Scientific sleuthing: a better evening in than watching Netflix?

“I’ve cancelled my Netflix subscription – this is more fun than any true crime documentary,” reflects neuroscientist Mu Yang on why she spends as much as 30 hours a week hunting for manipulated images in the scientific literature. “On the weekend or a holiday, I might spend seven or eight hours looking at papers – maybe more,” says Yang, whose day job is running the Mouse NeuroBehavior Core at Columbia University and who has recently become better known as part of what appears to be an ever-growing cadre of academic “sleuths”.

Since 2022, her reports have led to more than 300 retractions – most notably related to the work of brain researcher Eliezer Masliah, previously the head of the flagship division of neuroscience at America’s National Institute of Ageing, with its $2.6 billion (£1.9 billion) budget before an investigation by the National Institutes of Health “made findings of misconduct” against him in 2024.

Rather than any deep subject expertise, Yang says her eye for a duplicated image or a manipulated Western blot has been her greatest asset when fraud hunting.

“I’m a behavioural neuroscientist, watching mice run round in mazes, so I don’t have much subject knowledge of the chemistry papers that I’m looking at. But I’ve used very simple and basic tools to get results: often it is just clicking on an image and enlarging it,” she explains.

Even more straightforwardly, Yang’s scrutiny often identifies when images demonstrating the results of one study have been transplanted in full or in part into other papers authored years earlier, suggesting manipulation or fabricated results. In one case, Yang revealed that the same graph appeared in different forms in 72 separate papers, each supposedly representing different studies. “Some argue it’s an innocuous mistake – that a single graph doesn’t affect the paper’s findings. But science can’t tolerate having made-up nonsense in the literature,” she reflects.

But there is clearly lots of made-up nonsense there. A record of more than 14,000 research papers were retracted in 2023, with more than 9,000 in 2024 and more than 5,000 in 2025 by the end of August, according to a recent arXiv preprint. And the investigations that led to the retractions were often triggered by sleuth comments on the PubPeer website – which describes itself as “the online journal club”.

Sleuths’ impact has recently gone beyond the withdrawal of problematic research papers and the removal of senior positions from miscreants. In May, Harvard University fired behavioural scientist and dishonesty expert Francesa Gino for data fraud exposed by the Data Colada sleuth collective.

Moreover, while sleuths typically carry out their investigations without remuneration, last month, the Harvard-affiliated Dana-Farber Cancer Institute agreed to pay $15 million (£11.2 million) to settle a lawsuit initiated by UK-based sleuth Sholto David. As the whistleblower, David will receive $2.63 million (£2 million), with the lion’s share of the rest going to the NIH, which funded the research in question. David filed the claim in April 2024, shortly after writing a blogpost for For Better Science highlighting duplicated images in the work of senior Dana-Farber scientists (57 papers in total). Noting that the papers were used in grant applications, David used the US False Claims Act, which allows claimants to sue on behalf of government – triggering a Department of Justice investigation – and to receive up to 30 per cent of any damages awarded (David’s payout is 17.5 per cent of the total).

“Even after fees and taxes, it’s still a large amount of money,” admits David, who is in his early thirties and works in a laboratory at an Oxford biotech firm. However, “it’s certainly not enough to retire on” and it “certainly wasn’t why I began looking into these papers. I’ve written more than 6,000 comments on PubPeer and have been looking into mistakes and forged stuff since I started my PhD” (awarded by Newcastle University in 2019). So even after his windfall, he will “still be working through journals and papers in the same way”.

But he is pleased to have been recognised for his work in the Dana-Faber case and he hopes that he won’t be the last to receive some reward for all those unpaid hours: “Many things fell into place which meant I was recognised as the whistleblower even though I’m outside the institution. I hope more people make claims themselves if I’ve shown them that it can be done.”

Source: 

iStock montage

The “David v Goliath” win against one of America’s most prestigious research institutes – which has not admitted fraud but pledged to improve “error-checking” – may also help to raise the prestige and professional profile of sleuths, many hope. Those who spend their spare time hunting for suspect images are often labelled “failed scientists, frauds, conflicted, malicious [or] defamatory”, one sleuth told Science following the Dana-Farber settlement. In addition, sleuths – some of whom prefer to remain anonymous – are sometimes accused of overegging the egregiousness of minor errors that, in reality, have little bearing on the reality of a paper’s overall findings.

Accusations of jealousy and ill will have frequently been levelled at David by commenters on his humorous but scathing “Science Police” YouTube videos, often filmed from his bedroom or garage, which summarise his investigations. David happily admits that he isn’t as successful as those he criticises and that, at a certain level, this irks him.

“Would I like a research job? Of course I’d like to spend time tinkering in the laboratory [as an independent researcher], but I’ve accepted I won’t write a Nature paper. So maybe it does irritate me if someone producing manipulated images has a senior research job and I’m doing boring lab tasks,” says David. “I’m sure footballers in the lower leagues feel the same when they see their peers playing in Premier League. But things in science are either true or not true – it shouldn’t matter what motivates me in calling out something if I think it’s wrong.”

Similar ad hominem attacks were levelled against Purdue University’s David Sanders after he called out suspicious images in the work of internationally acclaimed cancer scientist Carlo Croce, leading to numerous retractions and Croce’s removal as departmental chair – though his university, Ohio State, found that lapses in his management of his lab did not amount to research misconduct. Croce sued Sanders for libel after the latter’s investigation was used as the basis for a front-page New York Times story, but Sanders won the case.

“It is true that my career has not been as illustrious as his,” Sanders concedes, “but I have made some important research contributions that are now included in textbooks.” And, more generally, he would “rather have a career based upon ethical behaviour and publications than one that relied upon repeated violations of ethical and scientific norms”.

His sense is that those who are “complaining the loudest about investigations of violations of scientific norms are those who have committed those infractions. But for critics with clear hands, his question is simple: “The self-correcting aspect of science has proven to be inadequate. How do critics of the sleuths propose that the task of ensuring scientific literature’s validity be accomplished?”

Sanders is among those who have called for sleuths to receive greater gratitude and professional reward. But even in its absence, David is surprised that more university-employed scientists aren’t interested in investigating and exposing poor practice and wrongdoing.

“Amateur sleuths are finding these [errors], but shouldn’t everyone be reading these papers in the same careful way?” he asked. “When I was thinking about becoming a scientist, I thought everyone would be reading papers voraciously, but it seems we’ve entrusted post-peer review to a handful of people.”

Source: 

iStock montage

Given the numerical mismatch with authors, even sleuths admit that despite their best efforts the vast majority of research fraud is going undetected.

“There is no way to keep up with the research paper mills,” concedes Columbia’s Yang, noting that artificial intelligence (AI) has made it “extremely cheap and easy” to churn out superficially passable journal articles that, nevertheless, have zero research merit.

Some observers have suggested that AI could also learn to detect fraud at much greater scale and potentially more accurately than sleuths can. For many sleuths, however, the focus on improving AI detection is a distraction from the real barrier to rooting out suspect papers from the literature.

“AI is just another tool that people are using,” notes Reese Richardson, a research fellow at Northwestern University who studies scientific fraud. “The publishing industry likes to talk about fixing the GenAI problem using AI, but we’ve demonstrated that the bigger issue is the industry itself, which has proven largely unresponsive to the problems we’ve been raising for many years.”

Publishers’ responses to “tortured phrases” flagged in their journals illustrate the industry’s inertia, says Richardson. Pioneered by the French sleuth Guillaume Cabanac, the bizarre renderings of scientific terms by AI or translation software – such as “fluffy rationale” for “fuzzy logic” or “shrewd devices” for “smartphones” – are increasingly recognised as a reliable indicator of problematic papers. Yet while some publishers are retracting 90 per cent of papers with tortured phrases, others have pulled only a tiny percentage, notes Richardson, referring to Cabanac’s latest dashboard on this topic.

“Most of the articles flagged for tortured phrases have not been retracted – it’s not very inspiring.”

Even when publishers accept that suspect images or nonsensical text have been identified, there is precious little redress, many sleuths insist. As David puts it, “There is very little sorrow that something has gone wrong or falling on swords”.

For Yang, the “incorrect correction”, which downplays the offence as accidental or trivial, is a particular bugbear. “Some corrections have clearly not [even] been read by an editor – they’ve just published exactly what the author has sent them,” she said. “I call it ‘fraud laundering’. I kick and scream on social media when it happens – I call publishers babies, spitting out the cases I’ve spoon-fed them – but it doesn’t make much difference.”

Source: 

iStock montage

That frustration has led some sleuths to conclude that a research integrity ombudsman, funded by publishers, is needed to force journals, universities and funders to step up their efforts. In Richardson’s view, the Committee on Publication Ethics (COPE) – whistleblowers’ only means of appeal if they are unsatisfied with a publisher’s handling of their complaint – ”lacks teeth” when it comes to tackling journals that fail to act on fraud reports, while David called the non-profit membership organisation “an industry-funded exercise in achieving absolutely nothing”.

In order to tackle the “real lack of accountability from everyone in the system we need someone with the authority to make a decision [on retracting]”, continued David. “Too often we are stuck in ridiculously circular email chains where a journal is waiting on an author who, after several months, then claims they’ve lost their laptop [or] they can’t respond because the institution is investigating and they’re waiting for the publisher to take action. You need someone with authority to cut through this at a certain point.”

That said, David would be wary of any official research fraud police being captured by the institutions they were regulating. “Paying sleuths to do this work has been discussed, but the field has come far on its own. Do we want to impose the failed and stodgy structures of academia on to the innovative world of sleuthing?” he asked.

Richardson is also sceptical. “The idea of a centralised ombudsman has some merit, but the industry would pressure it to become little more than window dressing,” he said, urging more national funders to invest in integrity assurance measures instead. The European Research Council’s six-year, €8 million (£7 million) project on “how, when and why does science fail to correct itself”, which ends in May 2027, shows how this work could be developed, he said.

Yang urges much more rigorous screening of journal submissions for plagiarism and image manipulation before they are even sent to peer reviewers. But the “toxic culture” of academia, which pressurises scholars to publish prolifically, also needs to change to reduce the incentives for research fraud, she said, bemoaning what she calls the “Barbie-ification of academia”.

“Barbie is an unrealistic standard of female beauty, but we’re selling a similarly unrealistic ideal of scholarly excellence to early-career researchers. It’s twisted,” she said.

“So I don’t feel angry about people who manipulate images or data. I don’t even like to use the word ‘fraudsters’. They are opportunists trying to survive in a very difficult system.”