When is self-plagiarism ok? - The Scientist - Magazine of the Life Sciences

When Robert Barbato of the E. Philip Saunders College of Business at Rochester Institute of Technology (RIT) heard he was being accused of plagiarizing his own work, he was a bit surprised. "I can't plagiarize myself -- those are my own words," he said.

And he is not alone in his views. Some scientists and publishers argue that it's "unavoidable" for scientists to re-use portions of their own text (not images or data, of course) from previous papers, and doing so may even be good practice. But others disagree, including many journals -- who have retracted papers in response.

"There are many ways you can say the same thing even when it comes to very technical language," said Miguel Roig of St. John's University, who has written extensively about plagiarism in academic literature. "It's a matter of what some have labeled poor scholarly etiquette."

In Barbato's case, the institutional committee formed to review the case unanimously decided to dismiss it. While the authors had reused some text in the introduction and methodology sections in two papers they had submitted simultaneously on gender differences in entrepreneurial business endeavors, the data were different and the papers reached vastly different conclusions. "Nobody saw anything wrong with this really," recalled Patrick Scanlon of RIT's department of communication, who served on the committee.

"Sometimes [text reuse] is just unavoidable," agreed Catriona Fennell, director of journal services at Elsevier. "Really, how many different ways can you say the same thing?" Because scientists tend to study the same topic over many years or even their entire careers, some aspects of their research papers, particularly the literature review and methodology, will be repeated. Once they've figured out how to word it succinctly and accurately, some argue, it's best left unchanged. "You're laying the groundwork for an ongoing discussion [so] making changes might actually be a bad idea," Scanlon said. "It would muddy the waters."

Indeed, even editors that tend to be on the strict side when it comes to text recycling make exceptions. Anesthesia & Analgesia recently pulled a paper due to the offense, as reported on the Retraction Watch blog, but the journal's Editor-in-Chief Steven Shafer said that the publication does not retract papers that only reuse text in the methodology section. "This is a very difficult area," admitted Shafer. While the recently retracted paper contained "multiple areas of duplicated verbatim or nearly verbatim text throughout," he said, not all cases are so straightforward, and each one "must be a judgment call."

With evidence that duplicate publications are on the rise, and estimates of more than 200,000 duplicates already archived in Medline, the scientific community is in dire need of better guidelines as to where to draw the line with respect to self-plagiarism -- and a better way of catching those that cross it.

"It's unfortunately a very gray area," said Jonathan Bailey, a copyright and plagiarism consultant and a writer for the website Plagiarism Today. "[When people] come to me asking what the lines are, I always have to say the same thing: 'You're going to have to talk to the publication you're submitting to.'"

The problem is that most publications don't have "hard and fast rules," Fennell said of Elsevier's journals. The most comprehensive guidelines with respect to self-plagiarism come from the Committee on Publication Ethics (COPE), but these guidelines refer only to truly "redundant publication," in which authors are attempting to pawn off old research as fresh and new. They contain no advice about scientists re-using their own text.

"There's nothing that says you can't have over 30 percent of your introduction being highly similar," said Harold "Skip" Garner, executive director of the Virginia Bioinformatics Institute, who has published several articles on plagiarism in scientific publishing. "There's nothing like that because it's impossible to calculate."

The good news is that with the bulk of publishing now done electronically and the advent of text similarity software to recognize possible cases of redundant publishing, identifying copied text is becoming a much less onerous task than it used to be. eTBLAST, for example, is a free text comparison program that searches the millions of abstracts archived in Medline, as well as a few other publically available databases. Once the publication spots a possible duplication, it's added to the Déjà vu database of highly similar citations, where scientists can evaluate and comment on the entries.

Probably the most widely used program to spot plagiarism in scientific publishing is Crosscheck, launched in June 2008 by CrossRef. A total of 119 publishers (nearly 50,000 journals) subscribe to the plagiarism detection program, including Elsevier, Wiley-Blackwell, and Springer, who donate their full text content to the database, which currently holds some 25 million pieces of scientific literature, and is "growing steadily," according to CrossRef Product Manager Kirsty Meddings. Crosscheck's subscribers can scan the database with the same iThenticate software used by Turnitin to check for possible duplications.

So far, the journals that have put the technology to use say it's working. Of the 60 papers flagged as having a high percentage of overlap with other publications in the first three months that the Society for Industrial and Applied Mathematics used Crosscheck (starting last March), "about 60 percent were self plagiarism," said David Marshall, the publisher at SIAM. "That is the majority of what we're uncovering."

"In my view, [having these programs] is one of the best things that ever happened because it puts scientists on notice," Roig said. Indeed, some journals have taken to explicitly announcing that they use Crosscheck in their instructions to authors, and/or post the Crosscheck logo on their website, hoping that just the threat of getting caught will act as a deterrent.

Even with these programs, however, editors must be careful, Bailey warned -- even high degree of text similarity can sometimes be legit. "It really is about context," Fennell agreed. "It's good software, but it doesn't replace human judgment."

The problem now is how to weed through the hundreds of thousands of suspected cases of duplicated publications currently in the scientific literature. "It's one thing to be a deterrent and preventative in the future," said Garner, but "who's going to clean up the mess that's already there?"