Correction (corrigendum)

Correction (used interchangeably with corrigendum by many journals) is one of five distinct post-publication editorial actions a journal can take, each triggered by a different kind of problem and carrying a different implication for the underlying paper’s reliability. COPE and ICMJE both frame the choice between them around one operative question: would a reader who knew about the error draw materially different conclusions from the paper? If no, a correction is appropriate; if yes, the paper needs a stronger action.

The five actions, in ascending order of severity: a correction (sometimes split into corrigendum, conventionally an author-originated error, and erratum, conventionally a publisher-originated production error, though many journals no longer distinguish the two in practice — see CASRAI’s dedicated erratum entry) fixes a bounded error, such as a wrong affiliation, a transposed figure, or a mislabelled axis, without touching the paper’s conclusions. An expression of concern (see CASRAI’s expression of concern entry) is issued when an investigation is under way and the outcome is not yet known, signalling readers to treat the paper with caution in the interim rather than asserting a specific finding. A full retraction (see CASRAI’s retraction entry) removes the paper’s standing as part of the scientific record because its central conclusions are unreliable, whether from an honest error serious enough to invalidate the findings or from misconduct.

Fault is a separate axis from severity, and the two are often conflated. ICMJE’s own framing does not split correction from retraction by who caused the error: a pervasive, serious honest error still warrants retraction, and a minor error introduced by the author rather than the publisher can still be handled with an ordinary correction. The conventional corrigendum/erratum split by origin is a naming convention some journals still use, not a severity distinction, and CrossRef metadata links a correction bidirectionally to the original article regardless of which of the five actions the journal ultimately took, so the published version and the notice always resolve to each other for readers and indexers alike.

COPE’s retraction guidelines and its correction and authorship-change flowcharts (see CASRAI’s overview of COPE’s flowcharts) set out the procedural steps for each action and the escalation path between them — for instance when an initial correction proves to have understated the problem, or when what looked correctable during investigation turns out to invalidate the paper’s conclusions. Editors are expected to match the action to the problem’s actual severity once understood, not the action that was easiest to issue at first report.