The Retraction Watch Database (RWDB) lets research offices, tenure committees and funders check whether a candidate’s published papers carry a retraction, correction or expression of concern. Used properly — combined with author, affiliation, article-type or date-range search fields, cross-checked against ORI case summaries, and read for the stated reason rather than the bare fact of a hit — it becomes a genuine due-diligence tool rather than a source of false alarms.
The Retraction Watch Database is a free, searchable index of scholarly retractions, corrections and expressions of concern, built by the Center for Scientific Integrity and distributed with Crossref. It is the largest curated source of retraction metadata available, but it indexes withdrawn papers, not people — why due diligence needs more than a name search.
- What is the Retraction Watch Database, exactly?
- How do you search RWDB for a specific candidate or grantee?
- What counts as a genuine red flag versus a false positive?
- How does RWDB compare with ORI research misconduct case summaries?
- What does a due-diligence screening workflow look like?
- Answer-first Q&A on retraction screening
- What this means for research offices and funders
- Where misconduct screening is heading next
What Is the Retraction Watch Database, Exactly?
RWDB is a structured dataset, not a blog archive: each entry records the original article, the notice, the stated reason, and the author/affiliation strings as printed on the paper. As of mid-2026 it logs more than 65,000 retraction entries. Crossref took on distribution in 2023, publishing the full CSV through a public repository rather than only the web form — the hosted interface at retractiondatabase.org suits one-off lookups; the download suits batch screening.
Critically, RWDB does not aim for completeness on corrections and expressions of concern the way it does for retractions. Its own user guide states EOCs and corrections are entered mainly “as they relate to existing retractions, blog posts, or high-profile studies” — a clean result there is not evidence of a clean record.
How Do You Search RWDB for a Specific Candidate or Grantee?
Search by author name first, narrow by affiliation, then confirm with PMID or DOI. Since 23 October 2024, RWDB has required every search to include at least one of: Article Type(s), an Original Paper date range, a Retraction/Notice date range, or a PMID/DOI — a blank author-only search no longer works, a change most existing search guidance predates.
- Author field: try name variants and the wildcard (*), e.g. *doe*, since journals list authors inconsistently.
- Affiliation field: free text only, matched against the original journal’s wording, not a normalised list.
- Article Type / date range: now mandatory as a search anchor; pair a rough employment period with the author name.
- PMID or DOI: the most precise route once a specific paper is identified.
Each search returns a maximum of 50 rows on screen, with a banner showing the true total — worth noting when a prolific or common-named candidate returns more hits than the interface displays.
What Counts as a Genuine Red Flag Versus a False Positive?
A retraction hit is a prompt to investigate, not a finding of misconduct. RWDB’s reason-code taxonomy (Appendix B of the user guide) separates honest error, authorship disputes and duplicate publication from deliberate fabrication — only the latter is relevant to a fitness assessment.
| Signal | Likely false positive | Likely genuine concern |
|---|---|---|
| Author role | Middle/minor co-author, no data or analysis role | First, corresponding, or last author |
| Reason code | Honest error, journal-initiated editorial correction | Data fabrication, image manipulation, plagiarism |
| Pattern | Single isolated retraction across a long career | Multiple retractions clustered in a short period |
| Notice type | Correction or expression of concern only | Formal retraction with a stated integrity reason |
A 2025 study via Taylor & Francis, indexed on PubMed, found metadata discrepancies between RWDB, PubMed and Web of Science for the same retracted articles — reason to cross-reference a second source before treating any record as final. RWDB also standardises the author field to “Editorial Staff” on journal-initiated notices, never to be misread as identifying the candidate.
Paper mills add a layer: outputs typically cluster by template, image reuse or tortured phrasing across unrelated author groups, a pattern the COPE–STM Paper Mills investigation has documented since 2022. A retracted paper matching paper-mill characteristics warrants closer scrutiny than an isolated retraction.
How Does RWDB Compare With ORI Research Misconduct Case Summaries?
RWDB and US Office of Research Integrity (ORI) case summaries answer different questions, and due diligence needs both. RWDB tells you whether a paper was withdrawn; ORI tells you whether a person was found, after federal investigation, to have committed misconduct — even where no retraction followed.
| Feature | Retraction Watch Database | ORI case summaries |
|---|---|---|
| Unit of record | A published article/notice | A named individual with a misconduct finding |
| Scope | Global, all disciplines and publishers | US Public Health Service-funded research only |
| Trigger for entry | A retraction, correction, or notable EOC is published | A formal ORI investigation concludes with a finding |
| Typical gap | Misses misconduct with no resulting retraction | Misses retractions outside PHS-funded, US-linked research |
Because ORI findings can precede, follow, or occur without a retraction, checking RWDB alone misses candidates sanctioned through supervision requirements or funding debarment whose flawed papers were never withdrawn. Hiring, tenure or funding decisions should run both checks, not treat either as a substitute.
What Does a Due-Diligence Screening Workflow Look Like in Practice?
- Confirm identity anchors — collect name variants, ORCID iD, and known affiliations before searching.
- Run the RWDB author search with a date range or article-type anchor as required, using wildcards for name variants.
- Filter to retractions specifically — the default result mixes in corrections and expressions of concern, which are not comprehensively indexed.
- Read the reason code for every hit rather than counting hits; separate honest error from fabrication, plagiarism, or image manipulation.
- Check ORI case summaries for the same name, independently, to catch misconduct findings with no associated retraction.
- Cross-reference a second metadata source (PubMed, Web of Science) before any hit informs a decision.
- Document the process and allow a response — record which fields were searched and on what date, and give the candidate an opportunity to explain any substantive finding before it affects the outcome.
Answer-First Q&A on Retraction Screening
What is the Retraction Watch Database?
The Retraction Watch Database is a free, searchable index of scholarly retractions, corrections and expressions of concern, distributed with Crossref. It records over 65,000 retraction entries with metadata on authors, journals, dates and stated reasons — but it indexes withdrawn papers, not verified findings against individuals.
Do retracted studies still get cited?
Yes. Citation-tracking studies confirm retracted papers continue to be cited after the retraction notice is published, often because citing authors are unaware of it. This is one reason due-diligence checks cannot rely on citation counts as a proxy for integrity.
What is the purpose of Retraction Watch?
Retraction Watch exists to track and report retractions as a window into how science self-corrects, publishing the underlying blog since 2010 and the structured database since 2018. Its purpose is transparency, not adjudicating misconduct — that sits with journals, institutions, and bodies such as ORI.
How do you check for retractions on a specific paper or author?
Search RWDB’s author or affiliation fields, or enter the paper’s PMID or DOI directly for the most precise match. Since October 2024 the search also requires an article-type or date-range anchor, so pair an author name with an approximate publication period.
What Are the Implications for Research Offices and Funders?
Institutions that skip a structured check within their research administration due-diligence process risk reputational and funding harm they could otherwise catch before an offer is made. Treating a single retraction hit as automatic disqualification risks penalising honest-error corrections with no integrity finding — both failure modes are avoidable with a documented, two-source workflow.
Authorship transparency makes this more tractable: contributor-role frameworks such as CRediT — originated by CASRAI in 2014 and now stewarded by NISO as ANSI/NISO Z39.104-2022 — let a research office consult authorship and contribution records to see whether a flagged co-author actually held a data-generating or analytical role, rather than a minor one, sharpening the false-positive filter above.
Where Is Misconduct Screening Heading Next?
Expect due-diligence practice to keep converging on multi-source verification rather than any single registry. As paper-mill detection tooling matures and Crossref’s stewardship of RWDB deepens, the advantage will sit with research offices that build a repeatable, documented workflow now — spanning RWDB, ORI case summaries, and contributor-role verification — rather than an ad hoc name search at the point of hire.
