When a paper is retracted the journal marks it withdrawn, but downloaded copies, secondary coverage and citations in other papers persist unchanged.
When a paper is retracted the journal marks it withdrawn, but downloaded copies, secondary coverage and citations in other papers persist unchanged. Studies of this find retracted work continuing to be cited positively for years afterwards, often by authors unaware of the status.
Papers are withdrawn for honest error, for a mistake found by the authors themselves, and for irreproducible results, as well as for fabrication. The notice usually states which, and the distinction matters when deciding what the underlying question now looks like.
A journal that suspects a problem but has not concluded its investigation may publish an expression of concern. The paper remains available and is flagged. Treating it as retracted overstates the finding; ignoring the flag understates it.
Major databases attach retraction notices to the record, and dedicated retraction indexes aggregate them across publishers. A search on the title or identifier surfaces the status in both places, and the two sometimes disagree on timing.
A claim that has circulated for a decade has had a decade in which the underlying paper could have been withdrawn. The more established a statement sounds, the more worthwhile the check, which inverts most people's instinct.
A retracted paper's citation count keeps rising, because citations accumulate from work already in progress and from authors citing a reputation rather than a reading. A high citation count is therefore not evidence that a paper is sound.
Many papers are corrected rather than retracted, with a figure replaced or an analysis redone. The corrected version is the one to read, and the correction notice sometimes changes the conclusion materially without changing the title.
Before relying on a specific study, search its title alongside the word retracted and check the publisher's page for a notice. It resolves most cases immediately and costs less than reading the abstract.
The claims on this page about withdrawn papers keep being cited for years, and checking takes a minute can be checked against sources outside this site: