Fabrication vs Falsification in Research Misconduct

Fabrication in research misconduct means inventing data or results that were never produced; falsification means taking data that were genuinely produced and altering, omitting, or manipulating them so the research record no longer reflects what actually happened. The distinction is not academic pedantry — it determines what evidence an institution must gather, which charge it can sustain, and how a finding is ultimately reported to a funder, journal, or regulator.

Research misconduct is formally defined, in the widely-used U.S. federal formulation, as fabrication, falsification, or plagiarism (FFP) “in proposing, performing, or reviewing research, or in reporting research results” — a definition set out under 42 CFR Part 93 and applied by the Office of Research Integrity (ORI). The same three-part structure underpins how UK institutions, journals, and funders describe misconduct, even though the UK has no single statutory FFP definition and instead relies on institutional policies aligned with the UK Concordat to Support Research Integrity.

• What counts as fabrication in research misconduct?
• How does falsification differ from fabrication?
• How do institutions decide which charge to frame?
• What do real case patterns show?
• Answer-first Q&A
• What are the implications for institutional policy?

What counts as fabrication in research misconduct?

Under the ORI’s regulatory definition, fabrication is “making up data or results and recording or reporting them.” The defining feature is that the underlying research event — the experiment, the survey, the observation — did not occur at all, or did not occur in the form reported. There is no raw dataset to falsify because none was ever collected.

Typical fabrication patterns include inventing survey responses from participants who were never recruited, reporting results from an assay that was never run, or generating a “representative” image for an experiment that has no corresponding physical record. Because fabrication leaves no authentic source data behind, investigators typically build a fabrication case around absence: missing lab notebooks, missing instrument logs, missing ethics-approval paperwork for claimed human or animal subjects.

How does falsification differ from fabrication?

Falsification is defined by ORI as “manipulating research materials, equipment, or processes, or changing or omitting data or results such that the research is not accurately represented in the research record.” Falsification presupposes that real research happened; the misconduct lies in the gap between what actually occurred and what was reported.

Common falsification patterns include selectively deleting outlier data points that undermine a hypothesis, digitally altering a western blot or microscopy image to remove an inconvenient band or feature, or reporting a sample size, randomisation method, or exclusion criterion that does not match the study’s actual conduct. In falsification cases, the original unaltered data — where investigators can recover it — is usually the single strongest piece of evidence, because it lets reviewers directly compare “what happened” against “what was reported.”

How do institutions decide which charge to frame?

Once an allegation reaches a formal inquiry, a research integrity officer (RIO) or equivalent lead within the institution’s research administration function must decide, on the balance of the available record, whether the conduct is better characterised as fabrication, falsification, or both. This framing choice governs what evidence the investigation committee needs to assemble and what standard of proof it must meet.

In the U.S. federal process, findings of research misconduct must be established by a preponderance of the evidence — that it is more likely than not that the conduct occurred, that it represents a significant departure from accepted practices, and that it was committed intentionally, knowingly, or recklessly. The Committee on Publication Ethics (COPE) applies a comparable logic for journal editors: COPE’s misconduct flowcharts direct editors who suspect fabricated or falsified data to refer the matter to the author’s institution rather than adjudicate guilt themselves, precisely because only the institution can access the underlying research record needed to distinguish the two.

In practice, the framing often turns on a single question: does a credible source record exist at all?

• If no notebook, dataset, instrument file, or ethics approval can be produced for a reported result, the case is typically framed as fabrication.
• If a source record exists but has been altered, cropped, reordered, or selectively reported, the case is typically framed as falsification.
• If an image or dataset from one experiment is reused and relabelled as the result of a different experiment, institutions frequently frame both charges together, since the record has been both fabricated for the second context and manipulated for the first.

What do real case patterns show?

Documented cases illustrate how the fabrication/falsification distinction plays out once an investigation examines the actual record.

The pattern across all four cases is consistent with the evidentiary logic above: wherever investigators found a complete absence of a traceable source record, they framed the finding as fabrication; wherever they found a record that had been altered or repurposed, falsification was added to the finding.

Answer-first Q&A

What is fabrication in research misconduct?

Fabrication is making up data, results, or research events and then recording or reporting them as genuine. It is distinct from falsification because no authentic experiment, survey, or observation ever took place — the research record has no real source to point to, which is why investigations focus on proving an absence rather than a discrepancy.

What are the three types of research misconduct?

The three recognised types are fabrication, falsification, and plagiarism — together known as FFP. Fabrication invents data, falsification alters or omits real data, and plagiarism appropriates another person’s ideas, words, or results without proper credit, under the ORI’s federal definition.

What is an example of fabrication in research?

A researcher who reports survey results from participants who were never recruited, or who claims an experiment was run when no instrument log, notebook, or ethics approval exists to support it, has committed fabrication. The Diederik Stapel case, where an entire dataset across dozens of psychology papers turned out never to have been collected, is a documented example.

What marks fabrication as unethical?

Fabrication is unethical because it is an intentional act of deception rather than an honest error — the researcher knowingly presents invented results as real findings. This intent requirement is why institutional and federal definitions explicitly exclude honest error or genuine differences of scientific opinion from the misconduct threshold.

What are the implications for institutional policy?

Getting the fabrication/falsification distinction right at the framing stage protects the integrity of the whole process. A misframed allegation can collapse under review if the evidence assembled does not match the charge — for instance, if an institution alleges fabrication but the respondent produces a genuine, if selectively reported, dataset, the correct charge was falsification all along.

This is also where authorship accountability becomes relevant: under ICMJE-aligned authorship criteria, every listed author is expected to be accountable for the accuracy and integrity of the parts of the work they contributed, which is part of why misconduct findings so often trigger parallel authorship disputes over who is responsible for a fabricated or falsified section. Institutions handling both strands together benefit from being precise about which finding — fabrication, falsification, or both — attaches to which contributor.

As research integrity offices mature globally, the practical guidance converges: build the case around the research record first, and let the record determine the charge — not the other way round.