What should I do when images in a publication appear to have been faked?

Question

I just came across a paper published in a journal (IF < 2), which has used a couple of images without mentioning the source. This itself is not necessarily a problem. But they seem to be from a commercial software product I'm familiar with (they haven't mentioned even the name of the software, though it generated the image), and based on my experience with this software, the image looks to me like it has been tampered with.

Specifically, it is an image where the software predicts properties of a compound which is used to verify the result. Now, let's say some parameter have a cutoff at 0.5% and they are getting 10.3%. To conform to their result they removed the '1', and it become '0.3%'.

Emails to the corresponding author came back empty.

Should I report that to the journal or leave it as it is (may ruin someone's career)?

They have used the image to prove the result of the experiment.

Answer 1

This could be a big deal and something you should report and, alternatively, it might not be a big deal or reportable. The difference comes in the context.

Are the editted images presented as legitimate data or results that are the raw output of the experiments/research? By which do they say something like "In image blah taken by an electron microscope you can see that the magical unicorn bonds have been created by our process." (Keeping in mind that I know nothing about bio-chem and 'magical unicorn bonds' is a stand in for some actual process). Or in CS something like "Here you can see the robot we built" Statements like this imply or outright state that the object or information in the image is not just representative data but actual results or output. Data like this should not be manipulated or edited except for clarity(circling a targeted area or adding minor labeling).

The other kind of image is a bit rougher. These images can demonstrate what was expected to be seen, abstract output from the research, conceptual information. These kinds of images often are edited or manipulated. Sometimes as a demonstration of what was expected("We would expect magical unicorn bonds to appear after our procedure but instead....") or as an explanation of something more abstract("the robot should follow the optimal path as shown here when it uses the stairs instead of running repeatedly into a wall"). These are things that are no reportable. They can be in poor taste and they absolutely should be caught by reviewers if they imply results beyond the scope of the actual research. But, in some fields, these are the best way to demonstrate expectations, abstract information or background information.

All that being said - when you say the image is from a "commercial software" that makes me wonder if you mean not that it was created with "commercial software" but is actually an image from some commercial source. In this case the image may be copyrighted and it may not be appropriate, at all, to use in this research. This will depend on the image, the source and the 'tampering'. As a counter example to this in the realm of computer vision every uses the standford bunny model in their publication. It's a thing. This is not inappropriate. Someone using an image from a text book, however, or a Google search that they do not own is inappropriate and should be first reported to the PI of the paper and, potentially, the publisher if no action is taken.

Answer 2

Manual alteration of figures aimed at deceiving readers (by fabricating, obfuscating, 'cherry picking' results, etc.) is a serious matter. It surprises me that people still try because it's very often quite easy to spot (other types of data fabrication are harder to catch). But it's sadly not uncommon, even in highly regarded journals and from researchers from reputable institutions.

For life sciences, according to one of pubpeer's moderators in a comment: 'Most of the life science reports involve image manipulation - a good majority are gels, with a bunch of duplicated specimen images as well. […] we see sometimes on PubPeer […] things like doctored NMR spectra in chemistry.' The latest seems to be related to your observation, although it's outside my area of expertise.

You can get a sense of the type of things that are reported on PubPeer reading this thread among many others. As an example, a close examination of this figure shows the use of copy paste to fabricate data:

Or this one, initially published, not in your average pay-for-publish shady 'open access' journal, but in Nature, that has a rather obvious copy-pasted middle panel:

(both the articles where retracted).

In your position, your first reaction is totally appropriate. Here is what I think is the best course of action:

1. Contact the author(s) it is a good way of showing that you are concerned but not necessarily interested in public shaming. If the authors do not react, then:
2. Contact the publisher (mentioning that you already contacted the authors to no avail) as suggested by @alarge in a comment. If it's a reputable publisher, the issue will be taken very seriously.
3. If all fails, you are left with public reporting of the issue, anonymously or not, via social media or the website listed above. Note that you are always at risk of putting yourself in trouble when reporting misconduct, the same as in any other field, so weight this risk if you intend to associate your name with the complaint.

EDIT:

Should I report that to the journal or leave it as it is (may ruin someone's career)?

I think that, as a scientist, you have a responsibility to report that sort of misconduct when you see it. What may ruin someone's career is their sloppy ethics, not your concern for integrity.

Answer 3

it may not be as bad as it seems. I often have to edit images from commercial software packages because the text is too small (or blurred) to read in the print version. Usually it's the axes that I have to fix because many packages make the fonts too small. The authors in this case may have simply overwritten with the same values, just in a larger text or (different/emphasizing font or color).

Answer 4

First you should discuss the matter with a trusted colleague or two to check that they agree with you. This is a heavy accusation and before doing anything you should check that you haven't missed anything and that there aren't other plausible explanations. You may very well be right, but it's also easy for one person to make an error without outside feedback.