What is an evidence map for a research paper?
For drafting a thesis or research paper, an evidence map is a working table that connects each planned claim to a source, an exact supporting passage, and a limit on what you can say. It answers a practical question: if a reviewer challenges this sentence, where would you look to check it?
A reference list identifies publications. A literature-review matrix helps compare studies. A claim-level evidence map adds the connection to your own argument: which paragraph needs the evidence, whether the source actually supports it, and what remains unresolved. You can maintain it in a spreadsheet or a Markdown table before writing polished prose.
This guide is about that drafting workflow, not a complete systematic evidence and gap map. Campbell evidence and gap maps are a distinct evidence-synthesis product with their own methods. Consult the Campbell guidance for producing evidence and gap maps if that is your research method; this checklist does not replace it.
Build the map before drafting the paragraph
1. Turn the outline into checkable claims
Start with a sentence you expect to defend, not a broad heading such as “machine learning.” “This dataset contains a particular number of records” needs a dataset record. “Our model performs better” needs your own evaluation under a defined protocol. Splitting these claims prevents a background citation from being used to stand in for an experiment.
2. Record the source and the passage separately
Give the source a stable ID. Record its title, authors or publishing organization, year, DOI or URL, and version. Then record the section, page, table, or dataset field that supports the claim. A working DOI confirms that a record exists; it does not confirm every number or interpretation attributed to it.
3. Preserve the evidence boundary
Write down what you actually accessed: metadata, abstract, full text, official dataset documentation, or an uploaded result file. If only the abstract was available, do not describe details that require reading the methods or results. Record contradictory sources rather than quietly dropping them.
4. Assign a drafting action
Use a small set of practical statuses: ready to use, needs verification, or unsupported. These are editorial labels, not a validated quality score. A row is ready only when the intended statement fits the evidence and its limitations. Otherwise, find a better source, narrow the wording, remove the claim, or wait for the missing result.
Worked example: a spam-classification paper
The public spam-classification example includes an evidence map. The following abbreviated teaching table shows how to read its source boundaries. It is not a new experiment or a fresh verification of every source in that project.
| Planned claim | Source to inspect | Boundary and action |
|---|---|---|
| Define the dataset used for spam classification. | UCI Spambase record (source S01 in the example). | Use the official record for dataset provenance and field definitions. It cannot establish the performance of your model. |
| Explain precautions against preprocessing leakage. | scikit-learn common pitfalls (S11). | Connect the implementation advice to your actual split and preprocessing code. Documentation is not proof that your code follows it. |
| Claim that one classifier is better than another. | Your evaluation outputs, split definition, metrics, and run records. | Do not infer a winning model from a dataset citation or a method paper. Wait for the comparison or describe it as planned. |
Notice that a single source does not make the entire paragraph verified. The dataset statement, the implementation choice, and the performance conclusion need different kinds of support.
A reusable evidence-map template
Create these columns in your own table: Claim ID; intended claim; source ID; bibliographic record or file path; supporting passage or result location; access level; limitations; verification status; target chapter or paragraph; next action and owner.
For a literature thesis, the source location might be an edition, chapter, and page of a primary text. Separate the quoted passage from your interpretation and from another scholar's interpretation. You do not need to invent an experimental dependency for a textual argument.
For an empirical paper, tie numerical claims to the actual result file and evaluation setup. An experiment plan describes work to perform; it is not evidence that the work has already produced a result. Keep illustrative or synthetic values out of claims about real findings.
Resolve gaps without inventing support
A missing row is a useful warning. Public literature and official statistics call for further retrieval and verification. Private records or unperformed experiments may require the author to supply material or complete the research. These are different dependencies: public information should not automatically become an upload requirement.
If you cannot obtain the needed passage, mark the claim as unverified. Replace an unsupported precise number with a supported narrower statement only when the source permits it. Adding more references at the end of a sentence does not repair a mismatch between the sentence and the sources.
Use the map during writing and review
Planning a degree thesis? Use the thesis-planning worksheet to connect chapter purposes and deadlines to the evidence you still need.
Before accepting a paragraph, check its factual statements against the mapped passages, compare the scope of the claim with the source, and check that the citation points to the correct record. After revisions, repeat the check: changing “may improve” to “improves” changes the evidence burden even when the reference number stays the same.
Paper Terminator uses evidence mapping within its scope-to-review writing workflow. Author review remains necessary, and missing evidence is not a guarantee of automated retrieval or verification. The public example lets you inspect the artifacts before starting a paid project.
When assembling chapters, keep source identity separate from display numbering. See unified references across chapters for that next step.
Check your draft length
Use the free academic word counter to count the full paper or exclude references.
Organize your bibliography
Use the reference converter and deduplicator for GB/T, APA and IEEE formatting.