What is retraction in research: meaning and how to check

A retraction is the formal withdrawal of a published paper from the scientific literature. The journal tells every reader that the paper's findings can no longer be relied on. The library guide at HKUST puts the purpose simply: retraction helps stop misleading information from spreading and makes sure that future research is not built on flawed or fraudulent findings.

What surprises most students is that a retracted paper usually does not disappear. The ICMJE recommendations, which many medical journals follow, say that retracted articles should remain in the public domain and be clearly labelled as retracted. Elsevier, for example, publishes a retraction note and puts a "retracted" watermark on the PDF of the original paper.

retractionnoun

A notice from a journal that a published article is withdrawn because its findings or conclusions cannot be trusted. The article usually stays online, clearly marked, alongside a notice that explains why.

Why keep it at all? Because other papers have already cited it. If it simply vanished, readers following those citations would find nothing and never learn what went wrong. A marked paper with a notice beside it tells the whole story.

Journals have three kinds of notice for problems found after a paper is published. Nandini had seen the words "erratum" and "expression of concern" on PubMed before and had never stopped to read them. Wiley's policy page explains each one clearly.

An expression of concern is a warning, not a verdict. The National Library of Medicine notes that the outcome of the concern may later require a retraction notice. So if you cite a paper with an expression of concern, check its status again before you submit your own work.

Nandini's first question was whether the authors had cheated. "Not always," said Dr. Varghese. "A retraction tells you the paper cannot be trusted. It does not, on its own, tell you why." Elsevier divides retractions into two broad groups.

Wiley's policy lists the same main grounds: clear evidence that the findings are unreliable, through misconduct or honest error; findings published before without proper cross referencing; plagiarism; and unethical research. The Retraction Watch Database uses a long list of more specific reasons, including paper mills, computer generated content and even "cites retracted work".

Plagiarism is on every list, which is one more reason to keep careful notes on your sources from the start. Our guide to types of plagiarism explains the forms it takes, including the accidental ones.

Dr. Varghese opened the paper on the journal's website. Across the first page of the PDF ran the word RETRACTED. At the top of the web page was a link to a short notice. These are the signs to look for.

• On the journal's page: a banner or link to a notice, often titled "Retraction: [article title]".
• On the PDF: a watermark or stamp saying the article has been retracted.
• In PubMed: the article is tagged "Retracted Publication" and shows a link that begins "Retraction in:", pointing to the notice.
• In reference lists and databases: the title is sometimes shown with "RETRACTED" in front of it.

Under the ICMJE recommendations, the notice should explain why the article is being retracted, include a full citation of the article, and carry its title in the heading. It should also be labelled in every form: the abstract, the full text and the PDF.

Nandini read the notice. The journal had found that several images of tissue samples had been duplicated and altered, and the authors could not supply the original data. The blood sugar results could not be trusted.

A paper can be retracted years after it was published, long after you downloaded it. So the PDF in your folder may not show what the journal's website now does. Check before you rely on a paper, and check your whole reference list again before you submit.

1. Open the journal's page for the article.Use the DOI link, not your saved PDF. Look for any notice of correction, concern or retraction.
2. Look it up in PubMed, for health and life science papers.Check for the "Retracted Publication" tag or a linked notice.
3. Search the Retraction Watch Database.It is free to search by title, author, journal, DOI and more, and it also lists many corrections and expressions of concern. Crossref acquired the database in September 2023 and made the data openly available.
4. Check the Crossref record.Search the DOI on Crossref and look for linked updates such as retraction notices.

The easiest habit is to let your reference manager watch for you. Since June 2019, Zotero has flagged retracted items in your library using Retraction Watch data, and its word processor plugin warns you before you cite one. It works for items that have a DOI or a PubMed ID, so fill those in. Our guide on how to use Zotero shows how to set up your library.

Nandini's next worry was her literature review. Three paragraphs leaned on the retracted study. "You cannot use it as evidence," said Dr. Varghese. "But you can mention it, if you say clearly what happened to it."

Elsevier's advice is to cite both the article and the retraction notice and to say in the text that the article was retracted. Editage suggests labelling it in the reference list, for example: Author AA, et al. RETRACTED: Title of article. Journal name, year, volume, pages. Follow your journal's or university's style if it gives a rule.

Retractions affect metrics too. As our guide to the impact factor explains, citations to and from retracted papers are now left out of the top number in the calculation.

Nandini rewrote her three paragraphs around two smaller studies that were sound. Her hypothesis became more modest, and much easier to defend.

Most students will never face a retraction of their own work. But mistakes happen in honest research too, and how you respond matters more than the mistake. If you find a serious error in a published paper, tell your co-authors and your guide, then write to the journal's editor and explain what you found.

• if the error is small and the conclusions still hold, the journal may publish a correction
• if the error changes the conclusions, the authors can ask for the paper to be retracted
• the editor decides which notice fits, and the notice explains what happened

The HKUST library guide notes that although retraction is usually seen as a bad thing, it also gives authors a way to take responsibility for their work and keep their credibility. A retraction you ask for yourself, for an honest mistake, reads very differently from one forced by an investigation.

The best protection is the boring kind: keep your raw data, record every step of your analysis, never alter an image beyond what the journal allows, and give credit to every source. Our guide on ethical considerations in research covers these habits in more detail.

Before her review meeting, Nandini ran her whole Zotero library through one more check. No other red banners. In the meeting, a committee member asked why she no longer cited the well known extract study. She explained that it had been retracted, and showed the notice. He nodded and moved on to the next question.