The fiction writer's voice audit must distinguish between the narrator's voice and the characters' voices — two different consistency problems that require different revision instruments. The narrator's voice is the prose itself: the sentences between dialogue and deep interiority, the descriptive passages, the narrative summary. This is the voice the audit is primarily assessing for consistency. Character voices are assessed separately, through the character revision's voice test from Week 3. A narrator's voice that goes generic in sections of narrative summary — the passages where the story is moving between dramatized scenes — is the most common fiction voice problem: the writer is in scene-mode for the dramatic scenes and in neutral-gear for the connective tissue, and the connective tissue suffers. The revision raises the connective tissue to the narrator's full standard: even the summary passages should sound like this narrator at their most specific and inhabited.
The Voice Revision — Consistency, Depth, and the Signature Sentence
Voice revision is the last of the structural passes before the line-level work begins. In Year Two you developed your voice consciously. Now you audit the thesis for consistency: where does the voice shift unexpectedly? Where does it go flat or generic? And where are the passages of extraordinary voice — the signature sentences — that can serve as the target standard for revision of everything weaker?
Voice as Architecture — What Consistency Means and What Variation Costs
Voice is the most difficult element of prose to analyze because it is the most pervasive. It is not located in a particular sentence or a particular passage — it is the cumulative effect of every choice the writer makes, at every level, across the whole manuscript. Sentence length, diction register, the ratio of abstraction to concreteness, the rhythm of the syntax, the relationship between the narrator and the reader, the characteristic way the prose handles time and approaches its subjects: all of these are voice, and none of them individually constitutes it. This is why voice is easy to recognize and difficult to teach — it is not a technique that can be applied but a sensibility that manifests through the accumulated pressure of hundreds of small decisions, most of them made without conscious awareness that a decision is being made.
The voice revision audit is not an attempt to make the prose more 'stylistic' or to impose a single tone on sections that are legitimately different in register or approach. Variation in register — the narrative that shifts between interiority and scene, between close rendering and distant summary, between the lyric passage and the plain sentence — is not a voice failure; it is a voice resource, available to the writer who can control it. The failure the voice audit identifies is something more specific: the unintentional shift, the passage where the voice has gone generic not because the subject called for plainness but because the writer's confidence or relationship to the material shifted — where the writing became careful rather than inhabited, where the prose is covering its ground without the specific syntactic energy that makes a voice recognizable as a particular person's rather than as adequate literary prose generally.
The signature sentence is the test. Find the three sentences in your thesis that are most fully yours — that could not have been written by anyone else. Post them near your writing space. When a passage's voice feels uncertain, return to those three sentences and ask: what does the writer who wrote those sentences know about how to make language move?
The confidence drop: sections drafted when the writer was uncertain — uncertain about what the scene was doing, uncertain about whether the material was working, uncertain about their own capacity to handle it — often produce prose that is technically adequate but tonally absent: it gets through the scene without the specific syntactic pressure that makes the voice present. The confidence-drop passage is identifiable by its generic quality — the descriptions that are observant but not particular, the interiority that is articulate but not specific, the dialogue that is functional but not inhabited. The revision restores the specificity the uncertainty suppressed: the concrete detail the writer reached past to get through the scene, the syntactic structure they defaulted away from, the register they flattened to avoid risk. The confident passages — the signature-sentence passages — are the model; the revision brings the confidence-drop passages toward that standard.
The relationship shift: some passages are drafted when the writer's relationship to the material is different from the relationship that governs the rest of the manuscript. A memoir drafted over two years will contain passages written from at least two different positions — the earlier position, when the material was still being worked through, and the later one, when the writer has arrived at the understanding the whole book is organized around. The relationship-shift passage has a tonal quality different from the surrounding material — sometimes more raw, sometimes more defended, sometimes more resolved than the manuscript as a whole has earned. The voice revision identifies these passages not to eliminate their specific quality but to assess whether that quality is serving the manuscript or creating an inconsistency it cannot sustain. Some relationship-shift passages are the manuscript's most powerful: the rawer sections of a memoir that has found its bearings in the later drafting carry an authentic urgency the more settled passages cannot replicate. The question is whether the shift is working or whether it is tonal static.
The research intrusion: sections that required substantial research — historical context, technical information, documented events — sometimes produce prose in which the research is visible as research: the register shifts toward the expository or the academic, the syntax becomes more neutral and informational, the voice recedes behind the obligation to be accurate and thorough. The research intrusion is most common in CNF and historical fiction; its effect is a flattening of the narrative voice at precisely the moments when the material's complexity requires the most distinctive rendering. The revision converts the research from exposition into scene: the information is not removed but translated into the mode the manuscript's voice is capable of sustaining — placed in a specific character's perception, filtered through a specific sensory context, embedded in the scene rather than presented in summary.
The genre convention override: some passages drafted early in the project — before the manuscript's specific voice was fully established — are written toward the genre's general conventions rather than toward this particular manuscript's particular approach. The passage that sounds like a memoir in general, rather than like this memoir, is the genre convention override: it is doing what memoirs do without doing what this memoir does. The revision identifies these passages by their relative interchangeability — they could be from another good memoir and would not seem out of place — and revises them toward the signature sentence's standard: the irreplaceable, the specific, the passages that could only have been written by the writer who wrote this manuscript.
Voice Consistency Across All Three Tracks
For screenwriters, the voice audit applies primarily to the action lines — the scene description that constitutes the screenplay's prose — rather than to the dialogue, which is character voice rather than writer voice. The screenwriter's action lines are where their distinct sensibility is most visible: the specific choice of what to render and what to omit, the rhythm and length of the descriptive sentences, the relationship between what is visible to the camera and what is implied. A screenplay whose action lines are generic — that describe what happens without the specific angle, the specific detail, the specific relationship to the story's emotional content that makes the description serve more than one purpose — is a screenplay without a writer's voice, whatever the quality of its dialogue. The voice audit asks of each scene description: could this have been written by anyone competent writing this genre, or does it carry the specific intelligence of this writer?
The memoirist's voice audit is complicated by the essay's requirement that the narrator's relationship to the material evolve over the course of the work. A memoir in which the narrator's voice is identical throughout — equally settled, equally certain, equally at the same distance from the subject — has a voice consistency of a kind that is actually a voice failure: it suggests a narrator who has not genuinely been changed by the investigation. The CNF voice audit should distinguish between intentional voice variation — the shifts in tone, proximity, and certainty that reflect the narrator's genuine movement through the material — and unintentional voice inconsistency: the passages where the voice has shifted not because the narrator's relationship to the material has shifted but because the writer's confidence or command has. The test: is this voice variation serving the memoir's account of the narrator's development, or is it serving the writer's inability to sustain the full voice throughout?
Phase 6, Topic 5 — Eliminating Adverbs and Strengthening Verbs
The adverb that modifies an action verb is almost always a symptom rather than a solution. 'She walked quickly' — the adverb is telling the reader how the walking happened because 'walked' is not doing the work of conveying the manner. The available verbs: hurried, rushed, strode, dashed, scurried, sprinted, charged, scuttled. Each of these carries the manner inside itself, without the external modifier. The adverb that tells how is usually the signal that the verb is too general for what the sentence is trying to do — that the writer has chosen the broad verb and then qualified it, when the specific verb was available and would have produced a sentence that is shorter, more precise, and more physical.
The adverb that modifies a dialogue attribution is the same symptom in dialogue: 'he said sadly,' 'she replied sharply,' 'he answered slowly.' The attribution adverb tells the reader how the speech was delivered because the dialogue itself, or the action surrounding it, has not conveyed the emotional content directly. The revision finds the element that will carry what the adverb was covering — the line of dialogue rewritten to carry its own sadness in the words' choice and rhythm rather than in the stage direction attached to the attribution; the physical action that carries the sharpness; the pacing or rhythm of the spoken lines that carries the slowness. The attribution adverb is almost always cuttable if the dialogue is doing its work.
The legitimate adverb — the adverb that is genuinely doing work that nothing else can do — is rarer than most drafts contain but exists. The adverb modifying an adjective ('deeply ambivalent,' 'barely visible,' 'almost unbearable') is often legitimate: the degree modifier is changing the adjective's meaning in a way that the adjective alone cannot convey. The adverb of manner that has no available single-word equivalent — where the verb carries the action but the adverb adds a specific quality that no other construction supplies — is legitimate. The test: after removing the adverb, does the sentence lose something specific and irreplaceable, or does it lose only the writer's anxiety about whether the reader has understood? If the latter, the adverb was covering the writer's distrust of their own prose, and the revision should remove the adverb and trust the prose.
The adverb audit: find twenty adverbs in the thesis draft — any word ending in '-ly' modifying a verb, adjective, or another adverb, plus the adverbs of degree (very, quite, rather, almost, nearly, barely, hardly) modifying adjectives. For fifteen of them, find the stronger verb, the more specific action, or the concrete image that does the adverb's work without the modifier. Keep five where the adverb is genuinely doing irreplaceable work. The fifteen-to-five ratio is a heuristic, not a rule — some drafts will have more legitimate adverbs, some fewer. But the exercise's discipline is the point: it develops the habit of questioning each adverb rather than accepting it, and most adverbs do not survive the question.
Find twenty adverbs in your thesis draft. For fifteen, find the stronger verb or specific image that does the adverb's work without the modifier. Keep five where the adverb is genuinely irreplaceable. Note in the session log: what patterns emerge — are the adverbs clustered around certain verbs, certain characters, certain types of scenes?
This Week's Texts
Your Year Two Synthesis Statement
You
Required. This is the primary text for the voice revision. The synthesis statement's account of your voice — the specific features you identified, the syntactic habits, the characteristic relationship to the reader, the specific elements of Phase 1–7 grammar you described as integrated versus still being worked on — is the standard the voice audit is working from. Read it before the audit. Have it open alongside the draft during the audit. The synthesis statement is not describing what you want your voice to be; it is describing what it already is. The voice audit's question is whether the full manuscript reaches that standard throughout.
Your ten strongest pages
You
Required. Before the audit: read the full thesis draft and identify the ten pages — not necessarily contiguous — where the voice is most fully yours. These are the benchmark. They may not be the most dramatically intense pages, or the most polished, or the pages whose content you are most invested in. They are the pages where the prose is most inhabited — where the specific syntactic energy that makes your writing identifiable as yours is most concentrated. Mark them. They are what the rest of the manuscript is being revised toward.
The Voice Consistency Pass
Read the thesis specifically for voice — not for plot logic, not for character consistency, not for thematic engagement, but for the quality of inhabitation in the prose. Mark every passage where the voice drops: goes generic, loses its syntactic specificity, sounds like adequate literary prose rather than this particular writer's prose. Mark every passage where the voice is fully present: where the signature-sentence quality is concentrated, where the prose could not have been written by anyone else.
From the marked passages, select the two weakest-voice sections — not the two weakest in content, not the two most in need of structural revision, but the two where the voice is most absent. These are the targets for this week's revision.
Revise both sections toward the voice standard established by the benchmark pages. The revision is not a stylistic overlay — do not impose the voice onto the content. Instead, identify the specific ways the benchmark pages produce their voice effect (sentence length and rhythm, diction choices, relationship between the abstract and the concrete, characteristic syntactic constructions from the Year Two synthesis statement) and ask, for each weak-voice sentence, what a writer with that specific voice would have written here instead. The revision preserves all content; it changes the specific syntactic choices, the diction register, the rhythm, the relationship to the material — everything that constitutes the voice rather than the story. Target: two substantive voice revisions of 300–600 words each.
Exercise #3: Voice Transplant — Applied to Revision
The Voice Transplant exercise from Year Two Week 14 was generative — it applied your voice to new material. This week it is applied to revision: the AI reads a weak-voice passage alongside a strong-voice passage from the same manuscript, identifies the specific features that distinguish them, and revises the weaker passage toward the stronger voice's standard. The value is comparative: the AI's revision alongside your own revision produces the clearest possible account of what the voice's specific features are and what is required to inhabit them.
1. Compare the AI's feature identification to the Year Two synthesis statement's account of your voice. Are the features the AI identified from the two passages consistent with the features you identified through the synthesis statement's analytical work? If the AI has identified features you did not name in the synthesis — specific syntactic patterns, specific diction tendencies — consider updating the synthesis statement with the new precision. The voice audit and the synthesis statement should be mutually informing documents; the audit makes the synthesis more accurate, and the synthesis makes the audit more targeted.
2. Compare the AI's revision of the weak-voice passage to your own revision from the writing exercise. Both revisions have the same goal — bringing the weaker passage toward the stronger voice's standard — but they will approach it differently. The AI's revision will be technically faithful to the features it identified; your revision will be faithful to the voice as you inhabit it from the inside. Where the AI's revision is more fully in the voice, note what it did that you did not: what specific syntactic or diction choice moved the passage closer to the standard. Where your revision is more fully in the voice, note what the AI's analysis missed — what quality of the voice is not fully captured by the features it identified.
3. The voice features that could not be captured in the AI's revision are the features that exist below the level of describable technique — the aspects of the voice that are genuinely the writer's own in the deepest sense, that cannot be reproduced by following a recipe however accurate. These features are the voice's irreducible core, and they are what the voice revision is ultimately serving. Note what they are in the session log, as precisely as possible: not 'something ineffable' but the specific quality — a relationship to uncertainty, a specific rhythm under the syntax, a particular way of being surprised by the material — that the AI's technically accurate revision could not replicate.
4. The voice transplant comparison is particularly useful as a benchmark for the remainder of the structural revision: carry the AI's feature list into the ongoing voice audit as a checklist — not to mechanically apply the features but to have a precise vocabulary for what the voice revision is doing and what it requires. The checklist should be working from the manuscript outward, not from the outside in; it names the features the voice already has, not features being imported.
This is the fifth and final structural revision pass: scene map (Week 2), character revision (Week 3), thematic audit (Week 4), voice audit (Week 5). After this week, the structural revision is complete and the line-level revision begins in Week 6. The transition from structural to line-level is a significant shift in scale and in cognitive mode — from the architectural view to the sentence-by-sentence view. The voice audit is the bridge between the two: it identifies the passages where the voice is weakest, which are the passages where the line-level revision will do the most work.
The Signature Sentence as Target
Identify the three sentences in your thesis that are most fully yours — sentences that could not have been written by anyone else writing a competent version of this book. Not the most beautiful sentences, not the most technically accomplished, not the most important to the plot or argument. The three sentences where the specific intelligence and sensibility of this particular writer is most concentrated and most irreplaceable. Post them near your writing space — literally, on paper, visible while you write. When a passage's voice feels uncertain, return to those three sentences and ask: what does the writer who wrote those sentences know about how to make language move? What specific things did they do that this uncertain passage has not done? The signature sentences are not a standard of achievement but an instrument of navigation: they tell you where your voice is when you can find it, so that you can find your way back to it when you have lost it.
The signature sentence test also reveals something about what the voice audit should prioritize. The sections of the manuscript that are furthest from the signature sentences in their syntactic and tonal approach are the sections whose voice revision is most urgent — not because they are the worst writing but because they are the writing most in need of the specific quality the signature sentences possess. The voice audit's blue/red marking, like the thematic audit's, is a map of what the revision needs to do. The signature sentences are the target the blue passages are aiming at.
Where the Voice Lives
When are you most fully yourself in the thesis? What subject, what moment, what type of scene brings your truest voice out — where the prose has a quality of inhabitation that you can feel while writing and that is visible in the draft afterward? It may be a specific kind of emotional register, a specific relationship between the narrator and the material, a specific type of scene (the intimate interior moment, the observed public scene, the scene that moves between past and present). Whatever it is: name it as precisely as you can. Then ask: how can you create more of those conditions in the remaining revision? Not through forcing the voice into territory where it isn't natural — voice cannot be forced — but through identifying what the conditions of your voice's fullest presence are, and arranging the revision to create those conditions more consistently. The structural passes are complete. The remaining work is to bring the full manuscript to the standard your best pages have established. What does your voice need to do that?
What You've Built — and What Begins Next Week
By the end of this week you should have: completed the full voice audit (weak-voice passages marked across the manuscript); identified the ten benchmark pages and the three signature sentences; completed both voice revisions (two passages of 300–600 words each); completed the AI Voice Transplant comparison with all four reflection questions; completed the adverb audit (twenty adverbs found, fifteen replaced, five keepers noted); updated the session log with the voice features identified across the AI exercise and the synthesis statement comparison.
Five structural passes are now complete: the triage (Week 1), the scene map (Week 2), the character revision (Week 3), the thematic audit (Week 4), the voice audit (Week 5). The manuscript's architecture is now as sound as the structural revision can make it. Week 6 begins the line-level revision: the sentence pass, the most demanding and most granular work of the revision sequence. Grammar: Phase 6 Topic 6 — zombie nouns and nominalizations. The sentence pass will take multiple weeks; the voice audit's findings about which passages are weakest provide its starting point.