CASRAI Dictionary

Tag: self-correction

  • Retractions and corrections: how the scholarly record self-corrects

    A retraction notice is often read as a verdict of disgrace. It is more accurate, and more useful, to read it as the scholarly record doing the one thing it is supposed to do: correcting itself in public. Science advances by being checkable, and a record that could never be amended would be a record nobody should trust. The mechanisms for amending it — corrections, expressions of concern, and retractions — are part of the basic infrastructure of trustworthy scholarship, and they sit squarely in the research-integrity domain. This article sets out how they differ, what governs them, and why a well-flagged record is a sign of health, not failure.

    Three different instruments for three different problems

    The single biggest source of confusion is treating “retraction” as a catch-all for any post-publication change. In fact there are three distinct instruments, each appropriate to a different situation, and using the wrong one does real damage.

    • Correction (sometimes erratum or corrigendum). The published work is fundamentally sound, but a discrete error needs fixing — a mislabelled figure, a transcription mistake in a table, a wrong affiliation, a miscalculated value that does not change the conclusions. A correction amends the specific error while leaving the article standing. By long convention an erratum denotes a publisher-introduced error and a corrigendum an author error, though many journals now use a single neutral “correction” label.
    • Expression of concern. An interim notice issued when serious questions have been raised about a publication but the matter is not yet resolved — for example, an investigation is under way, or the editors cannot obtain the underlying data. It alerts readers that the work is under scrutiny without prejudging the outcome. It is deliberately provisional, and should later be replaced by a correction, a retraction, or a notice that concerns were not substantiated.
    • Retraction. The strongest instrument, used when the findings are so unreliable that the published conclusions can no longer be relied upon — whether through honest error or misconduct. A retraction does not delete the article; the work remains visible and citable, but it is clearly marked as retracted so that no reader mistakes it for current, dependable science.

    What the COPE guidelines say

    The reference point for how all of this should be handled is the Committee on Publication Ethics (COPE), whose retraction guidelines are the most widely adopted standard among journals and publishers. COPE’s framing is important: retraction exists to correct the literature and protect its integrity, not to punish authors. The guidelines set out the grounds on which editors should consider retraction, including clear evidence that findings are unreliable (from misconduct such as fabrication or falsification, or from honest error), redundant publication, plagiarism, unethical research, or compromised peer review.

    Equally important are COPE’s expectations for the notice itself. A retraction notice should be linked to the retracted article in both directions, be freely available to all readers rather than paywalled, clearly state who is retracting and on what grounds, and distinguish — as far as possible — misconduct from honest error. COPE also stresses that retraction should be reserved for cases where the findings are genuinely unreliable; where the core results hold and only a part is affected, a correction is the proportionate response.

    Honest error is not the same as misconduct

    A persistent and damaging myth is that being associated with a retraction is always a mark of wrongdoing. It is not. A substantial share of retractions stem from honest error — a contaminated cell line, a coding bug in an analysis pipeline, an unreproducible result the authors themselves discovered and reported. Authors who proactively retract their own flawed work are practising good science, and the integrity system should reward that candour rather than penalise it. This is precisely why COPE asks that notices state the reason: a self-initiated retraction for honest error and a retraction forced by an investigation into fabrication are very different events that happen to share a label.

    Where retraction does intersect with misconduct — fabrication, falsification, plagiarism, or the industrial-scale fraud of paper mills — it becomes one of the most important tools the community has for cleaning the record. But conflating the two discourages exactly the honest self-correction the system most needs.

    Why a flagged record is a healthy record

    It can seem paradoxical that visibly marking work as unreliable makes the literature more trustworthy, but it follows directly from how science works. The alternative to a clear retraction is not a pristine record — it is an unreliable result sitting silently in the literature, being cited and built upon as though it were sound. A retraction that is properly linked, freely readable, and clearly reasoned stops that contamination from spreading. The danger is not retraction; the danger is unflagged error.

    This is also why persistence matters. A retracted article should not vanish: its DOI must continue to resolve, ideally to a version prominently watermarked as retracted, so that anyone who follows an old citation arrives at the correct status rather than a dead link or, worse, an unmarked copy. Crossref and similar infrastructure carry retraction status in metadata so that reference managers and discovery tools can surface it automatically — the record corrects itself not only on the page but in the machine-readable layer that downstream systems read.

    Crediting accountability, not just authorship

    Corrections and retractions also raise a question of responsibility, and here the structure of contribution metadata matters. The CRediT taxonomy records who did what on a paper, and one role — the corresponding author’s accountability for the integrity of the work — is precisely the locus of responsibility when something goes wrong. Knowing, in structured form, who curated the data, who ran the analysis, and who supervised the work helps an investigation establish where an error arose and who is answerable for it. Honest contribution records do not prevent error, but they make accountability legible when error surfaces, which is exactly when clarity is most needed.

    Where shared vocabulary fits

    “Retraction”, “correction”, “erratum”, “corrigendum”, and “expression of concern” are used loosely — and sometimes interchangeably — across journals, which muddies a reader’s ability to judge what a notice actually means. A shared, federated vocabulary that defines these instruments precisely — and points back to COPE for the governing guidelines — is what lets a retraction in one venue be understood the same way in another. Supplying that definitional layer is the role the CASRAI dictionary is designed to play; the relevant terms sit in the research-integrity domain.

    Related reading

  • Post-publication peer review and research sleuths: PubPeer and self-correction after publication

    It is tempting to think of peer review as a gate: a manuscript is scrutinised, it passes, it is published, and the matter is closed. But publication is not the end of scrutiny — or it should not be. Errors slip through review, problems become visible only once a paper is read widely, and occasionally misconduct is detected only after the work is in the literature. A scholarly record that could never be questioned after publication would accumulate its mistakes forever. Post-publication peer review is the practice of continuing to scrutinise work after it appears, and it has become an essential part of how the literature corrects itself. This article examines that practice and the people and platforms behind it, drawing on the research integrity domain of the CASRAI Dictionary.

    What post-publication peer review is

    Post-publication peer review is exactly what its name suggests: the evaluation of a published work by the community after it has appeared, rather than only by selected reviewers before it appears. It can take many forms — a published commentary, a letter to the editor, a structured review on a dedicated platform, or an informal note on social media. What unites them is the recognition that the few reviewers who assessed a paper before publication are not the last word on it, and that the wider community of readers can identify problems the original reviewers could not. It treats the published paper not as a closed case but as a claim that remains open to examination.

    PubPeer and the online commentary layer

    The best-known infrastructure for post-publication review is PubPeer, an online platform that allows readers to comment on published papers, identified by their DOI or other identifier. PubPeer functions as a kind of comment layer over the literature: a place where someone who notices a problem — an apparently duplicated image, an implausible statistic, a method that does not add up — can post their observation for others to see and discuss, often anonymously. Anonymity is a deliberate and important feature, because raising concerns about published work, particularly the work of senior or powerful researchers, can carry real professional risk. PubPeer has become a significant venue where concerns about specific papers are aired, examined and, in many cases, ultimately acted upon by journals and institutions. It has made post-publication scrutiny visible and persistent in a way that scattered private worries never were.

    Research sleuths

    Alongside the platforms is a community of individuals who have become known as research sleuths or integrity investigators: people who systematically examine the published literature for signs of error or misconduct. Some focus on image integrity, developing a practised eye for duplicated, manipulated or reused figures; others look for statistical impossibilities, tortured phrasing characteristic of manipulated text, or patterns suggesting paper-mill production. Their work is often painstaking, frequently voluntary, and sometimes carried out at personal cost. These sleuths have been responsible for surfacing a substantial number of serious problems that formal pre-publication review missed, and their findings — often posted on platforms like PubPeer — have triggered investigations, corrections and retractions. They represent a distributed, motivated form of scrutiny that complements the formal systems, catching things those systems were never designed to catch.

    From concern to correction

    Raising a concern is only the beginning; the question is what happens next. The journey from an observation to a formal change in the record typically runs through several stages:

    • A concern is raised — on a platform such as PubPeer, in a letter to the journal, or directly to an institution.
    • The journal or institution investigates. Editors may issue an expression of concern to flag that a paper is under question while inquiries proceed.
    • A correction is made where the problem is genuine but limited — a corrigendum that fixes the specific error.
    • A retraction is issued where the problems are serious enough to undermine the work’s reliability, formally signalling that the findings should not be relied upon.

    These mechanisms are how the published record is actually changed, and they are the formal counterpart to the informal scrutiny that surfaces the problems in the first place.

    The role of COPE

    The handling of these situations is not improvised; it is guided by established norms. The Committee on Publication Ethics (COPE) provides guidance to editors and publishers on how to respond to concerns about published work — how to investigate fairly, when an expression of concern is appropriate, how to handle corrections, and the proper grounds and process for retraction. This guidance matters because post-publication scrutiny, for all its value, must be handled responsibly: authors are entitled to due process, a concern is not the same as a proven fault, and the record must be changed carefully and transparently rather than reactively. COPE’s frameworks give editors a principled basis for turning a raised concern into a fair and proportionate response.

    Self-correction as a feature, not a failure

    It is worth stating plainly: a retraction or correction is not a sign that science is broken. It is a sign that the self-correcting mechanism is working. The alternative — a literature in which flawed work, once published, can never be challenged or corrected — would be far worse. Post-publication peer review, the platforms that host it, the sleuths who drive much of it, and the editorial processes that turn its findings into formal action are together the visible machinery of a scholarly record that takes its own reliability seriously. The willingness to revisit and correct published work is one of the things that distinguishes a healthy research culture from a complacent one.

    Describing integrity work consistently

    For concerns, investigations, corrections and retractions to be handled and recorded consistently across journals and institutions, the terms involved must mean the same thing everywhere — what counts as an expression of concern, a correction or a retraction, and how each is recorded. That consistency is what the CASRAI Dictionary supports: a shared vocabulary so that the status of a published work is understood identically wherever it appears, which underpins sound research administration. And because scrutiny after publication is itself genuine scholarly contribution, the work of reviewing and correcting can be described in the same framework used for every other — the CRediT taxonomy and its full set of contribution roles. The scholarly record is trustworthy not because it is never wrong, but because it has robust ways of putting itself right.