Look past the publisher’s chosen label
Some publishers reserve erratum for production mistakes and corrigendum for author mistakes; others use the single term “correction”. There is no dependable universal split. Ask whether the notice identifies the old and corrected information and explains the effect on interpretation.
A retraction does not erase the paper. It preserves the bibliographic record while warning readers that findings should not be relied upon. Retraction does not by itself establish intent or misconduct. An honest error can invalidate the main result, while a limited error need not invalidate an otherwise reliable article.
Measure consequences rather than embarrassment
A typographical error with no change of meaning may be corrected silently where policy permits. A wrong unit, reversed legend or missing competing-interest statement normally requires visible action because readers could be misled. The practical effect, not how awkward the error feels, drives the response.
Rerun affected analyses and trace consequences through abstract, tables, figures, discussion, data and conclusion. A local coding error may propagate widely. Record original value, corrected value, cause, validation method and interpretive effect. “The conclusion is unchanged” requires demonstrated analysis.
Where the primary finding is unreliable, an extensive correction can be harder to interpret than a retraction. COPE’s guidance includes clear evidence of unreliable findings, redundant publication and serious ethical breaches among reasons editors may consider. The journal must evaluate the full case.
Send the editor a complete error report
Give article DOI, title, precise location, description and corrected material. State discovery date, affected version, co-author notification and whether data, code or related publications are involved. Supply calculations or source records through an appropriate secure route.
Avoid minimising or assigning blame before evidence exists. “The legend reverses the treatment groups and affects the interpretation of Figure 2” is more useful than “a tiny typo”. Preserve disagreements among authors rather than manufacturing consensus. The editor may contact an institution where specialist investigation is needed.
The redundant-publication audit supports an overlap case. The AI responsibility matrix handles inaccurate authorship claims. The publication ethics hub joins related procedures. The single foundation route is the evidence-review guide.
A useful notice leaves the history readable
A correction identifies the article and DOI, lists material changes and says what happens to results and conclusions. It links to the article, and the article links back. The earlier state should not be overwritten so completely that readers cannot determine what changed.
A retraction notice gives the reason factually and separates unreliable results from unresolved questions about intent. It should be public, clearly labelled and connected to every electronic version. An expression of concern is not a substitute for an indefinite investigation; the journal should follow it with an outcome.
Carry the correction across linked research objects
Check the preprint, repository, dataset, code, conference abstract, press material and later papers. Correcting the journal alone is insufficient if the same wrong value remains prominent elsewhere. Use each venue’s versioning system and link back to the authoritative notice.
Review bibliographic databases for status display. Record correction requests without claiming that every downloaded copy or index has updated. Persistent DOI links and structured metadata improve discovery but cannot recall offline copies.
- Freeze the error and affected version.
- Reproduce analysis and downstream impact.
- Inform co-authors and editor.
- Let reliability guide the correction type.
- Publish a precise, durable, linked notice.
- Update connected research objects.
- Document outcome and remaining indexing limits.
Prompt correction is a component of sound scholarship. It protects the literature more effectively than preserving an incorrect result to avoid reputational discomfort.
Distinguish correction from updated interpretation
New evidence can change how a valid older study is interpreted without making the original article erroneous. A later commentary, replication or update may be the right scholarly response. Do not request a corrigendum merely because knowledge advanced. Conversely, do not call a wrong reported value a “new interpretation” to avoid correction.
Record the proposition as understood at publication and the newly discovered fact. This comparison helps editors distinguish a factual defect in the record from ordinary scientific development. Both deserve transparency, but they use different publication mechanisms.