Questionable research practices

Questionable research practices (QRPs) are behaviours in the design, execution, analysis, or reporting of research that distort findings or misrepresent how a study was actually conducted, without meeting the legal/regulatory bar for research misconduct (fabrication, falsification, or plagiarism — FFP, per the US federal definition at 42 CFR Part 93). QRPs operate in a grey zone: they are widely regarded as damaging to the reliability of the published record and to public trust in research, but — unlike FFP — they typically involve selective or biased choices applied to genuine data, rather than inventing or altering it outright.

What makes something a QRP

A practice is operationally a QRP, rather than ordinary methodological judgment, when it involves an undisclosed choice made after seeing data or results that is then presented to readers as if it had been planned in advance, or when a researcher exploits legitimate analytic flexibility specifically to manufacture a publishable or statistically significant result. The defining feature is non-disclosure: many of the same techniques (subgroup analysis, model comparison, exploratory hypothesis generation) are legitimate research activity when reported transparently as exploratory or post hoc. They become questionable when they are dressed up as confirmatory, pre-planned analysis.

Common forms

• P-hacking — trying multiple analyses, covariates, exclusion rules, or stopping points until a result crosses a statistical significance threshold (commonly p < 0.05), then reporting only the analysis that ‘worked.’
• HARKing (Hypothesizing After the Results are Known) — presenting a hypothesis that was formulated after looking at the data as though it had been the study’s original, a priori hypothesis.
• Selective outcome reporting — measuring several outcomes but reporting only those that reached significance or supported the desired conclusion, while omitting null or unfavourable results from the write-up (a practice that transparent prospective study registration is specifically designed to make detectable).
• Salami-slicing — splitting a single coherent dataset or study into the minimum publishable units needed to generate the largest possible number of separate papers, inflating a publication record without a proportionate increase in genuine new knowledge.
• Other recognised forms — cherry-picking data points, ‘fishing expeditions’ or data dredging without a stated hypothesis, inappropriate co-authorship (honorary or ghost authorship), and failing to report conflicts of interest.

Worked examples

Example 1 (HARKing + selective reporting): A team runs a study testing five secondary outcomes alongside one pre-registered primary outcome. The primary outcome is null. One secondary outcome reaches p = 0.04. The paper is written up presenting that secondary outcome as the study’s central hypothesis, with no mention that four other outcomes were tested and not disclosed — the reader cannot tell the ‘finding’ was one of six tests, not one.

Example 2 (salami-slicing): A single multi-site survey dataset is divided into five near-identical papers, each analysing one site or one subgroup with essentially the same methods and framing, submitted to five different journals without cross-referencing one another, in order to maximise publication count rather than because each subgroup analysis represents a distinct research question.

Counter-example (not a QRP): A researcher pre-registers a primary hypothesis, finds a null result, and then explicitly labels a subsequent analysis of the data as ‘exploratory’ in the published paper, disclosing that it was not pre-specified. Because the post hoc nature of the analysis is disclosed rather than concealed, this is legitimate exploratory research, not a questionable research practice — transparency about what was planned versus discovered after the fact is the operational line between the two.

QRPs vs. research misconduct

The distinction matters because it determines what kind of institutional response applies. Research misconduct, under the US federal definition (42 CFR Part 93) and comparable definitions used by bodies such as the UK Research Integrity Office (UKRIO) and ALLEA’s European Code of Conduct for Research Integrity, is narrowly scoped to fabrication, falsification, and plagiarism (FFP) — inventing data, altering data or results, or appropriating someone else’s ideas, results, or words. Misconduct findings trigger formal institutional inquiry/investigation processes and can carry funding, employment, and publication-record consequences.

QRPs sit below that threshold. They do not typically involve invented or altered data, and most misconduct-definition frameworks (including 42 CFR Part 93) explicitly exclude honest error and legitimate differences of scientific judgment. But QRPs are still recognised across the research-integrity literature and by bodies such as UKRIO as a serious threat to the reliability of the published record precisely because they are common, are rarely detected after the fact, and — unlike outright fabrication — can be normalised as ‘just how the analysis is usually done’ within a field. Addressing QRPs is generally handled through methodological and cultural reform (pre-registration, registered reports, open data/analysis code, reporting guidelines) rather than through formal misconduct proceedings.

Why it matters for research administration

Because QRPs fall outside formal misconduct definitions, they are not typically addressed through a Research Integrity Officer’s misconduct-inquiry process. Institutions and funders instead address them upstream, through requirements such as prospective registration of study protocols, journal adoption of registered-report formats, mandatory data/code sharing, and reporting-guideline compliance (e.g. CONSORT, PRISMA) — all of which are designed to make selective or post hoc analytic choices visible rather than concealable.

References

• UK Research Integrity Office (UKRIO), ‘Questionable Research Practices’ guidance
• US Office of Research Integrity, 42 CFR Part 93 (federal research misconduct definition)
• John, L. K., Loewenstein, G., & Prelec, D., ‘Measuring the Prevalence of Questionable Research Practices’ (2012)
• Simmons, J. P., Nelson, L. D., & Simonsohn, U., ‘False-Positive Psychology’ (2011)