The Father opens as a domestic drama and gives no sign of being anything else. A daughter lets herself into her father's London flat. He has driven off another carer, and he is certain the woman stole his watch. The daughter is tired and patient, and she tells him she is moving to Paris. The flat is solid, well lit and full of the furniture of a long life, and the scene is played straight: two people in a kitchen having an argument they have clearly had before. A viewer settles in for a film about an old man's decline, watched from a safe distance.

The next scene removes the distance. The father hears a sound, goes through to the sitting room and finds a man he has never seen, who tells him calmly that this is his flat and that he has been married to Anne for ten years. Then Anne comes back from the market with a chicken for dinner, and she is a different woman. A different face, a different voice, the same name and the same concern. The film does not cut away to show us which of the two is real, and it gives no sign that anything has gone wrong with the picture. We are exactly as lost as he is, and the film intends to keep us there.

Unreliable perception has been one of the dominant devices of popular fiction for a generation, and it usually arrives as a twist. The template is nearly a century old: a narrator the reader trusts, a lie the reader does not notice, and a late chapter that exposes it. The unreliability is a lie the book eventually corrects, and the correction is the payoff. The reader turns back through the pages, sees how the trick was done, and is given firm ground to stand on. The Father belongs to a smaller and older tradition with a different commitment. The audience experiences the unreliability while it is happening and is never given the outside vantage that would settle what was real. There is no confession, no second witness, no scene played from a position the troubled mind cannot reach.

The useful word here is focalisation, Gérard Genette's term for the consciousness through which a story is perceived, as distinct from the voice that tells it. The distinction is what lets a ghost story, a horror film and a film adapted from a stage play sit on the same shelf. The Turn of the Screw has a narrator and the two films have none, but all three hold to a single unreliable focaliser from the first scene to the last, and all three refuse to let anyone else's perception in.

Henry James got there first, in 1898. The Turn of the Screw ran in Collier's Weekly from January to April of that year, and it opens on a frame: a house party at Christmas, ghost stories by the fire, and a guest called Douglas who owns a manuscript written by his sister's governess, twenty years dead. The manuscript is the rest of the book. A young woman takes a post at a country house called Bly, in sole charge of two orphaned children, Miles and Flora, and begins to see figures: a man on the tower, the same man at the dining room window, a woman in black on the far side of the lake. The housekeeper, Mrs Grose, sees none of them, and the children never admit to seeing them either.

The evidence is balanced with great care. When the governess describes the man to Mrs Grose, with his red hair, no hat and long pale face, the housekeeper recognises Peter Quint, the master's valet, who is dead and of whom the governess had never heard. That is the strongest evidence for the ghosts, and it is strong. Nearly everything else in the book is the governess's interpretation of what the children's silences mean, and she interprets quickly and without doubt. When Flora is confronted at the lake with a woman the governess can see plainly, the child sees nothing and says so, and Mrs Grose sides with the child. Edna Kenton in 1924 and Edmund Wilson in 1934 argued that the ghosts belonged to the governess, and the argument has run ever since without resolution, because James made sure it could not resolve. In the preface he wrote for the New York Edition a decade later, he called the story "an amusette to catch those not easily caught".

The governess tells everything, including the moments that make her look frantic, and her candour is never in doubt. The narrator of a twist novel knows the truth and arranges the sentences around it. James's governess has nothing to arrange. Her perception is the unreliable instrument, and it is the only instrument the book provides, which is the difference between an unreliable narrator and an unreliable focaliser. It is also why a twist novel can end with the truth told and James's cannot. She believes every word she writes.

The book's boldest decision is the one readers notice least. The frame never closes.

The party by the fire, Douglas, the listeners who asked for the story: none of them comes back. The manuscript ends, and the book with it, on the governess holding Miles after the final confrontation, and on the report that "his little heart, dispossessed, had stopped." The whole argument comes down to one word, dispossessed. In one reading a possession has ended and the boy has been freed at the last moment. In the other he has been deprived, and her grip is what did it. The last word the book gives the child serves both readings equally, and no one from outside the manuscript is allowed back in to choose between them.

Black Swan, in 2010, carried the same commitment into a register James would not have recognised: psychological horror, with the body as its site. Nina, a dancer in a New York company, is cast as the Swan Queen and must dance both halves of the role, the innocent white swan and the black one who seduces the prince away from her. The camera stays close behind her for most of the film, often at her shoulder, and nothing reaches the audience that has not reached her first.

The ballet is itself a story of doubles, the enchanted swan and the sorcerer's daughter disguised as her, and the film lets the doubling leak into everything Nina sees. Her perceptions come apart one at a time:

• Her reflection moves a beat after she does.

• A woman passing her in a subway passage wears her face for an instant and then does not.

• She finds scratches on her back she does not remember making.

• She peels a strip of skin from beside a fingernail down to the knuckle, and in the next shot the finger is whole.

• She takes Lily, the rival dancer, home after a night out and sleeps with her, and the next day Lily says she went home with someone else.

The film does correct her, repeatedly, and the corrections are its cleverest device, because each one arrives through the same channel that produced the error. The finger is whole because Nina now sees it whole. Lily's denial is spoken to Nina, by a woman Nina has already cast as her enemy, in a scene Nina is perceiving. A correction delivered by the unreliable consciousness settles nothing, and the film knows it.

The largest correction comes at the end. On opening night Nina fights Lily in the dressing room, kills her with a shard of broken mirror and hides the body. She goes out and dances the black swan better than she has ever danced anything. Then Lily knocks at the dressing room door to congratulate her, alive, and the shard is in Nina's own stomach. It is the only correction the film grants in full, and it arrives too late to help. Nina dances the last act, falls to the mat, and tells the people gathering round her that she felt it, that it was perfect. The house is applauding. The screen goes to white before anyone can tell us whether she lives, or whether the performance she felt was the performance the audience saw. Every answer the film might give has been placed inside the one person unable to give it.

The Father is the most instructive of the three because it needs so little. It has no ghosts and no horror imagery: no blood, no transformation, no dread in the score. The film was adapted by its director from his own stage play, and it produces its entire effect with three plain tools:

1. Casting. Two actors play the daughter and two play her husband, and the film never marks the substitution.

2. Set dressing. The flat changes between scenes, a kitchen, the colour of a wall, a painting that hung there and now does not, so that the viewer begins to understand, well before the father does, that his flat and his daughter's flat and a third place are being laid over one another.

3. Editing. The same evening happens twice with different details, and the second time his son-in-law hits him.

Of the three, The Father comes nearest to supplying a floor. The last scenes take place in a care home, where a nurse tells him that his daughter lives in Paris and visits at weekends, and the film treats this as bedrock. But the nurse is played by the actor who earlier walked into the kitchen as Anne, and a male nurse who looks in on him is played by one of the men who claimed to be Anne's husband. The floor the film finally offers has been built from the same materials as everything above it.

Even the ending is his.

I have watched The Father three times. The second and third viewings were no easier than the first. I knew which evening would repeat and whose face would come through the door, and the knowledge did no good, because what holds the viewer is the absence of anywhere else to stand, and knowing the plot does not supply one.

A twist is used up by the first viewing. This technique is not.

The reason the sustained form is rare is that the twist offers something it cannot: a moment. A twist can be marketed, discussed, protected from spoilers and rewarded with a second reading in which everything looks different. The sustained form offers a condition instead, and a condition is harder to sell. A reader who finishes The Turn of the Screw has nothing to announce. What they have is a book that goes on arguing with itself, and a century and a quarter of other readers arguing alongside it.

Film has to fight its medium to work this way. The camera's default is objective, and an audience believes what it sees unless it is told otherwise, which is why both films labour so hard: recasting a daughter, repainting a kitchen, putting a shard in the wrong body. Prose has the opposite default. Every sentence of a novel is somebody's account, even in the third person, and a novelist does not have to fight the form to stay inside one mind. The effort goes into the ending, where the conventions of the thriller send for a second witness. The detective arrives, or the police file, or the other narrator, or the confession in the final pages, and the reader is handed back to solid ground. The contemporary novelist can decline to send anyone.

The form that refuses most naturally is the diary. The governess wrote her manuscript years afterwards, when she knew how it ended, and James let that knowledge colour every page. A diary has no afterwards. Each entry is written by someone who does not know what the next will say, for no reader, with no one on hand to object, and whatever corrections it contains come from the same hand on a later day, which is no more trustworthy than it was the day before. It is the form I chose for a novel of my own, for exactly this reason. A diary does not end when the matter is resolved. It ends on the last day its writer had anything to say, and nobody comes afterwards to explain.

Subscribe now

Gabriel Betteredge, house-steward of an English country estate, has been asked by his employer’s family to write down everything he remembers about the disappearance of a large diamond. Before he begins he consults his copy of Robinson Crusoe, which he has been consulting for thirty years as a guide to every difficult decision. He then spends some pages on his history with the book, the household, his daughter, his late wife, and his views on the relation between Sundays and shaving. The diamond does not appear for some time. The crime takes considerably longer.

Jack Reacher, a former military policeman travelling without luggage or a fixed address, is eating breakfast in a diner in a small American town when the local police arrive and arrest him. The first sentence puts him in the diner. The second brings in the threat. By the end of the first paragraph the reader knows where they are, who the protagonist is, and what is immediately at stake.

Both openings are the best of their kind, and each is close to the exact opposite of the other. What each one buys, and what each one costs, is a more useful question than which is better.

The eighty-page opening buys a relationship. The narrator’s reliability is constructed slowly, the world acquires texture, and the reader develops an interest in a particular consciousness. By the time the plot arrives, the reader has lived with the narrator for some hours and knows how to weigh what he says. This is structural insurance for everything that follows, paid for in advance.

What it costs is every reader who never reaches page eighty. The reader who decides at page thirty that nothing is happening is a reader the book cannot reach. In a reading economy where the sample chapter and the first ten percent decide most purchases, that is a real risk to take. The novel may be brilliant but the reader who closed it at page thirty will never find out.

Calling the long opening leisurely is wrong, because it is doing work. The Robinson Crusoe habit is not padding and it is not local colour. It is a test the reader runs on the narrator without being told they are running it. A man who consults a novel for guidance in every difficulty is absurd, and the reader registers the absurdity immediately. A man who admits to the habit in the first pages of a document he has been asked to write for a legal purpose is also, less obviously, transparent. By page ten the reader has learned that this narrator will report things that make him look ridiculous. That is precisely what the rest of the book requires of him, and there is no faster way to establish it. The reader watches him do it.

The one-paragraph opening buys the reader. Someone opening a thriller wants the contract settled immediately: this is the world, this is the protagonist, this is the stake. The contract is delivered, the reader is engaged, and the reader is still present in chapter ten when the plot becomes complicated. The opening has paid for the rest of the book.

What it costs is that insurance. A reader given a contract but no relationship is in the book for the plot rather than for the narrator, and if the plot weakens they will leave. If the narrator turns out to be complicated in ways the contract did not promise, they may feel misled. The fast opening makes promises the rest of the book has to keep, and those promises are about action and stake rather than depth.

The fast opening has its own version of the same trick, which is less often noticed. The arrest in the diner is not there only to start the plot. It establishes that the protagonist is the kind of man to whom this happens and who does not appear to mind very much, and it does that in the space it takes to describe a breakfast. Both openings are characterisation. One does it by accumulation and one does it by incident. The choice between them is a choice about how much of the reader’s patience the book can spend before it has earned any.

There is a middle position, and it is where most good contemporary novels sit. The Secret History opens with a confession of murder: the reader is placed inside the action, the voice is established, and specific promises are made about what the book will do. It does not run for eighty pages and it does not reach a crisis in three sentences. It uses the first chapter to set the voice and the central event, and then slows down and does the relationship work that the long opening would otherwise have done. Atonement manages something similar with its first long set piece: the dinner and the play and the visit to the fountain, which does contract work and relationship work at the same time. The reader is given enough plot to want the rest and enough texture to trust the writer. The calibration in both is very precise.

The standard explanation for the long Victorian opening is that it was a function of serial publication, and that no novelist paid by the instalment would have chosen otherwise. That is partly true, and the economics did support the long opening. But many of the great long openings were written by novelists with options. Eliot did not write the prelude to Middlemarch for the serial economics: she wrote it because it was doing structural work she could not do anywhere else. James, in the late novels, did not build his accreting openings because the magazines required them, but because the consciousness he wanted to render needed the time. The long opening is a tool that some writers chose when faster ones were available, because the slow one did something the fast one could not.

The mirror-image objection is that the one-paragraph thriller opening is shallow, and that the form does not produce literature. This is snobbery, and it does not survive contact with the books. A brisk first paragraph does not commit a novel to a brisk four hundred pages. Some of the best crime fiction of the past thirty years opens fast and then runs slow, and the slow running is where the literary work happens. The opening is a door. It tells you very little about the size of the house.

The contemporary novelist has both openings available, and choosing between them should follow from what the rest of the book needs. A novel that depends on the reader’s deep trust in a particular narrator needs more opening than the fast model allows. A novel that depends on the reader being in motion from the first line needs less than the Victorian model provides. Defaulting to either pole because it is what the form is expected to do is the only mistake.

One issue is that the fast opening has become close to mandatory for anything sold as a thriller. Publishers expect it, agents expect it, readers expect it: a manuscript that arrives with a long opening will be asked to compress. The compression is often right. It is also close to automatic, applied whether or not the long opening was doing structural work, and the writer who chose it deliberately is not always given the chance to say why. The reflex is a small flaw in the current state of the craft.

The remedy is mostly the writer’s. Anyone who chooses the long opening has to know precisely what it is buying and be ready to argue for it. That argument will sometimes be lost. It is still worth making, because the alternative is a form in which every book opens the same way regardless of what it needs.

I spend more time on openings and endings than on any other part of a book. The opening I am most pleased with took more than a dozen rewrites. It opens fast enough to make a contract and then slows deliberately in the next pages to build a deeper relationship with the narrator. Whether the calibration is right is not for me to say.

Betteredge and his Robinson Crusoe would not survive a contemporary editor. That is a loss, and we should be clear about what has been lost: not charm, and not leisure, but a particular kind of trust that can only be built at length. The fast opening cannot buy it. The slow opening cannot survive without a reader willing to wait. Most novels today are written for a reader who will not.

Subscribe now

The distinction is important, because most novels marketed as retellings cannot tell you which of the two things they do. Most readers have no particular reason to work it out on the publisher’s behalf.

A retelling earns its source. An update borrows from it. The earning is structural: a real retelling identifies the question the original could not finish asking, and writes the version that question takes now. An update only changes the surface — the period, the setting, the names, the idiom — but leaves the substance where it was.

The form attracts a great deal of enthusiasm and very little theory, which is how the shelves have filled with feminist retellings of Greek myth, gender-flipped Austen, Beowulf in suburban America, Macbeth as corporate thriller. Many are good books and some are excellent. A substantial number are borrowing while the jacket claims they are earning.

The test I use is a simple one. Imagine the source did not exist. Is the book still doing the same work? If the answer is yes, the source has served as scaffolding for a book that could have stood without it, and the retelling is decorative. If the answer is no, if the book depends on what the source argued and is writing the modern version of that argument, then it is the real thing. It needs its source. It could not have been written without it.

Wide Sargasso Sea fails that test in the right direction. Remove Jane Eyre and Rhys’s novel does not survive as itself. It argues with the original on every page. The mad first wife who was a Gothic device in 1847 is given the consciousness Brontë withheld, and the Creole inheritance that Brontë mentions and moves past becomes the central matter of the book. Rhys is not bringing Brontë up to date. She is finishing her.

James, published in 2024, does the same work on Huckleberry Finn. The novel is told from Jim’s point of view, and it demonstrates that Twain’s choices about whose consciousness his book could enter were the limits of what Twain could write rather than the limits of the story. Take Huckleberry Finn away and James does not survive as itself either. It is structurally dependent on its source, and it writes the version Twain could not.

Demon Copperhead stands in the same relation to David Copperfield. Dickens wrote a novel about a child inside a system that consumed children; the modern book relocates it to Appalachia during the opioid epidemic. The case for it as a real retelling has nothing to do with the rhyming names or the parallel chapters. It rests on the identification of the question Dickens was asking — what a particular kind of poverty does to a particular kind of child, and what the system around the child does or refuses to do. The retelling answers the same question under conditions Dickens never saw. Without Dickens there would have been a fine Appalachian novel. With Dickens there is a better one.

The Cromwell novels are the harder case to argue and the most instructive. Hilary Mantel did not retell a novel. She retold the chronicle, the historical record of Henry’s reign and Cromwell’s part in it. The retelling work lay in identifying what the chronicles could not give Cromwell: a consciousness. Had she simply written a historical novel about Cromwell, the books would be different in kind. They work because they argue back at a record with a particular silence at its centre.

The Greek myth wave deserves a fairer hearing than the test gives it at first glance, because myth has no fixed source text to argue with. There is no single Iliad-equivalent for most of these stories, only versions, and a novelist working from myth is joining a line of tellings rather than answering one book. The test still applies, but it has to be applied to the tradition instead of to a text: has the novel identified something the received versions could not say, or has it retold the received version with the camera moved? A book that gives Penelope or Briseis a voice and leaves every question of the tradition where it found it has moved the camera. A book that makes the tradition’s silence its subject has done the work. The distinction survives the absence of a source; it just becomes harder to prove.

Now the borrowing case. The Lion King takes the architecture of Hamlet — uncle kills king, son in exile, return and revenge — and builds something with no structural dependence on Shakespeare whatsoever. It does not argue with Hamlet. It does not finish anything Hamlet left undone. It uses the architecture as material and puts up a different building, which is an entirely respectable thing to do. Remove Hamlet and The Lion King is a slightly less tidy revenge story. Remove Jane Eyre from Wide Sargasso Sea and there is no book.

None of this is a judgement about quality. Both are good work of their kinds. The difficulty arrives when the borrowed kind claims to be a retelling. A novel that has borrowed an architecture and produced a coherent independent story does not need to be sold as a retelling in order to be worth reading, but once it is sold as one, the test applies, and a great many recent novels fail it.

The pattern I see most often runs like this. A novelist takes a classic with a famously underwritten female character and writes a book from that character’s perspective. The prose may be lovely and the execution elegant. But there is no structural argument with the source: the original character was thin because she was a function of the plot, and the new book gives her interiority without altering any question the original was asking. The same book could have been written about an invented character and would have worked identically. The classic’s name is on the cover for marketing.

I am not going to name examples, because the point is not to indict. The pattern is worth naming so that readers can see it, and it is widespread because publishers noticed correctly that retellings sell and concluded incorrectly that anything in the vicinity of a classic can be marketed as one.

Re Jane is the useful diagnostic in the other direction. It rewrites Jane Eyre as a contemporary novel about a Korean American woman working as a nanny in Brooklyn and later teaching in Seoul, which on the surface is transposition — the same story with different settings. But the structural work is being done. The novel asks Brontë’s question: how a woman without family, fortune, or standing builds herself in a world that intends to use her up, under the economic and cultural conditions of the contemporary diaspora. The answer changes the book. That is location rather than modernisation, and it is the distinction in one example.

The strictness of the test attracts the obvious objection, which is that the borrowing kind of book deserves the name too. I disagree, though mostly about vocabulary, and the vocabulary can be settled by people with more patience for it than I have. The structural difference is what I am arguing for. It determines whether a book has earned its source or merely used its name, whatever we agree to call the two cases.

Two things follow from getting it wrong. The first is that the retelling form is weakened by books which take another’s name recognition without doing the work. A reader sold three borrowed books in succession begins to mistrust the category, and the serious retelling that arrives fourth is mistrusted. The second is that the form, done properly, is one of the few ways the canon stays alive without being kept on a shelf. The great books contain questions their authors could not finish. The questions did not stop being live; the conditions for answering them changed. A real retelling is a novelist saying: here is what your question looks like written through to its end.

I apply the test to my own drafts, which is less comfortable than applying it to other people’s. One of them failed it at thirty thousand words. The book I was writing would have worked without its source, which meant the source was there for decoration. My options were to abandon it or to go back and find the question I had walked past. I went back. It cost half a year.

That is the discipline the form asks for, and it is why good retellings are rare. A borrowed book can be written by anyone with a shelf and an idea. An earned one requires the writer to find the thing the original could not say, which means reading the original closely enough to locate its silence, and then writing towards that silence rather than around it. It took Rhys twenty-one years to find the silence in Jane Eyre.

Lily Bart owes nine thousand dollars. The figure is named on the page. Edith Wharton does not approximate it. She does not write a substantial sum and does not gesture at a vague debt. She writes the number, tells you who Lily owes it to, what Lily’s annual income is, what a woman in Lily’s position can earn by writing letters or trimming hats, and what the distance between those figures means for the question of whether Lily survives.

The House of Mirth was published in 1905, and precision of this kind was not the only way to write a novel of social observation even then. Henry James, her friend and in some respects her model, handled money by implication. Wharton chose otherwise. She wanted the reader to understand the economic mechanism that was closing on her protagonist, and she did not believe a reader could understand a mechanism rendered as atmosphere.

She was right. Contemporary literary fiction has mostly chosen atmosphere, and the choice has cost it something.

When a contemporary novel about wealth establishes that a character is rich, it usually does so by description. The apartment, the watch, the school, the second house. Wealth arrives as texture. What rarely arrives is the income, the source, the stability or instability of the position, the question of what would happen if the money stopped on a Tuesday. Wealth in contemporary literary fiction is largely a vibe.

The nine thousand dollars is not texture. The reader who has been given that figure knows what Lily can and cannot do to discharge it, knows that the men in the novel who could clear it with a signature are watching to see what she will accept in exchange, and knows how little time she has. The plot is the plot because the figures are the figures. Atmospheric wealth cannot generate that particular catastrophe, because atmosphere has no edges and a debt does.

The debt is also a lesson in how to make money do plot work, because of where it comes from. Lily does not borrow the money. She is given it by a married man who tells her he has been investing on her behalf and has done well. She accepts it in that spirit, but it is not true. There were no investments. He has been handing her his own money and waiting, and the moment she understands this the sum stops being a number and becomes a claim on her. Wharton could have conveyed the arrangement atmospherically, as a vague sense of obligation curdling into threat. Instead she gives the reader the figure. That is what makes the obligation calculable, and calculable is what makes it frightening.

The Custom of the Country is in part a study of ascending price tags. Each of Undine Spragg’s husbands and near-husbands occupies a different financial scale. The scales are named, so the reader watches her recalculate with every new prospect. The novel does not make Undine grasping by adjective. It makes her grasping by ledger, which is harder to argue with.

The Age of Innocence works the same way. May Welland’s settlement is named. Newland Archer’s expectations are named. The Countess Olenska’s compromised inheritance is named precisely enough that her decision at the end is not an abstract act of renunciation. It is a reckoning made against a social system whose terms the reader has been given in full.

You may object that this comparison is unfair, because the contemporary reader lacks the financial literacy to follow figures of this kind and would be confused rather than informed by them. This seems to me condescending to people who handle numbers all day, at work, on their phones, in every news bulletin about interest rates. Readers would follow the figures without difficulty. Novelists have largely decided not to supply them.

The second objection is that the named sum is a Victorian and Edwardian convention which fiction has outgrown for good reasons. I think the reasons are bad ones, and mostly editorial: a vague sense that money on the page is gauche, or that specifying it is dull, or that the mention of a number breaks a spell the prose is trying to cast. What the convention has produced is a body of novels about wealth that have neutralised wealth as a subject. The reader is given the texture and denied the mechanism, which is what made Wharton’s novels work.

An objection worth taking seriously is that named figures date. A contemporary reader has no instinct for what nine thousand dollars meant in 1905, and no reader in 2140 will have an instinct for what a sum means now. This is true and it does not damage the technique, because Wharton’s figures are relational rather than absolute. The debt is given against Lily’s income, the income against the cost of the life she is expected to maintain, that cost against what she could earn if she worked. A reader who knows none of the values still watches the ratios, and the ratios do not date. The novelist’s obligation is not to supply a number the reader can convert. It is to supply enough numbers that the reader can see the distances between them.

The exceptions are easy to identify, because they are the books in which financial situations produce actual plot. The Cromwell novels do it: Hilary Mantel understood that Cromwell is a financier as much as a courtier, so she does not hide his accounts. The Goldfinch does it in stretches: whenever the question of the painting becomes a question of money, the money is specified. What these books have noticed is that the counting is not decoration. The counting is the story.

Trollope is the obvious Victorian comparison, since he named figures with a freedom that Eliot and Dickens did not always allow themselves. The Way We Live Now, published in 1875, is a novel about a financial fraud, and it is comfortable showing the reader the books. It is unsparing because the numbers are unsparing. The Custom of the Country belongs to that tradition. Both novels diagnose an economic system by walking through it with the ledger open.

The scene I keep returning to is late in The House of Mirth, when Lily, having lost almost everything, sits down and counts her money. It runs a few paragraphs. Wharton names what is in the purse and what is in the small account. She names the expenses coming towards her, and lets the reader watch her do the sum that the reader has been doing alongside her for four hundred pages. It is one of the most upsetting scenes in American fiction, and the distress comes from the counting: we know what every line item means, and we reach the total before she does.

Written atmospherically, the scene would still function. Lily counts, the money is insufficient, the insufficiency registers as sensation. It would be a sad scene rather than a devastating one. The devastation comes from precision, which is what specificity buys and what evocation cannot.

There is a deeper reason for the change, which is not literary. The Edwardian novelist wrote for a reader who took it for granted that money is a structural fact of a human life. The contemporary literary novelist writes partly for a reader trained to find the direct discussion of money faintly embarrassing, in the way that discussing salaries at dinner is embarrassing. The embarrassment is a real shift in manners. It is not one a novel should defer to. A novel’s business is to make the structural facts of a life legible, and a novelist who soft-pedals money to spare the reader’s manners has chosen the reader’s comfort over the book’s clarity.

In my experience, not every novel needs this. Some books are better served by atmospheric wealth. Some genres depend on it. A reader who wants the texture of money rather than its mechanics is not reading badly. The point is that the tool exists in the tradition and has been mostly retired for reasons that do not survive examination. The few contemporary novels that pick it up again do something the atmospheric ones cannot.

Basil Hallward has painted the best picture of his life and will not show it to anyone. Lord Henry Wotton, stretched along a divan in the corner of the studio, asks him why. Basil says he has put too much of himself into it. This is the first page of The Picture of Dorian Gray, and nothing has happened: two men in a summer studio, the smell of roses coming through the open door, a finished portrait on an easel. A reader who has never heard of the book already knows that something in the room is wrong.

Nothing in the scene will say what. Basil’s refusal is phrased as a principle about art, and it is a good principle, and it is also a confession he does not know he is making. Lord Henry answers in aphorisms that are funny in the way a man is funny when he has decided that nothing matters very much. The wrongness is in the arrangement rather than in any single line: a painter frightened of his own painting, a friend amused by the fear, and a young man who has not yet walked in and is already the subject of the conversation.

This is a technique, and its parts can be named. A scene establishes that something is badly wrong before anyone in it can say so, using details that each have an innocent explanation available. The reader assembles the alarm out of parts that are individually harmless. Get Out (2017) opens with the same technique, executed at close to the same level, and the two openings explain each other.

The film begins with a young man walking alone through an unfamiliar suburb at night, on the phone, lost, complaining about the street signs. A white car slows behind him. The car doubles back. The song playing from it is cheerful and from another era. He turns to avoid the car, says something about not wanting trouble, and is choked from behind and carried to the boot. The sequence takes four minutes and contains no supernatural element, no gore, and no obvious villain. It contains a man walking somewhere he has correctly registered he should not be walking.

The house sequence that follows is where the technique does its real work. Chris arrives with his girlfriend at her parents’ place, and the parents go out of their way to be welcoming. There is a tour. There is the father’s remark that he would have voted for Obama a third time. There is the mother, a psychiatrist, offering to help with the smoking, and later stirring a teacup with a small silver spoon. There is the housekeeper, whose smile is held about two seconds past where a smile ends. There is the groundskeeper, running full speed across the lawn in the dark, and then stopping. None of these is a threat. All of them are wrong.

The first rule of the technique is that every detail must survive an innocent reading. The father is trying too hard, which is what nervous parents do. The mother is a psychiatrist, and psychiatrists offer to fix things. The housekeeper is awkward around guests. The groundskeeper likes to run at night. Any one of these can be explained by a viewer who wants to explain it, and the film knows that the viewer wants to, because the alternative is to believe something that cannot yet be said out loud. Wilde runs the same mechanism in reverse: Basil’s fear of his own canvas is explained, plausibly and at length, as an artist’s scruple.

The second rule is that the details accumulate rather than escalate. Lord Henry’s remarks in the first chapter do not get darker as the chapter proceeds. There are simply more of them, and the reader begins to notice that a man who says everything charmingly has not yet said anything he believes. The party guests in the film do not become more offensive over the afternoon. They arrive at their level immediately and stay there, one after another, admiring Chris’s build, asking whether it is true what they say, mentioning Tiger Woods. Escalation would give the protagonist grounds. Accumulation gives him nothing to point at.

What this produces is not dramatic irony in the ordinary sense. Dramatic irony gives the audience a fact the character lacks: the audience has seen the letter, or knows the identity of the man in the mask. Here the audience knows nothing the character does not. They are watching the same tour and hearing the same aphorisms. What they have is a position, outside the room, where the social codes are legible as codes rather than as manners. The character is inside the room and has to be polite. That difference, and not information, is the entire source of the dread.

It also creates the clock, which is the part most often missed. Once a reader has registered the wrongness the protagonist has not, the question of the scene stops being what will happen and becomes how long this will take. Wilde makes the reader wait through an entire chapter of charm before Dorian appears, and then wait again while he is worked on. The film makes the audience wait through an afternoon of small talk. The waiting is the thing being made.

Then there is the third rule, which is what makes the openings watchable: someone else can see it. In the second chapter of the novel, Basil begs Lord Henry not to spoil the boy, and says that his life as an artist depends on him. He is entirely right about the danger but he says it to the wrong person, quietly, while Dorian is at the piano with his back to them turning the pages of some Schumann. The film gives this role to the friend on the phone, who hears an account of the weekend, correctly identifies what is happening, and is four hours away with no standing to do anything about it. In both, the person who understands the situation is outside the room where it is occurring, which prevents that person from being useful. It is a very old device and it still works, because the audience is in the same position as the friend: certain, and unable to act on the certainty.

Both also have to decide where deniability ends, and both place the collapse late and give it to an outsider rather than to the protagonist. In the film it is a photograph. A camera flash goes off in the face of another guest, and the man comes apart and screams at Chris to get out, the first moment in the afternoon when anyone says a true thing in a plain sentence. The room immediately covers it with a seizure and an apology. In the novel the collapse is Dorian’s own, in front of the finished portrait, where he says aloud that he would give his soul to trade places with the painting. Before that sentence the book could still be a comedy of manners about three men and a picture. After it, it cannot.

The fourth rule is the hardest to execute, and it is what separates both of these openings from imitations of them: the protagonist’s failure to act must be socially reasonable. Chris cannot accuse his girlfriend’s mother of anything. There is no sentence available to him. To say it aloud is to be the guest who accused the psychiatrist of hypnotism over dinner, and that costs him far more than another hour of small talk. Dorian is in the same position, and worse, because the man corrupting him is delightful and has not proposed anything. Both are trapped by courtesy, which is a stronger cage than a locked door and is harder to write well.

Politeness as a trap is also why both works can use pleasure as the delivery vehicle. Lord Henry is the funniest character in The Picture of Dorian Gray, and Wilde knows exactly what he is doing by making the corrupting influence the best company in the book. A reader who finds him delightful, which every reader does, has been placed inside Dorian’s error rather than above it. The film does the same with its party scene, which is often very funny, and the laughter is the point: a viewer who is enjoying the absurdity of these people is, for the length of the laugh, doing what Chris is doing: treating an emergency as an awkward afternoon.

The version of this technique that fails is the one where the wrongness is legible as wrongness. A character who says there was something strange about that house has blown the entire effect, because the alarm has been moved from the reader to the page. So has the writer who makes the detail undeniable, the smile that is described as sinister, the aphorism that is openly cruel. The whole apparatus depends on plausible deniability being maintained on both sides at once, for the character and for the audience, until the moment it collapses.

I learned this the slow way. An early draft of a chapter I had been working on had a character ask, in the third paragraph, whether everything was all right. It was a reasonable question and it destroyed the chapter, because the moment the question exists, the scene is about the answer. I cut the line and replaced it with a conversation about a dinner. The reader now had nowhere to put their alarm, which is where it belongs.

Neither of these openings is subtle, exactly. They are legible, deliberate, and quite loud once you know where to look. That is the mark of a technique executed well. What they share is a refusal to let the audience discharge the tension by naming it. A reader watching Basil refuse to exhibit the portrait, or a viewer watching a smile held two seconds too long, is given no way to be right yet.

The novel was published in 1890 and the film in 2017, under conditions that share almost nothing. They arrive at the same structure because they need it for the same reason: both are about a man being taken apart by people who are being nice to him, and neither can afford for him to notice early.

Each sequence ends the same way. A man decides to stay.