May 2009


Fabrication and Falsification are Widespread in Science

How common is fraud in published scientific research? Some scientists have asked this question in relation to plagiarism, finding a common indifference to plagiarism among technical medical publishers.

Plagiarism can sometimes be detected with software, and is readily examinable from published research. However, it is more difficult to determine the prevalence of scientific fraud that affects the outcome of research, i.e., fabrication and falsification, to the exclusion of plagiarism.

An obvious reason for this difficulty is that fraud is a sensitive issue; it is reasonable to suspect that people are unlikely to admit to having committed it themselves. Additionally, it is often difficult to determine if deliberate fraud has occurred; the only ones who know may be the original scientists themselves.

This may be why scientific analyses of the extent of non-plagiaristic scientific research fraud have yielded conflicting results. Nevertheless, it is important to determine its extent.

Daniele Fanelli (University of Edinburgh, United Kingdom) has systematically analyzed many previous scientific studies of fabrication and falsification in scientific research. In her study, scientists themselves were asked to admit to having participated in or having witnessed fraud.

How often to scientists admit to nonplagiaristic fraud?

Fanelli focused her efforts on 18 studies of relevant fraud (fabrication and falsification), mostly based in the United States. Most of the analyzed studies were based on bio/medical and clinical research.

On average, almost 2% of the scientists in these analyzed studies admitted to having fabricated or falsified their research findings. However, this fell by half when the words "fabrication" or "falsification" were included in the analyzed studies.

"Questionable practices," i.e., inconclusive falsification, was admitted by, on average, almost 10% of the scientists. These are only admitted cases, and since it is consequently reasonable to suspect that they are low estimates, the prevalence of nonplagiaristic fraud in bio/medical and clinical research is clearly quite high.

How often do scientists witness nonplagiaristic fraud?

Here we get to more solid evidence that fraud is more common than the 10% level. It is reasonable to suspect that people are more likely to admit to having witnessed fraud than having committed it themselves.

More than 14% of scientists have witnessed fraud. This only fell to a little over 12% when the words "fabrication" or "falsification" were included in the analyzed studies, possibly indicative of the less-sensitive nature of the question.

What are the consequences of nonplagiaristic fraud?

It is commonly thought that fraud will untimately be discovered and punished. Is this the case?

Unfortunately, no. Fanelli found that approximately half of the fraudulent scientists had any action taken against them.

This means that at least half of them got away with it. The remainder may or may not have been proven and punished in any meaningful sense.

My two cents.

For this blog, I originally wrote a long list of advice for faculty, postdocs, and students, regarding how to prevent and deal with fraud by others, based on my experience, because I have had to deal with it so often. After reading my writing, I figured that some of it would probably be taken as very controversial, so I ended up deleting it (it felt good to let it out though).

I will say that, in my experience, roughly 2/3 of students, postdocs, and faculty in chemistry ("research institutions") have routinely, knowingly committed serious scientific research fraud; some of their careers are based almost entirely on knowingly fraudulent results. I strongly concur that the 2% fabrication and falsification rate of admission is a low estimate of its prevalence, at least in the academic world.

for more information:
Fanelli, D. (2009). How Many Scientists Fabricate and Falsify Research? A Systematic Review and Meta-Analysis of Survey Data PLoS ONE, 4 (5) DOI: 10.1371/journal.pone.0005738