In 2010, Ivan Oransky, a medical editor and physician, co-founded the nerdishly delightful website Retraction Watch. The plan was simple: Track and publish all the retractions announced at scientific journals, making fraudulent and erroneous studies easier to spot.
Since then, by methodically shining a light on research that turns out to be bogus, Retraction Watch has become the overlord of data showing the rise in the number of retractions in journals.
Today RW released its annual count of the number of retractions for fiscal year 2015, and the data is pretty amazing:
Once again, there were more retractions last year than the year before. To be precise, the number of retracted articles shot up by 37 percent —from 500 in 2014 to 684 in 2015. In the same period, there was only a 5 percent increase in citations indexed at the biomedical research database Medline.
I've spoken to Oransky about this trend in the past. He said research fraud may be more pervasive, but detection of it is also getting better. That's the likely explanation for the rise:
There are at least two good explanations. One answer is that we’re better at finding problems: There are simply more eyeballs on papers nowadays. Whether open access or not, there are more people looking at papers online. There is plagiarism detection software. There’s some evidence that fraud and misconduct may be on the rise as well, the increasing pressure on scientists, decreasing funding lines from the National Institutes of Health.
We can comfortably say it’s because we’re better at catching fraud, but whether it’s on the rise is still an open question, though directionally it does seem to be the case.
Some have also argued that the uptick in retractions is a good sign for science: More people are rooting out bad research and correcting the record.
For now, the Retraction Watch team has promised to dig through this year's pulled papers to see if they can find any new trends so they can get to the bottom of whether "last year is an outlier or a sign of a new plateau."
