Omission and Negative-Evidence Audit¶

Negative-evidence audit — instantiates Textual Close-Reading Mode

Tests whether something absent is meaningfully missing by first establishing why it should have been present, then ruling out innocent explanations for the gap.

Most close reading works from what is on the page. The Omission and Negative-Evidence Audit works from what is not — but disciplined so that absence is never assumed to speak. Its defining requirement is a two-part gate: before a silence may count as evidence, the audit must (1) document the expectation that the missing element should have appeared — a defined enumeration, a parallel structure, a genre convention, a repeated practice — and (2) rule out the innocent explanations for its absence. An omission that clears both is an unmet expectation, which can be load-bearing; an omission that clears neither is just one of the infinitely many things a text does not say. That gate is the whole mechanism.

Example¶

A contract review turns to an indemnity clause whose carve-outs read: "except for fraud, willful misconduct, and violation of law." The reviewer's question is not what the list says but what it omits: gross negligence. The audit builds the expectation first. The counterparty's own template and the market-standard formulation of this clause routinely include gross negligence in exactly this enumeration — so its absence here is against both convention and the party's own prior practice. Then it tests the innocents: is gross negligence swept up by "willful misconduct" (no — they are distinct standards), or is this a drafting slip (unlikely, given a negotiated redline history)? Having passed both gates, the omission becomes a consequential reading: gross negligence is not carved out, and therefore remains indemnifiable — a conclusion a quick paraphrase of the clause would never surface. The scenario is illustrative; no real party is quoted.

How it works¶

The audit runs a fixed sequence that a presence-focused pass cannot: define the expected set, locate the site where the element would have appeared, rate how strong the expectation is, test the innocent explanations (covered elsewhere, genuinely out of scope, mere accident), and only then admit the silence as evidence. The convention-and-authority map does the load-bearing upstream work — it is what tells you the expected set in the first place. Without a stated expectation the audit has nothing to measure against and must stay silent; that refusal to speak from bare absence is the feature, not a limitation.

Tuning parameters¶

Expectation source — enumeration, genre convention, prior practice, or parallel text. The more explicit and binding the source, the stronger the claim absence can bear.
Expectation strictness — how complete the expected set must be before a gap counts. Strict lists catch real omissions but flag benign ones.
Innocent-explanation bar — how thoroughly "covered elsewhere / out of scope / slip" must be excluded before silence is allowed to speak.
Sweep scope — every conceivable absence versus only decision-relevant ones. Unscoped audits drown in things that merely could have been said.
Deliberateness weighting — how much intent to read into a gap versus treating it as accident.

When it helps, and when it misleads¶

Its strength is recovering meaning that presence-only reading structurally cannot see: exceptions, omitted voices, silently unaddressed edge cases, the item conspicuously dropped from a formula. It is most valuable exactly where expectation is sharp — enumerations, parallel clauses, repeated formulae.

Its failure mode is the oldest one in interpretation: the argument from silence is trivial to abuse, because every text omits infinitely much, so an unconstrained audit "finds" significance in any convenient gap.^[1] Run backwards, it is a conclusion in search of a hole: decide the reading, then point at whatever absence flatters it. The discipline is the gate itself — no meaning is drawn from silence until a defensible expectation has been stated and the innocent explanations have actually been excluded.

How it implements the components¶

recurrence_contrast_and_omission_register — it fills the omission facet: registering absences together with the reason each one should have been present.
domain_authority_and_convention_map — supplies the convention, defined enumeration, or genre norm that turns a bare gap into an unmet expectation rather than a non-event.

It covers the omission facet of the register only — recurrence and its variation are Motif Concordance and Recurrence Map, and cross-passage contrast is Parallel-Passage Comparison. It borrows a convention map to set expectations but does not reintroduce full historical and authorial context; that is Contextual Reentry Memo.

Instantiates: Textual Close-Reading Mode — the audit supplies the negative-evidence findings the archetype's warranting can build on.
Consumes: Parallel-Passage Comparison — comparable passages and their conventions are a primary way to establish the expectation an absence violates.
Sibling mechanisms: Motif Concordance and Recurrence Map · Parallel-Passage Comparison · Sequence and Placement Map · Micro-to-Macro Claim Matrix · Diction, Syntax, and Punctuation Trace · Line-by-Line Annotation Protocol · Close Reading Table · Paraphrase-Difference Test · Counterreading Stress Review · Contextual Reentry Memo · Interpretive Saturation Log · Edition or Variant Collation Table

Notes¶

The audit is only as good as the expectation it starts from: with no defined expected set, it has nothing to measure absence against and should decline to speak rather than manufacture a gap. An absence it does confirm is not yet a conclusion — it still has to be integrated into a warranted claim by the Micro-to-Macro Claim Matrix.

References¶

[1] The argument from silence (argumentum e silentio) infers meaning from what a source does not say. It carries weight only when the source would be expected to mention the thing and the silence is otherwise unexplained — precisely the two conditions this audit forces before an absence may bear any load. ↩