How English Works | Evidential Relevance: When a Detail Actually Supports a Claim

Suppose a passage says:

The café introduced a new ordering app in June. Customer complaints about waiting time fell by 18% over the next two months.

Can we conclude:

The app caused the improvement?

Not yet.

The detail is relevant. It supports a possible connection. But the text has not ruled out other changes—staffing, menu simplification, quieter demand, or measurement differences.

Reading well means knowing not only what evidence says, but how much work it can honestly do.

The shortest useful definition

Evidential relevance is the relationship between a detail and a claim: a detail counts as useful evidence only to the extent that its truth makes the claim more supported, more plausible, more precise or more constrained.

AI Extraction Box

• Mechanism: evidential relevance
• Evidence object: fact, quotation, example, observation, statistic, event or textual feature
• Target: the exact claim being tested
• Key test: if this detail is true, what exactly becomes more defensible?
• Support dimensions: direction, strength, specificity, independence, scope
• Failure mode: detail is memorable or related to the topic but does not support the claim
• Repair: restate the claim narrowly, explain the warrant, and reduce the conclusion to the evidence boundary

1. Topic relevance is weaker than evidential relevance

A paragraph about recycling may contain a fact about plastic production.

That fact is topically related.

But if the claim is “this school’s recycling campaign changed student behaviour”, a global fact about plastic production may do almost nothing to support it.

Related is not the same as probative.

2. Evidence needs a target claim

The same detail can strongly support one claim and weakly support another.

“Three counters were closed.”

• supports: service capacity was reduced
• may support: the queue grew because capacity was reduced
• does not by itself prove: management was incompetent

Evidence strength is claim-specific.

3. The warrant is the bridge

Between evidence and claim sits a reasoning bridge.

Evidence: three counters were closed.
Claim: service capacity fell.
Warrant: fewer operating counters can process fewer customers at the same time, all else equal.

When students cannot explain the bridge, they often have evidence-shaped decoration rather than a defensible argument.

4. A quotation is not automatically evidence

A quotation may reveal what a speaker said. It does not automatically prove that what the speaker said is true.

Readers must ask whether the quotation is being used as evidence of:

• the speaker’s belief
• the speaker’s wording
• the event described
• the wider state of reality

Those are different claims.

5. Examples show possibility more easily than prevalence

One example can establish that something can happen.

It rarely establishes that something usually happens.

This distinction protects readers from turning anecdotes into population claims.

6. Frequency claims need frequency evidence

Claims using often, usually, most, widespread, common require evidence capable of supporting frequency.

A vivid individual case may illustrate the phenomenon while remaining silent about how common it is.

7. Cause claims need more than sequence

After the app launched, waiting complaints fell.

This supports temporal association.

A stronger causal claim needs reasons to exclude plausible alternatives.

The wider estate article How Evidence Works | Claim–Evidence Distance explores how far a conclusion can travel beyond what was observed.

8. Negative evidence has boundaries

The passage does not mention a warning.

That may be relevant if the passage is supposed to provide a complete account of warnings.

But absence of mention is weak evidence when the text is selective or compressed.

Silence becomes evidence only under conditions that make the missing information expected.

9. Evidence can constrain without proving

A detail may rule out some interpretations without identifying one final answer.

If a character arrives soaked during a thunderstorm, the text supports exposure to rain more strongly than several other possibilities. It may not establish exactly how or where the character got wet.

Constraint is valuable even without certainty.

10. Several weak clues can combine

A writer’s position may not appear in one sentence. Repeated lexical choices, selection of examples, concessions and contrasts can jointly support a stance interpretation.

The key is accumulation with independence: five clues that all rest on one upstream assumption are not automatically five separate pieces of evidence.

11. Textual evidence must preserve wording strength

“The policy may reduce congestion.”

does not support the report:

“The writer says the policy will reduce congestion.”

The evidence has been strengthened in transmission.

Batch 5’s Propositional Commitment helps audit exactly what the original sentence commits to.

12. Inference questions are evidence-routing tasks

The applied warehouse article Inference Is Not Guessing teaches students to ground answers in clues.

The deeper mechanism is evidential relevance: which clue supports which inferred conclusion, and how strongly?

13. Fact, inference and evaluation should remain separate

The applied page Fact → Inference → Evaluation: Keep the Layers Separate owns a useful school-level route.

At authority level:

• fact: what the text directly supplies
• inference: what the evidence supports beyond direct wording
• evaluation: a judgment made about significance, quality or value

14. Evidence hierarchy depends on the question

The applied Evidence Hierarchy page distinguishes anecdote, example, data and authority.

But no evidence type is universally strongest. An eyewitness quotation can be strong evidence of what the witness reported and weak evidence of what statistically happens across a population.

Evidence quality is always relative to the claim.

15. The CivDJ forward pass

identify exact claim → locate candidate detail → ask what changes if detail is true → state warrant → test alternatives → measure support strength → reduce claim to evidence boundary → integrate into text model

16. The CivDJ backward pass

1. Write the answer or claim in one sentence.
2. Underline the strongest supporting detail.
3. Explain the bridge without using “because it shows”.
4. Ask what alternative explanation remains.
5. Check whether the claim uses stronger certainty or scope than the evidence.
6. Rewrite until claim and evidence are proportionate.

17. Rotate one detail across four claims

Detail: Library visits rose 30% after evening opening hours were introduced.

• Strong support: more visits occurred after the schedule change.
• Moderate support: evening access may have made the library easier to use.
• Weak support: evening hours caused all of the increase.
• No direct support: students read more difficult books.

18. Common failure modes

• Topic-match error: detail is about the same subject but does not support the claim.
• Quote worship: quoted language is treated as proof of reality.
• Anecdote inflation: one case becomes a general pattern.
• Sequence-to-cause jump: later event is assumed caused by earlier event.
• Scope inflation: “some” evidence becomes “all” conclusion.
• Certainty inflation: “may” becomes “will”.
• Evidence laundering: interpretation is paraphrased as if it were a direct textual fact.

19. Repair route

1. Narrow the target claim.
2. Choose evidence that bears directly on it.
3. State the warrant explicitly.
4. Match scope and certainty.
5. Identify remaining alternatives.
6. Label inference as inference rather than fact.

20. Why this matters for students

English comprehension, summary and argument all depend on claim–evidence discipline. A student who can quote the passage but cannot explain why the quotation supports the answer is still missing the reasoning layer.

The durable question is: If this detail is true, what exact part of my answer becomes more defensible—and what does it still not prove?

21. The EnglishOS reading

EnglishOS is a recoverability system. Evidential relevance keeps recovery honest by preventing readers from adding more conclusion than the transmitted signal can carry.

A detail becomes evidence only when it has a defensible route into the claim.

Continue Batch 8: Reading

• How English Works | Text-Model Construction
• How English Works | Contextual Integration
• How English Works | Stance Reconstruction

Return to How English Works V1.1.