An early-career researcher preparing data for sharing discovers that a variable was reverse-coded in a paper published last year. Flip the sign and one of the three main conclusions disappears. Her first thought is “does anyone know?” The better question is how to fix it as quickly and transparently as possible.
Journals and bodies such as COPE draw a clear line between honest error and misconduct. Reporting an error early usually leads to a correction, and occasionally to a voluntary retraction that peers respect. Staying silent until someone else finds it is the bad scenario.
Ways of correcting the scientific record
| Notice | When it is used | The original article |
|---|---|---|
| Correction (corrigendum, erratum) | The error does not change the main conclusions: a wrong value in a table, a missing funder, a mislabelled figure | Stays, linked to the correction |
| Expression of concern | The journal has serious doubts but an investigation is incomplete | Stays, with a warning attached |
| Retraction | Findings are unreliable due to major error or misconduct, duplicate publication, or ethical breaches | Remains online, marked as retracted with a notice explaining why |
| Retraction and replacement | Honest error changes the results, but the corrected study still has value | Replaced by the corrected version, with history preserved |
Terminology varies by publisher (some use erratum for publisher errors and corrigendum for author errors), but the logic is the same: the impact on the conclusions decides the form.
Your own error: assess the damage before writing
- Reproduce the original analysis to confirm the error is real.
- Rerun the corrected analysis and compare every result.
- Ask which conclusions change: only the numbers, the direction, or statistical significance?
- Tell your co-authors before you tell the journal. They share accountability and must agree to the notice.
Writing to the journal
Write to the handling editor, or the journal's general editorial address if that editor has moved on. Keep it short and specific:
- Article title, DOI and publication date.
- What the error is, how you found it and why it happened.
- The impact: old values, new values, and which conclusions change.
- The form you propose: correction or retraction.
- Confirmation that all co-authors agree.
- Attachments: original and corrected analyses, and revised tables.
There is no need for long apologies or personal explanations. Editors need information to decide, and clarity is itself evidence of good faith.
A sample correction notice
Correction notices are short, neutral and readable without the original article. For example: “In the original article, items 3 and 7 of the job satisfaction scale were reverse-coded in error. After recoding, the coefficient in Table 3, row 2, changes from 0.21 to −0.04 and is no longer statistically significant. The second conclusion in the Discussion is therefore not supported by the data. All other results are unchanged. The authors apologise for the error.” Three to five sentences, and readers know exactly what still stands.
If the journal proposes a different form of notice from the one you suggested, listen to the reasoning. Editors have seen far more cases than you have and usually follow publisher guidance.
When the error is in someone else's paper
You read a paper and the numbers do not add up, or an image appears to be duplicated. A responsible sequence:
- Check carefully that you have not misread. Many apparent errors come from rounding or a different calculation.
- Write politely to the corresponding author, set out the specific inconsistency and ask whether there is an explanation.
- If there is no reply after a few weeks, or the reply does not resolve it, write to the journal editor with the same details.
- Post-publication platforms such as PubPeer are another channel, but keep your tone factual and do not speculate about motives.
Distinguish “the numbers are inconsistent” from “the authors cheated”. You can observe the first. The second is for a formal process to establish.
Don't cite retracted papers without knowing
Retracted papers keep getting cited for years, usually by people who never noticed. To protect yourself:
- Look for the Crossmark button on the article page; clicking it shows the current status.
- Zotero can flag retracted items in your library using Retraction Watch data. Turn it on and pay attention to it.
- Before submission, check your most important references in the Retraction Watch database.
- If you genuinely need to mention a retracted paper, for example to discuss the case, say so in the sentence and cite the retraction notice as well.
What a retraction does to a career
Not all retractions are equal. A retraction initiated by authors who found an honest error is read very differently from one following a misconduct investigation. Notices usually say who initiated the retraction and why, so anyone reviewing your record will see the difference.
What does lasting damage is a pattern: several problematic papers, slow responses, denial in the face of evidence.
Prevention: habits that make corrections less likely
- Analysis code that runs from raw data to final tables, so coding errors can be traced.
- One co-author independently reproduces the key numbers before submission.
- Cross-check tables, figures and text: is the sample size the same everywhere?
- Keep unedited originals of images such as gels and micrographs.
A task for this week: switch on retraction alerts in Zotero and scan your library. If a retracted paper is sitting in the reference list of your next submission, you will find out before your reviewers do.
Câu hỏi thường gặp
What is the difference between a correction and a retraction?
A correction fixes an error that does not change the main conclusions, and the article remains valid. A retraction is issued when the findings can no longer be relied on, whether through serious error or misconduct.
Will retracting my own paper ruin my career?
It has consequences, but a voluntary retraction for honest error is viewed very differently from one caused by misconduct. Retraction notices usually record that the authors reported the problem themselves.
How can I check whether a paper has been retracted?
Use the Crossmark button on the article page, search the Retraction Watch database, or enable retraction alerts in Zotero.
What should I do if I find an error in someone else's paper?
Check carefully, then write politely to the corresponding author. If that does not resolve it, contact the journal editor with specific details and avoid speculating about motives.
Can I cite a retracted article?
Only when discussing the retraction itself, and you should state its retracted status and cite the notice. Do not use retracted work as evidence for your argument.
