From question to public record

Four-step process

1. Define the question, corpus identity, analytical view, and counting rule.
2. Test the pre-specified measurement and comparison.
3. Challenge the result with controls, alternatives, and prior-art review.
4. Publish methods, results, limitations, revisions, and negative findings.

Study types and evidence

Exploratory work identifies questions and possible patterns; confirmatory work fixes the hypothesis and decision rules before the relevant analysis. These labels do not substitute for evidence.

E0

Observation: a pattern has been identified.

E1

Structured pattern: the relationship is reproducible in the stated corpus.

E2

Statistical support: a documented model supports the relationship.

E3

Robust across controls: the result survives relevant alternative counting rules, corpus views, and methodological controls.

E4

Independently reproducible: an independent reproduction confirms the result.

Scientific status and novelty are independent axes: a result can be robust yet previously known, or potentially novel without being well supported. Prior-art searches are necessarily limited by the sources and search terms available, so a novelty label is never a claim of exhaustive proof.

Negative results are published when they clarify a tested claim or method.

What the result does not prove

Each important result separates measurement from interpretation.