Research and Evidence: Prefer Primary Sources, Show Verification Steps

Every factual claim in a manuscript must trace back to something you can check. The gap between what the source says and what the sentence asserts is the claim-defensibility gap. This hub routes every question about sourcing, verification and evidentiary integrity to the right guide. The contract is plain: prefer primary evidence, show verification steps, and distinguish a source's claim from the writer's conclusion. You need a method for tracing a claim to its origin, measuring the distance from evidence to assertion, and deciding whether a sentence earns its period.

• Section Contract: Prefer primary evidence, show verification steps, distinguish source claim from writer conclusion.
• Core Discipline: Forensic editing: slowing down, tracing claims to sources, identifying structural moves that give prose its persuasive force.
• Default Reader Stance: Null hypothesis default, non-acceptance until evidence moves you. Not disbelief, but a refusal to accept without warrant.
• Most Common Failure: Confusing a source's attributed claim with the writer's interpretive conclusion and passing both off as established fact.
• Who This Is For: Working editors, nonfiction writers, critical reading teachers, researchers translating complex findings, readers exhausted by manipulative discourse.

Start at the origin. A primary source is the thing itself: the study, the transcript, the dataset, the letter, the statute. A secondary source reports on, interprets or summarizes the primary source. The difference is not academic. Every time a claim passes through another layer of summary, the evidence-to-claim latency stretches and distortion compounds. A writer citing a news article that cites a press release that summarizes a study asks you to trust three intermediaries. Each had an editorial reason to select, compress and reframe. Preferring primary evidence means refusing that chain when verification matters.

Traceability Is the Test

For a source to be traceable, you must locate it and confirm it says what the writer claims. A citation to a named publication is not traceable if the publication sits behind an inaccessible paywall, is out of print or is cited for a claim that does not appear in it. Apply a simple ordinal scale before a manuscript leaves the fact-checking workflow: untraceable, traceable to a named source, traceable to primary evidence. Attribution specificity maps onto this directly. "Experts say" is pseudo-attribution, invoking authority without accountability. A named researcher with an accessible study is a verificatory path.

Separate the Source From the Conclusion

Check whether the writer distinguishes what the source claimed from what the writer concluded. This is the single most productive diagnostic in evidence-based editing. A source might report a correlation. The writer might report a cause. The writer's sentence might cite the source correctly and still overreach. Flag the gap. The Chicago Manual of Style offers guidance on citation mechanics, but no style guide enforces the intellectual honesty of keeping those two categories separate. That is a craft discipline, and it belongs to the editor.

Verification is a sequence of moves. Make them visible in the finished text or recoverable from the editorial record. The fact-checking workflow moves through distinct stages: source triangulation, author query and legal review.

Triangulate, Then Ask

Source triangulation means verifying a claim against multiple independent sources, not just the one the writer originally cited. A claim that appears in only one place is not necessarily false. It is unverified. The distinction matters. Author query means going back to the writer and asking: where did this number come from? What page of the report? What does "the data clearly demonstrate" actually mean under the analytical choices that produced it?

Show Your Work

A piece of nonfiction that asserts without showing its evidentiary path asks for trust it has not earned. The charitable reading principle requires engaging a text in its strongest plausible formulation before critique. The null hypothesis discipline requires accepting nothing until the evidence meets its burden. Hold both at once. Read charitably. Verify skeptically. The two are complementary halves of reading with integrity.

Legal Review Is Structural

Legal review is the final gate, and it is not only for libel. A legal reader checks whether a claim about an identifiable person or entity is supported by the record, whether the language of certainty matches the strength of the evidence, and whether the piece adopts a framing that implies an unproven assertion. The legal review is a structural audit. It catches the presupposition load that editorial reading often skims over.

Every nonfiction sentence that cites a source contains two potential claims: the one the source actually made and the one the writer makes by citing it. They are not the same. A study finds a statistically significant association between two variables under a specific model with specific controls. The writer writes, "Research shows that X causes Y." The source claim is the association. The writer conclusion is the cause. The warrant connecting them is unstated and almost certainly indefensible without substantial additional argument. This is the claim-defensibility gap. Purdue Online Writing Lab materials on argument teach students to distinguish evidence from interpretation. Measure the gap as the distance from what the evidence supports to what the sentence asserts. Smaller is better. A claim can be factually true, the study really did find an association, and still indefensible if the cited evidence does not reach the stated conclusion.

Kill "Research Shows"

The "research shows" construction is the most common carrier of this gap. The phrase often introduces a single study, a cherry-picked result or a finding the writer has not read in primary form. The plural "shows" creates an impression of consensus that may not exist. Edit it out. Replace it with what is actually known: "One 2024 study of 1,200 participants found…" The difference is the difference between evidence-based editing and evidence-flavored writing.

Diagnose the Gap Before You Fix It

When you route a verification question, ask what kind of gap you are looking at. Is the source misrepresented? Is the source accurately reported but insufficient to support the conclusion? Is the source absent entirely, with the prose relying on the writer's general knowledge? Each gap demands a different response: correction, qualification or attribution. Routing to the wrong fix wastes time and leaves the problem in the text.

Evidence-based editing is a professional standard. It requires that every factual claim in a manuscript be verified against a traceable source before publication. The editor audits the evidentiary skeleton.

Hunt Hidden Agency

Count the verbs turned into nouns per hundred words and flag where agency disappears. "A decision was made" hides who decided. "The data were analyzed" hides who chose the exclusion criteria. Nominalizations are choices. The editor's job is to ensure the choice was deliberate and the hidden actor can be recovered.

Shorten the Evidence Chain

Track evidence-to-claim latency. Count the sentences from the introduction of evidence to the claim it supports. When the latency is long, intervening material reshapes what the reader thinks the evidence established. A graph introduced on paragraph two and interpreted on paragraph eight has passed through six paragraphs of argument that color its meaning. Shorten the chain or signpost it explicitly: "The same data do not support the stronger claim that…"

Move Caveats Into the Sentence

The end-note problem is endemic. A text buries its caveats in footnotes or a final paragraph while the body speaks in unqualified declaratives. The reader's impression is shaped by the body, not the caveats. The fix is structural: move the qualification into the sentence that needs it. Do not let a paragraph assert what a footnote denies. The text must be true at the level of the sentence.

Task	Guide to Consult	Question the Guide Answers	Primary Diagnostic
Evaluating a source	Source Evaluation	Is this source traceable, and what does it actually say?	Attribution specificity and source traceability
Identifying a primary source	Primary vs Secondary Sources	What is the origin of this claim, and how many intermediaries stand between me and it?	Evidence-to-claim latency
Verifying a claim	Claim Verification Process	Does the cited evidence support the sentence as written, or is there a gap?	Claim-defensibility gap
Distinguishing source from writer	Source Claim vs Writer Conclusion	What did the source claim, and what did the writer conclude from it?	Warrant explicitness
Auditing an argument's structure	Argument Structure Audit	What are the unstated warrants, and do they hold?	Presupposition load and motte-and-bailey detectability
Checking for hidden agency	Nominalization and Agency	Who did what, and has the sentence hidden them?	Nominalization rate
Integrating counterarguments	Counterargument Integration	Are opposing views mentioned, summarized, or structurally incorporated?	Counterargument integration depth
Measuring certainty language	Qualifier Density and Certainty	Does the strength of the claim match the strength of the evidence?	Qualifier density

The table above is the map. Each row routes a specific task to the guide that answers it. Do not read the guides in order. Read the one that matches what you are trying to do. Holding a manuscript and wondering whether a statistic is real? Start with Claim Verification Process. Writing and want to ensure your own prose does not overclaim? Start with Source Claim vs Writer Conclusion. The guides are ordered by task, not by difficulty. Each assumes you arrived with a problem to solve.

Traceability First

Source traceability is the first filter. Before you evaluate a claim, ask whether the source can be found. If it cannot, the evaluation stops there. An untraceable claim is not a claim for editorial purposes. It is an assertion with a decorative citation. Route it to the Source Evaluation guide, which teaches you to distinguish traceable from untraceable, primary from secondary, and attribution from pseudo-attribution.

Measure the Gap

Once traceability is established, measure the claim-defensibility gap. This is the distance from what the evidence supports to what the sentence asserts. The Claim Verification Process guide walks you through the measurement: locate the claim in the source, locate the claim in the manuscript, mark the difference. The difference might be a single word. "Suggests" versus "proves." "Associated with" versus "caused by." One word can open a gap wide enough to fail legal review. The guide teaches you to spot those words and to know which ones carry the most weight in your genre.

Catch Structural Deception

For structural problems, the motte-and-bailey that shifts from a defensible narrow claim to a controversial broad one, the straw-person that misrepresents an opponent's position to make refutation easier, the Argument Structure Audit guide is the destination. These patterns are durable. The content changes. The structure remains. Learning to see the structure rather than the content is the skill that transfers across every domain you will edit.