A colleague forwards you a paper close to your own case series. By the third read something is wrong: the counts in Table 2 do not sum to the stated total, and a figure panel appears in a 2021 paper from another group.
You have no research office to hand it to, and the author is a professor. Here is the order of contact, the evidence to attach, and the language that keeps a report factual.
Decide which of two things you have before you write
An error report says: this published text does not hold together, and here is where. An allegation of misconduct says: a person did something dishonest.
They go to different bodies. Journals assess the literature and can correct, flag or retract an article; institutions assess conduct and cannot amend a publisher's record. Sending one to the other is a common reason a report stalls.
Almost everything visible from outside is an error report: you can see that two panels are identical, not why, and you do not need to know.
Check the problem is real before you write
Three checks, twenty minutes. Confirm which version you are reading, since a preprint and the version of record can differ. Resolve the DOI: check that it reaches the article you think it does, and read the publisher's page rather than an emailed PDF, because a correction notice should be linked to and from the article record there and a two-year-old PDF will not show one. Then re-derive the finding and note both numbers.
The order of contact, and why skipping a rung slows you down
Rungs 1 to 3 run in sequence: each can do something the rung below it cannot, and each is entitled to answer before you go past it. Rung 4 is not the step above rung 3 but a separate route, taken when what you have is an allegation about a person's conduct rather than an observation about the published text. The intervals are your own diary entries — no journal owes you a reply by a fixed date unless it has published one.
| Rung | What you send | What they can do | No useful answer | Diarise |
|---|---|---|---|---|
| 1. Corresponding author | The concern email, private | Explain the discrepancy, supply source data, or request the correction — the fastest route | No reply, or one that does not address the location you named | 21 days |
| 2. Journal editorial office | The same email, plus when you wrote to the author and what came back | Assess it, ask the authors for source data, and issue a correction, an expression of concern or a retraction — which the author cannot do | No acknowledgement, or a referral back to the author | 60 days |
| 3. Publisher | The same email, plus the dates and outcomes of rungs 1 and 2 | Direct the journal, and act where the journal will not — the right rung when the journal itself is your concern | No reply, or a form response that does not name the article | 60 days |
| 4. Institution | An allegation of misconduct only, to the research integrity office of the institution where the work was done | Investigate the conduct of a person it is responsible for and report to the journal | Receipt confirmed and nothing more — investigations are confidential | No rung above |
Starting at rung 2 is legitimate where writing to the author would be inappropriate — where the concern is about consent or ethics approval — but it is usually slower, because the journal's first move is to ask the author anyway.
What to attach, and what to leave out
An editor acts when they can find the problem without repeating your work: citation and DOI, exact location, what you observed, what you are asking for. Nothing about intent.
Dear Editor,
I am writing about "[Full title]", [Journal], [year], DOI [10.xxxx/xxxxx].
I have two observations about the published article.
1. In Table 2, the group counts (12, 18, 17) total 47, against a stated denominator of 52. The same denominator is used for the percentages in the Results, second paragraph.
2. Figure 1, panel C appears to overlap with Figure 3, panel A of [full citation and DOI]. I attach both panels at the resolution available to me.
I am asking whether the authors can clarify these points, and whether any correction to the record is required.
I have no competing interest in relation to these authors or this subject. [Or state it plainly.]
I consent to my name being passed to the authors. [Or: I ask that my name is not passed to the authors.]
Yours sincerely, [name and affiliation]
| What you have found | Attach | Do not assert | Ordinarily a matter for |
|---|---|---|---|
| Arithmetic or denominator inconsistency | The stated figures, your recalculation, and the location | That the data were altered or invented | The journal |
| Duplicated or overlapping images | Both panels, both DOIs, and where the overlap falls | That the duplication was deliberate | The journal, or the institution if referred on |
| A citation that does not resolve, or does not say what is claimed | The reference as printed, what it resolves to, and the sentence it supports | That the reference was fabricated | The journal |
| Ethics or consent statement absent or self-contradictory | The statement as printed, the study design, and any conflicting statement | That approval was never obtained | The journal, which will consult the institution |
| A statistical result that cannot arise from the reported data | The reported values, the test named, and your working | That the analysis was manipulated | The journal |
| Suspected duplicate or redundant publication | Both DOIs, and the overlapping sections, tables or cohorts | That the authors concealed the earlier paper | The journal, which will ask what was disclosed |
Two of those rows — duplicated images, and suspected duplicate publication — are covered in our guides to image integrity and duplicate publication.
Write about the paper, not about the people
Three sentence patterns carry almost any concern.
The article states [X]. From the data reported in [location], I calculate [Y].
[Element] in [location] appears to overlap with [element] in [full citation]. Both are attached.
I was unable to reconcile [statement] in [location] with [statement] in [location].
Sentences to delete before you send
- Anything asserting intent: deliberately, knowingly, in order to.
- Anything asserting motive: pressure to publish, funding interests, career incentives.
- Fabrication, falsification, fraud, and any variant of "made up".
- Characterisations: sloppy, careless, predatory, incompetent.
- Rhetorical questions: "How is this possible?"
- Comparisons with the group's other papers, unless a specific overlap is your concern and attached.
- Anything told to you in confidence that is not in the published record.
Each pattern can be checked by somebody else. A report you can evidence is acted on and an accusation you cannot evidence is not, regardless of which is true; it is also the most defensible thing to have written.
What a COPE-aligned journal is expected to do with your report
Our editorial policies align with COPE Core Practices, and COPE publishes guidance and flowcharts covering a reader's concern about a published paper: acknowledge it, assess whether it is specific enough to act on, put it to the authors, seek the underlying data. Four outcomes follow, and the first is a result rather than a failure.
- No change to the record. The concern is explained, or is not substantiated.
- A correction. The error is real and contained; the conclusions survive.
- An expression of concern. A well-founded concern that cannot yet be resolved, commonly while an institutional process runs.
- A retraction. The findings are no longer supported — see our guide to retractions.
How long this takes, and what silence usually means
Longer than seems reasonable. A journal may have to reach authors who have moved, request years-old files, and wait on a confidential institutional process, so months of silence can be work you cannot see. Keep a record of each email and what came back, and move up one rung when your interval passes. You are also allowed to stop: the work is slow, unpaid, and often ends with no public outcome.
What PubPeer can do, and what it cannot compel
PubPeer is a post-publication commenting platform where readers comment on published articles, under their own name or pseudonymously, subject to its own moderation rules. It is one way a wider audience learns that a paper is disputed. It cannot change the record: only the journal or publisher can issue a correction, an expression of concern or a retraction, and a comment starts no journal process. Going public early buys visibility and costs you control of the discussion.
Anonymity, and what it does and does not buy you
Raising a concern anonymously or pseudonymously is legitimate: the author may hold power over your training or your post. But an editor who cannot come back to you cannot ask the question that would let them act, so an anonymous report rests on what you attached and must be more complete, not less. Nor does it make you untraceable: a specific enough concern can identify its author.
Raising a concern about an article we published
Where to write. [email protected], using the email shape above: the DOI, the exact location, what you observed, what you are asking for.
What we publish. Our handling of corrections, expressions of concern and retractions is set out in Corrections, Retractions & Withdrawals, and our wider position in Publication Ethics.
What we do not publish, and therefore do not promise. No investigation timeline, no acknowledgement window, no case-tracking or reference-number system, no named integrity officer or committee. Do not assume any of them exists.
What any COPE-aligned journal is expected to do. Assess a specific, evidenced concern, put it to the authors, and resolve it through the public record where that record must change.
If your concern is about who is listed as an author rather than the content, see authorship disputes. Otherwise read Corrections, Retractions & Withdrawals and Publication Ethics.