If you’re anything like us, when you take a pill, drive over a bridge, or make other important decisions, you want to know those choices are based on solid evidence. You want to know the company making that pill has tested it in in good faith; that the bridge engineers knew what they were doing when they calculated the load capacity of the struts and spans; and that, in both cases, the people who developed the underlying data did so honestly, correcting any errors they found as soon as possible to minimize the downstream consequences.

Alas, that isn’t always the case. Our experience running a site called Retraction Watch for about a decade and a half tells us scientific misconduct is far more common than anyone wants to admit, and that, when serious problems are detected, researchers can take years to correct or retract their work—when they do so at all. Our database of retracted articles (the world’s most comprehensive) shows journals now retract thousands of papers each year, yet our best guess is the number should be at least 10 times higher.

What’s at the foundation of this problematic scene? Those who closely monitor scientific publishing agree a significant part of the reason for the lack of intellectual honesty about correcting the record is the academic reward system, which prizes publications above all else. Cheating, sloppiness, and scientific wishcasting are all symptoms of a regime in which quantity trumps quality.

Until roughly two decades ago, retractions were pretty rare, numbering only in the two-digit range per year. That has changed significantly, with the number growing past 10,000 in 2023. With the rise has come an increase in stigma around retraction, in part because people are aware that retractions often resulted from misconduct. It’s not surprising that many scholars now take steps—sometimes even hiring lawyers to block action by journals and publishers—to stop or obscure them.

Because they’re supposed to be understood primarily as a way to keep the scientific record healthy, retractions are not officially meant as a form of punishment. According to the Committee on Publication Ethics, a UK nonprofit membership organization for publishers and others, “The main purpose of retractions is to correct the literature and ensure its integrity rather than to punish authors who misbehave.”

With that framing in mind, Science editor-in-chief H. Holden Thorp—whose efforts in this area are important and, we think, well-intentioned—has suggested retractions be decoupled from determinations of who is at fault. His notion is a two-stage process in which possible retraction is considered relatively quickly–attending promptly to the soundness of the scientific record—and questions of misconduct are given separate, longer considerations. While we see the logic here, our experience tells us putting off misconduct investigations makes them even more secretive and may mean they never see the light of day. And, like retractions, misconduct investigations are a key part of keeping science healthy.

In theory, and sometimes in practice, scientists already sanction authors who retract papers, based on explanations of what went wrong. Economists studying the reputation effects of retractions have found that when researchers acknowledge honest errors in retraction notices, they do not suffer any downstream damage to their citation counts—a highly imperfect but often-used measure of career success. When such notices mention misconduct, however, citation rates fall.

But—unfortunately—retractions have become a punishment of last resort, because many of the institutions involved have abandoned their responsibility to safeguard knowledge. Universities see retractions as a stain on their reputations, likely to lead to decreased federal funding (and the overhead costs that come along with the core dollars) and decreased support from alumni and corporations. Publishers see them the same way, with publicly traded companies listing retractions and paper mills (shady companies that sell papers and authorship to desperate authors) as significant risks to their reputations in prospectuses and quarterly reports. These institutions have little economic incentive to do the right thing.

Retractions have become a punishment of last resort, because many of the institutions involved have abandoned their responsibility to safeguard knowledge.

What about applying government pressure? One would think the federal agencies that fund or depend on scientific research to advance human health—the NIH and FDA, for example—would aggressively push publishers to correct the record more quickly. The NIH, after all, has funded much of the shaky foundation beneath so many failed Alzheimer’s treatments. Most of the researchers involved in this specious science, however, have never been held accountable by the Office of Research Integrity, which requires but a handful of retractions each year as part of their agreements with those it finds have committed misconduct.

The only U.S. federal agency, in fact, that has ever sanctioned a publisher was the Federal Trade Commission. In 2019, the FTC won a $50 million settlement against OMICS, a publisher long known for dodgy practices, for having “deceptively claimed that their journals provided authors with rigorous peer review and had editorial boards made up of prominent academics. In reality, many articles were published with little to no peer review, and many individuals represented to be editors had not agreed to be affiliated with the journals.”

To be sure, discipline is not the only way to discourage misconduct. And it may not even be the most effective way. To make a dramatic difference, governments and universities must abandon the “publish or perish” incentives at the root of the problem.

But others have recommended that shift for decades, and while some important moves in the right direction have occurred, one has the sense such incentives are a barely movable object. So, in the meantime, we have retractions, which, when done properly, are public sanctions that also manage to correct the record. We can only hope the real solution is not a bridge too far.