3. Why a Scene Must Have a Job

3.1 Unity and necessity: the classical inheritance

The oldest form of the principle is Aristotle’s, and it is more exacting than its modern descendants.

In the Poetics (c. 335 BCE), Aristotle argues that a plot is not unified by having a single hero — the events of one man’s life do not compose a whole — but by the internal relation of its parts. The test he proposes in chapter 8 is a test of removability: the parts of a plot should be so arranged that if any one of them is displaced or removed, the whole is disturbed. Anything whose presence or absence makes no perceptible difference is not a part of the whole at all.

That is, in the vocabulary of this chapter, the extraction test (§7.1, test 4), stated two and a half thousand years earlier and applied to the incident rather than to the scene.

Two further Aristotelian claims bear directly on the model.

First, the necessary-or-probable connection. In chapter 9 Aristotle distinguishes the poet from the historian: the historian relates what happened, the poet what would happen, “according to necessity or probability”. The events of a plot must follow one another not merely after but because. Chapter 2 (Scene and Sequel) develops this into the causal-chain test — the requirement that each unit could be joined to the next by and so or but rather than by and then.

Second, and most usefully, the criticism of episodic plot. Aristotle names the worst kind of plot the episodic: one in which the episodes succeed one another without probability or necessity. He observes, drily, that bad poets write such plots through their own fault and good ones through the pressure of competition — an early acknowledgement that structural failure is often produced by working conditions rather than by incapacity. The modern draft full of well-written scenes that do not connect is an episodic plot in Aristotle’s exact sense.

Freytag (Die Technik des Dramas, 1863) supplies the nineteenth-century engineering version. His five-part structure — introduction, rising action, climax, falling action, catastrophe — is now known chiefly as a triangle in school textbooks, but Freytag’s own concern is with the scene as the place where the play’s forces are transmitted: he analyses the individual scene as a small dramatic movement with its own exciting force and its own turn, and he is explicit that scenes exist to raise or lower the tension by a measurable degree. The modern craft habit of asking whether a scene “goes up or down” is Freytag’s inheritance. So, less happily, is the tendency to think of narrative shape as a diagram — a tendency this chapter resists in §16.

3.2 Scene and picture: Lubbock, James, and a slogan’s decline

The vocabulary in which English-language writers now discuss this material was fixed in one book: Percy Lubbock’s The Craft of Fiction (1921).

Lubbock, systematising Henry James’s practice and James’s own prefaces (collected in 1934 as The Art of the Novel), distinguishes two modes of narration:

Lubbock’s celebrated formulation is that the art of fiction does not begin until the novelist thinks of the story as a matter to be shown — so exhibited that it will tell itself. James’s own instruction to himself in his notebooks was blunter still: dramatise, dramatise.

The vulgarisation. By the middle of the twentieth century Lubbock’s distinction had been compressed into a three-word rule — show, don’t tell — which is now the most widely transmitted and least examined piece of advice in anglophone creative writing. What went wrong in the compression was a conflation.

• The distinction between scene and summary (a matter of duration) was conflated with the distinction between presentation and commentary (a matter of narratorial stance) and with the distinction between the concrete and the abstract (a matter of diction). These are three separate variables, and a rule that collapses them cannot be applied intelligently. “She was angry” is telling in the diction sense; “she had been angry for a fortnight” is summary in the duration sense; “the narrator observes that anger of this kind is always a form of fear” is commentary. Only the second is what Lubbock meant.

Booth’s correction should be taught alongside the rule. In The Rhetoric of Fiction (1961), Wayne Booth demonstrated that the showing/telling opposition is incoherent as an absolute: all narrative is told, “showing” is one rhetorical strategy among others, the author’s voice is never actually absent but only disguised, and the modernist preference for effacement is itself a rhetoric with its own costs — chiefly, that a wholly dramatised narrative loses the ability to establish moral distance and can leave a reader without the means to judge what they are watching. Booth’s book is the reason a serious pedagogy of the scene cannot simply say dramatise everything.

3.3 Duration: what “dramatised” actually means

Genette analyses duration as the relation between story time (the notional time occupied by the events) and narrative time (the space the text gives them, measured in words or pages). Four canonical movements exhaust the possibilities:

This gives the definition of dramatisation that the chapter uses:

A scene is the mode in which discourse time approximates story time. It is not defined by the presence of dialogue, or of physical action, or of sensory detail — those are its usual carriers, not its criteria. It is defined by the rate at which the telling moves relative to the told.

Proust changed the ratios. Genette’s central corpus is À la recherche du temps perdu, and one of his most striking findings is that summary, the workhorse of Balzac and Fielding, has almost disappeared from Proust: the Recherche is built out of enormous scenes and vast ellipses, with iterative narration doing much of the connective work that summary used to do. This is the single best piece of evidence that scene/summary proportion is a historical variable rather than a constant of the form (§3.6).

3.4 Kernels, satellites, and the code of actions

Barthes, in the “Introduction to the Structural Analysis of Narratives” (1966), divides narrative functions into cardinal functions (noyaux, nuclei) and catalysers. A cardinal function opens, maintains or closes an alternative consequential for the story: the telephone rings, and it will be answered or not. A catalyser fills the space between nuclei — the character crosses the room, lights a cigarette, notices the rain — and can be removed without altering the story’s logic, though not without altering its texture.

Chatman, in Story and Discourse (1978), anglicises this as kernels and satellites, and adds the formulation that makes it usable in revision: kernels are branching points, moments at which the story could go one of several ways; satellites are amplifications of a path already chosen. Deleting a kernel destroys the narrative logic; deleting a satellite impoverishes the narrative without breaking it.

Barthes’s S/Z (1970) supplies the third instrument, the proairetic code — the code of actions, by which the reader recognises a sequence of gestures as an action with a name (“the Rendezvous”, “the Murder”, “the Departure”) and thereby knows what completion would look like. The proairetic code is the reason a scene generates expectation at all: the reader who recognises the opening moves of a confrontation knows the shape of a confrontation and reads forward to see whether it is completed, deflected or abandoned. A scene that activates no proairetic sequence has nothing for the reader to anticipate, which is the mechanism beneath the vague complaint that a passage “has no tension”.

Bal contributes the definition on which the value shift ultimately rests. In Narratology she defines an event as the transition from one state to another state. This is a stronger claim than it looks: on Bal’s definition, an occurrence that leaves every state as it found it is not an event at all, whatever activity it involves. “They talked for an hour” is not an event. “They talked for an hour and afterwards he would not look at her” contains one.

Bal’s other contribution is focalisation — the separation of the agent who perceives from the agent who speaks — which supplies the question the four domains cannot ask on their own: whose scene is this? A scene’s value shift is always a shift for someone, and a draft in which the value shifts for a character other than the focaliser is a scene told from the wrong side. Chapter 13 (The Viewpoint Shift) treats the general problem; the local form of it is test 1’s supplementary question (§7.1).

3.5 The cognitive strand: what readers actually store

Discourse psychologists hold that readers do not store sentences; they build a situation model, a mental representation of the state of affairs described. Rolf Zwaan and Gabriel Radvansky’s synthesis (1998) proposes that readers monitor a small set of dimensions and incur a measurable processing cost whenever one of them shifts:

The supportive inference. The dimensions readers track are states, and comprehension is the updating of those states. A passage that updates none of them gives the reader nothing to encode. This is the cognitive form of the scenery problem: the reader’s memory of a novel is substantially a record of its transitions, so a scene that produces no transition produces no memory. The correspondence with Bal’s definition of an event (§3.4) is exact, arrived at from a different discipline.

A third strand bears most directly on the scene. Work on event segmentation (Zacks and colleagues; Speer, Zacks and Reynolds) finds that perceivers and readers spontaneously divide continuous experience into discrete events, and that they do so at points of change — a shift in a character’s goals, location, or causal situation. These boundaries show distinctive processing signatures. If that is right, then a scene is not merely a convenient authorial container: it is a unit that corresponds to how experience is parsed. And its boundary is a change.

3.6 The historical strand: the scene–summary ratio is a moving variable

One qualification applies to every test in this chapter, and it is historical. The preference for dramatised scene over narrated report, which contemporary craft teaching presents as a technical fact, is a period convention whose dominance is roughly a century old and whose extremity is more recent than that.

Fielding’s practice is not incompetence. Tom Jones (1749) opens each of its eighteen books with an essay by the narrator, summarises months in a paragraph, and reserves full dramatisation for a small number of occasions. The effect is not slackness but authority: the reader is in the hands of a historian of the action who can move at will between the general and the particular, and the rationing of scene makes the dramatised occasions land with a force a wholly scenic novel cannot achieve. A book in which everything is a scene has no way to say this one matters more.

3.7 The first principle, stated precisely — and qualified

The value shift. A scene worth keeping moves some charged value — trust to suspicion, safety to danger, ignorance to knowledge, hope to despair, intimacy to distance — from one pole towards its opposite, so that something is true at its close which was not true at its opening.

The term and the formulation are Robert McKee’s (Story, 1997), and the attribution matters: it is a screenwriting concept, developed for a form of roughly two hours and a hundred and twenty scenes, in which the pressure on each unit is far greater than in a four-hundred-page novel. McKee’s own account is more careful than its transmission: he insists that the shift be charged — that the value be one the story has established as mattering — which is what distinguishes a value shift from mere incident.

First: the test measures charge, not size. A scene in which the only change is that a woman notices her husband has stopped asking about her day has shifted a value. A scene containing a car chase in which nothing is risked has not.

Second: the shift need not be visible to any character. In the great scenes of interiority — The Portrait of a Lady chapter 42 is the chapter’s principal example, at §9.8 — the shift is entirely in the protagonist’s understanding and entirely invisible from outside the room. In the great scenes of dramatic irony, the shift is in the reader’s knowledge and invisible to everyone in the book.

Third: some scenes fail the test and are indispensable. Melville’s “A Squeeze of the Hand” (§8.1), the iterative texture-scenes of Things Fall Apart, the dinner in To the Lighthouse (§8.10), the whole class of scenes whose function is to establish the normal against which a later departure will register — these are answerable to the book, not to the scene-level test. This is why §7.0 insists that the output of a test is a sentence, not a verdict.

Fourth: the test is a period instrument. It formalises the assumptions of a scenic tradition running from James through Hollywood to the contemporary workshop. Applied to Fielding, Sterne or Cervantes it produces nonsense — not because those writers were careless, but because they were not attempting the thing the test measures.

The correct standing of the principle, then, is this: it is not a standard a scene must meet, but an instrument that makes a writer’s choice visible to the writer. A scene that fails it may be right as it stands; what matters is that its author should know that it fails, and be able to say why the book needs it anyway. The Coda (§16) returns to this and states it as the chapter’s final position.