As dramatic as the Anversa case is, he is far from alone. This month, Anversa’s lab saw 13 papers retracted, but even if all journals honor the retraction requests, he won’t crack the top 10 for scientists who’ve had their articles pulled from the literature. Neither does Cornell University’s Brian Wansink, the food marketing researcher — and former media fixture — who experienced a similar fall over the past few years. The dubious honor for most retractions goes to Yoshitaka Fujii, a Japanese anesthesiologist who fabricated his findings in at least 183 papers, according to a 2012 investigation launched by journal editors and Japanese universities.
We have tracked these numbers for an organization we founded in 2010 called Retraction Watch. Over the past eight years, we have compiled a database of more than 18,000 retracted papers from the scholarly literature, the oldest of which dates to 1756. And although the rate of retractions seems to be plateauing, at somewhere south of one-tenth of 1 percent of papers published, it grew dramatically between 2000 and just a few years ago. Why? And what is the significance of these misadventures in science?
Every retraction tells a story. At least half the time, that story involves misconduct or fraud. Consider Diederik Stapel, a Dutch psychologist who crafted his splashy findings about human behavior out of whole cloth; and Andrew Wakefield, whose spurious claims about the risks of childhood vaccinations and autism helped foment disastrous anti-vaccine campaigns.
But sometimes retractions tell tales of science working just as it should, without misconduct. Although retractions are considered the nuclear option in scholarly publishing, they’re really a sign that science is operating as advertised. At its heart, science is the ultimate self-correction machine. Advance a hypothesis, take a stab at finding proof, then try to replicate the results if successful — or, if not, try again. When something goes awry with a publication, science is already well prepared for both the detection of problems and the means of dealing with them.
Retractions — and the growth in their numbers — are also a story of sleuths. People such as Elisabeth Bik, a microbiologist who reviewed more than 20,000 papers by hand and found that about 4 percent of them had evidence of inappropriately manipulated images. Or James Heathers and Nick Brown, who refer to themselves as “data thugs” and have devised simple but illuminating mathematical tools to quickly find evidence of potentially suspect statistics.
One truth of modern science is that technology — particularly software for editing images — makes perpetrating fraud easier than ever. But it makes being a detective easier, too. No reputable publisher goes to press with a manuscript today without first running it through a program to identify plagiarism.
The Anversa case highlights another trend worth watching: the use of the False Claims Act, also known as the Lincoln Law, to go after scientists who commit fraud using federal dollars. The False Claims Act allows whistleblowers to collect hefty rewards for reporting faked data in grant applications and reports. Columbia University settled a similar case for $9.5 million in 2016. And Duke University is poised to settle a case allegedly involving some $200 million in grants — and more than a dozen retractions so far — early next year.
Put another way, fraud fighters have many more weapons in their armory than they did even 20 years ago, and a growing army appears more willing to use those tools. Their allies are government watchdogs such as the U.S. Office of Research Integrity, which oversees government research grants, and the National Science Foundation’s Office of Inspector General. While both agencies do remarkable work, their resources and remits are constrained. That’s particularly true for the Office of Research Integrity, which lacks subpoena power and can only enter voluntary agreements with researchers to keep them from receiving additional funding, typically for periods of three years. Scientists have taken to fighting these sanctions in administrative courtrooms, dragging cases on for years.
While 18,000 retractions may sound like a lot, that amount is clearly just a fraction of the total number of papers that should be banished from the literature. We know that because some 2 percent of researchers admit, in anonymous surveys, to acts that are considered misconduct. And for every whistleblower who sees his or her work lead to a retraction, we hear from several who are met with silence or retaliation. The work is just beginning.