The published version of an article is referred to as the version of record. This is intended as the final version of the article after peer review, copyediting, typesetting and proofreading, and is meant as a permanent entry in the scholarly record. Formally speaking, the version of record includes the article itself, the abstract, any references, all images and tables, plus supplementary materials, if any. The key expectation is that the version of record remains unchanged in time, as this is what other researchers may cite or build upon. There are, however, some cases where it may be appropriate to issue a correction, an expression of concern, a retraction or, more rarely, a removal.

Corrections

A correction is typically issued when errors or omissions affect how the article is interpreted, but the integrity of the findings remains unaltered. It can also apply when errors or omissions affect only part of the findings. Examples where this is appropriate may include missing information (e.g. on funding or competing interests) or mislabelling of figures, tables or other forms of data.

Expressions of concerns

An expression of concern is considered when significant doubts about an article have emerged, for example, in terms of research or publication misconduct. It serves as a warning to readers that the article’s findings may not be reliable and that care should be taken when reusing or building on them.

Where investigations of the concerns raised provide conclusive evidence of misconduct or serious issues, corrections or retractions typically follow the expression of concern. However, the expression of concern may remain if investigations are inconclusive or if it is difficult or impossible to investigate the issues identified (e.g. limited cooperation from authors and/or their institutions).

Retractions

Retractions are typically issued when major errors in any part of the article make the discussion or conclusions unreliable, or when severe cases of misconduct have taken place. Examples of misconduct that can easily lead to retractions include fabrication, falsification, plagiarism, or manipulation of the peer review process. Articles that have been published in more than one journal should be retracted from all but the first venue of publication.

Importantly, retractions may arise from honest errors, too, for example, issues with analysis or measurement errors that have led to incorrect results. It should be noted that retractions are not appropriate if a correction is sufficient to address the issues under consideration or if there is not enough evidence to support claims of misconduct or errors. Authorship disputes should not lead to retractions either.

Removals

In general, retracted articles are not fully removed from the journal’s website or databases. Even if the full text is removed, the article metadata is not. Removals (including full text and metadata content) should occur only in very rare circumstances. For example, an article may be removed because of legal reasons if it includes materials that pose risks to the general public or protected animal or plant species.

Communicating concerns or changes to the Version of Record

The definitions and examples given above are based on COPE’s guidelines but journals may use these mechanisms differently. A journal’s policy around corrections, expressions of concerns, retractions and removals should be detailed on their website, likely in their publication ethics statement or journal policies page. When corrections, expressions of concern, retractions or removals occur, they must be clearly communicated to readers. Any of these actions must be promptly flagged on the original article’s webpage to warn readers and minimise harmful effects. The text describing a correction, an expression of concern, a retraction or a removal on the journal’s website should use neutral and objective language, avoiding inflammatory wording and focusing on facts. It is important to note that any notices of corrections, concerns or retractions should not be behind paywalls.

Furthermore, article metadata should be updated and reindexed to reflect changes, i.e the metadata should indicate that the article has been retracted or removed.

Crossmark is a helpful way to reflect editorially significant decisions: Crossmark members commit to informing Crossref of updates, such as corrections or retractions, and, optionally, to providing additional metadata about editorial procedures and practices. In all cases, article metadata should be retained (e.g. title, authors) to ensure the integrity of the scholarly record.

Lastly, these changes to the version of record should be followed regardless of when the article was published, even if it was published a long time ago or has not yet appeared in print versions of the journal.