The chapter needs only the theory its method uses. Each strand below starts from a question a writer asks in revision, then gives the source that answers it.
Why not simply cut the description?
Ian Watt’s The Rise of the Novel (1957) argues that the novel differs from earlier prose fiction less in subject than in a mode of presentation he calls formal realism: the convention that a novel is a full and authentic report of individual experience, and is therefore obliged to the particulars that romance and allegory could do without. Ordinary names with surnames, dated time, named streets, objects that are owned, priced and worn: Watt’s term for these is circumstantial particularity, and his thesis is that they are not decoration on the form but the form itself. Defoe’s inventories of Crusoe’s stores persuade because they behave like a ledger, not because anyone checked them. Two consequences follow for revision. A book that refuses particulars becomes another kind of book, a fable or a parable, so the advice to cut description is safe only as advice to cut unselected description. And particularity was from the start an effect rather than a fidelity: the detail authenticates more than it reports.
How much should I describe?
Erich Auerbach’s Mimesis (1946; translated by Willard R. Trask, 1953) opens by setting two ancient texts side by side. In Book 19 of the Odyssey the old nurse recognises Odysseus by a scar, and the poem stops for some seventy lines to tell how he got it, leaving nothing in shadow. In Genesis 22, the binding of Isaac, almost nothing is given: no place, no faces, no thoughts, no length of journey. Auerbach calls the biblical manner “fraught with background”, because the unsaid presses on the said and the reader is compelled to interpret. The two texts mark the ends of a scale on which every descriptive decision sits. The Homeric pole delivers presence and risks digression; the biblical pole delivers depth and risks obscurity. Modern taste prefers the second, and Auerbach shows why: a selected detail surrounded by unsaid alternatives produces depth as a by-product of its scarcity, which an exhaustive description cannot, because nothing has been suppressed. Depth is a function of omission. He is not recommending the biblical mode, though; both are complete styles, and the Homeric one is the economy of epic, of Balzac and of the catalogue.
Why does an odd angle make a reader see?
Viktor Shklovsky’s essay “Art as Device” (1917; translated by Lee T. Lemon and Marion J. Reis as “Art as Technique”) starts from habit: perception becomes automatic, so that we recognise objects instead of seeing them. Art restores perception by making the familiar strange, a process Shklovsky called ostranenie, usually translated as defamiliarisation, and his phrase for its purpose is “to make the stone stony”. His examples are Tolstoy’s, and they are descriptive rather than figurative: a flogging described without the word for it, the institution of property seen by a horse. The working rule is to describe the operation and withhold the name. It has a cost in speed, since estrangement lengthens perception, and it depends on contrast: a novel that estranges everything has adopted an idiom and estranged nothing.
What about the detail that does nothing?
Roland Barthes’s essay “The Reality Effect” (1968; translated by Richard Howard) is the strongest objection to this chapter’s claim. Analysing the opening of Flaubert’s “Un cœur simple” (1877), Barthes notes that in Mme Aubain’s room an old piano supports, under a barometer, a heap of boxes and cartons. The piano signals bourgeois standing and the boxes a declining house, but the barometer signals nothing: it belongs to no code and forecasts no event. Barthes argues that such details, having no meaning within the story, come to mean one thing only, that this is real. They produce what he calls the reality effect, the impression that the text was transcribed from a world rather than made. He offered this as a critique of realism’s rhetoric, not as advice, but its consequence for a writer is direct. A craft principle of no detail without a job treats the barometer as waste; Barthes treats it as what makes the room convincing.
Three replies are available. The first says that authentication is itself a job, which saves the principle at the cost of making it unfalsifiable. The second says that the two operate at different levels: Barthes describes an effect across a whole book, which needs a proportion of unrecuperable detail, while the craft principle judges a passage. The third says that the principle is a period preference and not a law. This chapter works from the second and keeps the third in view. Its practical form is that the question is not whether each detail is functional but whether the passage has a principle of selection, a reason that these details and not others were chosen. A paragraph of eight details of which two are inert reads as a world; a paragraph of eight inert details reads as a list.
How many details will a reader keep?
Psychologists of reading hold that readers do not store sentences but build a situation model, a running representation of what the text describes, tracked along a few dimensions: protagonist, space, time, cause and intention (Rolf Zwaan and Gabriel Radvansky’s synthesis, 1998). Spatial information is tracked less automatically than causal information and is updated when the story needs it, when someone must move, hide or be found. A room described so that a reader can find the exits is remembered as a room; a room described to be admired is remembered, if at all, as prose. The images readers form are also schematic: a gist filled with defaults and a few specified elements, so that past a low threshold each added detail competes with the last. Elaine Scarry’s Dreaming by the Book (1999) adds that imagined objects become vivid when the prose gives the imagination instructions, above all instructions to move something, so that an object lifted, turned or passed in front of another is easier to see than one held still and catalogued. This is one reason the strongest details are so often details of use: a thumb hooked in a cuff, a drawer that sticks. These findings describe what is easy to process. They do not show that sparse description is better, and Proust is not worse than a thriller for being harder to read.
The conventions have dates
The prescriptions in the rest of the chapter are calibrated to a convention roughly sixty years old. Balzac opens Père Goriot (1835) with several thousand words on a boarding house before the plot begins, and Hugo halts Les Misérables (1862) for chapters on Waterloo and the Paris sewers; these were the form working as its readers wanted. A literary novel now often introduces a room in two or three details, and the count norm, the number of details a convention allows for establishing a place or a person, has fallen by something like a factor of ten (a rough estimate, not a count). Photography and film gave readers a stock of images a single cue can summon, publishing stopped paying by length, and modernist and then workshop taste prized compression. Naomi Schor’s Reading in Detail (1987) shows that the detail’s present prestige reverses a much older hierarchy that held it trivial. Learn the convention well enough to write inside it, and know that it is one.
See also: the Reference theory, which stages the collision with Barthes in a table and develops each strand; The Opening Pages (Chapter 9), on the rules of notice that make a placed detail read as a promise.