Peer review is the assessment of scholarly work by people with relevant expertise, generally independent of its authors. In academic publishing, reviewers examine manuscripts and advise editors about their suitability for publication, often recommending changes to methods, analysis, or presentation. The term also covers expert evaluation of research grant applications. It is a form of organized scrutiny rather than a guarantee that findings are correct or that all errors have been detected. Publication decisions and funding decisions remain the responsibility of the relevant editors or institutions. (icmje.org)
Historical development
Scholarly refereeing developed gradually from editorial judgment, learned-society discussion, and consultation with specialists. The Royal Society launched Philosophical Transactions in 1665, but its early publication practices were not equivalent to modern external review. Initially, the editor controlled selection; a committee of Fellows assumed responsibility in the eighteenth century. In 1832, the Society introduced a more formal system of expert refereeing, with written reports informing publication decisions. (royalsociety.org)
The Society also experimented with publishing referee reports in 1831–1832. This illustrates that questions about transparency and reviewer responsibility accompanied the development of refereeing itself. Historical accounts therefore distinguish the emergence of scholarly journals from the later establishment of systematic review procedures. (royalsociety.org)
Editorial workflow
A typical journal process begins with editorial screening for scope, submission requirements, and basic publication standards. A manuscript may be rejected at this stage without external review. If it proceeds, an editor invites specialists whose expertise matches the subject and methods. Reviewers provide reports, sometimes accompanied by confidential comments to the editor, and recommend acceptance, revision, or rejection. (journals.plos.org)
Authors receiving a revision request usually submit a revised manuscript and a response explaining how they addressed each substantive concern. Editors may return the revision to reviewers or assess it themselves. Several rounds can occur. Reviewers advise rather than vote on a binding decision: an editor may disagree with their recommendations, seek additional expertise, or distinguish essential corrections from optional suggestions. Some journals also transfer reports between related publications to avoid duplicating work. (journals.plos.org)
What reviewers assess
Assessment commonly addresses whether the research question is clear, the methods are appropriate, and the conclusions follow from the evidence. For an experimental study, this may involve experimental design, controls, measurement procedures, and sufficient methodological detail. Quantitative work may require scrutiny of statistical analysis and uncertainty. Review can also examine supporting data, reporting quality, and conditions relevant to reproducibility. (journals.plos.org)
Publication criteria differ. Some journals emphasize originality, significance, and likely interest to their readership; others principally assess technical soundness and whether the conclusions are supported. Consequently, rejection does not necessarily mean that a study is invalid. It may reflect the journal’s scope or selection criteria. Specialist statistical or ethical advice may supplement subject-matter review when a manuscript requires expertise beyond that of its initial reviewers. (journals.plos.org)
Anonymity and transparency
Review systems vary in who knows whose identity and which documents become public:
- Single-anonymized review: reviewers know the authors’ identities, while their own identities are ordinarily withheld from authors.
- Double-anonymized review: author and reviewer identities are withheld from one another, although editors generally know both.
- Open peer review: an umbrella term for arrangements involving disclosed identities, public reports, or other forms of openness.
- Transparent review: reports and author responses accompany the published article, without necessarily disclosing reviewer identities. (journals.plos.org)
These features are separable. Publishing reports does not automatically mean publishing names, and a public review file need not contain confidential editorial discussions. Journals therefore specify which parts of the process are disclosed and whether participation is optional. (nature.com)
Responsibilities and integrity
Peer review depends on ethical obligations concerning expertise, timeliness, fairness, and confidentiality. Reviewers should disclose a conflict of interest, such as a relationship or competing interest that could compromise impartial assessment. They must not exploit unpublished ideas, share confidential manuscripts without authorization, or require irrelevant citations simply to increase their own citation counts. Criticism should concern the work rather than the author. (members.publicationethics.org)
Suspected plagiarism, fabrication, or other research misconduct should be reported to the editor rather than investigated independently by the reviewer. Confidentiality also affects the use of generative artificial intelligence: uploading an unpublished manuscript to an external service can disclose protected material. ICMJE guidance requires reviewers to obtain journal permission before using AI to facilitate review and emphasizes that generated assessments may be inaccurate or biased. (members.publicationethics.org)
Limits and alternative arrangements
Review assesses the evidence made available; it is not ordinarily a complete repetition of the research. Its conclusions remain dependent on reviewer expertise, the information supplied, and editorial judgment. Passing review should therefore not be equated with independent confirmation of every result. Reviewer availability and delayed responses can also prolong publication. (nature.com)
Alternative arrangements change the timing or visibility of evaluation. A preprint is typically shared before formal journal review; server screening is not equivalent to peer review. In post-publication peer review, invited assessment follows public release, as at F1000Research. Registered Reports move an initial review of the research question and methodology before outcomes are known, offering in-principle acceptance subject to specified conditions. This format is designed to reduce publication bias and selective reporting rather than make publication depend on striking results. (asapbio.org)