03 · What You Need to Know
An Honest Origin Does Not Eliminate a Later Duty to Correct
Separate the original mistake from what you do after discovering it
Suppose you accidentally use the wrong variable in an analysis. At the time, you genuinely believe the analysis is correct. Later, you discover the mistake.
There are now two distinct events.
The first is the original error. Under frameworks such as the U.S. Public Health Service research misconduct definition, honest error is distinguished from research misconduct.
The second event is your response after learning that the research may be wrong. At that point, you possess information you did not have when the mistake occurred. Continuing to communicate the known erroneous result without correction is therefore not simply a continuation of the original innocent mistake.
Making an honest mistake
A genuine error occurs without the conduct satisfying the applicable requirements for research misconduct.
Leaving a known consequential mistake uncorrected
Once the error is recognized, a new question arises about your responsibility to prevent an inaccurate or misleading research record from continuing to be used.
Not every typo creates the same correction obligation
Research integrity does not require treating every inconsequential typographical error as an emergency.
The appropriate response depends on what the mistake changes. An incorrect punctuation mark is different from a wrong sample size. A misspelled author affiliation is different from a mislabeled treatment group. A formatting error in a table is different from a calculation error that reverses the main result.
For published biomedical research, ICMJE distinguishes between errors that do not change the direction or implications of results and errors serious enough to invalidate the paper's results and conclusions. Journals and disciplines may use different editorial terminology, but the underlying principle is useful: the correction should match the significance of the error.
Materiality is more useful than embarrassment
A mistake is not consequential because it is embarrassing. It is consequential because it affects what others can reasonably understand, verify, or conclude from the research.
Ask whether the error changes numerical results, qualitative interpretations, figures, classifications, effect estimates, statistical inferences, limitations, participant information, attribution, or the central conclusion. Also consider whether the error affects compliance with an approval, reporting requirement, or other obligation.
An error can be materially important even if the main conclusion survives. Readers may still need corrected values or methods to reproduce the work accurately.
The duty becomes stronger once other people can rely on the error
An error discovered in a private draft that nobody else has seen is relatively easy to correct. Fix it, document the change where appropriate, and improve the workflow that allowed it to occur.
The situation becomes more consequential when incorrect information has already entered a manuscript under review, thesis, public dataset, conference presentation, funder report, repository, clinical or policy document, or published article.
At that point, other people may make decisions based on information you know is inaccurate. The responsibility is no longer confined to your private research files.
The appropriate correction depends on where the research is in its lifecycle
| Where the error exists |
Typical integrity response |
| Internal analysis or working draft |
Preserve relevant records, correct the error, rerun affected work, document consequential changes, and repair the workflow. |
| Manuscript being prepared |
Inform relevant collaborators and ensure the corrected evidence is used before submission. |
| Manuscript under review |
Inform co-authors and communicate a material error to the journal through the appropriate author rather than allowing review of findings known to be incorrect. |
| Published article |
Contact the journal with an explanation and corrected evidence so the appropriate correction mechanism can be determined. |
| Public dataset, repository, or research output |
Correct or version the affected record transparently according to the repository or platform's procedures. |
| Report to a funder, regulator, ethics body, or other authority |
Check the applicable reporting requirements and correct material information where required. |
The exact procedure varies, so researchers should verify the policy of the responsible institution, journal, funder, repository, or other authority rather than assuming that one mechanism applies everywhere.
Correction should preserve rather than erase the research trail
Correcting research does not mean making the original mistake disappear from history.
Preserve the original data, code, records, or files needed to reconstruct what happened. Create corrected versions with appropriate version control or documentation. If a public record changes, use the correction mechanism provided by the responsible platform or publisher.
This is particularly important when an error is serious enough that questions may later arise about its origin. Transparent correction helps distinguish responsible error management from undocumented alteration.
Authorship does not allow one researcher to decide privately for everyone
When an error affects collaborative work, relevant co-authors or project leaders generally need to know. A researcher should not quietly alter a shared analysis while allowing collaborators to continue presenting the old result.
Disagreements can arise over whether an error is material or what response is necessary. If so, document the corrected evidence and consult the relevant journal, institutional research integrity office, methodological expert, ethics body, or other appropriate authority.
One collaborator's reluctance to acknowledge the problem does not make the evidence disappear.
Correction is about the record, not punishment
Researchers sometimes resist correction because they fear that any public notice will be interpreted as an admission of misconduct.
That conflates two different functions.
Editorial corrections, retractions, and related mechanisms address the reliability and accuracy of the literature. Misconduct proceedings address researcher conduct under a defined standard. A serious honest error can require a substantial correction even when nobody committed misconduct.
This is why a researcher can be cleared of misconduct while the underlying research remains unreliable.
Correction can become more urgent when the stakes are higher
Not all research errors create identical risks.
An error in a little-read descriptive table and an error affecting a clinical recommendation, safety conclusion, public policy claim, or widely reused dataset may warrant different urgency even if both arose honestly.
Consider who may rely on the result and what consequences could follow from delay. Where human participants, patient care, safety, regulatory obligations, or other substantial interests are involved, additional reporting requirements may also apply.
Waiting for someone else to notice is not a correction strategy
Once a material error is confirmed, silence does not restore the original result.
The fact that no reader has complained, no reviewer noticed, or no replication has failed does not remove the underlying problem. Research integrity depends partly on researchers correcting the record because they know it is wrong, not only after someone catches them.
Watch Out
Do not confuse uncertainty about the exact correction mechanism with uncertainty about whether a known material error should be addressed. If you are unsure whether a correction, withdrawal, retraction, updated dataset, or another action is appropriate, disclose the problem to the responsible body and let the applicable process determine the mechanism.
Correcting the output is only half the job
After fixing the research record, ask why the error survived.
Was there no code review? Did multiple spreadsheets circulate without version control? Were samples labeled ambiguously? Was the student working without adequate supervision? Did nobody independently verify a key calculation?
If the same process remains unchanged, the correction addresses yesterday's error while preserving tomorrow's.
A responsible response therefore includes improving the procedure, training, documentation, or quality control that allowed the mistake to persist.
07 · A Quick Checklist
Does Your Honest Mistake Need to Be Corrected?
When deciding what correction is required, check:
Have you preserved the original records and independently verified that the error is real?
Does the error change data, methods, analyses, results, figures, interpretations, conclusions, attribution, or another consequential part of the research?
Could another researcher or decision-maker reasonably be misled if the error remains uncorrected?
Has the affected information already been shared with collaborators, reviewers, readers, funders, repositories, or other audiences?
Have you identified every output derived from the erroneous data or analysis?
What correction or reporting requirements apply at your institution, journal, funder, repository, or regulatory body?
Have relevant collaborators or responsible officials been informed where necessary?
Have you corrected the process that allowed the mistake to survive?
