Fiona Fidler, a metaresearcher at the University of Melbourne, was outraged. She had discovered that her appraisal of a submitted paper had been changed before being sent to the author, sometimes drastically. The words “very sympathetic” had become “generally sympathetic.” “This one is a good example” ended up as “this one still needs work.” Worst of all, she felt that the bottom line of her peer-review report to the journal Educational and Psychological Measurement, recommending that it accept the paper with minor revisions, was misrepresented in the editor’s rejection letter to the author.
“I had never experienced anything like this before,” Fidler says about the 2012 incident. She demanded explanations from the journal editor. And she later partnered with the snubbed paper author, Rink Hoekstra, a psychologist at the University of Groningen, to find out how widespread this practice was.
With colleagues, they’ve now surveyed 322 editors at high-impact journals across ecology, economics, medicine, physics, and psychology on when they think altering peer-review reports is justified. Published as a preprint earlier this year at the Open Science Framework and now under review at eLife, the survey reports that 91% of the respondents identified at least one situation in which they would edit a report. More than 80% said they would do so if a reviewer used offensive language or made inappropriate personal comments about the authors. But 8% said they would change the reviewer’s overall recommendation—even without their permission, a finding that shocked Hoekstra. He believes that’s probably an underestimate, given the stigma of admitting to such dubious behavior in a survey. “I think there are probably even more who actually do it.”
The survey comes amid a wider push for consistent guidelines. There’s little dispute that editors should intervene when peer-review comments are hostile. Such comments have a disproportionate impact on minorities and other marginalized groups, inducing self-doubt and harming productivity, according to a 2019 PeerJ paper. It’s not just ethical to edit such review reports—it’s essential, says Seth Leopold, editor-in-chief of Clinical Orthopaedics and Related Research.
Jane Alfred, director of Catalyst Editorial, which offers training on research integrity and publication ethics, thinks it’s best to return reports containing hostile language and ask the reviewer to change it. Reviewers are often grateful for the chance to revise ill-considered comments, she says. But Leopold says this may be impractical at fast-paced journals, and trying to educate reviewers who make unprofessional comments is likely futile. It’s better, he says, not to ask that person to review again.
Changing a reviewer’s recommendation is another matter, says Howard Browman, a council member of the nonprofit Committee on Publication Ethics, which is developing guidance on how to navigate the ethics of editing reviewers’ comments. “It’s so obviously something you wouldn’t do,” he says.
Yet it apparently happened to Fidler, who stumbled on the changes to her report by accident. Although the review process was double-blinded, she recognized the paper as Hoekstra’s because she had seen him present the work, on Ph.D. students’ statistical reasoning, at a conference in Slovenia. So after sending in her review, she emailed him to congratulate him and tell him that her review was nitpicky but positive.
That same day, Hoekstra’s paper was rejected by the journal. “The reviewers have spoken in nearly a single voice in their recommendation to me that I decline publication of the paper in its current form,” wrote the editor-in-chief, George Marcoulides, a research methodologist at the University of California, Santa Barbara. Although editors may override reviewer recommendations, the normal practice is to explain this in the decision letter. Hoekstra wrote back to Fidler, attaching the two reviews he got from the journal and asking for suggestions of other journals that might publish the work.
When she saw the reviews, Fidler cottoned on. “I think we have a controversy on our hands,” she wrote to Hoekstra. She downloaded her review from the online journal system, and compared it, word by word, with the review Hoekstra received. Her sentence saying her concerns were minor had been deleted. Other sentences criticizing Hoekstra’s methods had been added. She emailed Marcoulides, who replied that the journal computer system sometimes blended and distorted reviewer comments.
Fidler calls that explanation “preposterous,” saying it would be impossible for a technical glitch to create “perfectly grammatical sentences that are exactly opposite in meaning.” She reached out to contacts at SAGE, the journal’s publisher, who said they would investigate, but she heard no more from them. In an email to Science, a SAGE spokesperson wrote that the publisher “addressed the issue directly with the editor at the time” and that Marcoulides now double-checks reviewer comments for consistency.
Later in 2012, Marcoulides invited Hoekstra to resubmit his paper. This time, Fidler’s review made it through intact, and Hoekstra’s paper was accepted—pivotal for him winning a permanent job at Groningen. In an email to Science, Marcoulides wrote that the reviewer comments automatically attached to his decision letter were “distorted” and his edits were intended to clarify his interpretation of Fidler’s assessment. “In hindsight, I should have contacted her rather than attempting to resolve the problem on my own,” he wrote. He added that he still sometimes edits reports for clarity or to remove inappropriate language.
Few journals offer explicit guidance on when editing peer-review reports is and isn’t permissible. Alfred says they ought to, and should also allow reviewers to opt out from being edited. No matter how well-intentioned editors may be, she says, clear policies will ensure a transparent and unbiased process. Many journals have a safeguard: They share all reviews and the editorial decision with reviewers, allowing them to see how their comments were communicated to the authors. But about 20% of the editors in the survey report that their journals do not send out either the reports or the decision letters to reviewers.
Simine Vazire, editor-in-chief of Collabra: Psychology and a colleague of Fidler’s, argues for a bright line on the question: no edits without reviewer permission. Her journal has no policy on the issue, but she is considering proposing one. Without clear boundaries, she says, it becomes easy to rationalize changes, adding that journal editors have lots of power and little accountability. “There’s no one watching over editors,” she says. “I think it’s especially important that they have really hard and fast rules for themselves.”