← Program HomeMFA Year Three · Fall SemesterWeek 8 →

Week 7 of 36 · Fall Semester · Line-Level Revision

Line-Level Revision II — Dialogue, Interiority, and the Scene's Final Polish

The second sentence pass focuses specifically on dialogue and interiority — the two elements that most directly carry character and that most often need the most work in revision. Dialogue that was merely functional in early drafts can become genuinely alive in revision. Interiority that was over-explained can be stripped to its essential core. This is where the characters become themselves.

Commitment15–20 hrs

Program Week79 of 108

Craft FocusDialogue & Interiority Revision

GrammarPhase 6, Topic 7 — Register, Diction Levels, and Code-Switching

DeliverableThree Revised Dialogue Scenes

Craft Lecture

What Dialogue Does That Prose Cannot — and What Interiority Does That Dialogue Cannot

Dialogue and interiority are not decorative elements of prose fiction — they are, in many ways, its primary instruments. Dialogue is where character becomes action: where what a person says, and the gap between what they say and what they mean, and what they choose not to say at all, constitutes the most immediate available evidence of who they are. Interiority is where the reader inhabits a consciousness: where the character's relationship to the events they are experiencing — their interpretation, their self-deception, their private knowledge — produces the intimacy that prose narrative uniquely offers and that no other storytelling form can replicate. The scene that is merely described from the outside, without dialogue to generate its conflict and interiority to give it depth, is a scene the reader observes. The scene that uses both instruments well is a scene the reader lives inside.

The line-level revision's dialogue and interiority pass is not the same as the dialogue and interiority work of the structural passes. The structural character revision asked whether characters were distinguishable, whether their governing desires were legible, whether they earned their place in the architecture. The line-level pass asks something more granular: is every line of dialogue the right line? Is every sentence of interiority doing what only interiority can do, or is it doing what the scene's action has already done more effectively? The line-level pass is where the functional becomes the genuine — where the dialogue that gets the information across becomes the dialogue that could only belong to this character, at this moment, in this particular state of wanting and fear.

Read only the dialogue — skip all action and attribution. Can you identify the two speakers by voice alone? Does the scene's turn happen in the dialogue itself, or around it? If around it, the dialogue is not carrying what the scene needs it to carry.

— craft principle

Four Principles of Dialogue Revision at the Line Level

On-the-nose dialogue is the most common first-draft failure and the most urgent line-level target. On-the-nose dialogue is dialogue in which a character says precisely what they mean — where the subtext is made text, where the emotional content that should be carried beneath the words is instead stated on top of them. 'I'm angry at you because you never listen to me.' 'I wish you understood how much I love you.' 'I can't believe you would do that to me.' These are not bad sentences — they are accurate, they are emotionally honest, they communicate clearly. They are bad dialogue because real people in the grip of strong feeling rarely say exactly what they mean; they say what they can say, which is always something other than the feeling's full truth, and the gap between what they say and what they feel is where the scene's tension lives. The revision finds what this character can say here — the deflection, the displacement, the overstatement of a lesser concern because the real one is unspeakable — and replaces the direct statement with the oblique one. The reader, who can feel the subtext in the oblique statement, experiences the emotion more acutely than they would if it were named.

Compression: most first-draft dialogue runs longer than it should. Characters explain what they are feeling, provide context the reader already has, respond to what was said rather than to what was meant, complete their thoughts when interruption would be more true. The compression test: cover every other line of a dialogue exchange and read only one character's lines. Does the scene make sense? If yes — if one character's lines are responsive to the content of the lines they cannot see rather than to the specific words that preceded them — the exchange may be too evenly balanced, too much a conversation of equals working through information together, when the scene's reality is that each character is talking past the other toward their own need. Real dialogue frequently does not respond directly to what was said. The compression revision removes the lines that respond too directly, that resolve misunderstandings too efficiently, that complete thoughts the scene's tension requires to remain incomplete.

The action beat as dialogue instrument: the action beat — the sentence of physical action that accompanies or interrupts dialogue — is one of the most underused tools in first drafts and one of the most powerful in revision. The action beat does three things simultaneously that dialogue attribution cannot do: it places the character physically in the scene; it provides a specific action whose relationship to what the character is saying creates or deepens subtext; and it breaks the rhythm of the dialogue exchange in a way that controls the reader's pace through the scene. The revision adds action beats where the dialogue has been running as pure exchange — where the characters have been disembodied voices — and removes action beats where they are redundant (describing what the dialogue has already made clear) or distracting (pulling attention away from the exchange at a moment when the exchange should have the reader's full focus).

The said problem — and its solution: Elmore Leonard's rule that writers should use 'said' and 'asked' as their default attributions, reserving all other verbs for moments when they add information the dialogue and action cannot carry, is the single most useful piece of dialogue revision advice available. Most first drafts contain attributions that are attempting to do the dialogue's work: 'she replied,' 'he insisted,' 'she countered,' 'he sighed,' 'she snapped.' These attribution verbs are almost always signals that the dialogue is not carrying its emotional content — that the writer has used the attribution to tell the reader how the line was delivered because the line itself is not delivering it. The attribution audit (the editorial tip) identifies every non-said attribution in the manuscript and asks whether the dialogue and action can carry the information without it. If yes, 'said' replaces it. If the scene is stronger without the attribution at all — if the speaker is clear from the exchange and no attribution is needed — it is cut entirely.

Three Principles of Interiority Revision at the Line Level

Over-explanation is interiority's primary first-draft failure, as on-the-nose dialogue is dialogue's. The over-explained interiority tells the reader what the character is feeling rather than rendering the specific texture of that feeling from inside the character's consciousness. 'She felt angry. She had always felt this way when he dismissed her opinions, and now it was happening again.' This is interiority that explains rather than renders: it names the emotion, provides its history, and connects it to the current event. The revision finds the specific thought — the exact content of the consciousness at this moment — and renders that rather than the emotion's label. Not 'she felt angry' but the specific, idiosyncratic thought that anger produces in this particular person at this particular moment: the irrelevant memory it surfaces, the physical sensation it creates, the inappropriate impulse it generates, the specific way this person's anger is different from generic anger.

The filtering problem: filtering is the use of unnecessary perception verbs that create distance between the reader and the character's experience. 'She saw that the room was empty.' 'He noticed that her hands were shaking.' 'She felt that something was wrong.' Each of these sentences puts the perception verb between the reader and the perceived thing — the reader sees the character seeing rather than seeing directly. The revision removes the filter: 'The room was empty.' 'Her hands were shaking.' 'Something was wrong.' The character's consciousness is still present — the scene is still in their point of view — but the filter that was creating distance has been removed, and the reader inhabits the perception directly. The filtering problem is pervasive in first drafts; the line-level revision should search specifically for 'she saw,' 'he noticed,' 'she felt,' 'he realized,' 'she thought,' and assess each one: is the filter necessary (is there a reason to foreground the act of perception rather than the perceived thing?) or is it a habit creating unwanted distance?

The proportion test: how much interiority is the right amount? The answer depends on the scene's purpose and the manuscript's governing mode. A scene that is primarily exterior — that is doing its work through action and dialogue — needs very little interiority; too much slows the action and signals that the writer does not trust the scene to communicate without explanation. A scene that is primarily interior — that exists to render a character's relationship to events rather than the events themselves — needs sustained interiority, and the interiority revision should deepen rather than compress it. The proportion test asks of each scene: is the ratio of action to interiority appropriate to what this scene is doing? The over-interior scene tells the reader more than they need to know about what the character is thinking; the under-interior scene leaves the reader outside the character at a moment when the inside is the point.

Cross-Genre Note

Dialogue and Interiority Across All Three Tracks

Literary Fiction

Literary fiction's dialogue revision should pay particular attention to the scene's turn: the moment when something changes between the characters, when the power shifts, when the thing that was unspoken becomes impossible not to speak, when one character understands something about the other or themselves that they did not understand before the scene began. The turn should happen in the dialogue — in a specific line, at a specific moment — rather than in the action or the interiority that surrounds it. If the scene's turn is being narrated rather than enacted in the exchange itself, the dialogue revision's primary task is to find the line in which the turn happens and make sure it is the best line the scene is capable of producing.

Screenwriting & Playwriting

The screenplay and stage play have no direct interiority — what the character is thinking is not available to the audience unless it is externalized through action, dialogue, or (in theatre) the conventions of soliloquy and aside. The dialogue revision for dramatic work is therefore doing double duty: it must carry not only what the character means and what they are concealing, but also what in prose would be rendered as interiority — the character's relationship to what is happening, their private interpretation of events, their governing desire in this moment. The action beats in a screenplay are often the primary carrier of this interiority-equivalent: the gesture that reveals the private response, the pause that communicates more than speech would, the physical action that is the visible surface of the invisible feeling. The dialogue revision for screen and stage should be conducted alongside an action-beat audit: are the action beats doing the work of interiority, or has the script left its characters' inner lives to be inferred from words alone?

Creative Nonfiction & Memoir

Memoir dialogue operates under an ethical and epistemological constraint that fiction dialogue does not: the writer cannot know exactly what was said, and the dialogue they render is a reconstruction — accurate to the spirit of the exchange, to the relationship and the moment, but not a transcript. The memoir dialogue revision should assess whether the reconstructed dialogue is doing the work that the ethical conventions of the form permit: conveying the emotional truth and relational dynamics of the exchange rather than claiming more precision than memory can supply. The overly specific memoir dialogue — the exchange rendered with the exactitude of a transcript decades after the fact — raises readerly skepticism; the dialogue revision may need to adjust specificity in the direction of acknowledged reconstruction. At the same time, the memoir's interiority — the narrating self's commentary on what the experiencing self thought and felt — is often the form's primary strength, and the interiority revision should assess whether the narrating self's voice is doing its full work in the interiority passages.

Grammar & Style

Phase 6, Topic 7 — Register, Diction Levels, and Code-Switching

Phase 6 · Register — Every Character Lives at a Specific Level of Language

Register is the social and tonal level of language use: formal, informal, technical, vernacular, elevated, demotic. Every speaker occupies a characteristic register — a home base of diction and syntax that reflects their education, their social position, their relationship to the person they are speaking with, and the emotional state they are in at the moment of speaking. The first-year grammar curriculum established the two great lexical streams (Anglo-Saxon and Latinate); the diction register work builds on that foundation by adding the social and situational dimension: the same person speaking in their professional register and their domestic register and their register under emotional duress are speaking in syntactically and lexically different ways, and the failure to differentiate those registers in dialogue is the failure to render the full complexity of how real people use language.

Code-switching is the practice of moving between registers — which characters do constantly in real life, often without awareness. The lawyer who uses precise legal language at work and neighborhood vernacular with childhood friends; the first-generation immigrant who switches between the diction of their native language's influence and the register of their adopted culture depending on context; the character who speaks one way to their boss and another way to their children; the person who adopts a more formal register under stress as a form of emotional self-protection and drops it in moments of genuine ease. Code-switching is not affectation or inconsistency — it is how language actually works in a social world, and the manuscript that ignores it produces characters whose speech is unnaturally consistent across all contexts.

The register audit asks two questions of each character's dialogue across the full manuscript: Does this character have a characteristic register — a home base that is specific to them and distinguishable from other characters' home bases? And does the character code-switch realistically — do their register shifts reflect the changes in social context and emotional state that the narrative places them in? A character who speaks in the same register in every scene, with every other character, in every emotional condition, is a character whose language is not reflecting the full reality of how people use language. The revision identifies where the register is static when it should be dynamic, and where code-switching opportunities have been missed that would deepen the characterization and the scene's social texture.

Register contrast as characterization — the same character in two contextsAt work: 'I'll need to review the documentation before we can proceed to the next phase of implementation.' At home, to their child: 'I don't know, baby. I just don't know.' Same character; two entirely different registers — the first Latinate, abstract, institutional, self-protective; the second Anglo-Saxon, concrete, emotionally exposed, at the bottom of Hayakawa's ladder. The contrast between the two versions is itself characterization: the reader sees the gap between the public self and the private one, and that gap is more revealing than either register alone would be.

Code-switching under emotional stress — when the home register breaks downThe character who maintains a formal, controlled register in most circumstances and whose register breaks down under extreme emotional pressure — who starts with 'I find your behavior deeply problematic' and ends with 'just stop, just please stop' — is a character whose emotional journey the reader can track through the language rather than through the narrator's description of it. The register collapse is one of the most effective tools for the dialogue revision: the moment when a character can no longer sustain their characteristic register, when the vernacular or the raw Anglo-Saxon breaks through the formality, is almost always the scene's emotional peak.

This week's grammar exercise: take one scene with two characters and write each character's dialogue as it would sound in three different registers — formal, informal, and under emotional stress. The exercise develops the ear for register variation and produces material that can be assessed against the dialogue in the current draft: is the draft's dialogue reflecting the register complexity this exercise reveals is available, or is it working from a single register when multiple registers would be more true?

Take a scene with two characters. Write each character's key lines as they would sound in three registers: formal, informal, and under emotional stress. Assess the current draft's dialogue against these three versions: is the draft using the right register for each moment, and where are code-switching opportunities being missed?

Core Reading

This Week's Texts

Writing Fiction

Janet Burroway

The dialogue chapters, reread with Year Three revision eyes. Burroway's account of dialogue in Year One was a framework for drafting; the Year Three reading is a diagnostic tool applied to dialogue that already exists. Read specifically for her account of subtext — the gap between what a character says and what they mean — and apply the diagnostic to the three dialogue scenes selected for revision this week. Where is the subtext in each scene? Where is dialogue on-the-nose, delivering its content directly when the scene would be stronger if the content were delivered obliquely?

Required

10 Rules of Writing

Elmore Leonard

Widely available online. Leonard's rules are primarily dialogue rules — his account of how to write dialogue that sounds like speech without being transcription, how to handle attribution, when to trust the reader to know who is speaking without being told, how the action beat does the work the attribution verb was attempting. Read the full ten rules before beginning the attribution audit. Leonard's rule about 'said' is the attribution audit's governing principle; his rules about what to leave out are the compression test's practical guide.

Required

Two pages of dialogue from your most admired writer in your genre

Your choice

Required. Choose a writer whose dialogue you find most fully alive — not the writer you most admire generally, but the writer whose dialogue specifically produces the effect of inhabited speech: the sense that these lines could only belong to these people, at this moment, in this particular scene. Analyze every word choice: the specific diction register, the length of each line, the proportion of statement to question to non-answer, the action beats and where they fall in the exchange, the attributions and whether they are 'said' or something else and why. Two pages of close analysis of how the best dialogue in your genre works is worth more than twenty pages of reading.

Required

Writing Exercise

The Dialogue Revision — Three Key Scenes

Exercise

Identify the three most important dialogue scenes in your thesis — the three scenes where the exchange between characters is carrying the most weight, where the most is at stake, where what is said and what is not said is doing the most critical narrative work. These are not necessarily the longest dialogue scenes or the scenes with the most lines; they are the scenes where the dialogue is the scene's primary instrument.

For each of the three scenes, conduct the stripping test before revision: read only the dialogue, skipping all action beats and attributions. Ask three questions from the stripped-down read. Can you identify the two speakers by voice alone, without any attribution? Does the scene's turn happen in the dialogue itself — in a specific line — or does it happen in the action and interiority surrounding the exchange? Where is the dialogue most on-the-nose, delivering its emotional content directly when it should be carrying it beneath the surface of the words?

After the stripping test, revise each scene's dialogue toward greater specificity, distinction, and subtext. The revision has three objectives: make each speaker's voice more distinctly their own at the level of diction register and characteristic syntax; find the on-the-nose lines and replace each with the oblique version — the thing the character can say rather than the thing they mean; and ensure the scene's turn happens in the dialogue rather than around it. Target: three substantial dialogue revisions across the three scenes, each involving at minimum the compression of over-long exchanges, the replacement of at least two on-the-nose lines, and the attribution audit applied to every non-said verb.

Three substantial dialogue revisions: stripped-down test conducted, on-the-nose lines replaced, attribution audit applied, scene turns located in the dialogue

AI Workshop

The Dialogue Subtext Assessment

Tool: Your Perfect Tutor / Claude

Paste your most important dialogue scene — the one that carries the most narrative weight, where the most is at stake between the characters. The AI reads only the dialogue, assessing it for voice distinction, subtext, and on-the-nose moments, then rewrites the three most on-the-nose exchanges using subtext and implication.

Read only the dialogue in this scene — set aside the action and attribution for now, and read only the spoken lines. Assess: (1) Can you identify the two speakers by voice alone, without any attribution? What specific features of diction, syntax, and characteristic expression distinguish one speaker from the other? (2) Where is the dialogue most alive — where does subtext do real work, where is the emotional content carried beneath the surface of the words rather than stated on top of them? Identify the specific line or exchange where this is most effective and explain what it is doing. (3) Where is the dialogue on-the-nose — where does a character say exactly what they mean in a way that deflates the tension the scene is building? Identify the three most on-the-nose exchanges. (4) Rewrite those three exchanges using subtext and implication: find what each character can say here — the deflection, the displacement, the thing that is true but not the whole truth — rather than what they mean directly. Preserve the emotional intent of each exchange; change only the mode of delivery from direct to oblique.

1. The AI's assessment of voice distinction: does it match the voice test from the Week 3 character sheets? If the AI can identify the speakers by voice alone from the stripped-down dialogue, the voice revision has succeeded for this scene. If the AI cannot distinguish the speakers, the dialogue revision's most urgent task is the voice differentiation — finding the specific syntactic and diction features that make each character's speech pattern individual. The register exercise from this week's grammar section is the instrument: write each character's lines in three registers and assess which register the current draft is using and whether it is the right one for this moment in the scene.

2. The AI's identification of the dialogue's most alive exchange: compare it to your own assessment from the stripping test. If the AI identifies a different exchange as the most alive than the one you found most effective, the divergence is worth investigating. The AI's choice may reveal that what the writer finds most interesting in the scene — the exchange that is thematically richest, that carries the most narrative significance — is not the exchange that is producing the most immediate effect for a reader coming to it cold. The distinction between the writer's most invested exchange and the reader's most alive exchange is one of the line-level revision's most important findings.

3. The AI's three rewrites of on-the-nose exchanges: evaluate each against your own revision of the same exchanges. For each: which version — the AI's or yours — better serves the scene's specific dramatic situation, the characters' specific relationship, and the voice each character has been established as having? The AI's rewrites will be technically accurate to the principle of subtext over statement; your rewrites will be more faithful to the specific texture of these characters' speech. Where the AI's rewrite is more effective, note what it did that you did not. Where yours is more effective, note what quality of the characters' speech the AI's version could not capture.

4. Apply the attribution audit to the full revised scene after completing the dialogue revision: circle every attribution that is not 'said' or 'asked.' For each: is the verb adding information the revised dialogue and action beats cannot carry? If the dialogue revision has done its work — if the on-the-nose lines have been replaced by oblique ones, if the action beats are carrying the emotional content the attribution verbs were attempting — most of the non-said attributions will be unnecessary. Change them to 'said' or cut the attribution entirely where the speaker is clear. The attribution audit is the final pass on the dialogue revision; it confirms whether the line-level work has trusted the dialogue enough to release it from its stage directions.

The dialogue and interiority pass completes the line-level revision. After this week, the manuscript has been through the full revision sequence: structural (Weeks 1–5) and line-level (Weeks 6–7). Week 8 begins the image revision — a different kind of pass, focused on the governing metaphors and recurring images that carry the manuscript's deepest thematic weight. The line-level work will continue in the background throughout the remaining revision weeks; the sentence pass and dialogue revision are not done in two weeks but established as ongoing practices that the writer carries into every subsequent pass.

Editorial Tip

The Attribution Audit

🗣️

Every Attribution That Isn't 'Said' Is a Confession

Read through your thesis manuscript and circle every dialogue attribution that is not 'said' or 'asked' — every 'she replied,' 'he insisted,' 'she countered,' 'he sighed,' 'she snapped,' 'he admitted.' For each circled attribution: is the verb adding information — about the emotional tone of the delivery, the character's physical state, the social dynamics of the exchange — that the dialogue itself and the surrounding action beats do not already carry? If yes, the attribution may earn its place. If no — if the dialogue line is already delivering the sharpness, the insistence, the exhaustion that the attribution verb is trying to name — change the attribution to 'said.' The non-said attribution is almost always a signal that the dialogue is not doing its full work. The revision that makes the dialogue capable of carrying its own emotional freight will automatically make most non-said attributions unnecessary.

Elmore Leonard's formulation is the cleanest: 'If it sounds right, it is right.' The dialogue line that sounds sharp does not need 'she snapped.' The line that sounds tired does not need 'he said wearily.' The line that sounds like a question — syntactically or in rhythm or in its relationship to what preceded it — does not need 'she asked.' The attribution audit's discipline is the discipline of trusting the dialogue: the revision makes the line capable of carrying its own sound, and then the attribution steps back to 'said' because the line no longer needs to be told how to be read.

Journal Prompt

The Dialogue-Driven Scene

The Scene That Is Held Together by What Is Said

Which scene in your thesis is held together primarily by dialogue — the scene where the exchange between characters is doing the structural, emotional, and thematic work, where what is said and left unsaid is the scene's primary architecture? Name it. Then ask: is the dialogue strong enough to carry it? Apply the stripping test mentally: if you removed all the action and attribution and interiority from this scene and left only the spoken words, would the scene still exist as a scene — would its conflict and its turn and its emotional arc still be present? If yes, the dialogue is doing its work. If no — if the scene depends on what surrounds the dialogue to carry what the dialogue should be carrying — what would happen if you stripped all the action beats from this scene and left only the words? Not as a revision proposal but as a diagnostic: what would be lost, and what does that loss reveal about what the dialogue is currently not doing that it needs to do? Write for twenty minutes on the scene and the dialogue that holds it or fails to hold it, and what the revision needs to give it.

Week in Summary

What You've Built

· · ·

By the end of this week you should have: completed the stripping test on the three most important dialogue scenes; revised all three scenes (on-the-nose lines replaced, voice distinction improved, scene turns located in the exchange, attribution audits applied); completed the AI dialogue subtext assessment with all four reflection questions; completed the register exercise (each character's lines in three registers); analyzed two pages of dialogue from the most admired writer in your genre; updated the session log with findings from the attribution audit and the on-the-nose line replacements.

Looking Ahead to Week 8

With the line-level revision complete, Week 8 begins the image revision: finding and deepening the governing metaphors — the recurring images that arise from the writer's preoccupation and that, developed to their full potential, carry the manuscript's deepest thematic weight. Grammar: Phase 6 Synthesis — the full word-level toolkit applied simultaneously to four pages of the thesis. The image revision is one of the most generative of the revision passes; it often produces new material rather than simply refining what exists.