test 10. Compile the headings and last sentences and ask where the book actually changes. If there are three such places, there are four Parts; if there are none, the book may not want Parts, and the honest response is to say so rather than to impose them.

The exception. One Hundred Years of Solitude (1967; trans. Gregory Rabassa, 1970) has twenty chapters, no numbers, no titles and no Parts, and it is one of the most architecturally controlled novels of the century. The absence is functional: a novel about generations repeating, in which characters share names across a hundred years, deliberately denies the reader the apparatus that would let them keep their place. Disorientation is the represented experience, and a table of contents would undo it. The condition on the exception is the usual one — the reader must be able to feel that the absence is a decision. Márquez’s chapters are so distinct in movement that no reader mistakes the lack of numbering for carelessness.

8.13 Summary: symptom to mechanism