good practices

The specter of duplicate images

Issue # 245
jul 2016

Daniel Bueno

Microbiologist Elisabeth Bik, a researcher at Stanford University Medical School in the United States, spent two years analyzing images of Western blot tests—a method used in molecular biology to identify proteins—that appeared in 20,621 papers in the biomedical field published in 40 scientific journals between 1995 and 2014. She found a multitude of problems: there were duplicate images in 782 papers, or 3.8% of the total. In 230 articles, the images simply appeared twice in the same manuscript to illustrate different experiments—a sign they may have been duplicated in error rather than in bad faith, according to Bik. In the other cases, there are indications of deliberate manipulation. In 356 articles, some duplicate images were repositioned or inverted. And in 196 papers, there was clear evidence of fraud, such as pasting the same “band” from a Western blot test onto several images, as if it were rubber-stamped with a repeating pattern.

The results of the analysis were made available in an article in bioRxiv, a repository for manuscripts in the field of life sciences that have not been published in a journal. Bik used an artisanal approach to her research. First, she looked through each article for signs of duplication. When she came across a questionable situation, she loaded the images into a software program that can adjust the contrast in search of evidence of manipulation. Whenever a problematic image appeared, two researchers working with Bik—Arturo Casadevall and Eric Fang—reviewed her findings . At the time of the analysis, none of the problematic articles had been retracted. Bik made the decision to notify the editors of the publications about what she found. More than 700 reports were sent to journals. She also wrote to 10 institutions where there were recurring problems—in situations where at least three papers from the same research group contained duplicate images. The result of this initiative, she says, was six articles retracted and another 60 corrected.

The study shows that the more-cautious publications manage to prevent the publication of papers that contain adulterations. The Journal of Cell Biology, which since 2002 has scanned images from submitted papers to search for duplications, had only 0.3% of its articles questioned. But in the International Journal of Oncology, there were problems in 12% of the papers. Three countries stood out in terms of their numbers of suspicious articles. Papers from researchers in India were 1.93 times more likely to contain duplicate images than would be expected from the frequency of publication. Next on the list was China, with a 1.89 greater probability, and Taiwan, with 1.20.