How Do You Avoid Turning Suspicion About a Paper Into an Unsupported Accusation?

When Does a Legitimate Research Concern Become an Unsupported Accusation?

You notice something troubling in a paper. Two images look unusually similar. Numbers do not reconcile across tables. A result seems difficult to reproduce. Sample sizes change without an obvious explanation. Perhaps someone else has publicly questioned the study as well.

You may have identified a genuine problem. What you usually have not established from that observation alone is why the problem exists.

That distinction matters. Moving from "these values appear inconsistent" to "the authors falsified their data" changes the claim from an observation about the research record into an allegation about conduct. The second claim requires evidence that the first does not provide.

Responsible scrutiny does not mean staying silent about suspicious research. It means describing what you can substantiate, separating observation from inference, and allowing appropriate investigative processes to determine matters that the published paper alone cannot establish.

Describe the Evidence Before You Describe the Cause

A Problem in a Paper and Misconduct Are Not the Same Claim

Start With What You Can Actually Observe

Suppose a paper reports 214 participants in its methods section but a results table appears to contain 227 observations. That discrepancy is something you can identify and verify from the publication.

You can ask why the numbers differ. You can check supplementary materials for exclusions, repeated observations, missing-data procedures, or unit-of-analysis explanations. If the discrepancy remains unexplained, you can describe it as an unresolved inconsistency.

What you cannot infer from the discrepancy alone is that the authors deliberately invented participants or manipulated the data. Honest error, reporting error, an undocumented analytical decision, a misunderstanding on your part, or misconduct are different possible explanations. The paper may give you evidence of the discrepancy without giving you evidence sufficient to choose among those explanations.

Each statement makes a stronger claim than the one before it. Your evidence needs to become correspondingly stronger.

Research Misconduct Has a Specific Meaning in Many Formal Systems

Definitions vary by jurisdiction, institution, and research system, so you should use the definition applicable to the case rather than assuming that every questionable practice is formally classified in the same way.

For example, the U.S. Office of Research Integrity defines research misconduct in the Public Health Service context as fabrication, falsification, or plagiarism in proposing, performing, reviewing, or reporting research. Honest error and differences of opinion are excluded from that definition. A formal finding also requires more than showing that something in the research record is wrong: under the applicable federal framework, the conduct must represent a significant departure from accepted practices, have been committed intentionally, knowingly, or recklessly, and be proven by a preponderance of the evidence.

Those requirements illustrate why "I found an error" and "I established research misconduct" are not interchangeable propositions.

Do Not Infer Intent From the Error Alone

Intent is particularly easy to overstate. A duplicated image might result from deliberate manipulation, but it might also result from an assembly mistake, file-selection error, or another cause. An incorrect statistical result might arise from fabricated data, an erroneous formula, coding mistake, transcription error, or misunderstanding of the method.

This does not mean that every anomaly deserves an innocent explanation. It means that the anomaly itself may not establish which explanation is correct.

ORI's framework makes this distinction explicit by excluding honest error from research misconduct and requiring intentional, knowing, or reckless conduct for a formal misconduct finding in the research it governs.

Reliability Can Be Questioned Before Misconduct Is Established

You do not need to resolve the cause of an anomaly before asking whether it affects the evidence.

If an analysis cannot be reproduced from the reported information, you can say that reproducibility is unresolved. If two reported values conflict, you can say that the discrepancy needs clarification. If the journal has issued an expression of concern, you can state that the publication carries an expression of concern and explain what the notice says.

COPE's current guidance separates these issues in an important way: institutions investigate allegations of misconduct, while journals need to correct or retract findings that are invalid or unreliable regardless of whether the problem resulted from misconduct or honest error.

Use Primary Documentation When Discussing Serious Concerns

If you hear that a researcher is "under investigation," locate the underlying documentation before repeating the claim. Has the institution publicly confirmed an investigation? Has the journal issued an expression of concern? Is there a formal finding? Or are you looking at commentary that repeats another person's allegation?

ICMJE recommends appropriate procedures when misconduct is alleged or concerns arise about the integrity of published work. Editors may contact institutions and funders and may publish an expression of concern while relevant procedures are underway. If an institutional investigation later establishes misconduct, the journal can act on that finding.

That sequence is useful for researchers too. Match your language to the strongest status that authoritative evidence actually supports.

“Alleged” Does Not Make an Unsupported Claim Harmless

Adding words such as alleged, possibly, or suspected can accurately communicate uncertainty, but qualifiers are not a substitute for evidence.

"The authors allegedly fabricated their data" still repeats an allegation of fabrication. If all you actually observed is an unexplained numerical inconsistency, the more precise statement is about the inconsistency itself.

Similarly, writing "this appears fraudulent" does not become methodologically rigorous merely because appears has been inserted. Academic hedging has many uses; laundering an unsupported accusation should not be one of them.

Check Whether You Could Be Misreading the Paper

Before escalating a concern, try to falsify your own interpretation.

Read the methods and supplementary material again. Check whether sample sizes legitimately differ across analyses. Determine whether apparently duplicated images represent controls intentionally reused across panels. Verify whether a table uses participants, observations, clusters, or another unit. Rerun calculations where possible. Check for a published correction.

This does not require giving authors unlimited benefit of the doubt. It is ordinary error checking applied to your own critique.

If the concern survives those checks, you can describe it more confidently because you have ruled out some straightforward explanations.

Ask for Clarification Without Accusing

When appropriate, a neutral request for clarification may resolve an apparent problem quickly.

Instead of writing, "Why did you manipulate these data?" ask how a particular discrepancy arose or how a reported value was calculated. Specify the table, figure, variable, or analysis involved and explain what you cannot reconcile.

The response may reveal information missing from the article, confirm an error requiring correction, or leave the concern unresolved. None of those outcomes requires you to begin by asserting misconduct.

Serious Concerns May Need Formal Reporting

Some issues should not be handled merely through public discussion or direct correspondence, particularly when the evidence suggests potentially serious manipulation, when contacting the authors could interfere with an investigation, or when institutional or journal procedures specify another reporting route.

COPE emphasizes cooperation between research institutions and journals because they have different responsibilities and access to different evidence. Institutions are generally positioned to investigate researchers and underlying records, while journals are responsible for maintaining the integrity of the published record.

If you report a concern, provide the evidence that led to it. Identify exact pages, figures, datasets, passages, or calculations. Separate what you observed from what you think might explain it. Let the responsible process investigate beyond the evidence available to you.

Protect the Difference Between a Good-Faith Concern and a Proven Case

A concern does not become illegitimate simply because an investigation later finds insufficient evidence of misconduct. ORI notes that a good-faith allegation can remain good faith even when a subsequent investigation does not support a misconduct finding, provided it was made with a reasonable basis rather than knowing or reckless disregard of contrary information.

This distinction protects both sides of research integrity. Researchers should be able to raise evidence-based concerns without needing to prove the entire case themselves. At the same time, the people whose work is questioned should not be publicly treated as guilty merely because a concern was raised.

From Suspicious Numbers to a Defensible Research Concern

The discrepancy has not been ignored. In fact, it has been described more precisely. What has been avoided is adding a claim about fabrication or intent that the discrepancy itself cannot establish.

Common Ways Suspicion Turns Into Overclaiming

Make Your Claim No Stronger Than Your Evidence

A useful discipline is to ask what you could demonstrate to another researcher using the material currently available. State that first. Then identify what remains unknown.

This approach is not timid. A precise statement such as "the reported primary outcome cannot be reproduced from the published data and the discrepancy remains unexplained" can be quite serious. Its strength comes from being demonstrable rather than dramatic.

Before You Describe a Paper as Suspect or Misconduct as Possible

Questions About Raising Research Integrity Concerns Responsibly

Be Specific About the Problem and Modest About Its Cause

How to Cite This Guide

This guide is intended to be read, shared, and used in research, teaching, and academic work. If you draw on its ideas, explanations, or other content, please acknowledge the source by citing the guide. Doing so gives appropriate credit and helps your readers locate the original resource.