The ten tests are for revision, not composition: applied while drafting they produce self-conscious dialogue. Run them on one exchange at a time, not on a whole scene, which averages a failing passage away; and run them in order, since tests 1 to 4 (Removal, Want, Status and Cost) are structural and a failure there makes the rest pointless. Record a sentence, not a verdict: “Dov holds the upper hand at the start and a smaller share of it at the end” is useful, and “fails test 3” is not. Each test below is failed on purpose by a book worth reading.
Test 1 — Removal
Replace the exchange with one sentence of summary. What is lost?
Failure. Nothing: the summary is the better version.
Exception. An exchange kept for a voice the reader needs to hear at length.
Test 2 — Want
What does each speaker want from the other before this ends, in one sentence each?
Failure. One engine (a briefing) or none.
Exception. The interrogation, a one-engined form chosen as such.
Test 3 — Status
Who holds the upper hand at the first line and at the last, and was it contested?
Failure. Static status that nobody tried to disturb.
Exception. Status fiercely contested but held, or fixity that is the subject (Beckett).
Test 4 — Cost
What does each disclosure cost the person making it?
Failure. A briefing in dialogue’s clothes.
Exception. The naive listener, who needs what she is told.
Test 5 — Silence
What does each speaker choose not to say, and who else is in the room?
Failure. Everyone says what they think.
Exception. The climactic plain line, kept in reserve.
Test 6 — Implicature
Is each turn what was called for? Where it is not, what does the departure imply?
Failure. All turns cooperative: one channel.
Exception. Dostoevsky’s full self-accounts, whose gap is between the said and the case.
Test 7 — Transcript
Strip every tag and action beat and read aloud. Can you tell who is speaking?
Failure. Speakers told apart by content only, or not at all.
Exception. A declared uniform register (Wilde, Compton-Burnett).
Test 8 — Substitution
Could another named character say the exchange’s three best lines?
Failure. The line does no character work.
Exception. A chorus, where interchangeability is the point.
Test 9 — Attribution
How many turns carry a tag, how many tags are not said or asked, and what do the action beats do besides identify?
Failure. The exchange is being narrated at.
Exception. Period practice, where rejoined is the convention.
Test 10 — Interruption
Who takes the floor from whom, and where do interruptions cluster?
Failure. Formal alternation: a committee, not a contest.
Exception. Ceremony, where the turn order is fixed by rule.
Four refinements sharpen the tests. In test 1 (Removal), compare length as well as content: an exchange of nine hundred words that a sentence fully replaces has spent the reader’s attention on a transaction that did not need it. In test 2 (Want), the want must be directed at the other person (I want to understand the situation is not a want from anyone) and immediate, not the story goal. In test 3 (Status), name the currency (who holds the information, the physical position, the right to end the conversation, the right to ask), since these can move in opposite directions, and an exchange in which someone gains one and loses another is doing more than one in which everything moves together. In test 6 (Implicature), published fiction typically flouts in a minority of turns, by this chapter’s reading somewhere between a fifth and a half (an estimate, not a count), and the flouting turns cluster at the crisis.
Reading the results together
Failures cluster, and the cluster points to the first move. Test 1 (Removal) alone means a scene the writer needed in order to find out what happened and the reader does not need: use the summary. Tests 2 (Want) and 4 (Cost) together mean a briefing: give the facts to someone who does not want to say them. Test 3 (Status) alone means static status: find what one of them could concede, and make the other reach for it. Tests 5 (Silence) and 6 (Implicature) mean an exchange on the nose: apply the twelve mechanisms in The Mechanism to its three most direct turns. Tests 7 (Transcript) and 8 (Substitution) mean the ventriloquist exchange, and the instinct to make each speaker more distinctive by adding tics and catchphrases produces labels instead of people. Differentiation is structural: give each character a different average sentence length, a different relation to questions (who asks, who answers, who deflects) and a different default subtext mechanism, and distinguishable speakers follow without a single quirk. Failure on all of tests 1 to 5 (Removal, Want, Status, Cost and Silence) means the passage is not an exchange; cut it, or find the want that would make it one.
See also: the Reference protocol, with each test’s procedure and a table of combined failures; the domain audit, which marks each turn with the domains it serves.