Direct comparison
Falsification vs. Fabrication in Research
Falsification manipulates real data; fabrication invents data that was never collected. Compare the two under 42 CFR 93.103 with real adjudicated cases.
Side-by-side comparison
| Dimension | Falsification | Fabrication |
|---|---|---|
| Did the underlying research event occur? | Yes — the experiment, observation, or process actually happened | No — the reported result has no real underlying event |
| What's dishonest | The representation of real data or materials | The existence of the data itself |
| 42 CFR 93.103 definition | 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 | Making up data or results and recording or reporting them |
| Typical form | Selective exclusion of data points, undisclosed image adjustment, altered measurement values, misrepresented protocol steps | Invented survey responses, non-existent experiments, wholly synthetic datasets presented as real |
| Common detection methods | Image-forensic tools (Proofig, ImageTwin), statistical forensics (GRIM, SPRITE, Benford's law), failed independent replication | Failed independent replication, absent raw data/lab notebooks, whistleblower reports from collaborators with access to the real record |
| Federal citation | 42 CFR 93.103 | 42 CFR 93.103 |
| Requires intent to be adjudicated as misconduct? | Yes — honest error or differences of opinion are explicitly excluded | Yes — honest error or differences of opinion are explicitly excluded |
Common questions
FAQ
Can a single case involve both falsification and fabrication?+
Yes. Eric Poehlman (University of Vermont, 2005) fabricated and falsified data across roughly ten publications and 17 grant applications in the same body of work, admitting to 54 separate ORI and institutional misconduct findings — the two categories are independent, not mutually exclusive, and often appear together.
Is falsification worse than fabrication, or vice versa?+
Neither the regulation nor ORI ranks one as categorically worse — both are treated as equally serious forms of research misconduct under 42 CFR 93.103, and sanctions are determined case by case based on severity, extent, and impact, not which FFP category applies.
Does an honest mistake count as falsification?+
No. 42 CFR 93.103 explicitly excludes honest error and differences of opinion from the research misconduct definition. A finding requires the conduct to be a significant departure from accepted practice, committed intentionally, knowingly, or recklessly, and proven by a preponderance of the evidence.
Where does plagiarism fit relative to these two?+
Plagiarism is the third FFP component and is evaluated independently — it concerns appropriating another person's ideas, processes, results, or words without credit, not the honesty of the data itself. A case can involve any one, two, or all three FFP categories.
Going deeper
