TL;DR: Questionable research practices (QRPs) are not research misconduct. In the United States, research misconduct is a narrow, regulatorily defined category — fabrication, falsification, or plagiarism (FFP), per 42 CFR Part 93 — that must be committed knowingly, intentionally, or recklessly to trigger a formal investigation. QRPs sit in the much larger grey zone below that bar: undisclosed choices applied to real data (p‑hacking, HARKing, selective outcome reporting, salami-slicing, and similar practices) that distort what gets published without meeting FFP’s intent or falsity requirements. The judgment call this guide exists to support is not “did this happen” but “which side of the line does it fall on, and who handles it.”
| Dimension | Research misconduct (FFP) | Questionable research practice (QRP) |
|---|---|---|
| US regulatory definition | Yes — 42 CFR Part 93.103 (fabrication, falsification, plagiarism) | No federal regulatory definition; institution-defined |
| What happens to the data | Invented, altered, or stolen from someone else | Genuine data, but analyzed, selected, or reported in a biased or undisclosed way |
| State of mind required | Must be committed intentionally, knowingly, or with reckless disregard for the truth | Often habitual, normalized, or a product of incentive structures rather than deliberate deception |
| Who investigates | Formal inquiry/investigation process under the institution’s misconduct policy, reportable to funders (ORI for PHS-funded research) | Handled through mentoring, journal correction/retraction processes, peer review, or institutional research-integrity education — not a federal misconduct case |
| Typical outcome | Findings can include retraction, funding debarment, publication bans, employment action | Correction, methods clarification, editorial note, or no formal action at all if disclosed after the fact |
| Examples | Fabricated datasets, doctored images, plagiarized text/ideas | P-hacking, HARKing, selective outcome reporting, salami-slicing |
Why the line matters
The FFP/QRP boundary is not a matter of severity on a single sliding scale — it is a difference in kind, and institutions, journals, and funders route the two down entirely different processes. A research misconduct finding under the US Public Health Service framework (42 CFR Part 93) requires a formal inquiry and investigation, can be reported to the funding agency and the Office of Research Integrity (ORI) for PHS-supported work, and can carry federal consequences (debarment, funding restrictions, publication agreements). A QRP, by contrast, generally has no equivalent federal reporting pathway in the US: it is handled locally, through mentoring, methods training, editorial correction, or (in more serious or repeated cases) institutional research-integrity procedures that fall short of a misconduct finding.
That does not mean QRPs are harmless or that “not misconduct” is a defense. The empirical replication literature treats QRPs as a leading contributor to the reproducibility problems documented across fields — the Open Science Collaboration’s 2015 Science reproducibility project found that a majority of a large sample of psychology studies did not replicate on the original effect size, a result widely attributed in the methodological literature to undisclosed analytic flexibility rather than to outright fabrication. QRPs are the mechanism by which technically-true, technically-not-fabricated research can still mislead the published record at scale — which is exactly why they warrant a page of their own rather than being folded into the misconduct process.
The operational test: what actually makes something a QRP
The line researchers and integrity officers actually apply is not “was a shortcut taken” but two more specific questions:
- Was a choice made after seeing the data or results, then presented as if it had been planned in advance? Subgroup analysis, covariate selection, exclusion criteria, and stopping rules are all legitimate research decisions — when disclosed as exploratory or post hoc. They become questionable when the write-up implies they were pre-specified.
- Was analytic flexibility exploited specifically to manufacture a publishable or statistically significant result, rather than to better understand the data? This is the mechanism researcher degrees of freedom describes: the large number of defensible analytic choices available in any real dataset, any one of which can be selected after the fact to produce a “significant” result.
If the answer to either is yes and it was not disclosed, the practice is a QRP regardless of whether the underlying data are genuine. If a choice was made in advance and reported as such, or made after the fact but clearly labeled as exploratory, it is ordinary methodological judgment, not a QRP — even if the same technique (a subgroup analysis, a covariate adjustment) is involved in both cases. Disclosure, not the technique itself, is what moves a practice across the line.
Where a QRP can tip into reportable misconduct
Institutions and journals generally treat QRPs as escalating toward a formal misconduct referral when one or more of the following is also present, since these move the behavior from “biased judgment” toward the FFP definition’s fabrication/falsification territory:
- The underlying data were altered or invented, not just selectively analyzed or reported — at that point it is falsification or fabrication, not a QRP, regardless of the researcher’s stated intent.
- The practice was repeated and consistent, across multiple papers or grant applications, in a way that suggests a pattern rather than an isolated lapse in judgment.
- There is direct evidence of intent to deceive — for example, records showing a result was reported as confirmatory with knowledge that it was actually derived post hoc, or instructions to a collaborator to conceal an unfavorable result.
- Text or ideas were taken from someone else’s work without attribution — that is plagiarism under 42 CFR Part 93.103, not a QRP, even if it happens alongside other grey-zone practices.
Short of those triggers, most institutional research-integrity offices treat a single instance of undisclosed selective reporting, a p-hacked result, or HARKing as a matter for correction, retraction request, mentoring, or research-integrity training — not a misconduct inquiry. The exact threshold is set by institutional policy (many US universities model theirs on the ORI framework even for non-PHS-funded work) and, outside the US, by national frameworks such as the European Code of Conduct for Research Integrity, which explicitly separates “research misconduct” from “other research integrity violations” (its term for the QRP-equivalent category) and assigns them to different response mechanisms.
The most common QRPs, and how each is normally disclosed instead
| Practice | What it looks like undisclosed (QRP) | What legitimate disclosure looks like |
|---|---|---|
| P-hacking | Trying multiple analyses, covariates, or exclusion rules until one crosses p < 0.05, reporting only that one | Pre-registering the analysis plan, or reporting all tested specifications with a multiple-comparisons correction |
| HARKing | Presenting a hypothesis formed after seeing the data as though it were the study’s original, a priori hypothesis | Explicitly labeling the hypothesis as exploratory or post hoc, generated from the data rather than tested by it |
| Selective outcome reporting | Measuring several outcomes, reporting only the ones that reached significance or supported the desired conclusion | Prospective registration of all planned outcome measures before data collection, so omissions are detectable |
| Data dredging | Mining a dataset for any statistically significant pattern with no prior hypothesis, then presenting it as confirmatory | Framing the result explicitly as hypothesis-generating and calling for confirmatory replication |
| Salami-slicing | Splitting one dataset or study into the minimum publishable units to inflate a publication count, without disclosure | A single comprehensive publication, or clearly cross-referenced companion papers that disclose the shared dataset |
Frequently asked questions
Is a QRP the same as scientific misconduct?
No. In US federal research-integrity policy, “research misconduct” is a defined term limited to fabrication, falsification, and plagiarism (42 CFR Part 93.103), committed with intent, knowledge, or reckless disregard. QRPs involve genuine data handled in a biased or undisclosed way — they are addressed through institutional, editorial, or mentoring channels, not the federal misconduct process, unless the specific triggers described above are also present.
Can a QRP still get a paper retracted?
Yes. Retraction is a publication-integrity remedy, not a legal finding of misconduct, and journals can and do retract or correct papers over undisclosed p-hacking, HARKing, or selective reporting once identified — via COPE-aligned editorial processes — without that necessarily triggering, or requiring, an institutional misconduct investigation.
Does pre-registration eliminate QRPs?
It substantially reduces the opportunity for several of them — p-hacking, HARKing, and selective outcome reporting specifically become detectable, not merely less likely, once an analysis plan and outcome list are registered before data collection. It does not address salami-slicing, authorship-related QRPs, or misconduct that occurs outside the analysis stage, and a registered study can still deviate from its plan; what pre-registration changes is that undisclosed deviation becomes visible rather than invisible.
Who decides whether something is a QRP or reportable misconduct?
In the first instance, this is normally an institutional determination — a designated research integrity officer or equivalent applies the institution’s misconduct policy (in the US, typically modeled on the PHS/ORI framework in 42 CFR Part 93) to the specific facts. Journals apply a parallel but separate process for the published record, generally following COPE guidance, which can result in correction or retraction independent of whatever the author’s home institution concludes.
Related CASRAI resources
- Research Integrity Fundamentals: FFP, RCR, and the Misconduct Process — the misconduct side of this line: how FFP is defined, what RCR training covers, and how a misconduct allegation is actually investigated.
- P-hacking vs. HARKing — a direct comparison of the two most commonly confused QRPs.
- Questionable research practices (dictionary definition)
- Researcher degrees of freedom — the underlying mechanism most QRPs exploit.
- Salami-Slicing and Redundant Publication
- OSF Preregistration: How to Register a Study on OSF Registries
- Research misconduct
- ORI (US Office of Research Integrity)
- European Code of Conduct for Research Integrity
Last verified 2026-08-16 against 42 CFR Part 93 (ecfr.gov), ORI guidance (ori.hhs.gov), the European Code of Conduct for Research Integrity (ALLEA), and Open Science Collaboration, “Estimating the Reproducibility of Psychological Science,” Science 349(6251), 2015 (DOI 10.1126/science.aac4716). Institutional thresholds for escalating a QRP to a formal misconduct referral vary by institution and jurisdiction — the escalation triggers described above reflect common practice under the FFP framework, not a single universal rule; consult your own institution’s research integrity policy for the specific threshold that applies there.
