I keep on answering the same types of questions on Twitter, so here is a handy Frequently Asked Questions page where I answer the most common ones. Updated January 2023.
Why don’t you give the DOI of the paper featured in your #ImageForensics challenges?
Q: Why do you not post a link to the papers that contain the images with duplications that you post on Twitter? I would like to know which paper these photos were published in.
A: I usually do not post links to those papers for several reasons. On Twitter, that might quickly evolve into making fun of the authors or institutions. Twitter being Twitter means that this could turn nasty very quickly. These types of posts are about raising awareness of the types of problems present in the scientific literature, not making fun of the authors. Some exceptions might apply.
Do you find these cases by eye?
Q: How do you find those cases of image duplications that you write about or post on Twitter? Do you use software or find them by eye?
A: In the past, I found these by just looking at them – just using my eyes. Until around 2020, I just flipped through PDFs and looked at the images. I do use Mac’s Preview to make an image darker or lighter to bring out some details, if needed. Nowadays, I use ImageTwin, and less frequently FotoForensics, Forensically, or FigCheck to confirm my suspicions, or to analyze complex figures with lots of panels. Mac’s Preview is the program I use to “decorate” the image with colored boxes or arrows to point out the areas that look similar to each other.
Isn’t there any software that can do this?
Q: Someone should develop software to find these things. Software is going to be much better than the human eye.
A: I agree, and several groups are working hard on that, and several tools are now available. I have shared a dataset of 400 papers with image problems and 800 matched controls in which I did not find any problems with several teams that are developing such software. I have not heard back from most of them, so I assume this is much harder than most people assume.
Current available tools are:
- ImageTwin developed by Markus Zlabinger (subscription required), very good in overlapping microscopy photos, good in finding repeats within a photo, and reasonable finding duplicated Western blots. Handles one PDF or ~20 screenshots at the time, and has a database of open source papers to compare images to.
- Sherloq, by Guido Bartoli (requires some Linux skills to install, but free).
- FotoForensics – Free. I don’t use this very often, does not seem to work well for me.
- Forensically – Free. Will detect direct pixel-level clones within one image, but cannot detect rotations or resized copies.
- FigCheck – free to use, limited to 10 images per day
- Daniel Acuna from Syracuse University has written authored this promising study (bioRxiv, 2018).
I get variations of this question so often, that I have turned this into a Twitter drinking game. So if I reply on Twitter to you with “Drink!”- this is why. Don’t worry, I drink mostly tea.
You should write a book about this.
Q: Why don’t you write a book about this? I would buy it.
A: That would cost me months, and I would feel it slows me down too much. For now, I would rather focus on finding and reporting more of these cases, and on educating peer reviewers and editors on how to find errors and misconduct before papers get published. But if you want to help me write a book, please let me know!
Can you check a paper for me?
Q: I suspect there is image duplication in a particular paper or by a particular group. Could you please check?
A: Absolutely. Send me an email at elies bik at gmail period com. It helps if you send me the PDF, and not let me guess which paper or author you mean. I might not find anything, but if I do, I might post this on PubPeer. I will honor any request for anonymity, and not share your name, unless you want me to. Please do not send long lists of papers. I get a lot of these requests each day and won’t always have time to answer or follow up on them.
Are you paid to do the work you are doing?
Q: Who pays you to find these cases of image duplication? Are you paid by Big Pharma/ Big Publishing / Big University?
A: I occasionally will do some consulting work for scientific publishers or universities. In all cases so far, these “jobs’ involved looking at small sets of papers from a particular group of authors where the journal or institution needed a second opinion. I will not take on consulting requests on papers that I have reported on myself (e.g., on PubPeer), since that would not count as a second opinion.
Since I am not employed, I also ask for an honorarium for my time to give lectures at universities or other institutions.
But most of the time I work off of tips I receive from other scientists, or on following leads based on my previous work. All of the papers I talk about on Twitter, on this blog, ScienceIntegrityDigest.com, or on PubPeer are papers I have worked on on a voluntary basis – unpaid.
You can find more on my financial disclosures and conflict of interest here.
You are not an expert. You are not a scientist.
Q: I really don’t see these duplications. I do not believe you can find these cases of image duplications. All gels / western blots look similar, so how many false positives do you call? Did you write any scientific papers about science misconduct? Since you do not work at a university, you are not a scientist.
A: Well, after scanning 20,000 papers for our 2016 study, and another 80,000 or so after that, I can tell you that most images indeed are unique. But I also developed a good feeling for which ones are not. Together with other authors, I wrote three scientific papers about this work. Here they are:
- In our 2016 mBio study, my co-authors Ferric Fang and Arturo Casadevall and I describe a dataset of 20,000 biomedical papers with photographic images. I scanned all of these and found about 800 of them (4%) to contain inappropriately duplicated images. The paper describes the types of duplications (simple, shifted, and manipulated) and some stats related to journal impact factor and country of affiliations. We hypothesized that about half of the problems identified by us might be caused by an intention-to-mislead.
- In a 2019 follow-up paper, we looked at the risk factors associated with these three types of duplications, and found that academic culture and monetary incentives were among the factors that played a big role.
- We also focused on one particular journal, Molecular and Cellular Biology, and looked at a set of 960 papers published in MCB. This study was published in 2018. About 6% of these contained duplicated images, and we followed those papers to see what the reasons were behind that. About 10% of these were retracted.
I have a PhD in Microbiology and worked 15 years as a postdoc / staff scientist at Stanford University’s School of Medicine, and 2 more years in Industry. I am currently a science consultant, but still very much a scientist.
“She has never authored a first author paper containing a single western blot” – https://archive.md/35E3H
Wrong. I know it is an older paper, but here it is: Bik et al., “Genetic organization and functional analysis of the otn DNA essential for cell-wall polysaccharide synthesis in Vibrio cholerae O139“, Molecular Microbiology (1996).
“She has worked as a research associate (aka the lowest level)”
Yes, I worked 15 years at Stanford, and most of those years as a Research Associate. Although some universities or companies use this term for technicians or research assistants, at Stanford University, that was an academic staff research position, usually for people holding a PhD. In Summer 2016, the designation of Research Associate was changed to Research Scientist, Research Engineer, and Research Scholar.