6. Text and Subtext: The Two Channels

6.1 The mechanism

Every line of working dialogue runs on two channels at once.

Working mechanism — the two channels.

Text — the literal content of the line: the words a court reporter would transcribe.

Subtext — the want, fear or withheld fact driving the line, legible to the reader through context, contradiction or omission, but rarely stated by the speaker.

The distance between the two channels is where dialogue does most of its actual work, and where a flat exchange most reliably reveals its flatness, because in a flat exchange the two channels have collapsed into one.

Two clarifications are needed at once, because the term is used loosely in workshops.

Subtext is not ambiguity. An ambiguous line is one whose meaning the reader cannot determine. A line with subtext is one whose surface meaning the reader determines easily and whose real meaning they also determine, and the two do not match. Subtext is a doubling of certainty, not a subtraction of it. This is the discrimination that separates the technique from the coyness Chapter 9 (The Opening Pages) calls the withholding tease: the reader must be able to compute the second meaning, or nothing has happened.

Subtext is not the same as concealment. A character may be perfectly aware of what they mean and perfectly aware that their hearer knows it; the indirection is then a courtesy, a formality, or a way of leaving both parties an exit. Goffman’s face-work (§3.3) covers a great deal of what is loosely called subtext, and the two are worth distinguishing in revision, because they produce different effects: concealment generates suspense, face-work generates poignancy.

6.2 The Gricean account, formalised

§3.1 made the identification; this section states it operationally.

Subtext is conversational implicature. A character generates subtext by visibly flouting one of Grice’s four maxims while remaining, at a deeper level, cooperative. The reader computes the meaning that restores cooperation. That computed meaning is the subtext.

The value of the identification is that it turns an intuition into a generative procedure. There are only four maxims. There are therefore only so many kinds of subtext, and a writer who is stuck can work through them systematically:

Each produces a distinct effect, and each is realised by one or more of the named mechanisms in §6.4.

One refinement matters for fiction specifically. In conversation, an implicature is computed by the hearer. In fiction there are two hearers — the character addressed and the reader — and they need not compute the same thing. Three configurations are available, and choosing between them is one of the more consequential decisions in an exchange:

The reader-only configuration is the theme domain’s dramatic-irony carrier (§4.4). The character-only configuration is rarer, harder, and the source of much of the pleasure of rereading: an exchange whose second meaning is available only once the book has supplied the missing premise.

6.3 What the narrator adds

Prose fiction has a resource drama does not, and most craft advice on dialogue — imported wholesale from screenwriting — ignores it. A novel has a narrator standing between the speech and the reader, and that narrator can do four things a stage cannot:

1. Withhold interiority, so that the reader is in the same position as the other character. This is the dramatic mode, and it is what Hemingway’s story does.
2. Supply interiority selectively, exposing one speaker’s subtext and not the other’s. This is the ordinary condition of the third-person novel, and it converts an exchange into an asymmetry: the reader is inside one player’s hand.
3. Contradict the speech, by reporting an action, a physical detail or a thought that the words deny. A character who says take your time while checking the clock has been contradicted by the narration, not by another speaker.
4. Refuse to arbitrate, which is a positive act in the presence of the other three. A narrator who could tell the reader what a character means and does not has produced a silence of the narration, distinct from the silence of the character.

The third of these is the workhorse, and it deserves a name because it is the most transferable technique in the chapter: the contradicting action beat. It is cheap, it is invisible, it discharges texture and relation while doing it, and it is available on any line. A great deal of the difference between drafted and published dialogue is that published dialogue puts the counter-evidence in the room.

6.4 A taxonomy of subtext mechanisms

Twelve named strategies. Each is a way of producing a gap between the channels; each has a characteristic effect, and a characteristic cost. The middle column names the Gricean flouting where one is clearly involved.

Each is developed below with an exemplar, what it buys, and what it costs.

1. Evasion. The question is heard, its relevance is acknowledged — often explicitly, I know what you’re asking — and the turn produced does not contain an answer. It is the purest form of the mechanism, and the one most dependent on the surrounding norm: it registers only where an answer was clearly due.

Exemplar. Stevens, in The Remains of the Day, asked about matters touching his father, his employer’s politics or his own feelings, produces turns of impeccable length and total emptiness. Buys: characterisation and relational value simultaneously, since a refused answer is a status act. Costs: it becomes exhausting when sustained; a character who evades everything becomes unreadable rather than opaque, which is a real risk in Ishiguro’s novel and one he manages by varying the register of the evasions.

2. The Changed Subject. The most cited and least analysed of the mechanisms. A character asked a direct question replies about something else, and the reader computes the implicature: the true answer is unbearable, or damaging, or would concede too much.

Exemplar. Hemingway’s man and girl swerve between the drinks, the beaded curtain, the hills and the train, and every swerve is legible as a swerve. Buys: the maximum implicature for the minimum text; it is the technique’s cleanest instance. Costs: it can read as authorial evasion rather than character evasion if the reader cannot tell what is being avoided. The rule is that the reader must know what question was dodged, or the swerve is noise.

3. The Question Answered With a Question. A variant sharp enough to warrant separate listing, because its relational meaning is different: it does not merely decline the answer, it transfers the obligation. Where were you? — Why does it matter?

Exemplar. Elizabeth Bennet uses it repeatedly against Darcy and against Lady Catherine, and it is the mechanism by which she keeps footing against social superiors. Buys: status, immediately and visibly. Costs: it is combative, and two characters using it alternately produce a tennis match (§8.5).

4. Over-answering. A character supplies substantially more than was asked for. The implicature is anxiety, guilt, or the construction of an alibi — the disproportion signals that the question mattered more than it should have.

Exemplar. This is the mechanism of the guilty explanation in every detective novel worth reading, and Dostoevsky uses it at length in Crime and Punishment’s interview scenes, where Raskolnikov’s compulsive elaboration is the evidence against him. Buys: an unmistakable signal at no cost to plausibility. Costs: it lengthens the page, and it is close to self-announcing; readers spot it easily.

5. The Too-Precise Reply. A specialised over-answer: exactitude where approximation was due. Asked when he last saw her, a character gives the date and the hour. The implicature is that the matter has been thought about a great deal.

Exemplar. Stevens’s habit of dating and locating his own recollections with unnecessary precision, in a novel whose subject is the unreliability of his recollection, is the definitive case. Buys: interiority without interiority; the reader learns what the character has been dwelling on. Costs: almost none, which is why it is the most under-used mechanism on this list.

6. The Delayed Agreement. Assent arrives, but in the packaging that conversation analysis identifies as belonging to refusals (§3.2): a pause, a hesitation marker, a token agreement, an account, mitigation. The reader registers that a refusal was intended and defeated.

Exemplar. Miss Kenton grants Stevens’s professional requests in this format throughout; the agreements are technically complete and audibly reluctant. Buys: the most efficient rendering of reluctance available in the language, and one grounded in a documented feature of talk rather than in an intuition. Costs: it requires the writer to trust the reader to hear format rather than content, which drafts often do not.

7. The False Agreement. Assent given in bad faith to close the exchange, which both speakers know is not assent. A flouting of quality where the flouting is licensed and mutual.

Exemplar. Countless marital exchanges in James and in Eliot end this way, and it is the standard closing move of an argument that neither party can win. Buys: a way to end an exchange without resolving it, which is structurally invaluable, since unresolved exchanges are what generate subsequent ones. Costs: it can read as a fudge if the reader is not shown, later, that the agreement was hollow. A false agreement must be cashed.

8. The Third-Person Deflection. The speaker attributes their own position to an absent party, to a general practice, or to an institution: people are saying; Ma said you’d ask that; it’s not done. Goffman’s footing (§3.3) is the exact account — the speaker is animating a position while declining principalship of it.

Exemplar. Dov’s line in the household exchange at §9.9 — She said you’d ask that. She didn’t say I had to answer — is a double deflection, first to the mother and then to the absence of an instruction. Buys: deniability, which is a relational asset, and a great deal of information about who the speaker is protecting. Costs: it can flatten a character into a functionary if used as their only register.

9. The Ritual Formula. A courtesy, condolence or set phrase supplied where a substantive answer was owed. I’m sorry for your loss. That’s very kind. These things happen. The formula is unimpeachable and empty, and its emptiness is the message.

Exemplar. The condolences offered to Stevens on his father’s death, and Stevens’s own thanks for them, are a sustained instance; so is nearly every exchange in a nineteenth-century drawing room where the real subject is money. Buys: an enormous amount of relational information at no cost to decorum, and it is the mechanism that makes polite society legible in fiction. Costs: the reader must know the formula is a formula, which is a problem across cultural and historical distance — a real difficulty in teaching Austen to students who cannot hear which of her civilities are dead.

10. The Interrupted Concession. A speaker begins to yield — You’re right, I should have — and is cut off before the concession completes, so that it never becomes binding. The interruption may be hostile, may be a rescue, or may be the interrupter’s inability to hear a concession they had not expected.

Exemplar. The mechanism recurs throughout Wuthering Heights, where Nelly’s narration repeatedly reports speech broken off, and in James, where the interruption is often the more delicate of the two speakers preventing the other from saying something irrevocable. Buys: the appearance of movement without the fact of it, which is what a mid-book exchange often needs. Costs: it is a delaying device, and a reader who notices the delay twice will notice it every time thereafter.

11. The Significant Repetition. A word from the other speaker’s turn is returned, unchanged and uncommented. Did she. Not this key. Fine. The repetition performs receipt without agreement, and its flatness is the flouting of manner.

Exemplar. Elena’s Did she at §9.9. Hemingway’s characters do it continually; so, in a different register, do Pinter’s. Buys: the shortest possible turn that is nonetheless an act, and a way of marking that a word has landed hard. Costs: it is stylistically conspicuous, and in quantity it stops reading as speech and starts reading as technique.

12. The Unfinished Sentence. The turn stops before its point, and the hearer — and the reader — complete it. Distinct from interruption, in that no one else has done anything: the speaker has stopped.

Exemplar. Stevens’s sentences do this at the crises of The Remains of the Day, and the novel’s most famous effects are produced by his declining to complete a thought about Miss Kenton. Buys: the reader’s participation, which is the deepest form of investment; the sentence the reader finishes is a sentence the reader has written. Costs: it is the most easily overplayed device on the list, and the ellipsis is the most overused punctuation mark in draft fiction. Use it where the character cannot continue, not where the writer wishes to imply depth.

How to use the taxonomy: substitution testing

The taxonomy’s purpose is not classification but substitution. A flat exchange is very often the right material in the wrong mechanism. The productive exercise is to take one turn — usually the most on-the-nose one — and rewrite it four times, once each as an evasion, a changed subject, a too-precise reply and a delayed agreement. The four versions will differ in what they imply about the speaker, and the choice between them is a characterisation decision the writer did not previously know they were making.

Three practical notes on selection:

• Mechanisms are not interchangeable in status terms. Evasion and the changed subject are defensive; the question-for-a-question and the significant repetition are aggressive; the ritual formula and the third-person deflection are institutional, and characterise a speaker as someone with a position to protect.
• Mechanisms should vary between characters. If both speakers evade in the same way, they are the same person (§8.2). A useful assignment in revision is to give each major character a default mechanism and one they never use.
• Mechanisms should vary across a novel. A character whose evasions become predictable has stopped being evasive and started being a habit.

6.5 The gap-closing move

Everything above concerns widening the gap between the channels, which is the right move through most of the body of a book. The chapter’s central structural claim concerns the opposite operation.

Closing the gap — letting text and subtext finally say the same thing — is a climactic move. It is available once, perhaps twice, in a novel, and it is spent the moment it is used.

The mechanism is simple once stated. A reader who has learned across three hundred pages that this character deflects, hedges and does not finish sentences has been trained to expect the gap. When the gap closes — when the character says what they mean, plainly, once — the plainness is legible as an event, because the reader has a baseline against which to measure it. The line need not be eloquent. It generally should not be. Its power is positional.

This is why the prohibition on on-the-nose dialogue is not a rule against direct speech. It is a rule against direct speech everywhere, which spends the technique’s one great effect before the story has earned it. A novel in which characters routinely say what they mean has no gap-closing move available at its climax, because there is no gap to close, and writers in this position typically reach instead for volume, event or a declaration that the reader has no reason to find remarkable.

Three published instances, chosen because they are structurally identical and tonally unlike each other:

Note that in two of the three the gap is closed by one party while the other keeps theirs shut. That asymmetry is the more useful lesson: the move does not require mutual disclosure, and is frequently stronger without it.

A corollary for revision. If a draft’s climactic exchange is not landing, the first thing to check is not the climax but the two hundred pages before it. The move requires a trained reader, and the training is done in exchanges that seem, individually, to be doing very little.

6.6 Historical drift, and a caution

The two-channel model presented here is not a description of dialogue as such. It is a description of a particular and datable ideal.

Two inferences follow, pulling in different directions, which is the point.

• The subtext ideal is a real technical achievement and worth teaching. It engages the reader’s inferential capacity, it produces characterisation without assertion, and it is the reason a great deal of twentieth-century fiction rereads well.
• It is a preference, not a law, and its dominance is a fact about a literary culture. Dostoevsky’s characters say outright what they think, at enormous length, and are among the most powerful in the form. Wilde’s say things nobody has ever said aloud. Compton-Burnett’s say outright what they mean in a syntax no one has ever spoken. §16 develops the argument; it is stated here so that the taxonomy above is not mistaken for a specification.