7. Diagnostic Method — the ten tests

7.0 How to use the protocol

The ten tests below are for revision, not composition. Applied while drafting they produce paralysis of a specific and recognisable kind — the writer who cannot name a room’s furniture without justifying it stops being able to see the room. Applied to a finished passage they are the most efficient triage instrument in this chapter.

Five procedural notes.

• Apply them to passages, not to every detail. Running ten tests on every descriptive element in a novel is not revision; it is a hobby. Select the passages that a reader found slow, the ones the writer suspects, and one representative sample per major setting.
• Run them in order. Tests 1–3 are structural, and their results condition the rest. There is no point auditing modifiers in a paragraph that would not be missed if deleted.
• Record the answer, not the verdict. The output of each test is a sentence describing what the detail does, not a pass or a fail. “This detail supplies mood and nothing else” is usable. “Fails test 2” is not.
• A failed test is a question. Every test below is failed somewhere by a novel this chapter admires. The test’s job is to make a choice visible so that it can be confirmed as a choice.
• Two of the tests are counting exercises and should be done mechanically, with a pen, on paper, without interpretation. Tests 8 and 9 are worthless when performed by estimate, because a writer’s estimate of their own descriptive density is systematically wrong.

7.1 The ten tests

Test 1 — The removal test

Cut the detail. What is lost?

Delete the detail or the passage and read the surrounding page. Then name what the reader would now fail to know, feel or anticipate — a plot fact, a piece of characterisation, a sense of the world, a thematic note, a date, or the scene’s mood.

Interpretation. “Nothing” is a real answer and it appears more often than writers expect. But it has two very different diagnoses, and distinguishing them is the whole skill. If nothing is lost and the passage was doing texture work, the detail may be a legitimate reality-effect element (§3.4) and should be kept in some quantity — the question then becomes one of proportion, and moves to test 8. If nothing is lost and the passage was asking to be noticed — it is long, or figurative, or placed at a paragraph’s end — it is a wallpaper detail (§8.2), and the emphasis is the fault rather than the detail.

The removal test is the detail-level form of the value-shift test from Chapter 1 (The Purposes of a Scene), and it inherits that test’s discipline: name the change, do not gesture at it.

Test 2 — The denotation/connotation test

State plainly what the detail shows. Then state, separately, what it implies.

Write two sentences. The first uses no evaluative vocabulary at all: a coat is hanging on a hook and its left cuff is worn through. The second states the freight: someone who is not here used to hook his thumb there, and nobody has moved the coat.

Interpretation. If the second sentence cannot be written, or merely paraphrases the first, the detail means only itself. That is decoration rather than selection. If the second sentence is much longer and more interesting than the first, the detail is carrying more than the object can plausibly bear, and the reader may feel the symbol arriving (§8.12).

Test 3 — The genericness test

Could this detail describe any object or place of this general kind, in any book?

Read the detail with the surrounding proper nouns removed. Ask whether it would be equally true of a different house in a different novel by a different author.

Interpretation. A detail that survives the transplant is a category detail: it describes the type rather than the instance. A kitchen with a stove, a forest with trees, a hospital with a disinfectant smell, an old man with liver spots. Category details are not always faults — they orient cheaply, and a novel needs some — but a passage built from them establishes nothing, because the reader already had all of it. The threshold question is: did this tell me something about this particular place, or only confirm what I assumed on hearing its name?

Test 4 — The substitution test

Replace the detail with a different true one. What changes?

This is the test that most directly operationalises the chapter’s first principle, and it is the one that teaches fastest. Take the object as described, and write two alternative true things about it — not better things, merely other true ones. The chipped cup; the bone-white cup; the cup with an inch of cold tea in it.

Then read all three in place and describe, in a sentence each, the different book each implies.

Interpretation. If the three versions produce three noticeably different scenes, the original choice was load-bearing, and the writer can now confirm it deliberately. If the three versions are interchangeable, the detail is not doing selective work and the passage would be better served by a different object altogether — which is the move from compression to selection described at §5.1.

The strong form of this test is to run it at the level of the object rather than the attribute: not which true thing about the coat, but why a coat and not the shoes, the umbrella, the shopping list on the table. Substituting objects rather than adjectives is where the real gains are.

Test 5 — The focalisation test

Whose attention selected this detail?

Name the perceiver. Then ask whether the selection reveals anything about them, or whether any narrator could have made it.

Interpretation. Three outcomes. If the selection is characterising, the detail is serving two domains and the passage is working. If the selection is neutral but the narration is avowedly external or omniscient — a nineteenth-century narrator, a documentary voice, a panoramic chapter — the detail is legitimate and the test does not apply. If the narration is close third or first person and the selection is neutral, the detail has escaped its perceiver, and the passage is being written by the author over the character’s shoulder. That is the unfocalised detail (§8.11), and in this chapter’s experience it is among the commonest defects in competent contemporary description.

Test 6 — The plausibility-of-attention test

Would this person, wanting what they want, in the state they are in, notice this?

Distinct from test 5, and the distinction matters. Test 5 asks whose selection is this. Test 6 asks is the selection credible given the character’s condition at this moment.

Interpretation. Attention narrows under stress, urgency, pain, fear and desire, and it narrows towards the useful and the personally charged. A character fleeing a building does not register the cornices. A character in grief does not conduct an even sweep of a room; they will register two things with unbearable clarity and nothing else. A character who is bored, waiting or professionally idle can conduct a sweep, which is why so much good description is given to people with nothing to do.

The test has a corollary that solves a great many problems: if a passage requires description the character would not plausibly perform, change the character’s state rather than the description. Give them a reason to wait.

Test 7 — The recurrence test

Does this detail, or a variation of it, appear elsewhere in the manuscript?

Search the draft for the object. Record every appearance, and for each, record what has changed — the object’s state, who is in relation to it, the surrounding context, or the words used.

Interpretation. Three patterns and three remedies.

This test is the one that most often changes a manuscript rather than a paragraph, and it should be run late, on the whole draft, with a search function.

Test 8 — The count test

How many discrete descriptive elements does this passage contain, and how many will the reader keep?

Count them mechanically. A modifier is one; a named object is one; a reported attribute is one. Then mark the two or three the writer most wants retained.

Interpretation. The count norm for contemporary literary fiction is low — two or three elements to establish a room, one or two for a person on first appearance — and the norm is a convention rather than a law (§3.6). What the count test measures is not conformity to it but the ratio of elements to intended retentions. A passage with fourteen elements and three intended retentions is asking the reader to discard eleven, and the reader will discard eleven of their own choosing, not the writer’s.

This is the practical form of the cognitive finding at §3.5: readers construct sparse, functionally adequate models, and additional description past a low threshold competes with itself rather than accumulating. A high count is defensible when the passage is a catalogue with an ordering principle (§6.3, strategy 4) or a set piece the book has earned. It is not defensible as a default.

Test 9 — The modifier and sense audit

Two mechanical counts on the same passage.

(a) The modifier audit. Underline every adjective and adverb. Then, for each, ask whether it is doing work a better noun or verb would have done. Walked slowly and uncertainly against hesitated; a large, imposing, dark building against a warehouse.

Interpretation. Two or more modifiers on a single noun is a signal, not a fault — but it is a signal that the writer could not find the right noun and is triangulating (§8.3). A passage in which the modifier count approaches the noun count is over-modified regardless of how good the individual modifiers are.

(b) The sense audit. Tally the passage’s descriptive elements by sense: sight, sound, smell, touch, taste, and the body’s own report (temperature, balance, fatigue, pain).

Interpretation. Most drafts return a tally dominated by sight, with a scatter of sound (a rough estimate is four fifths or more sight; an observation, not a measurement). That is not automatically wrong — vision dominates human perception too — but it wastes the medium’s advantage over the camera (§4.5). Smell in particular is under-used and disproportionately effective, because it is the sense most directly bound to memory and the one film cannot deliver at all. The remedy is not to add one of each; it is to ask which single non-visual element would carry the most, and to let it displace a visual one.

Test 10 — The inventory test

Does the passage have a selection principle, or only an order?

Two procedures, run together.

(a) The shuffle. Reorder the passage’s descriptive elements at random and read it again. If nothing is lost, the sequence had no organising principle and the passage is an inventory rather than a catalogue (§6.3, strategy 4; §8.1).

(b) The principle statement. Complete the sentence: These details, and not others, because ___. Acceptable completions name a perceiver’s preoccupation (because this is what a woman avoiding one object in a hallway would see), a causal argument (because the rent explains the lodgers), a temporal record (because these are the things that have not been touched since March), or a formal pattern (because the list moves outward from the body). Unacceptable completions are because they were there and because the room needed describing.

Interpretation. This is the summary test, and the one to run if only one test can be run. A passage with a selection principle that can be stated can be improved by any of the other nine. A passage without one cannot be improved at all; it can only be rewritten from a different principle, and §6.4 gives the procedure.

7.2 Reading the results

A warning about the protocol as a whole. These tests are biased towards the purposeful, and applying them at full strength to an entire manuscript will produce a book in which every detail justifies itself — which is a describable and undesirable condition. §5.4 and §16 give the counterweight. The correct use is triage: run the tests where a reader has flagged a problem, and leave the rest of the book its air.