Researchers and journals need to do more to counter inappropriately manipulated figures in science.
Seven years ago, a cover of The Economist showed Barack Obama, head down on a Louisiana beach in front of an oil rig — the picture of lonely despair. The image perfectly encapsulated the news magazine’s story about the massive pollution caused when BP’s Deepwater Horizon platform exploded, and what the president of the United States could possibly hope to do about it.
But Obama was not alone when the picture was taken. His companions had been digitally erased — and so The Economist was criticized by other news media for altering the image to support an editorial narrative. Many news outlets have followed a code of conduct about image manipulation almost since the era of digital photography began: photos can be cropped, but may not be doctored unless the manipulation is flagged.
Science has in some ways been slower to come to this conclusion, but pressure is building. Websites such as PubPeer and Retraction Watch regularly expose manipulated images — from pictures of electrophoresis gels and western blots to micrographs of cells. The bands in a gel may be duplicated by cutting and pasting, for example, or one cell type might be digitally removed. Studies suggest that up to one in five published papers in the life sciences includes one or more manipulated images.
The latest high-profile case concerns the research group of cell biologist Karl Lenhard Rudolph at the Leibniz Institute on Aging in Jena, Germany. An independent investigation by the Leibniz Association, which runs 91 science institutions in Germany, identified 8 high-impact papers with manipulated images. On 15 June, the association issued a statement formally reprimanding Rudolph for failure to appropriately supervise his research group, and calling on him to publish errata and, in one case, a retraction. It also issued sanctions that include shutting him out of the association’s competitive funding for three years.
The Leibniz Association is to be commended for enforcing standards of technical integrity as a duty of research-group leadership. But is the life-sciences community doing enough to reduce the sheer extent of image manipulation? Few manipulations arise because a researcher wants to cheat. Many scientists simply want to display a key point more clearly by improving contrast. But too often they are taking this type of beautification too far.
Such is the perceived danger of image manipulation that some journals, including the Journal of Cell Biology (a pioneer in this effort) and those published by Cell Press, check images in all the papers they accept for publication. But many journals do no checks at all. (Nature, currently reviewing its image-checking procedures, runs random spot checks on images and other checks as specified by editors. It also requires authors to submit unedited gel images for reference.)
Studies suggest that up to one in five papers in the life sciences includes manipulated images.
The EMBO Journal checks all images and finds manipulation in 20% of papers accepted for publication — a number that remains stubbornly high despite the journal’s openness about its policy.
How can this widespread practice be stopped at source? The Leibniz Association requires that the scientists it hires agree in writing to adhere to good scientific practice, and the institutes under its umbrella regularly organize scientific-integrity seminars. Such measures are increasingly being adopted by many universities and research institutions around the world, but are evidently insufficient.
Primary responsibility for image integrity lies with principal investigators, who need to be aware of its importance and ensure that the young scientists in their teams, who came of age in the digital era, wield Photoshop tools appropriately.
But that can be time-consuming. Most approaches are still manual, requiring each figure in a paper to be pulled out individually and run through a series of programs to check whether blots, bands, cells or other debris are where they are supposed to be. In the simplest cases, a figure may be fed into Photoshop and filters adjusted so that manipulations are easier to spot by eye.
Researchers and companies are developing algorithms to make detection more sensitive and reliable. Some companies are trying to automate the process so that the computer extracts images from papers and then runs checks sequentially. This promises publishers and large institutions options for outsourcing image analysis. One research-integrity company already offers an automated process that costs €10–15 (US$11–17) per paper, depending on the volume of papers checked — a tiny proportion of the overall costs of publishing a paper.
By both human and technological means, research organizations, researchers and journals need to do more to counter the image-manipulation challenge.
Related links in Nature Research
Related external links
About this article
Cite this article
Image doctoring must be halted. Nature 546, 575 (2017). https://doi.org/10.1038/546575a