What proportion of the world’s scientific output contains fabricated or falsified data? Scientific integrity expert James Heathers of Linnaeus University in Växjö, Sweden, reviewed and compared 12 different studies that examined a total of more than 75,000 scientific articles from fields such as social sciences, medicine, and biology and concluded that one in seven published articles may be at least partially fake. According to Heathers, the studies he analyzed used online tools to estimate the number of papers that contained falsification, and their conclusions generally converged on around 14% of the total. “There’s a really persistent commonality to the results,” he told the Retraction Watch website. His estimate was published in an article uploaded to the Open Science Framework preprint repository and has not yet been reviewed by other researchers.
Heathers says he decided to carry out the study after questioning a frequently cited finding in studies on misconduct: that 2% of scientists admit to having falsified, fabricated, or modified research data at least once in their careers. “The problem is that this figure is out of date,” he said, noting that it is based on a 2009 study that evaluated data collected in 2005. Strictly speaking, the data from Heathers’s study cannot be compared with those from 2009—while his work analyzed articles that used online tools to identify problems of any nature, the previous study looked at articles that made their conclusions by asking the authors questions directly and focused on problems that distort scientific knowledge, such as data falsification, not including ethical issues that do not affect the results, such as plagiarism.
Daniele Fanelli, a researcher at Heriot-Watt University in Edinburgh, Scotland, and author of the 2009 study, disagrees with Heathers’s methodology. It is not correct, she says, to mix problems with different degrees of severity—from changes that do not compromise the conclusions to completely fraudulent work. Epidemiologist Gowri Gopalakrishna of Maastricht University in the Netherlands believes Heathers’s study oversimplifies a complex topic by grouping together articles from different fields of knowledge where the prevalence of fabrication and falsification is not the same. “If you want to get the attention of the government and show the scale of the problem, it is probably useful in that way, but I really do think that it’s important to drill down,” said Gopalakrishna, who reached a different conclusion to both Heathers and Fanelli in her own 2021 study on the matter. She found that from a sample of 7,000 scientists based in the Netherlands, 8% confessed to falsifying or fabricating data at least once between 2017 and 2020.
The story above was published with the title “Researchers reach different conclusions on the scale of data fabrication” in issue 345 of November/2024.
Republish