Plagiarism and text recycling

Plagiarism concerns presenting another person's work as one's own; text recycling concerns reusing material from one's own earlier work. Their boundaries, detection and handling matter both when preparing a manuscript and when editing someone else's draft.

Uncredited words and ideas

Plagiarism includes taking another person's thoughts, writings, inventions or other creations without due acknowledgment; it is not limited to copying words verbatim. ^{1, 2, 3} The AMA Manual of Style's eleventh-edition definition includes ideas, words, data and graphics, whether published or unpublished, reported as one's own without appropriate credit. ² Unintentional plagiarism can result from poor paraphrasing, forgotten citations, negligence or carelessness. ^{1, 2}

A narrower editorial usage confines plagiarism to unattributed text and calls uncredited reuse of ideas or content “lack of acknowledgement”; under that usage, knowingly using another's ideas without acknowledging them is still unethical. ^{4, 5} Credit may take the form of a citation to published work, an accessible preprint or a personal communication. ⁵ Independent discoveries make appropriation of ideas difficult to prove: several people developed evolutionary ideas before or alongside Darwin, and Hooke, Schwann and Schleiden all recognised the importance of cells. ⁵ An idea must therefore be sufficiently distinctive to establish that it was taken from a particular person. ⁵ Reused text and images should clearly acknowledge their source and original creators. ⁶ Borrowing wording found while searching the internet for inspiration can constitute plagiarism. ⁷

Patchwork plagiarism weaves text from multiple sources into a new article. ⁸ Patchwriting also weaves original and borrowed text, but a language-teaching account distinguishes it from plagiarism where there is no clear intention to reproduce another's original ideas. ⁹ It can be tempting when finding appropriate English words, paraphrasing and adopting academic style are difficult. ⁹ Its alternative is to understand the source, rework and integrate its meaning, and express it through paraphrase or summary; this takes more time and effort when writing in an additional language. ⁹

Editors working in the Middle East have described borrowing fluent wording as a response to limited English proficiency and a wish to preserve accuracy and quality. ¹⁰ They distinguish the widely condemned appropriation of ideas from the more contested reuse of scientific wording, and from literary writing whose originality resides in its expression. ¹⁰ Some researchers argue that scientific language primarily conveys content, that science asks for simple wording, and that reuse could therefore be tolerated, particularly for authors with limited English and little access to editorial assistance. ¹⁰ Whether plagiarism is more common among authors writing in an additional language is unclear, although an elegant copied sentence can be conspicuous within otherwise error-filled prose. ¹¹

Prevention and training

Manuscript preparation includes close examination of references and the authors' earlier publications, including passages authors add during revision. ¹² Medical writers should identify potential plagiarism and help authors decide between citing earlier information, paraphrasing it and obtaining permission to reprint display items. ¹² Plagiarism can lead to manuscript rejection, even when copied passages were introduced by a junior writer. ^{13, 12} In 2013, the University of Düsseldorf revoked German education minister Annette Schavan's doctorate after investigating plagiarism in her thesis; she resigned from office and challenged the university's decision. ¹⁴ The university said it had handled her case like any other and that its definition of plagiarism had not changed since her doctorate in the 1980s. ¹⁴ In another case reported in 2013, a journalist resigned from The New Yorker after fabrication of quotations, plagiarism and self-plagiarism came to light, but subsequently received $20,000 to speak about decision-making at the Knight Foundation. ¹⁵ Controversy over that fee led the foundation to express regret, despite having invited him knowing he would discuss his misconduct. ¹⁵

A 2019–2020 survey of 260 writers, editors and publishing professionals had its largest response groups in the United States, Canada and United Kingdom, with respondents also in Japan, Cyprus and Uruguay. ² Of its 38 medical or science writers, 84% thought losing their job would be plagiarism's biggest consequence. ² Publishers were less confident than writers and editors in those professionals' ability to produce plagiarism-free work. ² Among 79 medical and science communicators very or extremely confident in understanding and avoiding or identifying plagiarism, 26 (33%) had no training or training limited to intentional plagiarism. ² Although respondents were generally experienced, most recalled academic training, sometimes as early as grammar school, and little subsequent training for writers or editors, especially on unintentional plagiarism. ² For preventing plagiarism in new content, 71% of publisher respondents selected contractual clauses; for finding it in existing content, 82% relied on editors and 74% on detection software. ²

Five precautions were recommended to help writers avoid plagiarism. ²

• Keep copied verbatim text separate.
• Cite public-domain and unpublished material too.
• Summarise, analyse and synthesise.
• Remember that plagiarism includes more than words.
• Be careful when reusing your own previously published material.

²

Better education and training of researchers is a proposed remedy for plagiarism. ¹⁶ A 2014 proposal advocated undergraduate writing courses to prevent plagiarism and harsh penalties for plagiarism committed after that training. ¹⁷

Forms of self-plagiarism

Self-plagiarism has been used both for recycling published data and for reusing one's own published words, which attract different judgments. ^{18, 19} A taxonomy adapted from Miguel Roig's ethical-writing guidance distinguishes five practices. ²⁰

Practice	Meaning
Data augmentation	Publishing old data with new supporting data as a new study.
Duplicate publication	Submitting essentially the same article for publication in two different journals.
Redundant publication	Publishing previously published data, with or without new data, from a new angle or focus.
Salami slicing	Publishing different results from one study separately when they would be better presented together, generally assumed to maximise the number of publications from the study.
Text recycling	Reusing published text in a new publication.

²⁰

The first three practices are forms of data recycling. ²⁰ Changing a title or formatting, translating a paper, or republishing only part of its data can conceal reuse without making it new work. ²¹ Authors' ownership of data does not make parallel publication of identical data acceptable. ²¹ In 2015, the BMJ was described as requiring copies of previous work with more than 10% overlap before approving a submission. ²¹

Fractionated reports can repeat background, methods and similar conclusions while using journal space for several accounts of one study. ²² Publication-count incentives and journals' reluctance to publish long reports contribute to this practice. ²² In a review of 56 systematic reviews containing 1,131 main articles, 60 articles had been published twice, 13 three times, three four times and two five times. ²² Another pattern adds a little new data to substantial previously published results while rotating the first author and changing the title. ²³ In a plagiarism case exposed in 1997, a researcher reused another group's paper from 12 years earlier, replacing data for one disease with data for another. ^{24, 25}

The 2009 Lancet editorial condemned data recycling as unacceptable while describing reused words as less serious. ¹⁸ An ACS Nano editorial called data recycling fraud, citing deception of editors, reviewers and readers, wasted review effort, reputational damage and publication pressure. ²⁶ Deliberately presenting recycled data or large amounts of text as new can lead to retraction and submission bans. ²⁶

A stricter interpretation of the Lancet editorial treats extensive verbatim republication as redundant or duplicate publication, nearly identical introductions and methods with different results and conclusions as objectionable salami publication, and recycled reviews or opinion pieces as deceptive even without reused data. ²⁷ A competing view holds that communicating an important message can outweigh editors' and publishers' interests in fresh wording. ¹⁸ Nearly two-thirds of 195 US university health-education staff surveyed in 2001 considered including the same section in multiple articles acceptable. ¹⁸

Acceptability of recycled text

A review of 63 plagiarism editorials published in 2008–2012 found that 41% defined plagiarism, with substantial variation in definitions and no clear consensus on acceptable self-reuse. ¹⁶ Warnings about consequences and discussion of detection software dominated those editorials. ¹⁶ Both authors and editors would benefit from clearer guidance on avoiding problems and deciding when and how to act. ^{28, 16}

Descriptions of methods are a particular point of disagreement: authors may regard rewording an accurate existing description as wasted effort. ²⁰ An author who apologised to Anaesthesia for related but different articles with identical titles argued that editors needed to decide what counted as self-plagiarism. ²⁰ In 2012, Anesthesia & Analgesia allowed original authors to reuse methods descriptions verbatim, but not other text; other journals did not permit the same reuse. ²⁹ Briefly describing procedures and citing the earlier full account avoids repeating the text, but can inconvenience readers who must locate or buy the earlier article. ²⁹ The American Chemical Society guidance quoted in that 2012 discussion required appropriate citation and quotation marks to legitimise recycled text, while convention still ruled out a whole page of quoted methods. ²⁹ Roig recommended quotations and proper paraphrasing. ²⁹ ICMJE guidance then covered particular cases, such as publishing important medical guidelines in several journals to reach wider audiences, without covering every form of self-plagiarism. ²⁰

A journal editor's stricter view in 2023 was that citations to earlier rationale and methods remove the need to repeat blocks of text, while reusable Results or Discussion wording raises the question whether a manuscript is new at all. ⁴ Difficulties finding a second formulation can be especially severe for an author who struggled to describe the subject in English the first time. ²⁴

Historical numerical limits also differed: an expert poll found that many accepted 10% recycled text and none accepted more than 30%, while a 2004 WAME report described some editors treating overlap above one-third of a review as excessive. ²⁴ Earlier accounts also quoted limits of 10% and 30%. ²⁴ One journal editor's 2023 interpretation of COPE and software recommendations placed overall identity of 15% or less below the plagiarism threshold, 25% or more above it, and 15%–25% within editorial discretion. ³⁰ His own procedure used different operational thresholds. ^{4, 30}

1. Warn authors when overall identity exceeds 20%.
2. Reject for plagiarism when overall identity exceeds 70%.
3. Regardless of the overall score, warn when any paragraph exceeds 20% identity.
4. Reject when more than 20% of paragraphs exceed 20% similarity, or any paragraph exceeds 70% identity.

^{4, 30}

Low overall similarity can conceal one or more paragraphs with much higher identity, even 100%. ³⁰ That editor regarded penalties for very short phrases or organisational titles such as World Health Organization as grounds for strong protest. ⁴ For low-similarity manuscripts, he advised either making small changes to secure acceptance or politely challenging rejection using the relevant definitions; high-similarity manuscripts required an innocent explanation or substantial rewriting. ³⁰

Research on text recycling

The Text Recycling Research Project began in 2017 to investigate beliefs and practices and develop guidance. ³¹ It defined recycling of prose, equations or visuals through three conditions. ³²

1. Material is identical to, or substantively equivalent in both form and content to, the original.
2. It serves the same rhetorical function in both documents.
3. At least one author of the new document was an author of the original.

³²

Early organisational statements varied widely and rested largely on personal beliefs in a field with little research. ³³ Findings reported in 2020 included a survey of 316 journal editors and editorial-board members across science, technology, engineering and mathematics (STEM), quantitative social sciences, qualitative social sciences and humanities. ³⁴

• About 84% considered text recycling at least sometimes acceptable, with no substantial differences between disciplines.
• More than 95% considered reuse from a conference poster at least sometimes acceptable; reuse from grant proposals, conference papers and grant reports was also commonly accepted, and even reuse from journal articles was sometimes acceptable to more than 40%.
• Methods material was the most acceptable to recycle, and Results material the least.

³⁴

Interviews with 21 editors-in-chief nevertheless showed substantial differences: some thought any recycling prevented originality, many accepted unpublished text but differed on what “unpublished” meant, and some recommended rewording detected recycling. ³⁴ Thus agreement that recycling could sometimes be acceptable did not establish agreement about when. ³⁴

Preliminary text analysis found recycling in articles from approximately 25% of grants across biological sciences, engineering, mathematical and physical sciences, and social, behavioural and economic sciences. ³⁴ The analysis treated articles from the same grant as an authorship group's own work, rather than restricting reuse to one individual author, and the percentage was relatively consistent across fields. ³⁴ In 2020, the project's next phase was intended to develop guidelines, policies and educational materials from the findings and stakeholder input. ³⁵

Similarity detection

Similarity software generally compares wording: it cannot assess ideas and concepts and can miss plagiarised content that has been reworded. ³⁶ Word similarity does not by itself establish plagiarism. ³⁶ Shifts in writing style can draw an editor's attention to a borrowed passage. ¹⁶

After Anaesthesia introduced screening of every submission, its editor reported in 2010 that 4% had been rejected directly for plagiarism over the first year, without specifying the exact rejection criteria. ³⁷ New entries in the Déjà vu database of highly similar Medline articles fell by approximately half in relative terms between 2006 and 2008, but the change could reflect better detection or greater caution among authors. ³⁷

A 2012 test of free checkers using PubMed abstracts found that Article Checker struggled to identify their origins, Dupli Checker detected similarity in most cases but returned inconsistent results for identical searches, and Plagiarisma identified nearly every sentence as unoriginal, with free use limited to five searches a day. ²⁵ CrossCheck then charged members of CrossRef a per-document fee and an annual administration charge. ²⁵

Plagiarism in commissioned work

An editor who finds recycled passages in a commissioned review should alert the client, consider proposing a remedy and let the client approach the faculty author. ³⁸ Some faculty may not realise self-plagiarism is a problem; one experienced writer usually resolved it by rewriting. ³⁸ Another experienced writer recommended rewriting even if the client insisted that reuse was acceptable, then declining future work for that client. ³⁸ Clients should understand that journals screen for plagiarism and copyright infringement and that infractions could lead to blacklisting. ³⁸ An unusually polished first draft, even perfect punctuation, was a suggested warning sign prompting a search for one of its passages. ³⁸

Editors were advised to use five conversational steps when raising plagiarism with a writer. ²

• Prepare concrete examples.
• Start the conversation with a different word.
• Ask questions.
• Avoid assumptions.
• Open discussion before deciding next steps.

²

Repeatedly commissioned needs assessments should contain original material, even for the same client and topic. ³⁹ Familiarity with the disease state makes later work easier without requiring reuse of the earlier wording. ³⁹ An alternative is an explicit agreement to update needs assessments rather than create a fresh document each time. ³⁹