You check for overstatement by reading the full text's results and conclusions first, then comparing each claim in the abstract against what those sections actually report. The abstract is a summary written separately from the results, so the comparison has to run in that direction: build your understanding from the results and conclusions, then test the abstract against it.
- Read the results and conclusions before the abstract. Note the direction and size of each reported effect, the conditions under which it held, and any stated uncertainty. Expected result: a short list of what the study actually found, in the authors' own terms.
- List the abstract's claims one by one. Separate factual findings from interpretive statements such as implications or recommendations. Expected result: a claim inventory you can match against your list from step 1.
- Match each abstract claim to a specific passage in the results or conclusions. Mark it as supported, broader than the evidence, or absent from the full text. Expected result: a per-claim verdict rather than a general impression.
- Flag scope changes. Watch for an abstract that drops a population, setting, or condition the results specify, or that turns a qualified conclusion into an unqualified one. Expected result: a set of narrowed claims you can quote accurately.
- Check the conclusion section against the abstract's closing sentence. If the authors' own conclusion is more cautious than the abstract, treat the conclusion as the statement to cite. Expected result: the sentence you will actually use in your manuscript.
- Record the discrepancy before citing. Note the page or section where the full text qualifies the claim, so your citation points readers to the accurate statement. Expected result: a citation that survives a reader checking the source.
When the full text is not immediately available, use a literature database to locate an open-access version. Europe PMC supports fielded searches across life-science literature and links to available full text, and Unpaywall records describe open-access locations associated with a DOI, including the selected best open-access location. If neither yields the full text, describe the claim as unresolved and verify it against the published version before relying on the abstract.
