What a Scientific Correction or Retraction Means
Understand corrections, expressions of concern and retractions. Check publisher notices, Crossmark and PubMed before relying on a scientific claim.
Key takeaways
- Read the notice to learn exactly what changed.
- Retraction does not automatically prove fraud.
- Check the current publisher version and linked records.
In this article
A correction changes specified parts of a paper; a retraction warns that the publication should not be relied on as a valid report in the way originally presented. Read the actual notice before deciding what it means for a claim. Retraction does not automatically prove fraud, and a correction does not automatically invalidate the whole study. Honest errors can also require substantial changes.1
An expression of concern signals unresolved questions while a journal or institution investigates. It is different from a definitive retraction.2 Our evidence center explains why source status is only one part of evaluating a result.
What the different notices mean
| Notice | What to investigate | What not to assume |
|---|---|---|
| Correction / erratum | Exactly which text, figures, data or calculations changed | That the change is trivial, or that every conclusion is invalid |
| Expression of concern | The stated uncertainty and whether an investigation remains open | That misconduct has been proved |
| Retraction | Reason, affected findings, scope and any republication | That every author committed fraud, or that all related research is false |
| Corrected republication | Whether a new version survived editorial review and how conclusions changed | That the original and revised estimates are interchangeable |
ICMJE recommends dated correction notices, an updated version and preservation of earlier versions. Serious errors that invalidate results may require retraction; specified circumstances can allow retraction and republication of a corrected paper.1
Check the publisher first
Open the article using its DOI or publisher page, rather than relying on an old downloaded PDF. Look for a notice near the title, an updated version, linked correction or retraction, and revised supplementary files.
Read the notice itself. Record its date and whether it changes the exact result you were using. A corrected author affiliation and a corrected outcome estimate have different implications. The important question is not just “was it corrected?” but “what changed?”
Check Crossmark when it is available
A Crossmark button can show updates registered by a participating publisher, including corrections or retractions. It may appear on the HTML page or PDF. Crossref explicitly says its presence is not a guarantee; it helps readers inspect reported updates.3
Not every publisher uses it, and an absent button is not evidence of an untrustworthy paper. Likewise, no update displayed does not replace checking the publisher’s current article record and notices.
Check PubMed and linked records
Search the PMID or full title in PubMed. Inspect associated notices such as Erratum in, Retraction in, or Expression of concern in, when supplied and indexed. NLM documents these comment/correction relationships and relevant publication types.4
Follow the notice link rather than treating a label as the whole explanation. If the publisher and database appear inconsistent, record both observations and the access date; database updates may lag. A negative search is not a guarantee that every issue has been identified.
Decide what this changes for the claim
If a numerical result was corrected
Use the corrected estimate and explain any material change. Recheck confidence intervals, denominators and whether the direction or interpretation changed. Avoid continuing to share an attractive old number simply because it appears in earlier news coverage.
If the paper was retracted
Do not use it as unqualified positive evidence for a clinical benefit. You may discuss it to explain the publication history, clearly identifying the retraction and linking the notice. Look for reliable independent evidence supporting or contradicting the same claim.
If an investigation remains unresolved
Describe the expression of concern and the specific uncertainty. Avoid claiming guilt or dismissing the issue as irrelevant without evidence. A single contested study should not carry an otherwise unsupported treatment claim.
These decisions concern the evidence, not personal accusations. A study’s publication status should be accurately described even when you disagree with its conclusions.
A short record you can keep
For each important source, note the title, DOI or PMID, current publisher URL, notice type and date, affected result, and date checked. When a claim relies on multiple studies, check whether they are independent reports or analyses of the same participants.
Use our peer-review verification guide to distinguish publication formats. Peer review and an unchanged publication record do not certify that a finding is correct; the methods, uncertainty and broader evidence still matter.
References
- ICMJE: corrections and version control.
- ICMJE: scientific misconduct, expressions of concern and retraction.
- Crossref: Crossmark and publication updates.
- NLM: PubMed User Guide, comment/correction relationships and publication types.
Guidance and links checked 10 October 2026. This article explains verification; it does not allege a retraction or misconduct for any named peptide study.
