Friday, May 19, 2006

Possible fraud....

In the course of serving on a committee for a graduate oral presentation, I noticed something very strange looking in a Phys Rev Letter from a few years ago. While unlikely to be seen in a casual glance at the printed version of the journal, it was very striking when the figures were blown up to 4' on a side by a digital projector. Basically, it looks like someone used what I shall delicately term the "Photoshop operator" to massage their data. This paper has been cited 65 times since its publication, and has been milked heavily by its authors.

So, what is the right course of action? I've got no actual proof of fraud, just a very suspicious figure. I've now emailed the editors at PRL twice about this, and received no response from any human being - just the form letter generated by their mail system. Since this is circumstantial, I'm certainly not going to accuse anyone publicly. Next I'm going to call PRL on the phone. Updates as events warrant.


Peter Armitage said...

Hey Doug. Enjoying your blog which came referenced on CosmicVariance some months back.

I have often thought after the whole 'Schon Affair' about what would have happened had he not been so manic and published only 1/4 the number of papers he did. He still probably woulda have gotten the Max Planck directorship and still be a respected household name (among err... select households).

This then begats the question with how often such fraud happens on a much reduced scale, with say one or two doctored figures every once in a while. I'd be willing to bet that it is very very common.

What to do? I don't know. The system unfortunately cannot protect against such fraud.

Douglas Natelson said...

Hey Pete. I was going to ask how JHU was, but after googling you, I now realize you're not there yet. I hope Geneva is treating you well.

I agree with what you said. I worry that cheating does happen at a limited level a fair bit. There also do seem to be some cultural issues about plagiarism in certain countries that are troubling. It's very tiresome to remind foreign grad students every year in my courses that copy-and-paste is intellectual theft, not an homage.

I'm pretty disappointed at the complete lack of response from PRL. Either they don't care (unlikely), or they're so massively understaffed that they just ignore emails that don't have a manuscript number in the subject line.

Anonymous said...

Hi Peter,

Regarding your comment, why do you to assume that those people were cheating when they used photoshop.
Does the graph shows a very important quantitative result or just a trend (qualitative result) ?

In my particular case, I have used photoshop to alter some of the pics of my papers because, it would take long time to re-plot the data (sometimes days), when the suggested changed by the reviewer was for instance, to change the label font or something like that.
I once had to re-plot everything because , I once sed in the axis labels, teh comma instead of decimal points, so that I changed everything by hand using photoshop... Do you think I was cheating, too?

Douglas Natelson said...

Regarding the comment above.... Obviously using photoshop to change fonts or symbols in a graph is innocuous. The paper I'm talking about is an STM experiment. When your data is essentially an image of a surface, you can't just go in and use the "clone" tool to rearrange surface features. That would be like using photoshop to actually move data points on a plot. That would be scientific misconduct by any reasonable definition.

Douglas Natelson said...

One more remark about the anonymous comment above. Moving data points is inappropriate even if the plot only shows a qualitative trend that isn't altered by the move. Points on plots are presumed to correspond to either experimental results or the output of a calculation. If their position is arbitrarily determined by author whim or aesthetics, that's misconduct.