Scientific Method

Retraction Watch blog tracks published papers as they get pulled

On May 20, 2015 a blog called Retraction Watch broke some national news: A widely publicized study on altering public opinion regarding gay marriage, published in December 2014 by the eminent journal Science, was about to be retracted. UCLA doctoral candidate Michael J. LaCour had claimed to find that openly gay door-to-door canvassers had been able to persuade a significant number of people to change their minds and support same-sex marriage, contrary to the prevailing wisdom that it is extremely difficult to influence anyone's personal beliefs. The paper received international attention, including a lengthy segment on the popular public radio program This American Life. A few months later, three researchers from Yale and the University of California, Berkeley, announced finding numerous irregularities in LaCour's study, including missing data, misrepresented data, misrepresented incentives to participants, and falsely identified sponsors. Eight days later, the paper was indeed pulled at the request of Donald P. Green, a political scientist at Columbia University who had not been involved in the research but who had helped LaCour prepare the article and been listed as a co-author.

This was not the first story broken by the blog, but it was the biggest; it attracted so much attention that Retraction Watch's server crashed. The website was founded in 2010 by journalists Adam Marcus, A&S '96 (MA), and Ivan Oransky. Their intent was to monitor how the scientific community deals with wrongdoing and errors in published research; the blog describes itself as "tracking retractions as a window into the scientific process." It publishes under the auspices of the nonprofit Center for Scientific Integrity, also founded by Marcus and Oransky, which describes its mission as "to promote transparency and integrity in science and scientific publishing, and to disseminate best practices and increase efficiency in science." In December 2014, the MacArthur Foundation granted the center $400,000 to further its work.

The Committee on Publication Ethics, an organization representing more than 9,000 editors of scholarly journals, states that papers should be retracted when there is clear evidence that the findings are unreliable or involve plagiarism or unethical research. Ferric Fang, editor-in-chief of the journal Infection and Immunity and a member of the Center for Scientific Integrity's board of directors, estimates that of the nearly 24 million articles listed in PubMed's database, only one out of every 6,300 has been withdrawn. A 2012 study by Fang, published in the Proceedings of the National Academy of Sciences, found that most articles listed by PubMed as retracted were pulled because of misconduct, not honest mistakes, and the leading form of misconduct was fraud (followed by duplicate publication and plagiarism).

Marcus, whose day job is managing editor of Gastroenterology and Endoscopy News (neither he nor Oransky takes a salary from Retrac­tion Watch), notes that though the percentage of published studies that are eventually retracted is extraordinarily low, the incidence of retractions has increased at a much higher rate than the increase in the volume of published papers. "The number of retractions rose tenfold while the number of papers only doubled between about 2003 and 2013," Marcus says. "It's not clear why." He believes some of the increase may be due to plagiarism detection software like CrossCheck, which has made it easier to spot pirated research. Journal editors are also paying greater attention to scientific misconduct and addressing it more transparently, he says, and he speculates that the pressure to further one's career by publishing striking results in top-tier journals may be a factor.

Sometimes a retraction is just part of the process. "You might say, 'Well, all the retractions and lack of reproducibility mean science doesn't work,'" says Marcus. "We sort of feel like the opposite is true. They're an indication that science is working as it should—people are correcting the record. Although things we might have believed to be true [turn out to be] not necessarily true, at least people are taking the time to find that out."