Preprints: shared before review
A preprint is a paper posted publicly before, or instead of, formal peer review, on servers such as arXiv, bioRxiv and medRxiv. Preprints let results travel fast, which matters, and they are screened only lightly for scope and obvious problems, not checked for correctness.
Read a preprint as a claim that has not yet been examined by independent experts. Many go on to be published with changes; some change a great deal; some never appear in a journal at all. medRxiv, for its part, states on its site that its preprints should not be used to guide clinical practice.
Peer-reviewed: examined, not proven
A peer-reviewed article has been read by other experts, usually anonymously, and revised in response before a journal accepted it. Review catches many errors and weak arguments. It does not re-run the experiment, and it can miss problems, so peer review is a filter rather than a guarantee.
It is worth knowing which version you have. The preprint and the published article can differ, and it is the published version that counts when you cite.
Corrections, expressions of concern, retractions
Publication is not the end of a paper’s life. A correction (erratum) fixes a specific error; an expression of concern is a journal’s public warning while questions are investigated; a retraction withdraws the paper because its findings cannot be relied on, through honest error or misconduct.
Retracted papers usually stay online, marked as retracted, so they can still turn up in searches and reference lists. Before relying on an older paper, check its page at the publisher for notices, or look it up in the Retraction Watch database, which Crossref now makes openly available.
A one-minute check
Where is it posted: a preprint server or a journal? Is there a DOI, and does the publisher’s page show any correction, concern or retraction notice? If it is a preprint, has a published version appeared since? Three questions, and you know what kind of evidence you are holding.
