We are an affiliate: if you order from Growth Guys using this code we earn a commission, at no extra cost to you. We are not Growth Guys and have no ownership of or role in the company.
A reviewer reads a manuscript and advises an editor on whether the work is sound enough to publish: whether the methods support the conclusions, whether the analysis is appropriate, whether the literature is fairly represented.
A reviewer reads a manuscript and advises an editor on whether the work is sound enough to publish: whether the methods support the conclusions, whether the analysis is appropriate, whether the literature is fairly represented. That is a real filter and it removes a great deal of poor work. It is also a narrower filter than the phrase 'peer-reviewed' implies in marketing copy.
In the ordinary case a reviewer reads a description of what was done and a summary of what was found. They do not re-run the analysis, inspect the raw measurements or visit the laboratory. Peer review therefore cannot detect fabricated data, and is not designed to. It checks whether the reported work hangs together.
A single peer-reviewed result is a single result. The property that makes a finding dependable is that other groups, with their own samples and equipment, get the same thing. Peer review happens before publication; replication happens after, if it happens at all. Treating the first as a substitute for the second is the most common misuse of the term.
'Published in a peer-reviewed journal' covers both decades-old journals with demanding review and venues that will publish nearly anything for a fee. The phrase alone does not distinguish them, and a citation that names only the fact of peer review rather than the journal is relying on that.
A number of journals now publish the reviewer reports alongside the accepted paper. Where that is available it is worth reading, because the reviewers' objections and the authors' answers show exactly which parts of the result were contested and how they were defended. It is the closest a non-specialist can get to seeing the argument rather than its outcome.
A preprint is a manuscript posted before review. The major servers label them prominently, and the label is accurate. Preprints are genuinely useful -- they are fast, and in some fields they are where the work appears first -- but a preprint cited as though it carried review is a specific misrepresentation, and one that is easy to check because the server says so on the page.
Papers are withdrawn after publication for error, misconduct or irreproducibility, and the retraction does not always travel with the citation. A paper can continue being cited for years after it has been retracted. Checking whether a source is still standing is a separate step from checking that it exists, and it takes about the same time.
None of this makes peer review worthless; a peer-reviewed paper is a far better starting point than an unreviewed claim. It means the work has passed a trained reader's judgement, which is a real bar. It just is not a guarantee of truth, and nothing in the process claims to be.
The claims on this page about the specific thing 'peer-reviewed' certifies can be checked against sources outside this site: