Plagiarism, Self‐Plagiarism, and Text Recycling

Abstract

When a manuscript is submitted to Headache it is subjected to a number of tests and evaluations.1 Such processes are needed to determine both the veracity and validity of every submission. One such test, which sadly is very necessary, is the running of every full-length article through software called CrossCheck (powered by iThenticate) to determine the extent of potential text overlap with previously published articles. At its most extreme, the editorial office is looking for outright text plagiarism (as opposed to the passing off of original ideas as one's own, which I am sure with Artificial Intelligence will one day facilitate detection, but is currently, despite our obvious desire to be watchful, almost impossible to ascertain).2, 3 Equally, the editorial office is also attempting to detect overlap that has not been properly sourced or passages of text that should instead be presented inside quotation marks. As with many ethical issues, there is a sliding scale of “ethicalness” when we are confronted with problem manuscripts.4 Careful judgment is required as to the correct course of action. This brief editorial aims to provide transparency for readers and authors alike regarding Headache's protocols for checking manuscripts and also intends to educate all stakeholders on good practice. Plagiarism is a particularly egregious ethical transgression and one that is probably easy for everyone to agree upon as problematic: the appropriation of another's ideas as one's own. Perhaps less obvious, but no less problematic, is when wholesale passages of a previously published text are lifted, without citation. Other than the misappropriation of credit for those words and all the obvious implications associated with that, it also leads to unnecessary duplication in the literature and the undermining of a reader's faith in the tenacity of any given journal's peer review process to effectively validate papers. Along with fabrication, it can be academic deception at its absolute worst. On the sliding scale, for many, self-plagiarism does not seem to rate so badly. This is presumably based on a notion that the words used are theirs already and the only person being “cheated” is the author themselves. Even the World Association of Medical Editors prevaricates on the issue and simply relays what authors already know: there is no consensus on whether, or by how much, self-plagiarism constitutes scientific misconduct.5 While that is not necessarily untrue there are other issues to consider. One is the not insignificant issue that those words are usually copyrighted and if they are lifted from a book or journal, the copyright holder is not necessarily the author, but the publisher or the society-owner, such as the American Headache Society for Headache. Therefore, excessive self-plagiarism could become a legal issue of copyright violation. Journals that detect this in a submission will be very careful to avoid implicating themselves in a lawsuit and will take action, ranging from asking for rewrites to outright rejecting the paper if the author completely failed to cite their previous work or there is evidence an author willfully deceived a journal. Plagiarism and self-plagiarism are evidently problematic, but they are not the only concerns. There is also the issue of simple text overlap. Often the offending text is a generic or standardized description of a headache type, prevalence rate or statistical/methodological design. Judgments on the unethicalness of such “borrowing” really then come down to the context of (re)use. Generally, overlap detected in the methods section is more tolerable. There really may be a limited of number of ways to describe a study design. The paper may well be a second or third report from a trial and so elements of text describing the trial can probably be acceptably recycled. Rather than the odd phrase, it is complete passages of text, certainly more than a couple of sentences that raise the alarm. A common anecdotal perception is that non-native English speakers are more prone to lifting phrases, presumably because this relieves the burden of writing in another language somewhat. While it is true that Headache has certainly seen cases of this, we have also unearthed poor practice from native-English speakers in the United States. In all but one circumstance of excessive text overlap or potential plagiarism in 2017, the authors were unknown to us, with no prior history of submitting to Headache. It is hard to determine why that might be. It could well be a class of low-quality authors chancing their arm, so to speak, to see if they can get away with publishing in a journal of good repute. Equally, some of the authors may simply be inexperienced or unskilled in the art of effective writing for publication. Of course, all of this concerns the actual use of words. It is perfectly possible to copy the thoughts and concepts of others and simply rewrite them. At present there simply is no tool available to journals that could detect such deception. In the past, journals were significantly handicapped in detecting plagiarism. In fact, detection usually only occurred following publication and typically surfaced only when an alert reader, or the wronged authors, raised concerns with the editorial office. CrossCheck launched back in 2008 as a tool that scans all published articles and indeed a variety of other sources to determine where text overlaps. Results are presented to the user showing percentage overlap with the original source material and the offending text is highlighted in a proof of the manuscript. To aid in the interpretation of the results CrossCheck returns, the user is able to access the original source article. Initially Headache used the tool sparingly as it was cost prohibitive. Around 2012 we switched to randomly selecting every 10th submission for checking. At the start of 2017 we determined to finally check every article and the results were alarming. During the course of the year we were compelled to reject 11 papers (2.8% of manuscripts tested) with similarity scores over 50%. In the most egregious case, a 95% similarity score was returned. Perhaps we were naïve to have been so surprised by such prevalence. Other journals have also noted either a higher than expected prevalence of problem manuscripts or, alternatively, have detected increasing occurrence year-on-year.6 By way of a comparison, a study of three medical journals conducted by the publisher Taylor and Francis reported that the percentage rejection rates due to possible plagiarism were 10%, 6%, and 23%, respectively, though the rejection criteria were not disclosed.7 Typically when an article scores a similarity score of 30% or greater once the references have been removed and strings of text of 10 words or less have been eliminated, we will automatically look deeper. However, the percentage score is not necessarily the only barometer of concern. Some papers score under 20% overlap but contain entire paragraphs lifted from other sources. This seems to be typical in self-plagiarism cases. If we see lots of snippets coming from two or three sources, we may well actually switch tack with our investigation and instead look to see if the authors are possibly engaging in redundant publication or salami slicing, with the similarity being generated by the fact that the authors have split their submission up beyond the minimum publishing unit.8 Our approach is to always handle a case with sensitivity, regardless of however blatant the problem is. We may ask for a rewrite but that has been exceptional. If we are immediately rejecting the paper because it is not something we care to publish, in addition to communicating that fact, we will also point out we detected a problem. In doing so we will also include a copy of the report. In the most egregious cases, of which there were three in 2017, we not only warned the authors of their behavior and included a copy of the report, we in turn notified their institution. Not that such action yielded any follow-up from the authors' institutions. However, we believe to simply reject the paper and not comment is to simply pass the problem on to someone else. And that other entity may be doing a poor job of paying attention and thus risking the potential of the authors getting away with deception. Just one case of plagiarism or excessive text overlap is unacceptable and the fact that we handled so many cases in 2017 is worrisome. Simply put, this journal wishes to communicate that authors must take this matter seriously. Outright plagiarism, self-plagiarism, and significant text overlap are problematic and must stop. As more journals use CrossCheck on every submission they receive, it will become increasingly difficult to get away with such unethical behavior. The solution is simple. Certainly beyond the methods section of a paper, text recycling should be avoided at all costs. That is a simple step to take. Another is to properly quote and cite material pulled from other sources. If there is still uncertainty, as with so many things in journal publishing, simply ask the editorial office for advice. Author confusion is in part generated by the fact that there are no standard rules and journals apply their own preferences. Finally, as there are many plagiarism detection tools available, it should be understood that there is no “magic number” for the similarity score below which you are safe. Again, every journal applies the rules differently. And as already pointed out in this editorial, it is not so much the percentage overlap but the extent of a consecutive string of words that proves problematic. Jason Roberts, PhD Headache Editorial Office

Ähnliche Arbeiten