Research misconduct, rather than error, is the leading cause of retractions in scientific journals, with the problem especially pronounced in more prestigious publications, a comprehensive analysis has concluded.
The analysis, described on Monday in PNAS, the Proceedings of the National Academy of Sciences, challenges previous findings that attributed most retractions to mistakes or inadvertent failures in equipment or supplies.
The PNAS finding came from a comprehensive review of more than 2,000 published retractions, including detailed investigations into the public explanations given by the retracting authors and their journals.
The project was intended to explore the types of errors that typically lead to retractions, said one author of the PNAS paper, Arturo Casadevall, a professor of microbiology and immunology at the Albert Einstein College of Medicine.
“And what we got blown away by was the fact that the retraction notices are wrong, in a lot of the cases,” said Dr. Casadevall, who produced the study along with Ferric C. Fang, a professor of laboratory medicine and microbiology at the University of Washington.
Research misconduct was found more prevalent in articles published by leading journals, including Nature, Science, and Cell, and its unexpectedly high rate should be taken as yet another warning that universities and grant-writing agencies are relying far too heavily on publication rates as a measure of scientific performance, Dr. Casadevall and Dr. Fang said.
“Right now we’re incentivizing a lot of behavior that’s not actually constructive to science,” Dr. Fang said. ...
The result is that of the 2,047 retractions, 67 percent were attributable to misconduct, Dr. Casadevall, Dr. Fang, and Mr. Steen wrote. Only 21 percent of the retractions were attributable to error, they said. The cases of misconduct often involved leading scientific journals, they said, matching previous research that suggested a correlation between fraud and a journal’s impact factor, which is a measure of how often its articles are cited by subsequent articles.
Why would anyone think that someone dishonest enough to fabricate their data would be honest enough to admit it with their retraction? I guess that may have been the motivation for the PNAS paper.
Note: I've tried to make up CVM data for a classroom example and it is very difficult. So difficult that I gave up and used real data. I wonder why the fabricating researchers never thought of that?