How peer review actually works

Every published research paper carries an invisible history: at least two specialists somewhere read it critically, wrote down what was wrong with it, and an editor decided whether the authors had answered them adequately. Peer review is that step. It is the mechanism by which a claim moves from being one research group's assertion to being part of the formal scientific record — and yet most researchers first encounter its mechanics only when their own manuscript is in it.

The path a manuscript takes

The process is more structured than the phrase "sent for review" suggests:

• Submission and technical check. The journal verifies scope, formatting, ethics approvals, declarations and plagiarism screening before anything scientific is assessed.
• Desk screening. The handling editor reads the manuscript and decides whether it merits review at all. A large share of submissions — at selective journals the majority — are desk rejected here, usually for being out of scope or too incremental. It is fast, and it is not a judgement on validity.
• Reviewer selection and invitation. The editor finds two or more independent experts without conflicts of interest. This is often the slowest stage, since most invitations are declined.
• Review. Reviewers assess the manuscript and submit a report with a recommendation: accept, minor revision, major revision, or reject.
• Editorial decision. The editor weighs the reports — they are advice, not votes — and decides. Outright acceptance on first submission is rare; major revision is the common good outcome.
• Revision and re-review. Authors respond point by point, and the editor either decides directly or returns the paper to the original reviewers.

What a reviewer is actually asked

Reviewers are not asked whether they like the result or believe the conclusion. They are asked whether the work supports what it claims: is the question clearly posed, is the method appropriate and described in enough detail to be repeated, do the statistics fit the design, are the data adequate, do the conclusions follow from the results rather than overreach them, is prior work cited fairly, and are the ethical requirements met.

Just as important is what review cannot do. Reviewers rarely see raw data, almost never repeat experiments, and cannot reliably detect fabrication by a determined author. Peer review filters for competence and coherence, not for fraud — a distinction routinely lost in public argument about what a "peer-reviewed" claim guarantees.

Who knows whose name

• Single-blind — reviewers know the authors, authors do not know the reviewers. Still the most common model, and the one most criticised for allowing prestige and affiliation bias.
• Double-blind — identities concealed both ways. It reduces bias linked to gender, institution and country, though authors are often identifiable from self-citation and subject matter.
• Open — reviewer identities disclosed, and sometimes the reports published alongside the paper. It improves accountability and gives reviewers credit, at the risk of blunting criticism of senior figures.
• Post-publication review — evaluation continues in public after release, as with preprint commenting and platforms such as PubPeer, which have surfaced integrity problems that formal review missed.

Peer review does not certify that a paper is true. It certifies that people who understand the field could not find a reason it must be wrong — which is a weaker claim, and still the best filter the record has.

Where the system strains

Submission volume has grown much faster than the pool of willing reviewers, and reviewing is unpaid, largely invisible work competing with everything else a researcher owes their time to. The result is slow turnaround, reviewer fatigue and uneven report quality.

Structural problems compound it: reviewers are poor at catching statistical error and fabricated data, there is a documented bias toward positive and novel findings over replication and null results, and predatory journals now imitate the outward form of peer review — an APC, a fast acceptance, no actual scrutiny — which is why checking a journal's indexing, editorial board and process before submitting is now part of a researcher's basic hygiene.

Reforms under trial include registered reports, where the method is reviewed and provisionally accepted before results exist; reviewer recognition through ORCID and services that record review activity; transparent publication of reports; and preprints, which decouple rapid dissemination from formal certification.

Why it matters for students and researchers

Peer review governs what counts as knowledge, what gets funded and how careers advance, and it is one of the few parts of research practice most people learn only by being subjected to it. Reviewing well is a trainable skill and one of the fastest ways to understand what separates a strong paper from a weak one. Current research on the system itself covers reviewer bias, AI-assisted screening and its risks, the rise of paper mills, and models that reward review as a first-class contribution. Following the peer-reviewed literature — and reading how journals describe their own processes — is how students and early-career researchers learn the norms before their first submission depends on them.

Frequently asked questions

What is peer review?

Peer review is the evaluation of a research manuscript by independent experts in the same field before publication. They assess whether the methods are sound, the evidence supports the conclusions and the reporting is complete, and they advise the editor to accept, request revisions or reject.

How long does peer review take?

It varies widely by field and journal, commonly from a few weeks to several months. The slowest stage is usually finding reviewers who accept the invitation, since most invitations are declined and the work is unpaid.

What is the difference between single-blind and double-blind review?

In single-blind review the reviewers know who the authors are but not the reverse. In double-blind review both identities are concealed, which reduces bias related to author reputation, institution, country or gender, though authors can sometimes still be inferred.

Does peer review detect fraud?

Not reliably. Reviewers typically work from the manuscript alone, rarely see raw data and do not repeat experiments, so deliberate fabrication can pass. Peer review checks competence, coherence and completeness; detecting misconduct usually depends on later scrutiny, replication attempts or post-publication review.