TL;DR

• Data fabrication means inventing results that never existed.

• Data falsification means manipulating real data to change outcomes.

• Both are forms of research misconduct that can end academic careers.

• Journals and institutions use detection tools to catch both practices.

• Honest reporting of negative or inconclusive results is always acceptable.

You have spent weeks on your research. The results are not what you hoped for. The data does not support your hypothesis cleanly. In that moment, the temptation to adjust a number, remove an outlier, or simply invent a cleaner dataset can feel small. It is not small. What is data fabrication and falsification is one of the most consequential questions in academic research, and the answer affects every researcher at every level, including high school students submitting to peer-reviewed journals for the first time.

These are not abstract ethical concepts reserved for professional scientists. Journals that accept student submissions, such as the Journal of Emerging Investigators and the International Journal of High School Research, operate under the same research integrity standards as any university publication. Misconduct discovered after publication leads to retraction. Misconduct discovered before publication leads to rejection and, in some cases, a permanent record with the journal.

Understanding what these terms actually mean, how they differ, and what the consequences look like in practice is the clearest protection a new researcher has. That understanding starts here.

What Is Data Fabrication in Research

Data fabrication is the act of inventing data, results, or findings that were never actually collected or observed. A researcher who records experimental measurements they never took, or who creates a dataset from imagination rather than real trials, is fabricating data. It does not matter whether the invented numbers seem plausible or whether the researcher believes the results would have turned out that way. If the data did not come from real observation or measurement, it is fabricated.

The U.S. Office of Research Integrity (ORI), which oversees federally funded research in the United States, defines fabrication as "making up data or results and recording or reporting them." This definition is widely adopted by journals, universities, and research bodies globally. The key word is "making up." No degree of good intention changes that definition.

For student researchers, fabrication often appears in subtler forms than outright invention. Recording an experiment you planned but never ran, filling in missing data points with guesses, or generating fake survey responses all fall under fabrication. The scale does not reduce the seriousness. A single fabricated data point in an otherwise honest paper still constitutes misconduct.

What Is Data Falsification and How It Differs

Data falsification involves manipulating real data, methods, or results in a way that misrepresents what actually happened. Unlike fabrication, the researcher starts with genuine observations. The misconduct lies in altering, omitting, or distorting those observations to produce a more favourable or cleaner outcome. Removing inconvenient outliers without scientific justification, adjusting a graph's axis to make a trend appear stronger, or selectively reporting only the trials that supported a hypothesis are all forms of falsification.

The ORI defines falsification 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." The phrase "not accurately represented" is the core of it. Falsification does not require a researcher to invent anything. It requires only that they present real data in a way that deceives the reader about what the data actually shows.

This distinction matters practically. A student who runs an experiment, gets mixed results, and then removes the trials that did not work before reporting averages is falsifying data. The original measurements existed. But the reported findings no longer reflect them honestly. Understanding what makes a research paper publishable includes understanding that honest, complete reporting of results is a non-negotiable standard, not a stylistic choice.

If you are working toward your first journal submission and want structured guidance on avoiding these pitfalls before you submit, joining the Publication Compass waitlist puts you first in line for an AI platform built specifically to support student researchers through the submission process.

Why Data Fabrication and Falsification Are Treated as Serious Misconduct

Research builds on itself. Every published paper becomes part of the foundation that future researchers cite, replicate, and extend. When fabricated or falsified data enters that foundation, it corrupts everything built on top of it. Other researchers waste time and funding trying to replicate results that were never real. Clinicians in medicine may make decisions based on false evidence. Policy may be shaped by manipulated findings. The damage is not limited to the original paper.

The Committee on Publication Ethics (COPE), which sets standards for journal editors worldwide, classifies fabrication and falsification alongside plagiarism as the three core forms of research misconduct. COPE guidelines require editors to investigate allegations of misconduct even after publication, and to issue retractions when misconduct is confirmed. Retraction Watch, a publicly accessible database, records thousands of retracted papers, many for fabrication or falsification, and the entries remain permanently searchable by name.

For a high school student, the consequences operate on a smaller but still significant scale. A retraction from a student journal is a matter of record. Some journals notify school supervisors or institutional contacts. College application essays that reference a retracted publication carry risk. The reputational cost arrives before a career even begins.

How Journals Detect Fabricated and Falsified Data

Detection methods have become substantially more sophisticated over the past decade. Journals and peer reviewers now use several overlapping approaches to identify suspicious data.

1. Statistical analysis of distributions. Real data collected from experiments tends to follow recognisable statistical patterns. Fabricated data often does not. Tools like GRIM (Granularity-Related Inconsistency of Means) and SPRITE can identify datasets where the reported means or distributions are mathematically impossible given the stated sample size and measurement scale. These tools are freely available and increasingly used by reviewers at student-facing journals.

2. Image integrity analysis. For research involving photographs, gel images, microscopy, or any visual data, journals use software to detect duplication, cropping, or manipulation. The journal Science and many others employ dedicated image integrity reviewers as part of their editorial process.

3. Cross-referencing with raw data. Many journals now require authors to submit raw data files alongside their manuscripts. Reviewers compare reported results against the underlying data. Discrepancies that cannot be explained by rounding or aggregation raise immediate flags.

4. Peer reviewer expertise. Experienced reviewers in a field develop intuition for results that seem too clean, too consistent, or implausibly aligned with the hypothesis. Unusually low variance, perfectly linear relationships, or results that replicate prior findings with suspicious precision can all trigger closer scrutiny.

None of these methods are infallible, but together they create a meaningful barrier. The assumption that a student journal will not look closely is not a safe one.

What to Do When Your Data Does Not Support Your Hypothesis

The most common pressure point where fabrication and falsification begin is a results section that does not deliver what the researcher expected. This is worth addressing directly, because the honest path forward is both available and publishable.

Null results, meaning results that show no significant effect or fail to confirm a hypothesis, are legitimate scientific findings. Journals including PLOS ONE explicitly evaluate papers on the soundness of methodology rather than the direction of results. A well-designed study that finds no effect contributes to the literature by preventing future researchers from pursuing the same dead end. Understanding what a research gap is and how to find one can help reframe inconclusive results as a contribution rather than a failure.

When results are mixed or inconclusive, the correct approach follows a clear sequence.

1. Report all results, including those that do not support the hypothesis, in full.

2. Distinguish between pre-planned analyses and any exploratory analyses conducted after seeing the data. Label exploratory findings as such.

3. Discuss limitations honestly in the discussion section. Reviewers expect limitations. A paper with no acknowledged limitations raises more suspicion than one that identifies them clearly.

4. Consider whether the findings, even if inconclusive, reveal something about methodology, sample characteristics, or measurement challenges that is worth communicating.

Honest reporting of imperfect results is not a consolation prize. It is the standard. Understanding what peer review involves and what happens to your paper during that process makes it easier to see why reviewers value transparency over tidy narratives.

The Difference Between Misconduct and Honest Error

Not every inaccuracy in a paper is misconduct. Honest errors happen. A calculation mistake, a mislabelled figure, or a transcription error from a lab notebook to a spreadsheet are not fabrication or falsification. The defining element of misconduct is intent. The ORI and COPE both distinguish between honest error and deliberate misrepresentation.

If you discover an error in your own work after submission or publication, the correct response is to contact the journal editor promptly and issue a correction. Journals have formal correction processes for exactly this situation. A correction is not a retraction. It does not carry the same implications. Proactively reporting your own error is treated very differently from having misconduct discovered by a reviewer or reader.

Publication Compass is built to help student researchers catch structural and methodological issues in their drafts before submission, which reduces the chance that honest errors make it into the final manuscript. The platform does not replace the researcher's judgment, but it provides a structured review layer that many first-time submitters do not otherwise have access to.

Frequently Asked Questions

Is removing outliers from a dataset considered data falsification?

Removing outliers is falsification if it is done without scientific justification and without disclosure. Outlier removal is legitimate when the criteria are defined before data collection, applied consistently, and reported transparently in the methods section. Removing data points after seeing results, because they weaken the conclusion, meets the definition of falsification under ORI and COPE guidelines.

What is data fabrication and falsification in the context of student research specifically?

For student researchers, data fabrication and falsification carry the same definitions as in professional research. Inventing survey responses, recording experiments that did not happen, or altering collected measurements to improve results all constitute misconduct. Student journals apply the same standards as professional publications, and some share editorial infrastructure with university-level review bodies.

Can AI tools help a researcher fabricate or falsify data without detection?

AI tools can generate plausible-looking synthetic data, but statistical detection methods are increasingly effective at identifying non-human data distributions. Journals that require raw data submission make AI-generated datasets easier to identify. Beyond detection risk, using AI to generate fake data is fabrication regardless of the tool used. Understanding what GPT-assisted research is and where it is permitted clarifies the boundary between legitimate AI assistance and misconduct.

What happens if fabrication or falsification is discovered after a paper is published?

The journal issues a retraction notice, which is permanently attached to the paper's record and indexed in databases including PubMed and Retraction Watch. The authors are typically notified, and in cases involving institutional affiliation, the institution may be informed. The paper remains visible but is marked as retracted. Citations to a retracted paper carry reputational risk for those who cite it.

Does a preprint count as a publication for misconduct purposes?

Preprint servers such as bioRxiv and arXiv have their own moderation policies and can remove content found to contain fabricated or falsified data. Misconduct in a preprint can be reported to the server and, if the author is affiliated with an institution, to that institution. Knowing what a preprint is and whether you should upload before submitting helps researchers understand the visibility and accountability that comes with early posting.

Conclusion

Data fabrication and falsification are not edge cases. They are the most direct way a researcher can undermine their own work and the work of everyone who might build on it. The pressure to produce clean, positive results is real, especially for students who feel their findings need to impress a journal editor or a college admissions reader. That pressure does not change what honest research looks like. It makes understanding these definitions more important, not less.

Report your results as they are. Acknowledge your limitations. If your data is inconclusive, say so and explain what that means. That approach will serve you across every paper you ever write. For more guidance on the full publication process, the Publication Compass blog covers each stage from drafting to acceptance in plain language built for researchers who are just getting started.