In September 2024, Eliezer Masliah, the head of the National Institutes of Health’s (NIH’s) neuroscience division, stepped down from his post after Science found evidence of addition, deletion, and mirroring of images in 132 papers he authored between 1997 and 2023. Prior to the discovery of his misconduct, Masliah was a preeminent researcher in the fields of Alzheimer’s and Parkinson’s diseases, having published over 1,100 papers that racked up more than 170,000 citations—enough to make him one of the top 10 cited scientists in both fields.
Masliah’s research has served as the theoretical underpinning for an entire generation of Alzheimer’s and Parkinson’s drugs, most of which have failed. Therapeutics such as prasinezumab, which targets a particular protein that Masliah identified as linked to Parkinson’s, have registered null (no better than placebo) results in clinical trials after years of painstaking research. By repeatedly doubling down on theories he knew had no sound scientific basis, Masliah marginalized alternative research pathways that were unable to gain funding due to widespread acceptance of his hypotheses. It is impossible to know precisely how much money was spent on Masliah’s dead ends or how many lives were lost while scientists wasted time and attention on false mechanisms of action. But the NIH alone spends $3.8 billion per year on Alzheimer’s research—much of it motivated by Masliah’s “insights”—and the disease kills 120,000 people annually in the United States, to say nothing of research or mortality elsewhere.
Few scientific fraudsters of Masliah’s ilk ever receive meaningful criminal penalties. The state does not shy away from severely punishing financial or other kinds of fraud, even when those frauds impact small segments of the population and have few knock-on effects. To restore integrity and public trust in science, the federal government must put scientific fraud on an equal footing with other types of fraud and prosecute research misconduct assiduously.
Good, reliable science is of vital public importance. For governments, social science informs policy proposals on everything from efficient taxation to environmental policy. For private actors, otherwise risky moonshot projects—like the development of a novel Covid-19 vaccine under extreme time constraints—can make fiscal sense if the state contributes research funding and expertise. When private actors succeed thanks to government support, they share the wealth through both taxation and positive spillover effects. It should therefore come as no surprise that highly developed states dedicate huge sums to subsidizing research and development (R&D): The US federal government spent over $185 billion on R&D in 2022, and OECD governments as a whole spent over $540 billion.
But when the state funds scientific projects, it does so under the assumption that scientists use the money responsibly and report their results honestly. Failure is a normal part of the scientific process—but fraud is not. When unscrupulous researchers such as Masliah instead use state funds to produce fake research that serves only to advance their careers, they are simultaneously guilty of violating the public trust and of simple fraud—the appropriation of government money under false pretenses.
The clearest indication of science’s growing unreliability problem is the “replication crisis,” or the growing body of scientific work whose conclusions cannot be replicated when researchers copy the exact methodology provided by a paper’s original authors. Unscrupulous researchers looking for landmark results often employ a toolkit empowering them to transform nonsignificant results into lucrative blockbusters. Their lineup includes techniques as straightforward as changing a 0 to a 1 in an Excel spreadsheet or photoshopping images to show effects where none exist and as subtle as p-hacking, a form of misconduct in which researchers add or delete data points until their claimed finding is “significant”—that is, having a 1 or 5 percent chance of occurring by random chance (the most common significance thresholds used in nearly all disciplines). Papers from all scientific fields tend to have significance values that cluster around the 1 and 5 percent thresholds—a clear sign that fraudsters are p-hacking their data to achieve marginal significance.
The replication crisis would be bad enough if it manifested only in relatively new findings that have not been built upon by later researchers, but a distressing proportion of foundational results have been rendered non-replicable. The situation is particularly dire in fields like psychology. The world-famous Milgram Experiment, which has been cited for over 60 years as evidence that people will commit immoral acts under even marginal pressure from an authority figure, was recently tarnished by new accounts proving that Milgram selectively presented his data in sensationalist ways. Other baseline experiments in psychology, such as the Asch conformity experiments and the Zimbardo prison experiments, have been either fundamentally misinterpreted or outed as fraudulent.
Though psychology has probably been hit hardest by the replication crisis, fields traditionally perceived as more quantitatively rigorous have not been spared. A 2021 economics paper by Adrien Matray and Charles Boissel, which purported to show that hiking dividend taxes increased firm investment, was retracted after replication researchers found clear-cut data manipulation. At even more rarefied heights, all-star behavioral economics researcher Francesca Gino was placed on administrative leave from Harvard Business School after a graduate student who was unable to replicate Gino’s results found substantial evidence of data manipulation.
The rot even extends to disciplines like medicine, in which image manipulation is the technique of choice. The Dana-Farber Cancer Institute, an ultra-prestigious research institution affiliated with Harvard, was forced to correct more than 30 papers earlier this year after independent researchers uncovered pervasive image manipulation linked to Dana-Farber scientists—to say nothing of the Masliah case.
Research misconduct at academia’s highest levels is worsened by a pervasive culture of silence and a total lack of self-regulation. When the graduate student who uncovered inconsistencies in Gino’s research took her findings to other professors, she was told to drop the allegations or face career suicide. Her PhD advisors refused to approve her thesis unless she stopped investigating Gino. Though Masliah was removed from his position as head of neuroscience at NIH, it is unclear whether he was retained by the agency in another capacity. For their part, young scholars know that success in their fields is highly dependent upon publication in prestigious journals—even if they must torture the data until it confesses the desired result.
Academic journals owned by prestigious publishers such as Springer and Elsevier have been caught waving through articles with sections obviously written by ChatGPT—with some peer reviewers leaving comments authored by the same source. A paper in Frontiers in Cell Development and Biology was retracted after readers spotted plainly AI-generated images that featured spelling errors and massive scale distortion. One of the article’s peer reviewers, reached for comment by Vice, claimed that image review was not his responsibility.
If scientific institutions are unwilling to self-regulate, then governments must step in. Though some worry that criminalizing scientific fraud will have a chilling effect on research, there is little reason for honest scientists to fear the lawman’s grasp. In the same way that no prosecutor would bring charges against a financial advisor for merely losing a client’s money, researchers need not fret about facing a criminal penalty for an experiment that does not work. But if researchers accept state funds and manipulate their data, fraudulently add or delete data points, and intimidate others into silence, they must be held accountable.
Scientific fraud statutes should therefore clearly designate the research practices that cross the line into fraud—data addition or deletion, image duplication or manipulation, withholding of relevant results, and statistical manipulation designed to make nonsignificant results significant—and prescribe penalties at least as severe as those applied against limited financial frauds, which tend to have localized effects that entail little social disruption. The first step toward restoring public trust and integrity in science is rehabilitating the reputation of scientists, and the first step toward that rehabilitation is exposing and sequestering those who have caused science to fall into disrepute. Otherwise, we may be doomed to a future of fraudulent results and failing trust.