The peer-review process is at the heart of the success of scientific publishing. As part of our commitment to the protection and enhancement of the peer-review process, Elsevier has an obligation to assist the scientific community in all aspects of publishing ethics, especially in cases of (suspected) duplicate submission or plagiarism.

Our journal editors frequently ask us how we can support them by using software to detect plagiarism. We take their concerns regarding plagiarism issues very seriously, and we would like to address this need, but it is important to point out that simple or imminent solutions do not exist. We are currently investigating new ways of detecting plagiarism and are exploring the effectiveness of plagiarism-detection software. For example, we are one of the first participants in the CrossCheck pilot, as described below.

Simple solutions do not exist

There are a number of reasons for textual overlap of submitted articles, including incorrect citations, self-plagiarism, reverse plagiarism (when an author copies an earlier manuscript but is published first), or, last but not least, acceptable re-use. In other words, duplications of text or references may, or may not, be legitimate. So, you may ask why simple solutions for the detection of real plagiarism or duplicate submissions do not seem to exist. The contributing factors include:

• Software quality. The capabilities of plagiarism-detection software are limited and are generally restricted to textual comparisons only--photographs, tables, and formulae are excluded. In practice, only verbatim comparisons can be made, providing percentages of text similarity.

• Human follow-up is required. Any software, used even at optimal levels, requires human involvement and follow-up. The software provides percentages of similarity in text; however, such similarities could relate to extensive citations with full reference to the original literature sources. In such cases, someone will have to filter through false positives and eliminate them before looking at more suspicious parallels. The human involvement required to make the final judgment proves to be a limiting factor in rolling out plagiarism-detection software.

• Early detection issues. Early detection, prior to peer review, is desired; however, this proves to be a technical challenge. For instance, some journals still have a paper-based peer-review process or may not have fully migrated to an online peer-review system. Also, keep in mind that online peer-review systems have firewalls for data protection reasons, making text comparisons during peer review more difficult, even within research submitted to, or published by, the same publisher.

• Pool of available articles. The issue here is that plagiarism-detection software requires making comparisons of (submitted or published) text similarities in the largest possible database with pre-indexed content. The absence of a universal database of published articles is a critically limiting factor, and thus, a solution for plagiarism detection will require publishers to create such a universal article database. This, apart from issues of scale, is also the main answer to the question why plagiarism detection in the scientific community appears to be less successful than that in colleges and universities where plagiarism-detection software is used. Students who copy from other papers usually do so from within free Internet domains.

• Infrastructure and editorial workflow. Plagiarism detection needs to fit into the specific workflow of individual journals. The different workflows and online peer-review systems of individual journals means that implementation of plagiarism detection requires solutions tailored to each journal, complicating the roll out to the more than 7000 peer-reviewed scientific journals that exist.

Investing in exploring the effectiveness of plagiarism-detection software

Cases of suspected plagiarism (or duplicate submission) are rarely limited to the same journal or publisher. More often than not, multiple journals and publishers are involved. Software solutions, therefore, ideally require cooperation between (all) publishing houses. We, together with five other publishers, recently entered into collaboration with CrossRef to develop and pilot a system that allows publishers to verify the originality of submitted and published works. The service, when it is launched, will be called CrossCheck. This pilot will involve software from   iParadigms, known for providing plagiarism software to the academic world. The idea is that the CrossRef members will create a single database of published articles (perhaps later also to include submitted articles) against which checking can take place.

  Click here for the press release about CrossCheck.

We hope that this pilot will give useful insights into the effectiveness of plagiarism-detection software. We expect to be able to assess meaningful ways of integrating this system into the editorial workflow and to determine the consequences on the workloads of the editorial teams involved in the pilot. A successful pilot will also help us to decide whether to scale up: that is, whether putting hundreds of thousands of articles through a plagiarism-detection workflow would generate the desired benefits--prevention of verbatim plagiarism and duplicate submissions. We will inform you about the results and engage you in the options ahead.

Finally, at Elsevier, we devote many resources to supporting our editors in following up on suspected plagiarism, duplicate submissions, or falsifications. We provide support for more than 400 cases per year. We are committed to continuing this level of support, irrespective of the outcomes of the CrossCheck pilot. Although we estimate only 0.1% of submitted articles are ever suspect--with considerable variation of occurrence between different academic areas--a workable software solution for plagiarism detection can lower the burden on our editors and pay off in terms of resources required to detect plagiarism. We recognize this burden and are committed to searching for the best solutions.

One of Elsevier's core tasks is supporting our editors by supporting and fostering the peer-review process. We have an ongoing commitment to safeguard the ethics of publishing and a responsibility to strive to prevent or minimize all aspects of unethical publishing behavior. Whether software can help in the detection and prevention of plagiarism is the core question that the CrossCheck pilot seeks to answer. We will keep you informed as we explore and discover new ways and means of solving this important issue.