The Narrative Presupposition Audit (NPA) — Version 0.2

 

The Narrative Presupposition Audit (NPA) — Version 0.2

v0.2 — Adds AUTHOR IMPORT to the named failure modes. No other change.

Theoretical foundations: Grant C. Sterling (Eastern Illinois University). Analysis and synthesis: Dave Kelly. Prose rendering: Claude (Anthropic). 2026.

---

Object

A single work of narrative fiction, examined for what it depicts as true about the person, his agency, and the moral order — at the pre-argumentative layer, meaning the layer at which such questions are answered by depiction before any character or narrator argues them.

Relation to Existing Instruments

The Classical Presupposition Audit (CPA) audits what a named figure argued. The Classical Field Audit (CFA) audits a field’s governing practice. The Cultural Displacement Audit (CDA) audits a domain’s behaviour. The NPA audits what a work depicts. Where a work argues, the argued material leaves the NPA and goes to the CPA.

---

Step 0 — Protocol Activation

State each of the following before any text is opened:

• (a) Work, edition, translation. Named exactly. For translated works, the translator is named and recorded as a limit on all subsequent findings.
• (b) Selection rule. The mechanical rule by which passages will be chosen, fixed now and not revised after results appear. Permitted rules: fixed interval; token-density ranking on a stated search term; complete work. A rule that names chapters by their expected yield is not permitted.
• (c) Disconfirming result. State what finding would count against the expectation held going in. If no such finding can be named, the run does not proceed.
• (d) Prior conclusion check. None stated or implied, or the prior conclusion named and held as hypothesis.

Self-Audit — Step 0:

• Edition named: ✓
• Selection rule mechanical and fixed: ✓
• Disconfirming result stated: ✓
• No prior conclusion: ✓

---

Step 1 — Layer Test

Determine for the work, and separately for any passage, whether the material asserts or depicts. Assertion includes authorial essays, prefaces, and sustained character argument. Depiction is everything shown without being claimed.

Assertion is out of NPA scope and is routed to the CPA. A work may be split: Zola’s novels depict and Le Roman expérimental asserts; Dostoevsky depicts throughout and argues in Ivan’s mouth.

Self-Audit — Step 1:

• Layer determined for the work: ✓
• Asserted material identified and routed: ✓

---

Step 2 — Passage Register

Open each selected passage in full. No finding may rest on a snippet, a summary, or a critical account.

Log each passage with a voice register:

• N — narration
• A — authorized character (one the work endorses)
• T — tested character (one the work stages in order to answer)
• R — reported at second hand (a character’s position as described by another)

R-tagged material may not ground a finding about the work.

Self-Audit — Step 2:

• Passages opened in full: ✓
• Voice register assigned to each: ✓
• No R-only findings: ✓

---

Step 3 — Commitment Findings

For each of C1 through C6, issue one finding on this scale:

• Carries — the work depicts a world in which the commitment holds, in the corpus’s form.
• Divergently Affirms — the commitment holds, but grounded or known in a way the corpus excludes.
• Non-Operative — the work does not engage the commitment. Not a deficiency finding.
• Displaces — the work depicts a world in which the counter-commitment holds.
• Split — voices within the work differ. The constituent values and their voice tags are named.

Each finding carries its voice tags and a quoted or paraphrased warrant from an opened passage.

Standing instruction at C2. Character formed by conditions is compatible with assent originated by the agent. A work is not Displacing at C2 merely for depicting formation. The finding turns on whether the work presents the moment of assent as the agent’s own — most visibly where a character pleads that he could not have helped it and the work displays the resolutions that produced the act.

Self-Audit — Step 3:

• Six findings issued: ✓
• Each warranted from an opened passage: ✓
• Voice tags carried: ✓
• Formation/origination distinction applied at C2: ✓

---

Step 4 — Work Profile

State the pattern across the six, and what the work is available to the corpus for — the specific use, not a grade. Note any commitment on which the work is the strongest available witness.

Self-Audit — Step 4:

• Profile derived from the findings: ✓
• No finding introduced at this step: ✓

---

Step 5 — Author Profile — STUB

Not available until two or more NPA runs on the same author are complete. A finding constant across the runs belongs to the author; a finding that varies belongs to the work. Specification deferred.

---

Named Failure Modes

• SELECTION-DRIVEN SAMPLE — passages chosen because they were expected to yield.
• VOICE CONFLATION — a character’s position attributed to the work.
• AUTHOR IMPORT — a finding about the work carried across to a claim about the person who wrote it. The NPA takes a work as its object and can never reach the author. Even a completed Step 5 reaches only the author’s practice across works — what she consistently depicts — which remains a fact about the works. Claims about the person belong to the CPA, on the argumentative record, or to personality analysis, on the letters and the recorded conduct. A term found in a character’s mouth is not evidence about the author’s temperament.
• SECOND-HAND ATTRIBUTION — an R-tagged report treated as the position it reports.
• LAYER CONFUSION — asserted material audited as depiction.
• FORMATION COLLAPSE — depicted formation read as displaced origination.
• TRANSLATION DICTION CLAIM — a finding resting on wording in a translated text.
• NULL PROFILE — five or more Non-Operative findings. The run is reported as out of scope, not as a profile: the framework is foreign to the work, which is a fact about the fit, not about the work.
• SNIPPET FINDING — any finding from a source not opened in full.

Ordinary Language Standard

Greek terms, philosophers’ English, and corpus coinages glossed in plain words at first use in every run.

---

Open Stubs

No confidence-level apparatus is specified. The CFA and CDA carry one and this instrument may need it, but not until a run shows where the uncertainty actually falls.

No rule is yet set for how many passages constitute an adequate sample. The first run records what it used, and the number is set by experience rather than in advance.

---

Theoretical foundations: Grant C. Sterling (Eastern Illinois University). Analysis and synthesis: Dave Kelly. Prose rendering: Claude (Anthropic). 2026.