Extended definition
Peer review is the central validation mechanism in contemporary scientific publishing: external reviewers with expertise in the topic evaluate a submitted manuscript, recommending acceptance, revision, or rejection to the editor, who makes the final decision. The practice has roots in the 18th century (Royal Society’s Philosophical Transactions) but was formalized and generalized only after World War II. Four modalities coexist today: single-blind (reviewers know author identities; authors do not know reviewers — still dominant in many fields), double-blind (bilateral anonymity — adopted in social sciences and humanities to reduce affiliation and gender bias), open peer review (identities revealed, reports published — adopted by F1000Research, eLife, BMJ Open, optional at Nature Communications), and post-publication peer review (continuous review after preprint or journal publication — model of PubPeer and variants). Lee et al. (2013) synthesized literature on structural biases (gender, affiliation, nationality, language); Tennant et al. (2017) offered a multidisciplinary review of ongoing innovations.
When it applies
Peer review is a near-universal requirement in formal academic publishing (indexed journals, top-tier computer-science conferences) and is what distinguishes validated literature from gray literature (preprints, reports, blogs). It applies in career decisions: peer-reviewed output counts fully in institutional evaluation systems. It applies in systematic reviews: standard inclusion criterion is peer-reviewed literature (with separate treatment for preprints when included). In applied research with critical timing (public health during pandemic), preprint + post-publication review has gained traction as complement — not substitute — of classical review.
When it does not apply
It does not apply in informal communication, science journalism, institutional blog posts, talks. It does not apply fully in smaller conferences where review is light or absent. It does not apply to editorials, invited commentary, and letters — categories that typically pass through editorial triage but not full peer review. In some humanities, the tradition of editor-reviewed monographs (not peer-reviewed anonymously) coexists with peer review — a hybrid system with its own legitimacy. It does not replace independent replication: review validates method and reasoning; replication validates result.
Applications by field
— Health and biomedical sciences: single-blind or double-blind dominant; ICMJE defines principles. Growing open peer review in BMJ journals. — Social sciences and humanities: double-blind as norm to reduce biases; registered reports in experimental psychology. — Computer science: top-tier conferences (NeurIPS, ICML, CVPR, ACL) with double-blind review + rebuttal; short timelines. — Physics and mathematics: arXiv preprints often discussed before formal review; journal review is subsequent validation.
Common pitfalls
The first pitfall is treating peer review as an absolute quality filter — gender, institutional, national, and language biases are documented (Lee et al., 2013); review is incremental improvement, not certification. The second is confusing acceptance with validity: peer-reviewed work can be wrong, and typically is partially wrong (methodological, data, interpretation limits). The third is assuming reports are always rigorous: research shows average review time is highly variable (from 1h to >15h per report), with a long tail in superficial reviews. The fourth is failing to distinguish journal prestige from per-article review rigor: top-tier journals have non-trivial retraction rates. The fifth is confusing peer review with endorsement: reviewers usually disagree partially with the work; acceptance reflects that the work meets the journal’s threshold, not that reviewers agree with all conclusions.