How Corrections, Expressions of Concern, and Retractions Change a Paper

A paper's publication date is not the end of its history. Journals can add a correction, warn readers with an expression of concern, retract findings that are no longer reliable, or in rare circumstances remove access to the article. The current notice and version matter more than a citation copied years ago.

Key points#

• Correction, erratum, and corrigendum often overlap in meaning; journal terminology varies.
• A correction should identify the error, show what changed, and leave the central findings reliable.
• An expression of concern is an interim alert while serious uncertainty remains unresolved.
• Retraction tells readers not to rely on an article's findings, whether the cause was honest error, misconduct, or another fundamental problem.
• Clear links, labels, watermarks, and metadata are necessary so notices follow the article across databases and reference managers.

Correction terms are less standardized than they look#

Some publishers historically used “erratum” for a production error and “corrigendum” for an author error. Others use “correction” for both. The label alone may not reveal who introduced the problem or how important it is.

A useful correction notice tells you the original article, the exact location and nature of the error, the corrected information, whether results or conclusions changed, who requested the correction, and when it was issued. The notice should have its own persistent identifier and link in both directions with the article.

Examples include a mislabeled axis, incorrect author affiliation, transcription error in a table, omitted conflict disclosure, or numerical error that changes a secondary analysis but not the main conclusion. The article may remain fully usable once you apply the correction. Silent alteration is the weaker practice: if a published version changes materially without a dated notice, you cannot tell which version was cited or whether the amendment affects your reading of it.

An expression of concern is a warning during uncertainty#

An expression of concern tells you that serious questions exist but the evidence or investigation is not yet sufficient for a final decision. The process may involve an institution, funder, regulator, publisher, or multiple parties. Journals do not always control its speed.

Possible triggers include unresolved doubts about data integrity, ethical approval, image authenticity, authorship, or the reliability of a central analysis. The notice should explain the concern as specifically as fairness and confidentiality allow.

The status is provisional. An investigation may clear the article, lead to a correction, or result in retraction. A later notice should close the loop and remain linked to the earlier warning. Treating an expression of concern as proof of misconduct overstates what it means. Ignoring it and citing the article as ordinary evidence understates it.

Retraction protects the record from unreliable findings#

COPE guidance frames retraction as a mechanism for correcting literature and alerting readers to publications whose findings are unreliable. Causes can include a major analytical error, fabricated or falsified data, plagiarism, unethical research, unauthorized publication, duplicate publication, or compromised peer review.

Retraction is not a single moral category. An author who discovers an honest coding error and asks for retraction has helped correct the record, and a notice should distinguish what is known about the cause without using retraction itself as shorthand for intent.

The original article usually remains accessible but is prominently labeled and linked to the retraction notice. Keeping the record visible preserves the history of citations and prevents an unlabeled copy from circulating as though nothing happened. The notice should identify who is retracting the article and why, avoid defamatory speculation, and be freely accessible.

Partial retraction is controversial because readers may struggle to determine which claims remain valid, and many publishers instead use a substantial correction when a separable part can be repaired and full retraction when central conclusions cannot be trusted. The decision depends on the extent and consequence of the problem, not simply the number of errors.

Removal is rarer than retraction#

Removing the article file entirely is generally reserved for exceptional situations, such as a serious legal order, a grave privacy breach, or content that poses an immediate and substantial risk if left accessible. A tombstone page should remain where possible, explaining that the content was removed and linking to a notice. Removal should not be confused with withdrawal of an unfinalized manuscript or preprint, and publisher vocabulary differs, but the principle holds either way: the scholarly record should remain traceable even when access to the content must be restricted.

Why metadata determines whether a correction works#

A notice on one journal page is not enough. Articles travel through PubMed, Crossref, library catalogs, indexing services, systematic-review databases, reference managers, repositories, and downloaded PDFs. If the status does not travel with them, you can unknowingly build an argument on outdated or retracted findings.

The National Information Standards Organization published its CREC recommended practice in 2024 to improve communication of retractions, removals, and expressions of concern. It describes responsibilities and metadata for publishers, hosts, aggregators, libraries, and other systems. Recommendations include prominent labeling, persistent links between the notice and affected object, and status information that machines can exchange.

Crossref likewise supports relation metadata connecting versions and updates. A visible journal banner helps a person reading online; structured metadata helps a search system, reference manager, or evidence pipeline detect the same change.

A reader's status check#

Before relying on an important paper:

1. Open the article through its DOI or journal page, not only a downloaded PDF.
2. Look for correction, update, concern, or retraction banners.
3. Follow every linked notice and note its date.
4. Check whether the HTML and PDF show the same status.
5. Search the title and DOI in PubMed and Crossref.
6. Determine which result changed and whether it affects the claim you plan to use.
7. Check whether a systematic review included the original or corrected data.
8. Cite the corrected version and notice when the history is relevant.

A correction does not automatically disqualify a paper. It can be evidence that the journal's repair process is working. The important question is whether the remaining claim is supported after the change.

For an expression of concern, avoid using the paper as the sole support for a consequential claim. Seek corroborating evidence and state the unresolved status. For a retracted article, do not cite its findings as valid evidence. It may still be cited to discuss research integrity, historical influence, or the retraction itself, with the status explicit.

Responsibilities after a notice appears#

Authors should alert the journal promptly, preserve relevant records, cooperate with review, and avoid minimizing a central problem as a typo. Editors should use proportionate notices, distinguish evidence from allegation, and update the article and metadata. Institutions should investigate matters within their authority and communicate conclusions needed to correct the literature.

Systematic reviewers and guideline groups should monitor included studies for status changes, because a corrected effect estimate may require reanalysis and a retracted pivotal study may change certainty or recommendations. Version monitoring is part of evidence maintenance, not merely a publishing-office task.