Corrections and retractions
How the published record is put right when something in it is wrong.
The principle
A published article is a permanent record. It is never quietly edited, replaced or removed. Where something in it is wrong, the correction is itself published, is linked to the article from both directions, and carries its own DOI, so that anyone who has the original also finds the correction.
Correcting the record is a sign that a journal is working, not that it has failed. Web of Science states that journals are not penalised for issuing retractions. We would rather publish a retraction than leave a wrong result standing.
The four instruments
Correction
For an error that affects a part of the article but not its conclusions. A typographical error in an equation, a mislabelled figure, a missing funder. The article is amended, the amendment is described, and the date is recorded.
Expression of concern
For a serious concern that has been raised but not yet resolved, where readers should be told while an investigation proceeds. It is withdrawn or replaced by a correction or retraction once the matter is settled.
Retraction
For findings that are unreliable, whether through honest error or misconduct, and for plagiarism, redundant publication, unethical research, or unauthorised use of material. The article remains available, clearly watermarked as retracted, with a retraction notice stating who retracted it and why.
Removal
Used only where leaving the article available would itself cause harm or break the law, for example defamation or a serious breach of privacy. The metadata and a note remain so the citation still resolves.