How many scientific papers are not original?

Is plagiarism afflicting science? In PNAS, Citron and Ginsparg (1) count the number of authors who are submitting articles containing text already appearing elsewhere. They report disturbing numbers of authors resorting to copying, particularly in some countries where 15% of submissions are detected as containing duplicated material. I am on the editorial board of an Institute of Electrical and Electronic Engineers (IEEE) magazine, which also finds it useful to run all of the submissions through a plagiarism filter. What can be done about this? In 1830, Charles Babbage deplored unreliable science. He discussed hoaxes, forgeries, data trimming, and “cooking” (selecting data to match a theory) (2). Today, doubtful papers may be plagiarized, invented, or mistaken. This paper documents problems at one extreme: straightforward plagiarism within one publisher. More complex deceptions can be found at the site retractionwatch.com, which includes, among other examples, invented or fraudulent data. Mistaken research was highlighted in an important study by Begley and Ellis, who found that it was impossible to replicate 47 of 53 oncology studies that they attempted to repeat (3). At a time when important scientific questions are under attack, we need to improve confidence in our publications. How can we increase our level of trust in the scientific literature? In 2012, more than 2 million papers were published (4). They appear in publications ranging from highly competitive and prestigious journals such as Nature, Science, Lancet, and this journal, down to the predatory publishers listed in scholarlyoa.com who will print pretty much anything for a fee. University faculty, in particular, are encouraged to publish because the reward systems often depend on publication and citation counts as ways of evaluating merit. The h-index is the modern equivalent of the old saying “Deans can’t read, they can only count.” In some countries, having a paper accepted in a top journal can mean a cash bonus, with Zhejiang University offering a $30,000 payment to an author who publishes in Science or Nature (5, 6). Given the incentives, it is hardly surprising that some authors are attempting to exploit the system. This can be surprisingly easy. Delgado et al. (7) explain how they created a half-dozen fake papers, with several hundred citations. One of the authors saw his citation count go up by a factor of 2 and his h-index increased from 10 to 15. Fans of bicycle racing may smile on reading that the fake papers were attributed to Alberto Pantini-Contador. Refereeing, at least for some journals, is pretty shaky. As cited by Citron and Ginsparg, Bohannon (8) submitted a fake article to more than 300 open access journals, and more than half accepted it. Following up, he found that one of these journals had plagiarized its own description from a