Most guidance on research misconduct — including CASRAI’s own guide to how an investigation actually proceeds — describes the process in the abstract: assessment, inquiry, investigation, finding. This page does the opposite. It looks at a small set of real, already-settled cases — findings published by the U.S. Office of Research Integrity (ORI), by the institutions themselves, or by courts — and asks what each one actually reveals about how fabrication and falsification happen, how they get caught, and what changes afterward.
Every case below is a matter of public record: an institutional investigation report, an ORI case summary, a journal retraction notice, or a court judgment, cross-checked against at least one independent secondary source (news coverage from Science, Nature, Retraction Watch, or equivalent). Nothing here is an allegation — these are closed cases with formal findings, most of them decades-settled and taught routinely in responsible-conduct-of-research (RCR) training precisely because the record is so thoroughly documented.
What these cases have in common — and why they’re rare
Confirmed research misconduct findings against a named individual are uncommon relative to the volume of published research — ORI closes on the order of a few dozen cases a year against a backdrop of millions of papers. That rarity is itself informative: the cases that do reach a public finding tend to share a specific profile — a high-status researcher, unusually clean or unusually productive results, and a whistleblower (frequently a junior colleague, student, or postdoc) willing to raise a concern that was, in every case below, initially costly to raise. None of these cases involves a single “smoking gun” moment; each involves a slow accumulation of statistical, methodological, or reproducibility anomalies that someone else eventually pieced together.
Fabrication: inventing data that was never collected
Fabrication — making up data or results — is the starkest of the three FFP categories because there is no underlying experiment to argue about.
Diederik Stapel (Tilburg University, social psychology)
Stapel, a prominent Dutch social psychologist, was found to have fabricated the data underlying dozens of published studies rather than running the experiments he described. A joint committee chaired by Willem Levelt, convened by the three Dutch universities where Stapel had worked, published its final report in 2012 after a 13-month investigation that reviewed all 137 of his papers and interviewed more than 80 people. The committee’s finding: Stapel had committed data fraud in at least 55 papers and 10 PhD theses he supervised. He was fired by Tilburg University in 2011 and reached a settlement with Dutch prosecutors in 2013 that avoided a prison sentence in exchange for community service and forfeiting certain benefits.
What broke the case open was not a single reviewer catching an error — it was three junior researchers in his own department noticing that his datasets were, across multiple studies, implausibly clean: means and effect sizes were suspiciously exact, and raw data he claimed to have collected himself, alone, in schools and other field sites, could never be produced or reproduced.
Hwang Woo-suk (Seoul National University, stem cell research)
Hwang’s 2004 and 2005 papers in Science claimed the creation of patient-specific human embryonic stem cell lines through cloning — a result hailed at the time as one of the most significant achievements in biomedical research. A Seoul National University investigation committee determined in December 2005 that all 11 claimed stem cell lines were fabricated; Science retracted both papers in January 2006. Hwang was subsequently convicted in Korean courts on charges of embezzlement and violation of South Korea’s Bioethics Law; the fraud charges specifically were not sustained. South Korea’s Supreme Court finalized the conviction in February 2014 with an 18-month suspended prison sentence — Hwang did not serve time.
The fabrication was exposed through independent DNA fingerprinting analysis that contradicted Hwang’s claims about the stem cell lines’ origin, combined with allegations first raised by junior members of his own lab about how the human eggs used in the research had been obtained.
Falsification: altering or selectively reporting real data
Falsification — manipulating existing data, materials, or results so the record no longer accurately represents what happened — is often harder to detect than outright fabrication because real underlying work does exist.
Jan Hendrik Schön (Bell Labs, condensed-matter physics)
Schön published an extraordinary volume of high-profile results in molecular electronics and superconductivity between 2000 and 2002, including papers in both Science and Nature. An investigating committee convened by Bell Labs concluded in September 2002 that Schön had committed scientific misconduct in at least 16 of the 25 papers it examined — reusing identical data graphs (including identical noise patterns) to represent results from what were claimed to be different devices and different experiments. Bell Labs fired him immediately; the University of Konstanz revoked his PhD in 2004; roughly two dozen of his papers across Science, Nature, and Physical Review journals were retracted.
The case is a textbook example of a specific detection method: other physicists trying to build on Schön’s results noticed that graphs published in separate papers, describing separate experiments on separate materials, were statistically identical down to the noise. Real experimental data essentially never repeats that precisely — the anomaly was visible from the papers themselves, without access to his raw lab notebooks.
Anil Potti (Duke University, oncogenomics)
Potti and colleagues published research claiming that gene-expression “signatures” from tumor samples could predict which chemotherapy regimen an individual patient would respond to — a result that moved directly into clinical trials at Duke, assigning real patients to treatments based on the predictor. Biostatisticians Keith Baggerly and Kevin Coombes at MD Anderson Cancer Center, attempting to replicate the published results, found that the underlying data and code contained basic, reproducible errors — samples mislabeled as both “sensitive” and “resistant” to the same drug, and results attributed to the wrong drug entirely. Duke suspended the trials in 2010; ORI’s finding, published in the Federal Register in November 2015, concluded Potti had included falsified data in a grant application, a submitted manuscript, and nine research papers, and barred him from US Public Health Service funding for five years.
This case is frequently cited in reproducibility discussions specifically because the errors were, in retrospect, ordinary spreadsheet-level mistakes — but the results built on top of them were never independently checked before being used to assign real patients to real treatments in a clinical trial.
When misconduct causes direct patient harm
Paolo Macchiarini (Karolinska Institute, regenerative medicine / surgery)
Macchiarini was internationally celebrated for implanting synthetic tracheas seeded with a patient’s own stem cells — an experimental procedure performed on patients, several of whom were not in the immediate life-threatening situations his published accounts described. Most of the patients who received the implants died. Four of his own colleagues raised formal misconduct complaints in 2014; an external reviewer Karolinska itself appointed, Bengt Gerdin, found evidence of misconduct — and Karolinska’s leadership initially rejected that finding and publicly stood by Macchiarini regardless. It took a 2016 Swedish television documentary series re-examining the case, and a change in institutional leadership, before Karolinska reopened the investigation, fired Macchiarini, and — in a final 2018 review — found him and several co-authors guilty of scientific misconduct. Sweden’s courts separately convicted him of causing bodily harm; an appeals court increased his sentence to two and a half years in prison in 2023.
Unlike the other cases here, Macchiarini’s is as much a case study in institutional failure as individual misconduct: the same evidence that eventually produced a guilty finding was available — and was formally raised by whistleblowers — years before the institution acted on it.
Grant fraud and the limits of self-report
Eric Poehlman (University of Vermont, aging and metabolism research)
Poehlman fabricated and falsified data across research on aging, menopause, and obesity, using it to support roughly $3 million in federal grant funding. Per ORI’s published case summary, he admitted to 54 separate findings of research misconduct and agreed to retract or correct ten publications. In 2005 he became the first academic researcher in the United States sentenced to prison specifically for research misconduct — a one-year term — and was permanently barred from applying for US federal research funding.
The case began when a junior member of Poehlman’s own lab noticed that data he presented did not match the raw values she had recorded during data collection — again, a junior researcher’s direct observation, not an external audit, as the initial trigger.
What these cases teach about detection
Read together rather than individually, a few patterns recur across cases that otherwise span physics, psychology, medicine, and biomedical statistics:
- Whistleblowers are almost always junior, and almost always take on real personal risk. In every case above, the first person to notice a problem was a student, postdoc, or junior colleague of the person eventually found responsible — not a senior peer reviewer or an institutional audit.
- Statistical and reproducibility anomalies, not confessions, break most cases open. Implausibly clean data (Stapel), identical results claimed for different experiments (Schön), and irreproducible published datasets (Potti) were each identified by someone independently trying to use or replicate the work — which is also the argument for open data and code as a misconduct-deterrence mechanism, not just a transparency one.
- Institutional first response is not always correct the first time. Macchiarini’s case is the clearest example: a properly commissioned external review found misconduct years before the institution acted on it. This is part of why federal oversight bodies like ORI exist as a check on institutional findings, and why whistleblower protections matter as more than a formality.
- A misconduct finding and a retraction are related but separate events, on separate timelines. Journals retract specific papers; institutions and funders make misconduct findings against a person. See CASRAI’s guide to how a retraction actually happens for the editorial-side process — Schön’s papers, for instance, were retracted by multiple journals independently over several months, on their own procedural timelines, not simultaneously with the Bell Labs finding.
How to read a real ORI case summary
ORI publishes its own case summaries directly at ori.hhs.gov/case_summary, organized by the year each case was closed, and lists only cases with a currently active administrative action (older cases whose sanction period has expired drop off the public list). Each entry names the respondent, the institution, the specific finding (fabrication, falsification, and/or plagiarism), and the administrative action imposed — typically a defined period of funding debarment, supervision requirements, or exclusion from PHS advisory roles. This is the primary source CASRAI’s own ORI dictionary entry and the investigation-process guide both point back to — it is the closest thing to a live, ongoing register of settled US federal misconduct findings, and a more reliable source for current examples than any third-party summary, including this page.
Plagiarism findings follow a somewhat different pattern than the fabrication/falsification cases above: they are more common in raw volume but are more often resolved at the journal or publisher level — correction, retraction, or an editorial note — without necessarily triggering a formal ORI or institutional misconduct finding, particularly where the plagiarized material is not tied to federally funded research. COPE’s own guidance and case database (referenced in CASRAI’s COPE guidelines guide) document this pattern at the editorial level in far greater volume than ORI’s federal caseload reflects.
Frequently asked questions
What are some famous examples of research misconduct?
The cases with the most extensive public record and the clearest formal findings include Diederik Stapel (fabrication, social psychology), Hwang Woo-suk (fabrication, stem cell research), Jan Hendrik Schön (falsification, physics), Anil Potti (falsification, oncogenomics), Paolo Macchiarini (misconduct plus criminal conviction, surgery), and Eric Poehlman (fabrication and falsification tied to federal grant fraud, the first US researcher imprisoned for research misconduct) — all summarized above with citations to the underlying institutional, federal, or court findings.
What is the difference between research misconduct and an honest mistake, in a real case?
The Potti case is a useful illustration of where the line sits: the underlying data errors that biostatisticians first identified were, individually, the kind of thing that can happen through carelessness. What moved the case from an error to a formal misconduct finding was ORI’s determination that the errors were not corrected once identified internally, and that falsified data was knowingly included in a grant application and manuscripts afterward. CASRAI’s investigation-process guide covers the intentional/knowing/reckless standard that formally separates misconduct from honest error in more depth.
How is research misconduct usually discovered, based on real cases?
Overwhelmingly by people close to the work rather than by external audit: junior lab members, co-authors, or independent researchers attempting to build on or replicate the published results. See “What these cases teach about detection” above.
Does a misconduct finding automatically retract every paper the researcher published?
No. A misconduct finding is a determination about a person’s conduct on specific, named papers or grant applications; retraction is a separate decision made independently by each journal for each specific paper. In several of the cases above, retractions extended over months after the underlying misconduct finding, and not every paper a researcher ever published was necessarily retracted — only those the investigation specifically identified as compromised.
Where can I look up real, current ORI case summaries myself?
ori.hhs.gov/case_summary is ORI’s own public register, described above under “How to read a real ORI case summary.”
Do these cases represent how common research misconduct is?
No — they represent the small subset of cases that reached a full public finding, usually because they involved a funding agency, a high-profile publication, or (in Macchiarini’s case) direct patient harm serious enough to draw criminal investigation. Most integrity concerns are resolved with less visible outcomes — corrections, quiet retractions, or institutional findings that are never made public in this level of detail — so this page should not be read as a representative sample of misconduct overall, only as a set of the most thoroughly documented examples available.
