When a published paper changes: corrections, concerns and retractions

The record of a published article can change. A correction, an expression of concern, a retraction or another editorial notice may be added after publication. The notice may change a detail, qualify the evidence or indicate that the findings should no longer be relied upon.

For readers, the task is to establish the current status and reconsider the specific claim for which the paper is being used. For authors, it is to report an error clearly and help the journal determine the appropriate action. Neither task is served by treating every notice as proof of misconduct or every correction as inconsequential.

Read the notice before interpreting its label

A correction can address an error while leaving the main findings reliable. An expression of concern alerts readers to a serious unresolved issue; it is not a completed finding of misconduct. A retraction signals that the article should not be relied on in the way an unretracted research report ordinarily would, for the reasons stated in the notice.

The precise notice categories and thresholds vary by journal. PLOS, for example, describes editorial notes as well as corrections, expressions of concern and retractions. Its policy explains the circumstances for each and links notices to affected articles. 2

Read the full notice and any linked updates. Identify what is affected: data, analysis, interpretation, authorship, permissions, editorial process or another issue. Determine whether the notice describes a resolved error, a continuing investigation or a final editorial action.

Do not infer more than the notice establishes. A statement that data could not be verified is not automatically a finding that a named author fabricated them. Equally, the absence of an allegation about intent does not make unreliable findings safe to rely on.

Reliability and misconduct are different questions

COPE’s retraction guidelines explain that unreliable findings can arise from honest error as well as misconduct. The purpose of retraction is to correct the literature, not to punish authors. The guidelines also recognise that a limited problem may be addressed through correction when the main findings remain reliable. 1

This distinction matters when discussing a paper publicly or deciding what to cite. You can explain that a finding is no longer dependable without making an unsupported claim about an author’s intent. You can also recognise that a correction preserves a usable result without dismissing the importance of the error.

Avoid a numerical rule such as “one wrong figure means retraction” or “a small percentage of errors means correction”. The significance depends on what the affected material does in the argument. A single error in a central analysis may matter more than several peripheral typographical mistakes.

When the notice does not resolve your question, describe the uncertainty. Do not fill it with speculation about motives or institutional reputation. Ask what evidence is needed for the claim you are making and whether the current record supplies it.

Check the publisher-maintained record

Start from the article’s DOI or the publisher’s current landing page, not only a saved PDF. Look for linked notices, status banners, version information and updated supplementary files. Open the notice itself rather than relying on a brief search result.

Where Crossmark is available, activate it and read the status information. Crossref explicitly warns that the presence of the button alone does not mean the copy is up to date. 4

Crossref’s update guidance explains that substantial changes are commonly represented by a separate notice with its own DOI and a relationship to the original article. Those links help connect records, but not every publisher uses exactly the same presentation. 3

Record what you checked and when. If you downloaded the paper earlier, compare it with the current version and linked notices before using it for a consequential claim. A familiar PDF in a reference manager may predate a correction.

Do not interpret the absence of a visible status tool as proof that nothing has changed. Inspect the publisher’s page and search the article title or DOI with the relevant notice terms if necessary. If records conflict, retain the links and ask the publisher or library for help.

A reference-checking procedure for readers

Use this sequence for papers central to an argument, especially before final submission of your own manuscript:

1. Identify the exact work. Match title, authors, year and DOI or another stable identifier. Distinguish a preprint from the journal version.
2. Open the current article page. Look for notices, corrections and version history.
3. Follow linked updates. Read the full notice and identify affected claims, data or files.
4. Check the cited proposition. Ask whether the particular statement in your manuscript is still supported.
5. Revise your use. Correct a value, qualify the statement, replace the evidence or remove reliance as appropriate.
6. Record the decision. Keep the date, notice link and reason for the change in your reference notes.

This is not a guarantee that every undiscovered problem will be found. It is a reproducible check of the available record. Focus effort first on studies that carry the most weight in your argument, while maintaining normal reference checking throughout the manuscript.

If you are conducting an evidence synthesis, follow its protocol and applicable methodological guidance when deciding how changed or retracted reports affect eligibility and analysis. Do not make an undocumented exclusion simply because a notice appears late in the process.

Reconsider the claim, not just the reference formatting

A correction can require more than replacing a citation string. Check whether the corrected value affects a comparison, calculation, figure or conclusion in your own work. If you extracted data, revisit the extraction record and any analysis that used it.

For a paper with an expression of concern, identify whether the unresolved issue bears on the proposition you are citing. Consider whether stronger independent evidence is available. If discussing the paper remains necessary, make its status clear and avoid presenting uncertain findings as settled support.

A retracted article may still be relevant as the subject of a discussion about the literature, research history or publication ethics. That is different from citing it as unqualified evidence for its original findings. Identify the retraction and its reason where relevant, and link the notice so readers can assess the context.

Do not assume every paper citing the affected article is therefore invalid. Trace the dependence. Did the later work use its data, rely on its central claim or merely mention it as background? The answer determines what needs reconsideration.

Keep this distinction visible in collaborative projects. A reference manager’s warning flag should prompt a substantive review, not an automatic deletion that leaves an unsupported sentence behind.

If you discover an error in your own paper

Preserve the original files and document what you found. Do not overwrite the evidence with a silently corrected dataset. Record the affected article, location, nature of the error and how it was detected.

Assess the scope with the relevant co-authors. Which tables, figures, statements or supplementary files are affected? Can the corrected analysis be reproduced? Do the findings or conclusions change? If the answer is not yet known, say so rather than reassuring the editor prematurely.

Notify the journal through its published process with enough information to understand the issue. PLOS, for example, asks authors reporting an error for the article citation and DOI, a description, proposed correction and supporting files where relevant. 2

A prompt factual report need not wait for a perfectly polished notice. Explain what is established, what remains under investigation and when further information can be supplied. Consult institutional support if the issue involves research integrity, sensitive data or a disagreement among authors.

Do not decide privately that an error is too embarrassing to report because the overall conclusion seems unchanged. Give the journal the evidence needed to assess whether readers require a correction or other notice.

An error-report template

This is a blank reporting structure. Use the journal’s required form or contact route if it has one, and do not place confidential data in an unsecured message.

Subject: Potential error in [article title and DOI]

We are writing about [full citation]. On [date], we identified [specific error] affecting [section, table, figure or supporting file]. The published material states [brief description]; the checked record indicates [correction or uncertainty].

Our current assessment of the effect is [affected results and conclusions, or what remains to be established]. We have reviewed [scope of checks completed] and still need to examine [remaining work].

We can provide [supporting data, code, source files or corrected figure] through [appropriate route]. Our proposed correction is [text or description], subject to your assessment. The co-authors have been informed as follows: [accurate status].

Please advise on the next steps and any additional material needed to evaluate the issue.

The template separates observation from interpretation. It also prevents an author from implying unanimous agreement or a completed audit when either remains outstanding. Replace every bracket with a checked fact, and remove statements that do not apply.

Keep a copy of the report and subsequent correspondence. If the issue affects a related preprint, dataset or software release, identify those records for the journal and relevant repository administrators so the relationship can be handled coherently.

If you find a problem in someone else’s paper

Describe the concern precisely and support it with locations and evidence. Distinguish an apparent inconsistency from a demonstrated error. If you are unsure whether a method or result has been misunderstood, ask a question that can be answered from the record.

Use the journal’s stated route for concerns. Include the article identifier, affected passage or figure and the basis of the concern. Avoid speculation about personal motives, nationality, seniority or institutional reputation.

Do not conduct unauthorised investigations into private systems or expose confidential material to make the report more dramatic. If the concern comes from information you hold under an obligation of confidence, obtain appropriate advice on reporting it responsibly.

Keep your own account factual if discussing the issue publicly. A journal’s lack of an immediate response does not establish that it endorses the paper’s findings or has ignored the report. The evidence available to you may not reveal the state of an inquiry.

Understand the institution’s and journal’s roles

COPE’s cooperation guidance distinguishes institutions’ responsibilities for researcher conduct from journals’ responsibilities for editorial processes and the integrity of the published record. An institutional investigation and an editorial notice therefore answer related but different questions. 5

A journal may need to alert readers while a wider investigation remains unresolved. An institution may have information about conduct that is not visible in the article itself. Do not assume one process must produce the same wording or timetable as the other.

For authors, cooperate with legitimate requests and retain the records needed to explain the work. If you disagree with a proposed notice, respond with specific evidence and use the journal’s appeal or complaint process. Disagreement does not justify ignoring questions about reliability.

For readers, use the current public notice to qualify reliance while recognising what it does and does not establish. Avoid treating an expression of concern as a lesser certainty of guilt; it is a warning that a relevant issue remains unresolved.

Keep a small status log for important references

Field	Entry to complete
Article and identifier	___
Claim for which it is used	___
Current publisher page checked	URL and date: ___
Notice or update found	Type, date and link: ___
Affected content	___
Consequence for our work	Retain / qualify / recalculate / replace / remove reliance
Action completed by	___
Related files or citations updated	___

Recheck central references when the manuscript is revised substantially or before a consequential reuse of the evidence. Update the manuscript, not just the log. A corrected understanding that remains in a private note does not help the reader encountering an outdated claim in the published text.

The purpose is an accurate chain from source to claim. A visible correction is part of that chain. Reading it carefully allows you to use what remains reliable, identify what is uncertain and stop relying on what the current record no longer supports.