Retractions and Corrections in Scientific Journals: Causes and Consequences

When a paper disappears from the scientific record — or gets a prominent notice stamped across its abstract page — it tends to make people uncomfortable in ways that a rejection letter never does. Retractions and corrections are the formal mechanisms journals use to amend or withdraw published research, and they sit at the intersection of error, accountability, and the slow, sometimes painful process of science correcting itself. This page covers how those mechanisms are defined, how the process unfolds, what typically triggers each outcome, and where the lines between them get drawn.

Definition and scope

A retraction is a formal withdrawal of a published article from the scientific literature. The paper is not deleted — it remains accessible, typically with a retraction notice overlaid or appended — but its findings are officially disavowed by the journal, the authors, or both. A correction (sometimes called an erratum or corrigendum, depending on where the error originated) is a published amendment that acknowledges a specific mistake while leaving the paper's core conclusions intact.

The distinction matters more than it might appear. Retraction Watch, a database maintained by the Center for Scientific Integrity, tracked over 10,000 retractions by 2023, a figure that reflects both genuine growth in misconduct and improved detection mechanisms. The broader landscape of research ethics and publication standards shapes how editors and institutions navigate these decisions.

The Committee on Publication Ethics (COPE), which publishes the most widely adopted retraction guidelines, defines retraction as appropriate when findings are unreliable due to misconduct or honest error, when the work constitutes plagiarism, or when it reports unethical research (COPE Retraction Guidelines).

How it works

When a concern is raised — by a reader, a reviewer, a co-author, or an automated plagiarism tool — the journal editor initiates an inquiry. The sequence typically runs as follows:

1. Concern flagged — internally by editorial staff or externally via post-publication correspondence.
2. Author notification — the corresponding author is contacted and asked to respond within a defined window, often 30 days.
3. Institutional involvement — for concerns involving data fabrication or research misconduct, the journal notifies the author's home institution, which conducts its own investigation under federal guidelines (for US-funded research, the Office of Research Integrity at HHS oversees misconduct inquiries).
4. Editorial decision — based on author response, institutional findings, and COPE guidance, the editor decides between no action, correction, expression of concern, or full retraction.
5. Notice publication — the retraction or correction notice is published, linked to the original article, and submitted to indexing databases including PubMed and Scopus.

An expression of concern occupies a middle ground — it signals that an investigation is ongoing without reaching a final conclusion. Journals issue these when resolution is expected to take substantial time, typically when an institutional investigation runs longer than 6 months.

Common scenarios

Not every retraction involves fabrication. The causes divide roughly into two categories: honest error and research misconduct.

Honest error includes cases where:
- A calculation or statistical error invalidates the main finding
- Image processing introduced unintended alterations
- A labeling error in a figure cannot be corrected without undermining the result
- Contaminated reagents or equipment faults affected experimental outcomes

Misconduct includes:
- Data fabrication or falsification
- Image manipulation to misrepresent results
- Plagiarism and duplicate publication, including self-plagiarism where the same dataset is published in two separate papers without disclosure
- Undisclosed authorship disputes or forged author consent

A 2012 analysis published in PNAS by Fang, Steen, and Casadevall found that roughly 67% of retracted biomedical papers were attributable to misconduct rather than error — a proportion that inverted many assumptions about the field. The peer review process does not reliably catch either category before publication, which is part of why post-publication scrutiny carries so much weight.

Corrections, by contrast, are almost exclusively associated with honest error. A transposed number in a results table, a missing unit of measurement, an incorrectly labeled axis — these are the everyday imperfections of research. When a correction leaves the paper's conclusions substantively unchanged, it functions as a mark of integrity rather than failure.

Decision boundaries

The judgment call between correction and retraction is not always clean. COPE's guidance frames the decision around one central question: are the primary findings still reliable?

If the answer is yes — even after acknowledging an error — a correction is appropriate. If the answer is no, or if the error cannot be disentangled from the core findings, retraction is required.

The harder cases arise when:

• Partial data problems exist — one of four experiments is compromised, but the other three still support the conclusion. Some journals issue corrections with revised figures; others retract the full paper and invite resubmission with clean data.
• Authorship is disputed — if an author listed did not contribute, or a contributor was excluded, some journals issue authorship corrections; others consider the submission fundamentally flawed.
• The error is minor but the conduct behind it was not — a small numerical correction that surfaces during a misconduct investigation may still result in retraction if the investigation reveals broader integrity problems.

Journals indexed in PubMed are required to propagate retraction notices through the National Library of Medicine, which maintains standards for how these notices appear in MEDLINE records. Failure to propagate notices has historically led to retracted papers continuing to be cited, a phenomenon documented repeatedly in the literature and one reason the broader scientific journal ecosystem has invested in better cross-database coordination.

The gap between a paper being retracted and that retraction being absorbed into the citing literature can span years — studies have found retracted papers continuing to accumulate citations at rates comparable to non-retracted papers for 24 months or more after notice publication.

References

• Committee on Publication Ethics (COPE) — Retraction Guidelines
• US Office of Research Integrity (ORI), HHS
• National Library of Medicine — MEDLINE Retraction Policy
• Retraction Watch / Center for Scientific Integrity
• Fang, Steen & Casadevall (2012) — "Misconduct accounts for the majority of retracted scientific publications." PNAS 109(42)