Identify exactly what the sentence asserts
A long sentence may contain a definition, an empirical result and an inference. Split these before assigning evidence. A verifiable claim is a statement for which a reader can reasonably ask, “How do you know?” Transitions and clearly signposted decisions do not always need external authority, but your contribution must remain distinguishable from sourced information.
Copy the current wording into the matrix and classify it: background, definition, method, number, comparison, causal statement, recommendation or interpretation. This affects the evidence required. A definition may need an authoritative disciplinary source; an effectiveness statement requires a suitable design; a methodological choice needs a transparent rationale and relevant standard, not merely a webpage on the same topic.
Turn evidence status into a useful diagnosis
Use a short controlled vocabulary. Directly supported means the passage carries the complete statement. Partially supported flags wording that extends population, period, direction or strength beyond the source. Interpretation identifies your own conclusion from cited material. Evidence needed is an open task. Unsupported by this source means the recorded item does not carry the claim.
A green status cannot come from a search snippet, abstract or similarity match alone. Open the original, navigate to the locator and read its context. Establish whether the sentence reports the study’s finding, an assumption, another publication or a limitation. Without that check, a secondary citation can be misrepresented as the paper’s own primary result.
Add a concrete action to every non-final status: “supply locator,” “limit claim to sampled group,” “retrieve primary study,” “check contrary evidence,” or “mark as author inference.” The table then directs revision instead of becoming a colourful inventory.
Audit each claim in a fixed sequence
- Mark the smallest independently verifiable part of the sentence.
- Open the recorded source at the exact locator.
- Compare subject, population, period, direction and strength.
- Determine whether the passage is a result, interpretation or cited third-party claim.
- Record limitations material to your wording.
- Assign a status and an executable next action.
- Revise the sentence, then run the comparison again.
For example, “Theme X appeared among twelve interviewed volunteers” cannot silently become “Students display X.” The shorter version drops both design and selection. Repair it by restoring scope or by finding additional appropriate evidence. Not every limitation must occupy the same sentence, but the paragraph as a whole should not leave a stronger impression than its sources warrant.
Assign functions to multiple sources
One claim may require several references. Give each a purpose: definition, primary result, replication, divergent finding, methodological qualification or context. Three citations placed together do not show that evidence is independent. Reviews may include the same primary studies, and separate papers may analyse one dataset.
When findings conflict, the newest or largest source does not win automatically. Compare questions, designs, measurements and populations. Present a range or unresolved disagreement where a single synthesis would be misleading. The matrix can use several rows for one claim or link to a separate evidence table.
Flag secondary citations explicitly. Consult the original where possible. If it remains inaccessible, follow the required citation convention and do not imply that you checked the primary text. For translations, retain original wording, your translation and a locator so that another reader can review the choice.
Close open evidence states before submission
Filter first for “evidence needed,” “partially supported” and “unsupported.” Do not close an item by merely changing its colour. Narrow the wording, replace the source, signpost interpretation or remove the sentence. Give special attention to numbers, universal quantifiers, comparisons such as “higher,” and causal verbs.
Version the matrix alongside the manuscript. A material sentence change reopens its evidence check. For final quality control, sample claims and trace each from manuscript to original page. Record recurring errors so the next revision can target the underlying problem rather than repeat a general proofread.
The literature matrix organises evidence at study level. The source log captures quotation and paraphrase while reading. Find further tools in the research methods hub, and consult the evidence-review foundation for matching findings to source locations.