Main

Anyone following the media cannot have escaped the seemingly endless reports of scientific misconduct. Two high-profile cases, with former Massachusetts Institute of Technology immunologist Luk Van Parijs and the now-discredited South Korean stem cell researcher Hwang Woo Suk, involved image manipulation. In a more recent case, Ban-Yang Chang and colleagues inappropriately processed original data affecting several figures in their paper that appeared 20 October, 2006 in Cell.

Although cases of deliberate fraud are thankfully more the exception than the rule, innocent attempts to make images 'cleaner' can in fact amount to misrepresentation and are more common than previously thought. Particularly eye-opening are the statistics released by the Journal of Cell Biology, which show that 25% of accepted papers contain at least one inappropriately altered figure for which an adequate replacement figure could be supplied by the authors when contacted. Such highly publicized incidents and harrowing statistics have stimulated discussions among researchers, journal editors and institutions about acceptable standards for image manipulation and who should be held accountable for monitoring this situation.

Whether image manipulation is truly on the increase or whether the scientific community has simply been made more aware of it than before is unclear. Nevertheless, the digital era has certainly made image manipulation easier. Before imaging software such as Photoshop was available, adjusting image data in the darkroom would have required concerted effort and skill. Thus, data acquired at the bench were almost identical to the data published, blemishes and all. The advent of Photoshop and other imaging software has changed the landscape considerably, making it very easy, and perhaps tempting, to modify images. To a degree, this stems from a 'generation gap' between older scientists, who do not necessarily understand the possibilities of imaging software, and younger scientists, who acquire the original data and prepare the images for publication.

What constitutes unacceptable data manipulation? Moving, adding, removing, enhancing or obscuring features or sections of an image clearly count. 'Cleaning up' background or removing 'nonspecific' bands from a gel may seem innocent enough, but such changes may in fact alter useful information. To help ensure that all data figures are truly representative of the research, Nature Publishing Group (NPG) and other scientific publishers such as Rockefeller University Press and the American Association for the Advancement of Science have instituted new guidelines in an attempt to eliminate inappropriate image manipulations.

First, all NPG journals encourage author contribution statements at the end of the paper (http://www.nature.com/authors/editorial_policies/authorship.html). Such information helps to clarify the accountability and responsibilities of all authors. Although this policy is voluntary at present, we at Nature Immunology are pleased that most authors choose to supply such statements.

Second, with the help of technical experts, the NPG family of journals has developed new guidelines explicitly outlining which image manipulations are appropriate and which are not (http://www.nature.com/authors/editorial_policies/availability.html). No manipulations that selectively affect only a portion of an image, remove information or add data obtained in a different experiment can be made. Authors should detail as supplementary information the instrument settings and software manipulations used to generate the images. In short, images submitted for peer review should be minimally processed. Manipulations that blatantly misrepresent data, such as the elimination of 'background noise', are obviously not acceptable without acknowledgement. The editors strongly encourage authors to thoroughly review the guidelines and will be happy to answer any queries regarding the new policy.

In addition, Nature Immunology will soon include a statement in our checklist, supplied on conditional acceptance of the paper, stating that figures provided for publication accurately represent the original data. This must be signed by the corresponding author before acceptance of the paper.

Finally, and importantly, the Nature family of journals is training our production editors to detect 'tell-tale' signs of image manipulation. We now do 'spot checks' of all images from one paper chosen at random per issue. We hope that this may act as a deterrent, although similar checks scrutinizing every paper accepted for publication in the Journal of Cell Biology have, somewhat disappointingly, not lead to a reduction in the frequency of cases.

NPG will continue to investigate the best ways to improve the screening process. Computer scientists are developing algorithms that can detect specific types of image manipulation. These will likely be packaged with Photoshop and other image-processing software in the near future. In this way, not only editors but also referees and laboratory heads could review images more easily. Nature Immunology will be keeping a close eye on the development of such software.

It is important to remember that responsibility for the images in a paper lies with the authors of the paper, particularly the corresponding author. Journal editors do not wish to function as 'data police', because to fulfill such a role could contribute to an adversarial atmosphere. Instead, we want to aid the publication of high-quality science. This has necessitated some policy changes to keep up with the changing world. We hope that our policies will help guide the scientific community and promote discussion of the issues at hand.