Credit: bpawesome/Getty
The replication of experimental results is important for building credible science, but replication studies are often difficult to find. A team of researchers wants to change that. They have started putting replication studies on the post-publication peer-review platform PubPeer, together with links to the original studies.
There have been numerous attempts to replicate scientific experiments over the past decade or so, and many of these have not been able to reproduce the original results. But the inability to find these replication studies easily wastes research time and resources. If scientists aren’t aware of a replication study, particularly one that was unsuccessful, they might cite the original research without knowing it has issues.
“If PubPeer can help people learn about replications, both successful and unsuccessful, that will speed the progress of science by facilitating scientific self-correction,” says Don Moore, a social psychologist at the University of California, Berkeley. A 2012 study by Moore on overconfidence1 was successfully replicated this year2.
The team behind the new project says it plans to create discussion threads on PubPeer for around 2,400 replication studies in batches of around 100 at a time. The replication studies are indexed in the FORRT Library of Reproduction and Replication Attempts (FLoRA), an initiative of the Framework for Open and Reproducible Research Training (FORRT), which is a community-driven effort to teach researchers about open science and reproducibility.
FORRT launched a database in 2024 to better catalogue replication studies. But the database’s creators at the Münster Center for Open Science in Germany say it’s still not easy to find replication studies because they are typically not linked to the original studies in many databases.
“Replications are generally less cited and less visible than original studies,” says Josefina Weinerova, a psychologist at Birkbeck, University of London, who is co-leading the PubPeer project.
Out of roughly 2,400 studies currently catalogued, almost 1000 have been successfully replicated independently, and 865 attempts at replication failed. For the remainder, researchers managed to replicate some of the outcomes.
Weinerova and her colleagues plan to inform the authors of all the papers that they have created threads for on PubPeer. Weinerova says that her team has already reached out to authors of the first batch of studies. Most didn’t reply, she says, and of the ones that did, none asked for comments not to be posted.
Because PubPeer is often used to flag potential issues or errors in papers, some researchers have complained about their papers being mentioned on PubPeer when no errors have been found.
But Weinerova says researchers should be pleased if studies confirming the validity of their work are highlighted on PubPeer. “Successful replication is still useful for other authors to know about,” she says. “While there may be perceived stigma with comments on PubPeer, there are also many comments that are not necessarily about pointing out major issues with papers.”
