How Peer Review Works: The Steps, the Checks and the Blind Spots
Peer review sends a study to independent experts before a journal publishes it. See each step, what reviewers check and what the process often misses.

Peer review is the check a scientific journal runs before it publishes a study: an editor sends the manuscript to a few independent researchers in the same field, who judge whether the methods are sound and the conclusions follow from the data. It is a filter for obvious errors and weak reasoning. Peer review does not rerun the experiment, does not inspect the raw data in most cases and does not guarantee that a published result is true.
That gap between what readers assume and what the process actually does explains a lot of confusing science headlines. A study can pass peer review, get wide coverage and later fail when other labs try to repeat it.
What happens to a paper after it is submitted?
A manuscript goes through a series of gates, and it can be stopped at any of them. The usual path at a research journal looks like this:
- The authors submit the paper, often with the underlying data, code or extra figures as supplementary files.
- An editor reads it first. Many papers are rejected at this stage, without outside review, because they fall outside the journal’s scope or look too preliminary. This is called a desk rejection.
- If the editor sees promise, they invite outside researchers to review it. Most journals look for two or three reviewers with no close ties to the authors.
- Reviewers send written reports and a recommendation: accept, minor revisions, major revisions or reject.
- The editor weighs the reports and makes the decision. Reviewers advise; the editor decides.
- If revisions are requested, the authors answer each comment point by point and resubmit. A paper can go through several rounds.
The whole cycle often takes months. Reviewers are usually unpaid volunteers who fit the work around their own research and teaching, which is one reason the process is slow.
What do reviewers actually check?
Reviewers judge the logic of the study from what the authors put on the page. Typical questions include:
- Does the design answer the question the authors ask? A study of whether a drug works needs a comparison group, for example.
- Is the sample large enough, and were subjects chosen in a way that avoids obvious bias?
- Are the statistical tests suitable for the data?
- Do the conclusions stay within what the results show, or do they reach further?
- Does the paper cite earlier work fairly, including results that disagree?
- Is the method described clearly enough that another lab could repeat it?
The last point matters most for the long run. Science corrects itself mainly through replication, when independent teams repeat a study and see whether they get the same answer. Peer review can only ask whether replication would be possible.
Who are the reviewers, and do they know who wrote the paper?
Reviewers are working scientists in the same specialty, and journals use several systems for deciding who knows whose name. In single-blind review, the most common model in many sciences, reviewers know the authors but the authors do not know the reviewers. In double-blind review, neither side is told, which aims to reduce bias toward famous names or prestigious institutions. In open review, names are shared, and some journals publish the reviewer reports next to the paper.
Each model has a trade-off. Anonymity lets junior researchers criticize senior ones without fear, but it also lets a careless or hostile reviewer avoid accountability. In small fields, authors and reviewers can often guess each other’s identity anyway.
What does peer review miss?
Peer review works best at catching flawed reasoning and worst at catching problems hidden in the data. The common blind spots are:
- Fraud: Reviewers generally trust that the numbers are real. Fabricated or manipulated data can pass if it looks plausible.
- Errors in raw data and code: Most reviewers do not have the time, or the files, to recheck every calculation.
- Results that do not replicate: A well-designed study can still produce a fluke result, especially with a small sample.
- Selective reporting: If authors ran many analyses and reported only the ones that worked, reviewers may never see the rest.
- Publication bias: Journals have historically preferred positive, striking results, so studies that found nothing are less likely to appear at all.
- Reviewer limits: Two or three people cannot be expert in every method a large, multidisciplinary paper uses.
When a serious problem surfaces after publication, a journal can issue a correction or retract the paper. A retraction means the journal no longer stands behind the findings. In the United States, the Office of Research Integrity handles misconduct cases involving research funded by the Public Health Service.
How are preprints different from peer-reviewed papers?
A preprint is a manuscript posted publicly before peer review, on servers such as arXiv for physics and mathematics or bioRxiv and medRxiv for biology and medicine. Preprints let results circulate in days instead of months, and other scientists often comment on them openly. They have not been through a journal’s formal check, though, and some change substantially or never get published. News reports should say when a finding comes from a preprint.
How can a reader judge a study?
Treat “peer reviewed” as a minimum standard rather than a verdict. A few habits help:
- Check whether the study was in people, animals or cells. Results in mice often do not carry over to humans.
- Look at the sample size and whether there was a comparison group.
- Prefer findings that several independent teams have reached.
- Give more weight to systematic reviews, which pool many studies on one question. A Yale review of 51 supplement studies carries more weight than any single trial in it, and a review of 350 animal studies on protein and aging still describes animals, not people.
- Search the paper’s title on PubMed, run by the National Library of Medicine, to find the journal record and any linked correction or retraction notice.
Common questions
Does peer review mean a study is correct?
No. It means a few independent specialists saw no obvious flaw in the methods and reasoning as presented. Confidence in a result grows when other teams replicate it.
Are reviewers paid?
Usually not. Most journals rely on researchers who review as part of their professional duties, which is one reason reviews can take weeks or months.
Why do retracted papers still get cited?
Some authors cite a paper without checking its status, and copies saved before the retraction carry no warning. Journal websites and databases such as PubMed mark retracted papers, so checking the official record is the safest step.
Site editor: Nikita Sheikin



