Research integrity statistics show that problems in science extend beyond proven misconduct. Surveys report fabrication, falsification, and questionable research practices; replication projects reveal substantial barriers to checking results; and retraction studies show that corrections do not always travel through the literature. The figures below cover different countries, disciplines, samples, and periods, so they should be read as evidence about specific populations rather than as one global prevalence estimate.
Contents
- How often research misconduct is reported
- Questionable practices and publication pressure
- Reproducibility and replication outcomes
- Data, code, and methodological transparency
- Peer review, authorship, and publication ethics
- Retractions and the scientific record
How often research misconduct is reported
Research misconduct is difficult to measure because surveys capture admissions or observations, while official case counts capture allegations handled by a particular organization. Those measures describe different parts of the problem.
In the Netherlands National Survey on Research Integrity, which covered the previous three years and was published in 2022, 4.3% of 6,813 academic researchers reported fabrication and 4.2% reported falsification. In the same survey, 51.3% reported frequently engaging in at least one questionable research practice. These are self-reported survey results, not findings that every reported behavior was independently established. Prevalence of questionable research practices, research misconduct and their potential explanatory factors (published 2022).
A systematic review and meta-analysis of survey studies published from 1987 to 2008 estimated that 2.9% of researchers admitted committing at least one form of fabrication, falsification, or plagiarism. The same analysis estimated that 15.5% had witnessed someone else commit at least one of those forms of misconduct. For questionable research practices, 12.5% admitted at least one practice, while 39.7% were aware of someone else using at least one. These estimates combine survey studies from different settings and years, so they are not a current worldwide rate. How Many Scientists Fabricate and Falsify Research? (survey studies published 1987–2008).
Official case statistics are narrower still. The U.S. Office of Research Integrity received 713 research-misconduct allegations and opened 117 cases in calendar year 2024. ORI recorded 119 total case closures that year: 39 administrative closures, 31 closures with no finding, 6 findings of misconduct, and 43 declines to pursue. The six findings comprised five falsification-or-fabrication findings and one combined falsification, fabrication, and plagiarism finding. CY 2024 ORI Annual Report.
| Measure | Result | Scope |
|---|---|---|
| Dutch researchers reporting fabrication | 4.3% | Previous three years; survey published 2022 |
| Dutch researchers reporting falsification | 4.2% | Previous three years; survey published 2022 |
| Researchers admitting fabrication, falsification, or plagiarism | 2.9% | Survey studies published 1987–2008 |
| Researchers witnessing one of those forms | 15.5% | Survey studies published 1987–2008 |
| ORI allegations received | 713 | U.S., calendar year 2024 |
Questionable practices and publication pressure
Questionable research practices can alter what is reported without necessarily meeting a formal definition of fabrication or falsification. In a survey of 807 ecologists and evolutionary biologists published in 2018, 64% said they had failed at least once to report results because they were not statistically significant. Another 42% said they had collected more data after checking statistical significance, and 51% said they had presented an unexpected finding as though it had been hypothesized from the start. Questionable research practices in ecology and evolution (survey published 2018).
The Dutch survey found an association between publication pressure and more frequent engagement in at least one questionable research practice, with an odds ratio of 1.22. It also found that perceived likelihood of detection by reviewers was associated with less research misconduct, with an odds ratio of 0.62. Odds ratios describe associations in that survey; they do not by themselves establish that pressure caused a behavior or that review detection produced the lower rate. Prevalence of questionable research practices and research misconduct (previous three years; published 2022).
A Danish survey of 3,402 people, published in 2024, reported that 9 out of 10 respondents had used at least one of nine surveyed questionable research practices in recent publications. The median respondent reported using 3 of the 9 practices. Across the Danish and international comparison samples, self-reported use was roughly 3 times higher than reported frequency of use. Is something rotten in the state of Denmark? (survey conducted in Denmark; published 2024). The Danish findings and the ecology survey use different questions and populations, so their percentages should not be directly combined.
Reproducibility and replication outcomes
Reproducibility statistics distinguish concern about the problem from the outcome of an attempted check. In Nature’s 2016 reproducibility survey, 52% of 1,576 researchers called the reproducibility problem a significant crisis. More than 70% said they had failed to reproduce another scientist’s experiment, and more than 50% said they had failed to reproduce their own experiment. Less than 31% believed that failure to reproduce a published result meant the result was probably wrong, while less than 20% had ever been contacted by another researcher who could not reproduce their work. 1,500 scientists lift the lid on reproducibility (survey published 2016).
The Reproducibility Project: Cancer Biology provides a more specific replication record. It completed 50 experiments from 23 papers and compared 158 effects with the originals; the replication work was conducted from 2013 to 2020, and the findings were published in 2021. For positive effects, the median replication effect size was 85% smaller than the median original effect size, and 92% of replication effect sizes were smaller than their corresponding original effect sizes. These outcomes concern the selected cancer-biology experiments, not all biomedical research. Investigating the replicability of preclinical cancer biology.
The same project illustrates why transparency affects replication. It completed 50 of the 193 experiments originally planned, or 26%. Only 4 of the 193 planned experiments had publicly accessible data sufficient to compute effect sizes and power analyses. Original authors could not provide requested data for 68% of the experiments, and 0 of 193 experiments were described in enough detail in the original papers to design replication protocols without contacting authors. Reproducibility in Cancer Biology: Challenges for assessing replicability (assessment published 2021).
Data, code, and methodological transparency
Access to data and code is one practical condition for checking scientific claims. In a survey conducted in late 2017, only 14.4% of public-health analysts said they had shared code, data, or both. In that survey, 33% said their data would be difficult for colleagues to find if they left their institution, and 43.2% said their code would be difficult to find. Data privacy was a barrier for 49.8% of analysts, while lack of time was a barrier for 41.7%. Use of reproducible research practices in public health (survey conducted late 2017).
Evidence from published biomedical and medical articles shows a similar gap between availability claims and usable materials. A review of 453 biomedical manuscripts published from 2016 to 2021 found that 50.1% failed to share analytical code, and only 10% organized their code in a structured and reproducible manner. Analytical code sharing practices in biomedical research.
Across medical articles published from 2016 to 2021, 8% declared that data were publicly available, but only 2% actually shared data. Declared code sharing was 0.3%, compared with 0.1% actual code sharing. Assessment of transparency indicators across the biomedical literature.
Systematic-review practices show another dimension. Among survey respondents connected with reviews published from January through April 2021, 13% reported sharing analytic code with their most recent review, and only 3% reported sharing all eight categories of review material asked about. Among respondents who shared materials, 89% cited transparency as a reason and 78% cited reproducibility. The survey was published in 2025. Systematic reviewers’ perspectives on sharing review data, analytic code, and other materials.
Peer review, authorship, and publication ethics
Publication ethics includes who receives authorship credit and how manuscripts are evaluated. A survey of corresponding authors found honorary or ghost authorship in 21.0% of articles published in six high-impact medical journals in 2008. Honorary authorship occurred in 17.6% of articles and ghost authorship in 7.9%. By article type, honorary authorship occurred in 25.0% of original research reports, 15.0% of reviews, and 11.2% of editorials. Honorary and ghost authorship in high impact biomedical journals (articles published 2008; survey published 2011).
An exploratory peer-review survey, published in 2009, found that 61.8% of respondents had encountered incompetent review at some point and 50.5% had encountered reviewer bias. A further 22.7% reported reviewers requiring unnecessary citations to their own publications. Smaller but consequential shares reported a breach of peer-review confidentiality (6.8%) or unauthorized use of ideas, data, or methods (5.0%). Perceptions of Ethical Problems with Scientific Journal Peer Review. These are reported experiences in an exploratory survey, not a prevalence estimate for every journal or discipline.
Retractions and the scientific record
Retractions are a correction mechanism, but their scale and downstream handling matter. A Web of Science analysis identified 8,466 retracted articles published from 2003 through 2022. They came from 109 countries and 2,347 journals. In that analysis, data and results issues ranked first among retraction reasons, followed by plagiarism and duplication. Retracted articles in scientific literature.
Retraction notices do not always prevent continued citation. Among 13,252 post-retraction citation contexts in biomedicine, covering citations and contexts from 1960 to 2020, only 5.4% acknowledged that the cited paper had been retracted. Continued use of retracted papers.
The potential effect on evidence synthesis is measurable. A systematic review of high-impact medical systematic reviews published from January 2013 through April 2024 found 61 reviews whose meta-analyses included at least one retracted study. Retraction occurred after systematic-review publication in 74% of the 50 reviews with extracted data. After excluding retracted studies, statistical significance changed in 11% of 166 recalculated meta-analyses, and 42% of 64 recalculated primary-outcome meta-analyses changed their effect estimate by at least 10%. Inclusion of Retracted Studies in Systematic Reviews and Meta-Analyses of Interventions.
An orthopaedic-literature search conducted in 2023 identified 207 retracted articles: 50.2% were clinical-science studies and 49.8% were basic-science studies. Reported reasons included plagiarism in 39 articles, intrinsic errors in 33, duplication in 30, fraud in 25, and peer-review manipulation in 20. Trends in Retraction of Orthopaedic Research Articles. The figures describe that database search and should not be treated as a census of all retractions in orthopaedics.