Corrections, retractions, and expressions of concern

There is an error in my published paper. What happens now?

The journal is told, and it decides which mechanism applies. The ICMJE frames the starting point without blame: "Honest errors are a part of science and publishing and require publication of a correction when they are detected."¹

What follows is a notice rather than a silent edit. The recommendation is that the journal publish a correction notice as soon as possible, detailing the changes from the original publication and citing it.¹

The electronic record is handled by versioning rather than replacement. Earlier electronic versions should carry a prominent note that a more recent version exists, and the citation should be to the most recent version.¹

Authors sometimes assume a small error can be fixed quietly because the file is electronic. The opposite convention holds, and it exists so that anyone who read or cited the earlier version can find out what changed. Age is the one thing that shifts the calculation: the ICMJE notes that for a paper ten or more years old, an editor may reasonably decide a correction is not warranted.¹

What goes in a correction notice, and who writes it?

The notice states what changed and points at the original. Its content is a matter for the journal, and the author usually drafts or reviews the wording.

That draft is a small document with a high cost of ambiguity. A notice saying a figure was replaced, without saying which panel and what was wrong with it, leaves a reader unable to work out whether their own use of the paper is affected.

Where a correction changes a number, the notice generally has to give both the old value and the new one. An editor can check that the values in the notice match the values in the corrected article, and that the notice describes every change the corrected version actually contains.

When does an error become a retraction instead?

When the finding cannot stand. COPE's Retraction Guidelines, in version 3 of August 2025, list the circumstances editors should consider, and the first is "clear evidence of major errors, irregularities in the data or images, or any form of misrepresentation."²

The other grounds are not all about error. They include findings compromised by fraud, identity theft, or fictitious authorship; material published elsewhere without proper attribution, permission, or justification; and work involving unethical research practices, compromised peer review, or undisclosed conflicts of interest.²

Honest error can also lead to retraction with republication. The ICMJE describes that route for cases where an honest error significantly alters the results or conclusions, the underlying science remains valid, and the changed paper survives further review and editorial scrutiny.¹

Is a retracted paper deleted?

It stays where it is. The ICMJE states that retracted articles should remain in the public domain and be clearly marked as retracted.³ The record is annotated rather than edited.

COPE's guidance describes the notice as the working part of the mechanism. A retraction notice should link to the retracted article and identify it clearly by title and authors. It should also be published promptly and made freely accessible to all readers, which rules out putting it behind the journal's paywall.²

The reasoning is practical. A paper cited two hundred times cannot be made uncited by removal, and readers arriving from those citations need to find the retraction rather than a missing page. Removal would also make the earlier literature harder to interpret rather than easier.

What is an expression of concern?

An interim notice used while something is being looked into. The ICMJE describes an editor initiating appropriate procedures where misconduct is alleged, and choosing to publish an expression of concern while those procedures run.³

It carries the same publication requirements as other notices. An expression of concern should be prominently labeled and should appear on an electronic or numbered print page. It should also be included in the table of contents, so that it travels with the issue rather than sitting where only a determined reader would find it.³

An expression of concern is not a finding. It records that a question exists and has not yet been resolved, which is a different statement from the one a retraction makes, and the difference matters to everyone named on the paper.

Who investigates a disputed paper?

Usually the authors' institution rather than the journal. The ICMJE describes the editor as seeking to discover the outcome of an institutional investigation and notifying readers of that outcome where appropriate.³ The journal publishes; the institution investigates.

Its definition of scientific misconduct covers four things: data fabrication, data falsification including deceptive manipulation of images, purposeful failure to disclose relationships and activities, and plagiarism.³

None of those four determinations is an editing question, and none of them is one we make. Where a client is preparing correspondence in a matter of this kind, the wording is ours to work on and the substance is not.

Timelines here are slower than authors expect. An institutional review can run for months, and a journal will generally wait for it rather than act on the allegation alone.

What can a language editor usefully do here?

Work on the notice and the correspondence around it. These documents are short, they are permanent, and they are usually written by someone under stress and in a hurry, which is the combination that produces ambiguity.

Consistency between documents is the most common fault. A correction notice, an author's letter to the editor, and the corrected article itself have to describe the same change in the same terms, and they often do not.

The other useful pass is completeness. Where a letter states that four changes were made, an editor can confirm the letter lists four and the corrected file contains four. Every document a client sends us is treated in confidence, and correspondence about a disputed paper is held on the same terms as an unpublished manuscript.

What we will not do

Draft an explanation of what went wrong in your data, characterize a coauthor's conduct, advise on whether a correction or a retraction is the right route, or write to a journal on your behalf. Those four decisions sit with the authors, the institution, and the journal, and an editor who takes them on is no longer editing.

Reporting what COPE and the ICMJE have published, as this page does, is a different act from advising on your case. Both organizations put their guidance online in full, and the notes below link to the pages every statement here was taken from.

The part that is ours is narrow and worth having anyway. A notice that says exactly what changed, in words that match the corrected article, is easier to write with help than alone.