The Dutch publisher Elsevier is investigating hundreds of researchers whom it suspects of deliberately manipulating the peer-review process to boost their own citation numbers.
The publisher is looking into the possibility that some peer reviewers are encouraging the authors of work under review to cite the reviewers’ own research in exchange for positive reviews — a frowned-on practice broadly termed coercive citation.
Elsevier’s probe has also revealed that several of these reviewers seem to be engaging in other questionable publishing practices in studies that they have themselves authored. The Elsevier analysts who uncovered the activity told Nature that they “discovered clear evidence of peer-review manipulation” and of academics publishing the same studies more than once. Elsevier said that their investigations will lead to some of these studies being retracted.
But it said it won’t be necessary to retract any studies found to be affected by coercive citation because the authors aren’t responsible for the problem, and citation manipulation doesn’t affect the research.
The suspicious activity was uncovered by Elsevier analytics experts Jeroen Baas and Catriona Fennell. The pair, based in Amsterdam, looked at the peer-review activity of almost 55,000 academics who review for Elsevier journals, in a bid to find how often these researchers’ work is cited by the papers they review. The study, which hasn’t yet been peer reviewed, was posted on the SSRN repository for social-science research on 6 September.
That analysis was prompted by a case at the Elsevier journal Geoderma. In 2017, Artemi Cerdà resigned from the journal’s editorial board after being accused of using his power to boost his own citation count and that of the journals he edited. (Cerdà, a soil scientist at the University of Valencia in Spain has denied accusations of citation manipulation; the publishing arm of the European Geosciences Union, where Cerdà also did editorial work, and Elsevier investigated the accusations and both concluded that he had manipulated citations). Elsevier has since amended its editor guidelines, editor contracts and reviewer guidelines to warn against the practice.
Fennell and Baas’s study found that, in most cases, reviewers’ own studies are not cited in work that they have assessed. Around 98.5% of the reviewers in the study’s sample are cited in fewer than 10% of the papers they reviewed.
But a small minority of reviewers — less than 1% of the nearly 55,000 examined — consistently seem to have their own work referenced in studies they have reviewed.
Because reviewers and authors tend to work in the same field, some overlap in citations is expected. But consistently seeing a reviewer’s work referenced in work they have assessed can be suspicious and could indicate instances of coercive citation. The practice is widely known to occur, but its extent is unknown, partly because the data underlying peer review usually remain confidential.
Elsevier is now approaching journal editors to ask whether the references in question are relevant. Fennell says that the company has completed its investigation into the most suspicious cases, but is still interrogating less-serious ones.
“Something needs to be done,” about citation manipulation, says Jonathan Wren, a bioinformatician at the Oklahoma Medical Research Foundation in Oklahoma City who is an associate editor for the journal Bioinformatics. Earlier this year, Bioinformatics banned a referee from reviewing for the journal after an investigation found that the researcher, whom Wren declined to name, had requested an average of 35 additional citations per review, 90% of which were for papers they had co-authored1.
Wren is currently writing an algorithm that would automatically flag unusual patterns in papers — including excessive citations to one particular author. “If we start policing it after it gets published,” he says, “then what do you do with the extra references?”
One idea that Elsevier is considering is the retraction of individual references in studies, a move that would be unprecedented. Another option, Fennell says, might be to issue corrections. “We’re still working out the best way forward,” she says.
Nature 573, 174 (2019)