What does peer review check before a research paper is published?

A research paper presents a claim together with methods, observations and reasoning. Peer review asks other people with relevant expertise to examine that account. It can identify missing information, weak inferences or methodological problems before publication. It does not turn the paper into unquestionable truth. Reviewers usually examine the reported work rather than independently repeating the entire study, so review and replication answer different questions.

Nature describes an editorial process in which editors first assess submissions and send selected papers to specialist reviewers. Reviewers provide arguments and criticisms; editors make the publication decision. Revision may be requested, and reviewers can disagree because they examine different aspects of the work. The decision is not simply a vote count. This is one publisher's process; journals differ in their criteria, review formats and degree of openness.

Imagine an invented paper claiming that a new coating reduces heat loss. A reviewer might ask whether coated and uncoated samples were measured under comparable conditions. Another might notice that the conclusion says 'all buildings' while the study tested one small sample. A third could ask for uncertainty estimates. These questions address different links between the measurements and the claim. A paper can be clear about its procedure yet still make a conclusion broader than its evidence supports.

The authors then have to respond. They may supply missing details, correct an analysis, add evidence or narrow the claim. A reasoned disagreement can also be explained. Merely adding more confident language would not answer a methodological objection. In the fictional coating example, changing 'all buildings' to a carefully bounded statement about the tested conditions could improve the paper even without changing the measurements. Precision about scope is part of scientific quality.

Review has limits. Reviewers can overlook errors, share assumptions or lack access to information needed to check a claim fully. Publication also depends on a journal's aims, so rejection does not by itself prove a result false, and acceptance does not prove it universally applicable. Later criticism, corrections and new studies remain important. The published paper becomes available for wider examination rather than being removed from scrutiny.

The diagram therefore contains a revision loop and separates reviewer advice from the editor's decision. The useful output is a stronger, more explicit account of what the evidence does and does not support. Peer review works as one checkpoint in a continuing process of inquiry. Its value comes from informed criticism and accountable responses, while the durability of a scientific conclusion also depends on what happens when other evidence and other investigators test it further.