How to Build a Claim-to-Evidence Table Before Submission

A claim-to-evidence table is a working document that lists every major claim in your draft and, for each one, records the source, the exact passage, the type of evidence, your confidence level, and the next action. Build it before final editing so that unsupported or thinly supported claims surface while you still have time to fix them. The table is not a summary of your paper; it is a diagnostic map that tells you which sentences can stand and which need more work.

This guide gives you a complete sequence: how to extract claims, how to locate and record passages, how to classify evidence and confidence, how to run checkpoints, a worked example under a deadline, a reusable table, and stop rules for when to escalate to a supervisor. It ends with a concrete submission decision.

Editing for style before checking evidence wastes effort. If a claim collapses under scrutiny, the paragraph around it usually has to change too. Rigorous research practice, as described in NIH guidance, treats careful assessment of prior knowledge, strengths, weaknesses, and knowledge gaps as core to the work. A claim-to-evidence table operationalizes that assessment at the sentence level.

For review-type work, the same logic applies to reporting. PRISMA 2020 provides a reporting checklist and expanded explanations covering search, selection, and synthesis methods. You do not need to run a systematic review to borrow the discipline: record what you searched, what you selected, and how each claim connects to a source.

Work section by section. Read each paragraph and write one row for every sentence that asserts something a reader could challenge: a finding, a comparison, a causal statement, a prevalence figure, or a methodological justification. Descriptive sentences about your own procedure usually do not need a source row, but claims about what others found do.

A practical rule: if a reviewer could write "says who?" in the margin, it belongs in the table. Aim for the major claims first. A first pass that captures the load-bearing claims in your introduction, methods rationale, results interpretation, and discussion is more useful than an exhaustive pass that you never finish.

Checkpoint A: Count your rows. If you have fewer rows than major claims you can name from memory, reread the discussion and conclusion, where overreach tends to concentrate.

For each claim, record the full citation and the specific passage that supports it, with a page or section locator. "Smith 2021" is not enough; you need the sentence or table that carries the claim. Copying the passage verbatim into the table forces you to check that the source actually says what you think it says.

When you need to locate or confirm sources, a scholarly search interface helps you find candidate papers quickly. WisPaper's academic paper search is built for finding papers through a scholarly search interface, and the Scholar Search entry point explains the same capability for readers new to the tool. Use search to fill gaps the table exposes, not to browse aimlessly.

Cochrane guidance on searching recommends planning searches from eligibility criteria, combining free-text terms with subject headings, using multiple sources, and documenting the process. Even a course paper benefits from writing down your search terms and databases in the table's notes column, so you can rerun or defend the search later.

Checkpoint B: For every row, can you point to a page or section? If not, mark the row "passage missing" and move on rather than guessing.

Label what kind of evidence supports each claim. Useful categories include: primary empirical study, secondary analysis or review, theoretical or conceptual argument, methodological precedent, and background or textbook statement. The label matters because it tells you how much weight the claim can bear. A single small primary study supports a narrower claim than a synthesis of several studies.

If you are working with a reference list you inherited or assembled quickly, screening citations is worth a pass. TrueCite checks BibTeX against real academic databases to flag hallucinated references, which is useful before you trust a bibliography you did not build line by line.

Checkpoint C: Scan the evidence-type column. If a strong causal claim rests only on a conceptual source, flag it for narrowing.

Confidence is your judgment about how well the evidence supports the exact wording of the claim. Use three levels and define them for yourself at the top of the table:

• High: the passage directly states the claim, and the source type is appropriate.
• Medium: the passage supports a narrower version of the claim, or the source is indirect.
• Low: the passage is tangential, the source is weak for this claim, or you have not yet read the full text.

Confidence is not a verdict on the source's quality in the abstract. It is a match score between this claim and this passage. A landmark paper can still be low confidence for a claim it never made.

The action column is what makes the table useful. Every row gets one concrete next step: keep as is, narrow the wording, find a better source, read the full text, move the claim to a limitation, or delete. Vague actions like "check" stall the process. "Replace with the 2019 cohort study already in My Library" is an action you can complete.

When you have many papers to track, organizing them helps. My Library lets you save or upload papers and organize them in the WisPaper library, so the sources behind your table stay in one place. If you need to ask questions across the papers you have collected, Library QA supports asking questions based on papers uploaded to My Library or added from search results.

Checkpoint D: Every low-confidence row must have an action other than "keep." If it does not, you have found the claims most likely to be challenged.

Suppose your course paper is due in five days, the rubric weights "evidence and argument" heavily, and you have limited access to paywalled journals. Your supervisor has returned comments on the introduction but not the discussion.

Day 1 — Extract and triage. Pull the major claims from the introduction and discussion. You find twelve. Four are descriptive, eight need sources. Two of the eight already have passages you can locate; six do not.

Day 2 — Fill gaps. For the six missing passages, run targeted searches using terms drawn from your claim wording. Save promising papers to My Library so you do not lose them. Where full text is unavailable, record the abstract as the passage and mark confidence as low until you can read more.

Day 3 — Classify and score. Assign evidence types and confidence. You discover that your strongest-sounding claim, a causal statement in the discussion, rests on a single conceptual source. Mark it low confidence with the action "narrow to an association and cite the empirical study already in the table."

Day 4 — Act. Rewrite the two lowest-confidence claims, delete one you cannot support, and move one to the limitations section. Rerun your search once with a broader term to check you did not miss an obvious source, consistent with Cochrane's advice to rerun searches before publication.

Day 5 — Verify and decide. Recheck every high-confidence row against its passage. Confirm that the reference list matches the table. Then make the submission decision described below.

Add two columns if they help your workflow: a search-notes column recording terms and databases, and a reviewer-comment column linking each row to supervisor feedback. Keep the table in the same document family as your draft so it travels with the project.

Set rules in advance so you do not spiral. Stop and escalate to your supervisor when:

• A load-bearing claim in your central argument has no source after a genuine search effort.
• Two credible sources directly contradict each other and your argument depends on one of them.
• You cannot access the full text of a source your argument relies on, and the abstract is ambiguous.
• A supervisor comment conflicts with what the evidence in your table shows.

Stop editing and submit when every major claim is either high or medium confidence with a completed action, and remaining low-confidence rows are confined to acknowledged limitations. Do not keep polishing prose to avoid the table; the table is the faster path to a defensible draft.

Return to the table one last time before you submit. Count the rows by confidence. If your central claims are high or medium and every low-confidence row has been narrowed, moved, or deleted, submit. If a central claim is still low confidence, do not submit yet; either fix it or discuss it with your supervisor, because a rubric that weights evidence will find it too.

For longer or more complex projects, a structured workspace can keep the table, sources, and draft aligned. Research Projects provides the WisPaper Research Projects area where you can create a new project, and Scholar Agent supports longer research tasks in areas including literature review. Use these when the claim map outgrows a single document, not as a substitute for reading the passages yourself.