by The Proceedings of the Royal Society B: Biological Sciences says it won’t withdraw a paper on clownfish behavior, even though a lengthy university investigation found it was fabricated.
An independent investigative panel at the University of Delaware (UD) concluded in a draft report last year that “inconsistencies and issues” with the 2016 study “constituted fraud.” But the magazine said in an editor’s note on February 1 that its own investigation had not found sufficient evidence of fraud, in part because a correction by the authors had solved the paper’s main problem.
Deakin University fish physiologist Timothy Clark, part of an international group of whistleblowers who found problems with the paper, called the decision “annoying”.
The paper, authored by marine ecologists Danielle Dixson of UD and Anna Scott of Southern Cross University in Australia, is one of 22 studies published between 2008 and 2018 that Clark and his fellow whistleblowers claim are fraudulent. The complaint specifically focused on Dixson and Philip Munday, Dixson’s Ph.D. Supervisor at James Cook University in Australia. Both have denied wrongdoing.
An independent UD panel examining Dixson’s work was “impressed by a serial pattern of sloppiness, poor record keeping, copying and pasting into spreadsheets, [and] Errors in many papers examined,” said a heavily redacted draft report obtained from Science. It also concluded that several articles involved scientific misconduct. UD said it asked journals to retract three articles.
For one of them, published in Science In 2016, Dixson did not have enough time to perform the huge number of experiments described in the paper, the panel wrote, and an Excel file that allegedly contained the study’s raw data contained more than 100 unexplained duplicates, showing that it couldn’t be real. Science withdrew the paper in August 2022. (ScienceThe ‘s news team is independent of its editorial board.)
Very big discrepancy
The Procedure B Paper suffered from similar timeline issues, according to the panel. The paper’s conclusion – that anemonefish can “smell” whether coral reefs are bleached or healthy – was based on a series of experiments in which fish were placed in a laboratory device called a “choice flume,” which forces them to choose which way to go they swim.
Dixson collected the data for the study, which the draft report included about 1,800 individual trials, each 9 minutes in length. If it had used a single channel, the trials would have taken 22 12-hour days to complete, “through continuous work without breaks or prep work, recalibration, cleaning, bucket changes, etc.,” the committee wrote. (Meanwhile, Dixson would also have had to carry 4,000 gallons (1,800 liters) of seawater in and out of the experiment room.) But the newspaper said the experiments ran from October 12 to October 24, 2014, a period of just 13 days.
Although the university requested a revocation last year, it did not share the committee’s report Procedure B, writes the magazine’s editor-in-chief, Spencer Barrett, in an email. “I wanted to know, of course, what evidence they have for this request,” says Barrett. “I tried twice in August and October [2022] and in both cases the report was said to be confidential.”
Instead, the magazine had three editors investigate the case over 6 months with the help of independent experts, resulting in a 59-page report, Barrett writes. Meanwhile, Scott and Dixson released a correction to the paper in July 2022, saying that the experiments actually took place over 33 days between October 5 and November 7, 2014, and clarifying that Dixson used two flumes simultaneously, thereby doubling the observation time. (Dixson reported using two flumes in other studies as well, although the UD panel “was at a loss to understand” how she could record fish behavior in two experiments simultaneously every 5 seconds.)
One of the whistleblowers, Josefin Sundin of the Swedish University of Agricultural Sciences, says the journal appears to have been too gullible in making the correction. “Why would someone run an experiment for 33 days but accidentally write the methods and data as if it were run for 12 days?” she asks. “That’s a very big discrepancy.”
says beret Procedure B Investigators have not investigated the veracity of the new time slot. “I agree that changing the timeline and the number of channels used was odd, but we accepted that as part of the fix,” he says.
Along with the correction, Dixson and Scott also uploaded the raw data for the study, which was missing, although the paper said it was available online. This dataset “raised a second set of problems,” according to the editor’s note. This appears to relate to an analysis of the Excel file by the whistleblowers, which shows that it suffered from some of the same problems as those for the Science Paper, including duplicating data across columns and numbers that don’t add up properly.
However, the journal’s investigation found that there are other possible explanations for suspicious patterns and that some problems with the data “are more likely to be the result of errors or poor data maintenance and correcting them would not change the conclusions,” the note said.
It is “stunning” that the Journal’s investigation looked at the newspaper in isolation “when it suffers from the same problems as the one now withdrawn Science Paper,” says Dominique Roche of Carleton University, another member of the whistleblower group. But Barrett says the journal’s process is “to examine the specific articles that we publish ourselves, and not articles in other journals. Each article is treated independently.”
Dixson and her attorney did not respond to a request for comment on the editor’s note. According to a notice of absence, Scott is currently on vacation.
UD has not publicly identified the third paper to be withdrawn, but the panel’s draft report suggests it is a paper published by Munday, Dixson and others nature climate change in 2014. The report noted that its data file was identical to that in the Science Paper, had “serious problems with the data sheet files, with patterns of copying and pasting data sheets in both signatures of manufacturing and falsification of data.” It is unclear if this paper is currently under investigation nature climate change.
