03 · What You Need to Know
How to Respond Without Either Overreacting or Quietly Going Along
Start by making the request explicit
Research integrity problems often hide inside vague instructions.
"Clean up the data."
"Make the figure clearer."
"Try a few more analyses."
"Don't complicate the discussion."
Each could describe perfectly legitimate research work. Each could also conceal something problematic depending on what is actually being requested.
Ask for specificity. Which observations should be removed, and according to what criterion? What exactly should change in the image? Which analyses are being proposed, and how will exploratory analyses be reported? Which limitation should be omitted, and why is it considered irrelevant?
Turning a vague request into a precise methodological decision often makes the integrity question much easier to evaluate.
Ask for the methodological or ethical rationale
Do not begin with "Are you asking me to falsify the data?" unless the evidence genuinely warrants that level of accusation.
Try to understand the reasoning first.
A collaborator may know something you do not. There may be a legitimate exclusion criterion, established disciplinary convention, instrument limitation, ethics requirement, or analytical reason for the request. Alternatively, asking for the rationale may reveal that the only justification is that the current result is inconvenient.
The relevant question is whether the proposed action can be defended independently of the outcome it produces.
Separate methodological disagreement from integrity concerns
Researchers legitimately disagree about methods. Two defensible statistical models may answer slightly different questions. Qualitative researchers may disagree about coding approaches. Collaborators may reasonably debate exclusion rules, theoretical framing, or how prominently a limitation should be discussed.
A disagreement becomes more concerning when the proposed action depends primarily on producing a preferred result, conceals what was actually done, misrepresents evidence, violates an approval or policy, unfairly appropriates credit, or exposes participants or others to inappropriate risk.
Legitimate methodological disagreement
Competing approaches have defensible methodological rationales, and the research can transparently represent what was done and why.
Research integrity concern
The proposed action cannot be adequately justified under relevant methodological, ethical, professional, or policy standards or would cause the research record to misrepresent what actually occurred.
Check the authority that actually governs the decision
Do not rely entirely on what people in the team remember the rule to be.
Depending on the issue, consult the approved protocol, preregistration, data-management plan, institutional policy, ethics or regulatory approval, funder conditions, disciplinary standards, journal instructions, professional code, collaborative agreement, or applicable law.
If the question is methodological rather than regulatory, consult authoritative methodological literature or an appropriately qualified independent expert.
The aim is to move the discussion from "I feel uncomfortable" versus "I've always done it this way" toward a defensible standard.
Changing a research decision is not automatically questionable
Research rarely proceeds exactly as anticipated. Researchers discover measurement problems, encounter unforeseen data structures, revise models, modify protocols through appropriate processes, or recognize that a planned analysis was unsuitable.
Changing course can be responsible.
The integrity problem arises when researchers hide consequential changes, falsely represent post hoc decisions as predetermined, choose methods solely because they produce a preferred result, or make changes that violate applicable requirements.
Transparency can therefore be as important as the methodological choice itself.
Outcome-dependent reasoning is a warning sign
Suppose a collaborator wants to exclude three observations. Ask what criterion justifies the exclusion.
If the criterion was established independently of the observed outcome and applies consistently, the exclusion may be entirely defensible. If the reasoning is simply "those three observations make the result nonsignificant," the request deserves much greater scrutiny.
The same logic applies to repeated analyses, subgroup choices, variable transformations, image selection, qualitative quotation selection, and other research decisions.
A useful test is whether the collaborator would recommend the same procedure if it produced the opposite result.
Do not assume that questionable automatically means formal misconduct
A collaborator can propose something methodologically weak, ethically problematic, professionally inappropriate, or inconsistent with responsible research without necessarily committing formal research misconduct.
The distinction between research misconduct and questionable or detrimental research practices matters because the evidence, procedures, and appropriate response may differ.
You do not need to prove misconduct before declining to participate in a practice you cannot responsibly defend.
Document consequential decisions, especially when disagreement persists
Documentation protects both the research and the researchers involved.
If the team agrees to change an exclusion rule, analysis, protocol, authorship decision, or other consequential element, record what changed and why. Where relevant, preserve the earlier version and note whether the change occurred before or after examining the results.
If you raise an integrity concern that remains unresolved, keep an accurate, professional record of the issue and relevant communications. Do not secretly edit correspondence, manufacture a paper trail after the fact, or create inflammatory notes intended to characterize colleagues rather than document events.
Record facts, decisions, rationales, and dates.
You can say no without immediately accusing anyone of misconduct
Researchers sometimes comply because they believe the only alternative is a dramatic confrontation.
There is considerable space between silent compliance and accusing a colleague of fraud.
You can state that you are not comfortable implementing the proposed change without a documented methodological rationale. You can ask that an independent statistician review the analysis. You can recommend reporting both analyses transparently. You can decline to put your name on a manuscript that represents methods or findings in a way you believe is inaccurate.
The objective is to prevent questionable conduct while leaving room for legitimate disagreement to be resolved through evidence.
Seniority does not turn a weak justification into a strong one
Power differences can make these situations considerably harder.
A doctoral student, junior researcher, research assistant, or untenured academic may face genuine career consequences for challenging a senior collaborator. Research integrity guidance that simply says "refuse" without acknowledging those conditions is not especially useful.
If direct disagreement feels unsafe or has failed, seek confidential advice through an appropriate independent route. Depending on the institution, this might include a research integrity office, ethics office, ombuds service, graduate school, department leadership, compliance office, trusted senior researcher outside the immediate reporting line, or another designated channel.
A senior researcher's assertion that "everyone does it" should be evaluated against actual standards rather than hierarchy.
Escalation should match the seriousness of the concern
Not every disagreement requires a formal misconduct allegation.
A disputed statistical model may be resolved by consulting a methodologist. An authorship disagreement may be handled under the institution's authorship or dispute-resolution process. A proposed deviation from an approved protocol may require ethics or regulatory review.
Credible evidence of fabrication, falsification, plagiarism, serious participant risk, concealment of major noncompliance, or another substantial breach may require formal institutional reporting.
Use the process designed for the type of problem you actually have.
Do not participate simply because someone else will take responsibility
A collaborator may say, "Just do it. If anyone asks, I'll take responsibility."
That offer does not erase your own role.
Research integrity responsibility follows individual conduct as well as team roles. If you knowingly perform an action that misrepresents research, another person's willingness to accept blame does not make your participation disappear.
The same principle explains why another team member's misconduct does not automatically become yours, but your own participation can create separate responsibility.
Watch Out
Do not alter, delete, or destroy research records because a collaborator asks you to "clean things up" after a concern has arisen. Preserve relevant evidence and seek appropriate advice if the requested action could obscure what actually happened.
If the request affects participants or safety, the response may need to be faster
Some questionable requests primarily threaten the accuracy of a manuscript. Others can expose people, animals, communities, confidential information, or researchers themselves to harm.
If the concern involves immediate safety, participant welfare, serious confidentiality breaches, or significant regulatory noncompliance, follow the relevant institutional emergency, ethics, safety, or regulatory procedures rather than treating the issue as an ordinary methodological disagreement.
The correct route depends on the research and jurisdiction.
