Plagiarism
Providing guidance on the Cochrane Library editorial policy on plagiarism.
Note that Cochrane's editorial policy on plagiarism relates to the Methodological Expectations of Cochrane Intervention Reviews (MECIR) reporting standard 22.
Tables and figures:
Detecting plagiarism with Crossref Similarity Check
Crossref Similarity Check is a service provided by the membership organization Crossref in partnership with plagiarism technology from iThenticate. When a document is checked in Similarity Check, it is compared with the content of the iThenticate database comprised of millions of documents, including material behind journal paywalls. Editors are able to use Similarity Check via the licence held by Cochrane’s publisher, Wiley.
What and when to check
Editors are encouraged to, at minimum, check at least a portion of text for all protocols and reviews (including updates) when initially submitted. Also see the Editorial Manager Knowledge Base, Similarity checks.
Additional Similarity Check screenings could occur at different stages in the editorial process (see Table 1). Editors may wish to screen more than once, or at particular timepoints, such as before peer review, or whenever it is identified that the writing style varies within a single document.
Table 1. Stages in the editorial process where Similarity Check screening could occur
Stage | Document | Recommended sections to screena |
---|---|---|
Proposal | All Review Proposal Forms | All text excluding references |
Protocol | Initial submission of protocol |
|
All resubmissions of revised protocols |
| |
Substantively updated protocols (i.e. new citation version) |
| |
Final version for publication | Screening not recommended at this stage | |
Review | Initial submission of review |
Omit
|
All resubmissions of revised reviews; or review ‘amendments’ | Where changes have been made to the text | |
Final version for publication | Screening not recommended at this stage | |
Update
| Initial submission of review update |
Omit
|
All submissions of revisions to updates | As at initial submission | |
Final version for publication | Screening not recommended at this stage |
a While it is possible to check an entire document for similar text, sections of a Cochrane review such as Methods, Characteristics of studies tables, and References are likely to give a high similarity score due to the nature of their content.
b Some Cochrane Review Groups may recommend the use of template text for the Background or Methods section. If so, the authors should have made a note of this within the protocol or review. See ‘Special circumstances for Cochrane Systematic Reviews’ for more information.
c It is possible to do this in Similarity Check; see Table 3.
How to check
Similarity Check provides a similarity score, indicating the total amount of text that matches text in other sources. Using Similarity Check involves:
- an automated step in which Similarity Check runs the online comparison; and
- a manual step in which a Cochrane editor interprets the report results and decide on next steps (see Table 2).
These two steps combined can take from 5 minutes to 2 hours, but it is usually around 15 minutes.
Table 2. Overview of Similarity Check process
Automatic process | Similarity Check finds and highlights overlapping text between manuscript and published material |
A similarity score is generated | |
Manual process | Similarity Check report reviewed |
Determine severity of plagiarism | |
Decide on action to be taken |
Once logged into Similarity Check, there is the option to submit different file types for screening. It is not recommended to submit the full version of the document because it may be very long and will include sections that have little value in being screened (e.g. references); see Table 1. Therefore it may be easier to select specific sections of the protocol or review to be screened. There are three possible approaches:
- Prepare a new document by cutting and pasting specific sections of text into a new document and save as one of the following file types: plain text, MS Word, PDF, RTF, PostScript, HTML, or XML.
- Use the Cut and Paste upload option in Similarity Check.
- From RevMan, using software installed to print to PDF, select the required sections and print and save as a non-RevMan PDF file type. Similarity Check does not accept RevMan file types (i.e. *.rm).
By default, Similarity Check includes the optional settings to exclude quotes (i.e. text within quotation marks), reference lists, and/or “small matches” of text to avoid false positives in the similarity index. However, while it is possible to request references to be excluded from comparison using Similarity Check, this does not always happen and it is preferable to upload a file without this section. It is not always advisable to exclude “small matches” to text because small matches could be direct quotes that need quotation marks and citations.
Similarity Check has an option to include a simultaneous Internet search (called “websearch” in Similarity Check) in addition to the standard iThenticate database search. This extends the Similarity Check comparison to include content not included in the iThenticate database, such as Wikipedia, and presents the collated results. This option should be used routinely.
When matches are identified in a report, Similarity Check has an option to exclude one or more matching sources. As described in Special circumstances for Cochrane Systematic Reviews, a high percentage of overlap would be expected between a protocol and review, and a review and an update. This functionality allows the user to exclude the protocol or original review, for example. This functionality may become less useful as the number of times a Cochrane Review is updated as the number of exclusions that need to be made increases. See Table 3 for the types of Similarity Check reports where this functionality is available.
The results of the report, and any action to be decided upon, should be considered by a Managing Editor or other senior editor.
Similarity Check reports
There are different modes of reporting in Similarity Check (see Table 3) some of which display different information. The Document Viewer is the default setting and shows the best matches for text in a submitted document (see Figure 1 and www.ithenticate.com/training/dv-walkthrough). The Document Viewer report has two pieces of information that will guide the editors either to have no cause for concern or to decide if any action is needed:
- It highlights any overlapping text, shows where it comes from, and shows how many words are overlapping in each instance (number of words is more informative than the percentage overlap, which is also provided). Editors can review all instances within the document.
- The Document Viewer report will include a similarity index score. Similarity Check’s similarity index should not be used as an absolute measure of whether significant overlap exists, but rather as a signal to have a closer look at the text. The score is a percentage of text that it has identified as an overlap with one or more other sources. A low score means less overlap and a high score means more overlap.
For further information on using Similarity Check, please see the official Similarity Check user manual (via TurnItin). For further information about the similarity score, see the iThenticate website.
Table 3. Types of Similarity Check reports
Document Viewer | Default report; a detailed report that uses colour coding to compare texts, and hyperlinks to allows user to review matches. You can exclude particular sources in this mode. |
Similarity report | Displays matching sources side-by-side with sampled text. You can exclude particular sources in this mode. |
Content tracking | Enables users to see if matches were manually excluded, or if there are more than one match for the sample, and ranking of proportional match in the report. You can exclude particular sources in this mode. |
Summary report | Same information as the similarity report, but it displays matching sources above the document |
Largest matches | Ranks sample according to the word count and percentage of words that match a string of words. |
Figure 1. Similarity Check Document Viewer example
This Similarity Check Document Viewer report shows the document being checked on the left side, highlighting matching text (in this example in red, blue and green), and the context of the matching text in the match document (Spirit MJ et al) on the right side. In this example the highlighted text in red and green match other sources than the text in blue and are not shown. The “Document Viewer” is the chosen reporting mode. Clicking on the “Text-Only Report” button will change the display to other reporting modes, which are detailed in Table 3. The “Similarity Index” applies to the entire document being checked and indicates the percentage of text from the entire document which overlaps with identifies sources (matched documents) and is shown in the upper right hand side of the report. |
Figures and images
Editors should be aware that Similarity Check will not identify any plagiarized figures or images, such as line drawings and photographs.
Special circumstances when similarity may be expected
The editorial policy on plagiarism lists the following circumstances that may generate high levels of text similarity, noting that high levels of similarity will not always be considered plagiarism. The policy also sets out requirements for correct citation and acknowledgement of template text, previous publication versions, and other instances where text is re-used.
Table 4. Special circumstances that may generate high levels of text similarity between Cochrane reviews, versions of Cochrane reviews, and other articles
Special circumstance | Guidance for editors |
Similar descriptions of methods used | Cochrane reviews can be expected to have a high percentage of overlap in the Methods section because of standardized methods. This is unlikely to cause concern unless text is copied verbatim and without correct citation |
Cochrane-specific templates used for text in one or more sections | If an author uses template text for one or more sections (e.g. Background, Methods), and states that a template has been used, a high percentage of overlap would be expected and should not cause concern |
Similarity in text between protocol and review, and review and update | A high percentage of overlap would be expected between certain sections of these versions (e.g. background, methods) and should not cause concern |
Similarities with published studies (e.g. trials described in the characteristics tables/risk of bias tables) | Some overlap would be accepted here. Authors should follow the best practice of correct citation and acknowledgement required by Cochrane's editorial policy to avoid the possibility of plagiarism |
Two or more reviews based on one protocol | A high percentage of overlap would be expected between certain sections of the protocol and the reviews that follow the protocol (e.g. background, methods). This should not cause concern, but it should be clear to the reader that the same text is used across a series of linked reviews |
Split reviews and merged reviews, e.g. one review split into multiple reviews, or multiple reviews combined into one review | Some overlap would be expected between the different reviews. This should not cause concern, but it should be clear to the reader that the same text is used across a series of linked reviews |
Co-publication of a Cochrane review or protocol, or republication of the text in official Cochrane journals or derivative products | A high level of overlap would be expected. This should not cause concern as long as the co-publication was agreed according to the policy |
Conversion of a non-Cochrane systematic review to a Cochrane review | A high level of overlap may be expected |
What to do in cases of suspected plagiarism
In submitted articles
The Committee on Publication Ethics (COPE) has published guidance, in the form of a flowchart, on how to deal with suspected plagiarism. This flowchart has been adapted, with permission, to Cochrane's editorial process. Editorial teams with a case of suspected plagiarism should follow the process outlined in the Flowchart. As shown in the flowchart, there is no arbitrary threshold that should be used to signify plagiarism, rather the nature of the duplicated material is as important as the incidence.
As described in the flowchart, once overlapping text has been identified, the severity of overlap will dictate the action to be taken. Common reaction from authors when confronted with accusations of plagiarism can range from indifference to anger and panic. Make your decisions thoughtfully. Sharing the similarity report can be useful for discussion with authors if it adds value to the discussion and understanding of the issue. Be educational rather than punitive.
It is good practice to ensure that the senior editors for a review are informed in cases of overlapping text and correspondence with authors. When authors make changes in response to an editor's feedback, a member of the editorial team should check the revised manuscript when resubmitted to confirm the revisions are sufficient.
Flowchart 1. What to do if plagiarism is suspected
Adapted with permission from COPE from the flowchart (2013): “What to do if you suspect plagiarism: Suspected plagiarism in a submitted manuscript”. |
In published articles
If editorial teams are alerted to suspected plagiarism in articles published in the CDSR, refer to the COPE flowchart for 'Suspected plagiarism in a published article' (see publicationethics.org/resources/flowcharts) and inform the Editor in Chief. The Editor in Chief may withdraw or retract the publication as a result.
Substantial and/or repeat instances of plagiarism
Cochrane editors should follow certain steps if they identify one of the following:
- high levels of clear plagiarism within one review;
- repeated instances of plagiarism at different stages of a review;
- repeated instances of plagiarism from the same author(s) in different reviews.
As noted in the Flowchart, it may be appropriate to report author(s) to academic institutions. This action, and any other very serious consequences, must be discussed and undertaken in consultation with a senior-level editor who will inform the Editor in Chief of such cases. The Editor in Chief will, in coordination with the senior-level editor, consider such situations on a case-by-case basis and decide on an appropriate course of action.
Recording information about cases of suspected plagiarism
Store Similarity Check reports
Editors can record similarity scores with notes of what was checked. If relevant, copies of similarity reports may be saved in the workflow files.
Record actions taken
Editors should consider whether to record as a note any action against an author in person record, sharing the note within their entity or with a specific administrative role. Notes should be as factual as possible, avoiding claims that cannot be substantiated: note what has occurred and the actions taken. For example, it would be appropriate to write, “A paragraph of text was copied verbatim from a separate article without acknowledgement of the original text. The author was asked to explain the reason for this and make appropriate changes before resubmitting.” It would be inappropriate to write, for example, “Author often plagiarises text”.
Monitor instances of serious plagiarism
Following the escalation of instances to the Editor in Chief, the senior-level editor and the EiC should monitor whether cases occur with the same authors or groups of authors, and take appropriate action.
Support resources
- List of resources for using the software, provided by Similarity Check
- COPE e-learning module on plagiarism. Requires COPE login to access (available to all editorial staff)