6.1 The decision, and its historical shape
The central craft decision in this territory is not whether to include the past. Every novel includes the past; a narrative whose characters have no history is not a narrative but a diagram. The decision is which register the past is delivered in, and the wrong choice is felt by readers who could not name what is wrong.
The two characteristic errors are symmetrical and equally common.
- A flashback used where backstory would do reads as bloated. The reader is asked to live through a moment that only needed to be reported, and the felt effect is that the book has lost confidence in the reader’s ability to take a fact on trust.
- Backstory used where a flashback was needed reads as flattened. A moment the reader should have felt has been summarised into a fact, and the emotional weight it was supposed to carry never arrives — so that when the present scene depends on that weight, the dependency fails silently, and the writer diagnoses a problem in the present scene.
The second error is the harder to detect, because its symptom appears somewhere other than its cause.
The historical position. Percy Lubbock’s The Craft of Fiction (1921) gave this decision its first sustained English treatment under the opposition of scene and picture — the dramatised moment against the summarised stretch — and made the case, following Henry James, that the novelist’s essential art lies in knowing which the material wants. Lubbock’s version is now dated in its confidence that scene is superior, an assumption the twentieth century’s fondness for show, don’t tell inherited and hardened, and one this chapter rejects: the two registers are a fork, not a hierarchy.
Genette’s treatment of duration supplies the more precise vocabulary. His four canonical relations between story time and narrative time — ellipsis (story time passes, no text), summary (much story time, little text), scene (approximately equal), and descriptive pause (no story time, much text) — describe a continuum along which any passage sits. Backstory occupies the summary band; flashback occupies the scene band. Chapter 5 (The Transition) uses the same apparatus for movements in the present; here it is being applied to the past.
The distinction between the two chapters’ use of Genette needs stating, because it is a common source of confusion. Chapter 5 (The Transition) asks how much text to spend crossing a gap. This chapter asks how much text to spend on the far side of it. A hard cut into a two-page flashback and a hard cut into a one-sentence summary are the same transition and radically different cutaways.
6.2 The fork
Backstory — the past, told. Costs little present-tense momentum. Right when the reader needs to know a fact.
Flashback — the past, dramatised as a scene. Costs real present-tense momentum. Right when the reader needs to feel a moment, not merely be informed of it.
Two elaborations, both practical.
A flashback is subject to every rule that governs a scene. Chapter 1 (The Purposes of a Scene) applies in full: someone wants something, meets resistance, and something is different at the end. This is not a stylistic preference. A past passage without a want and a resistance has no internal engine, which means the reader is being asked to spend attention on material that is not going anywhere, at a moment when they have already spent attention getting there. A flashback without a value shift is the most expensive form of summary available.
The fork is not a spectrum to compromise between. A passage that is neither a proper summary nor a proper scene — three paragraphs of half-dramatised past, with a fragment of dialogue and a stretch of précis — is the commonest form the failure takes, and it is worse than either pure option. It has the length of a scene and the engagement of a summary. In revision, the first move on a suspect cutaway is often not to improve it but to push it decisively to one end of the fork and see which version the book prefers.
6.3 The one-sentence compression check
The operation that resolves the fork is a single test, and it is the most reliable instrument in this chapter.
The one-sentence compression check. Compress the cutaway to one sentence of summary. Read the surrounding material with that sentence in place of the original passage. If nothing of consequence is lost — no experience, no felt shift, nothing the reader needed to live through — the material was backstory that got over-dramatised. If the compression flattens something essential, the passage was earning its place as a flashback.
Three notes on running it properly, because it is easy to run badly.
Write the sentence out. The check does not work as a thought experiment. A writer imagining the compressed version imagines it carrying everything the full version carried, because they can still see the full version. The sentence must exist on the page and be read in place.
Read it in context, not in isolation. The question is not whether the sentence is good. It is whether the passage two pages later still works with only that sentence behind it. If the later passage depends on an emotional weight the summary cannot supply, the compression has failed and the flashback is earned.
Expect the check to produce a third answer. Some passages compress to one sentence with nothing whatever lost — including the reason for their existence. That is not a verdict of “make it backstory”; it is a verdict of “cut it”. The check has three outcomes, not two, and the third is the most common in first drafts.
A rough restatement, for use when the check is inconvenient:
| If the weight lives in… | Use | Because |
|---|---|---|
| Knowing the fact | Backstory | The reader needs the information, not the occasion |
| Experiencing the moment — a decision made, a line said, a value shifting in real time | Flashback | A summarised value shift is not a value shift; it is a report of one |
| Neither | Cut | The passage is answering a question nobody asked |
6.4 A taxonomy of the registers of the past
The backstory/flashback fork is a binary, and a binary is too coarse for practical use, since it describes only the degree of dramatisation. What follows is a working repertoire of fourteen registers — recognisable ways of putting the past on the page, each with characteristic returns and characteristic costs. They are not mutually exclusive; strong novels braid several, and one of the marks of an assured book is that it can be seen choosing.
| # | Register | Description | Exemplar | Buys | Costs |
|---|---|---|---|---|---|
| 1 | The clause of implication | A subordinate clause or phrase that implies a history without narrating it | Mrs Dalloway (the illness, the war, in passing); The Remains of the Day (Darlington Hall “of those days”) | Enormous density; zero momentum cost; the reader does the work | Nothing can be felt; unusable for material that must land |
| 2 | The participial aside | An appositive or participial construction carrying an antecedent inside a present sentence | Middlemarch’s narrator; Austen’s introductions | Delivers biography without stopping; keeps the narrator’s authority visible | Requires a narrator with licence to comment; alien to close third |
| 3 | The summary paragraph | A stretch of narrated antecedents in Genette’s summary band | Père Goriot; Middlemarch Book I | Covers long extents cheaply; the natural home of a long reach | Reads as a report if extended; no experience delivered |
| 4 | The iterative habitual | The repeated action narrated once — every Sunday he would… | Proust’s Combray; the servants’ hall in The Remains of the Day | A whole era in a paragraph; characterises by pattern rather than incident | Cannot carry a singular event; a habit has no value shift |
| 5 | The dramatised flashback | A full past scene, moment by moment | Beloved (Sweet Home); The Remains of the Day (the father’s death) | The reader lives it; the only register that can transfer emotional weight | The most expensive unit in the repertoire; suspends the present entirely |
| 6 | The embedded oral tale | A character narrating the past at length, to a listener, within the story | Wuthering Heights; Season of Migration to the North; Frankenstein | The teller’s bias becomes content; a listener supplies motive and interruption | Two narrators to sustain; the frame must justify the occasion |
| 7 | The found document | Letter, ledger, diary, transcript, newspaper clipping | Dracula; Frankenstein; The Handmaid’s Tale’s closing apparatus | Authenticity; a different diction; the past as evidence rather than memory | The document must be plausible as itself; hard to sustain at length |
| 8 | Involuntary memory | The past arriving unbidden, triggered by sensation | In Search of Lost Time (the madeleine) | The past as something that happens to a character; unmatched authority | Not motivated by plot; hostile to the workshop’s trigger requirement (§6.6) |
| 9 | The repeating analepsis | A return to already-narrated material, to recolour it | The Good Soldier; Beloved | Each pass reinterprets; builds meaning by accumulation | Highest deficit in the taxonomy; every pass must differ visibly |
| 10 | The braided timeline | A second strand in the past, alternated with the present on a declared rhythm | Beloved; The Forgotten Garden (2008) | Sustained parallel argument; each strand can end on tension | Both strands must progress; the weaker strand becomes the skipped one |
| 11 | The reconstructed past | The past assembled from conflicting testimony, by narrators who did not witness it | Absalom, Absalom!; The Good Soldier | Makes the act of reconstruction the reader’s activity; epistemology becomes plot | Difficult; the reader must hold several accounts; not for material that must be certain |
| 12 | The unreachable past | The past approached and refused; the scene withheld | Beloved (the shed, deferred); The Sound and the Fury (Caddy, never focalised) | The gap becomes the subject; maximum curiosity in Sternberg’s sense | Only works if the refusal is legible as a refusal; one step from coyness |
| 13 | The false memory corrected | A past account established, believed, then overturned | Great Expectations (the benefactor); Emma | Recruits the reader into a misreading and then charges them for it | Requires a specific false belief to have been planted, not merely a blank |
| 14 | Prolepsis | Narration in advance of a later event | One Hundred Years of Solitude; The Death of Ivan Ilyich (the death announced first) | Converts the whole intervening book into a completing structure (§3.3) | Can neutralise suspense; the promised event must be worth the wait |
Notes on the registers that need them
Registers 1 to 4 are the underused half of the taxonomy. In a typical draft, the past arrives almost exclusively in registers 3 and 5 — the summary paragraph and the dramatised scene — with the other twelve unattempted. The clause of implication in particular is close to free: it costs no momentum, breaks no frame, and a novel can carry dozens of them without any accumulating sense that it keeps going backwards. A useful drafting discipline is to attempt every piece of past material at register 1 first, and promote it only when register 1 demonstrably fails.
Register 4, the iterative, deserves its own paragraph because it is the one most often missing and the one with the highest ratio of return to cost. Genette’s analysis of narrative frequency distinguishes the singulative (telling once what happened once), the repeating (telling several times what happened once), and the iterative (telling once what happened many times) — and identifies the iterative as one of Proust’s signature instruments. Every Sunday he would open the shop an hour early and count the till twice delivers a character, an era, a set of values and a rhythm in one sentence, and it does something the singulative cannot: it establishes what was normal, which is the necessary background for any departure from it to register. A novel with no iterative past has no baseline, and its dramatised past scenes therefore have nothing to be exceptions to.
Register 8 complicates the rest of the chapter, and §9.1 develops it. Proust’s involuntary memory is not the workshop’s motivated trigger. The taste of the madeleine has no plot relation to what it recovers; the connection is contingent, private, and the narrator himself cannot at first identify it. The device works by being unmotivated in the causal sense and overwhelming in the experiential sense. A writer who has learned that every cutaway needs a motivated trigger has learned a rule that would forbid the most famous cutaway in European fiction.
Register 12 is the hardest to execute and the most misused. Withholding a scene the reader wants is only powerful when the reader can tell they are being refused, and can tell why — because the material is unspeakable within the represented consciousness, because no witness survives, because the teller will not. Chapter 9 §8.6 makes the distinction under the heading of the withholding tease: is the gap in the character, or in the author? The temporal version of the question is identical.
Registers 11 and 13 are frequently confused. A reconstructed past is offered as uncertain; the reader is told, continuously, that this is an assembly. A false memory corrected is offered as certain and later withdrawn. The first makes the reader a collaborator; the second makes them a mark. Both are legitimate; they produce entirely different relationships with the narrator, and choosing between them is a decision about reliability, not about plot.
6.5 How to use the taxonomy: substitution testing
The taxonomy’s purpose is not classification. Naming a passage a register-5 dramatised flashback tells a writer nothing they did not know. Its purpose is substitution testing, on the same principle as the gambit substitution described in Chapter 9 §6.5.
The procedure:
- Take one cutaway that is not working, and identify its current register.
- Rewrite the same material in three other registers, chosen from the list. At least one should be considerably shorter than the original and at least one considerably longer.
- Do not evaluate until all four versions exist.
- Read each in place, with the surrounding pages, not in isolation.
Three substitutions that reliably teach something:
| From | To | What it reveals |
|---|---|---|
| Dramatised flashback (5) | Clause of implication (1) | Whether the scene was carrying weight or merely information. Most drafts discover the clause is sufficient |
| Summary paragraph (3) | Iterative habitual (4) | Whether the material was an event or a pattern. Summaries of patterns are almost always improved by the iterative |
| Any register | Embedded oral tale (6) | Who, in the present, would have a motive to tell this — and whether anyone would. If nobody would, ask why the reader is being told |
The third is the most revealing and the least used. Converting a piece of past material into something a character says aloud, to another character, for a reason, forces two questions the flashback register lets a writer avoid: why now and why to this person. A great deal of unmotivated backstory turns out, under this substitution, to be material nobody in the book has any reason to raise — which is diagnostic.
6.6 Triggers and returns
Having chosen a register, a writer faces two further decisions specific to the backward crossing: what motivates the departure, and what condition the present is in on the return. Chapter 5 (The Transition) owns the mechanics of the crossing itself — the anchor, the drift, the orientation budget — and this section handles only what is peculiar to going backwards.
The taxonomy of triggers
A trigger is the element in the present scene that motivates the cutaway. Its function is not decorative: it converts an authorial decision into a represented event, so that the reader experiences the departure as something happening in the story rather than something happening to the manuscript.
| Trigger | How it works | Strength | Risk |
|---|---|---|---|
| The object | A physical thing present in both timelines | Strongest available; the object continues to exist after the return, carrying the past with it | Becomes mechanical if every cutaway has one |
| The sensation | A smell, taste, sound, quality of light | Powerful and pre-verbal; the Proustian case | Clichéd in its common forms (baking, perfume, a particular song) |
| The phrase | Something said in the present that was said before | Cheap; can be exact; sets up verbatim repetition later | Requires the earlier use to be memorable enough to register |
| The question asked | Another character asks about the past | Motivates the content as well as the departure | Tends to produce the spoken register (6), which may not be wanted |
| The place entered | A location that holds the past | Automatic and reliable; the place does the work | The most over-used trigger in draft fiction after the photograph |
| The silence | An absence, a pause, a thing not said | Subtle; suits interiority-led fiction | Hard to make legible; the reader may not notice the motivation at all |
| The date | An anniversary, a season, a recurrence | Motivates without an object; useful when nothing physical survives | Can feel contrived if the calendar does too much work |
| The document | A letter, ledger or record read in the present | Delivers register 7 and its own trigger simultaneously | The reading must itself be motivated |
| The resemblance | Someone present who recalls someone absent | Efficient; carries emotional charge automatically | Risks sentimentality; overworked in family fiction |
| None | The narrative simply goes | The choice of confident omniscient narration, of Tolstoy, and of Proust’s involuntary case | Reads as authorial convenience unless the narration has established the authority to do it |
Three points about the list.
The object is the strongest because it survives the return. A sensation is over when it is over; an object is still in the character’s hand on the far side of the cutaway, which means the return sentence has something to work with. This is why the running example at §9.9 is built on a pen rather than on a smell.
The requirement of a trigger is convention, not law, and §3.8 explains where the convention comes from. The last row of the table is not a failure mode; it is the practice of most fiction written before about 1900 and of a great deal written since. What a confident narrator can do without a trigger, a close third-person narration usually cannot — not because of any rule, but because close narration has promised to stay inside a consciousness, and an untriggered departure breaks that promise. The trigger requirement is a consequence of narrative distance, not an independent rule, which is why it binds contemporary close-third fiction so tightly and classic omniscient fiction not at all.
Over-triggering is a real defect. A trigger that is too neatly apt — the woman looks at a broken clock and remembers the day time stopped for her — announces the machinery. The best triggers are slightly oblique: the object is connected to the memory by something private to the character, and the reader infers the connection rather than being handed it.
The return
The return is the least revised part of a cutaway and the one that settles its debt, and its requirement can be stated as a single rule.
The present must be changed by the time the narrative comes back to it. Not merely resumed — changed, in a way the reader can locate.
There are four kinds of change, and they are not equally strong.
| Kind of change | Description | Strength |
|---|---|---|
| Changed knowledge | The reader now knows something that recolours the present scene | Weakest on its own; the change is in the reader, not the story |
| Changed understanding | The viewpoint character now understands something they did not | Stronger; the character has moved |
| Changed question | The present scene now has a different question in it from the one it had before | Strong; converts the cutaway from a pause into a turn |
| Changed action | The character does something different from what they were about to do | Strongest; the cutaway has altered the plot |
The changed question is the most generally available of the four and the most underused. It requires nothing dramatic: the character simply comes back to the present with a question they did not have, which the present scene now has to carry. In the worked example at §9.9, Nadia returns from four lines of the past holding a question about her father — what did he think anyone would need to check — that the scene in the flat did not contain before she went. Nothing has happened; everything has changed.
On re-anchoring, and to avoid duplicating Chapter 5 (The Transition): the return needs an anchor as any transition does, and the anchor should generally be an element of the present the reader already has — the coat, the doorway, the object still in the hand. What is specific to the backward crossing is that the anchor may need to do less work than after a forward gap, because the reader never lost the present; they were holding it in suspension. Over-anchoring on the return — a full re-establishment of a scene the reader was in two pages ago — is the temporal equivalent of the throat-clearing bridge in Chapter 5 (The Transition), and it is common because writers correctly sense that a return needs marking and incorrectly conclude that it needs explaining.
6.7 Historical drift, and a caution against universalising the present
The conventions governing the past in fiction have moved substantially, and a student should be able to historicise the advice in this chapter rather than receive it as timeless.
| Period | Dominant handling of the past |
|---|---|
| Classical epic | The excursus and the character’s recital; the past fully illuminated, no suspended background (Auerbach, §3.8) |
| Early modern prose | The interpolated tale; the found manuscript; formal detachability taken for granted (Don Quixote) |
| 18th century | Documentary frames; the editorial apparatus; the biography-from-birth as the default shape (Tom Jones, Tristram Shandy satirising it) |
| 19th century | Preliminary and summarised exposition; the narrator’s authority to supply antecedents directly; the long oral chronicle (Wuthering Heights); the genealogical opening |
| Modernist | Chronology dissolved; memory as psychological rather than calendrical (Proust, Ford, Faulkner, Woolf); Mendilow’s subject |
| Mid-century to present | The cinematic flashback as default: dramatised scene, motivated trigger, marked entry and exit, changed return |
| Contemporary | The braided dual timeline, common in commercial and literary fiction alike (The Forgotten Garden, 2008; Where the Crawdads Sing, 2018); the past as a second book |
Two inferences follow, and they pull in different directions, which is the point.
The prohibitions in §8 describe a convention, and a recent one. The requirement of a motivated trigger, the suspicion of nesting, the demand for a changed return, the hostility to the long recital: none of these is a fact about narrative. They are the working rules of a particular literary culture — broadly Anglo-American, broadly since 1960, heavily influenced by screen conventions and by the workshop’s preference for the dramatised scene. They are worth knowing because a writer publishing into that culture is read by editors trained in it.
And the works that violate them are not tolerated exceptions. Wuthering Heights nests three deep. In Search of Lost Time triggers on a biscuit and stays away for a thousand pages. Absalom, Absalom! is almost entirely reconstruction by people who were not there. The Remains of the Day makes its present timeline deliberately the thinner of the two. These are not books that succeed despite their handling of the past; their handling of the past is what they are.
The properly critical position is therefore double, and identical in form to the one Chapter 9 reaches about openings: learn the current conventions well enough to break them deliberately, and read historically well enough to know that they are conventions.