There is no published figure for how many retracted papers are cited in UK doctoral theses, and no UK body collects one. What does exist is an open, free, machine-readable register of retractions covering roughly 50,000 records at the point it was consolidated in 2023 — which means that unlike almost every other integrity question in a doctorate, this one you can simply check.
Most candidates never do. The purpose of this page is to give you the evidence that exists, be honest about the evidence that does not, and leave you with an audit you can run on your own bibliography this afternoon.
What is actually evidenced
The single most important development is administrative rather than scientific. On 12 September 2023 Crossref announced that it had acquired the Retraction Watch database and made it a public resource. Its own arithmetic at the time was precise about scale and about overlap: “Crossref retractions number 14k, and the Retraction Watch database currently numbers 43k. There is some overlap, making a total of around 50k retractions.”
Two things follow, and both matter to you.
The register is open. The full dataset was released through Crossref’s Labs API, “initially as a .csv file to download directly”, and subsequently in a git repository, with the arrangement committing Retraction Watch “to keep the data populated on an ongoing basis and always open, alongside publishers registering their retraction notices directly with Crossref”. You do not need a subscription, an institutional login or a commercial tool to find out whether something you have cited has been withdrawn.
Neither source was complete on its own. Crossref held 14,000; Retraction Watch held 43,000; the union was around 50,000. That gap is the useful lesson. A retraction is recorded when a publisher registers it or when a curator notices it, and before 2023 those two streams were separate and each was partial. If your literature search predates the merge, it was run against a less complete picture than the one available now.
What is not evidenced, and will not be found
Three questions people expect this page to answer, which the published record does not support.
There is no UK national retraction statistic. The UK Research Integrity Office promotes good practice, supports investigations and publishes a procedure for the investigation of misconduct in research, and the Concordat to Support Research Integrity is described as the UK’s national policy statement on research integrity — but none of that amounts to a published national count of retractions or misconduct findings by UK institution. If you see a “UK retraction rate” quoted, ask where the denominator came from.
There is no measured rate of retracted citations in theses. Studies of retracted-citation persistence exist in the journal literature, but no sector body samples UK doctoral bibliographies, so any figure presented as “X% of PhD theses cite a retracted paper” is not drawn from a UK collection.
A retraction is not a finding of misconduct. Papers are withdrawn for honest error, unreproducible results, duplicate publication, authorship disputes and publisher error as well as for fabrication and plagiarism. Treating every retraction as fraud misreads the register badly, and it will make you sound naive if an examiner asks. This is a different object entirely from student academic misconduct, which is governed by your institution’s own disciplinary procedure rather than by the scholarly record.
Why this reaches a doctoral thesis specifically
A taught-degree essay cites recent, well-known work over a few weeks. A doctorate does something structurally riskier: it accumulates a bibliography over four or more years, much of it read in the first eighteen months, and then submits it as a single document long after the reading happened.
That creates three exposures a shorter piece of work does not have.
- Time. Retraction typically lags publication by years. A paper you read and stored in your first year can be withdrawn in your fourth, and nothing will tell you unless something is watching your library.
- Inheritance. You rarely cite a fraudulent primary study directly. You cite a review, or a meta-analysis, or a methods paper that rests on it — and the retraction propagates into your argument without ever appearing in your reference list.
- Examination. Your thesis is read closely by two people chosen because they know this literature better than you do, and an external examiner who has followed a controversy in their own field will recognise a withdrawn study on sight. This is exactly the kind of specific, checkable question that an examination probing your original contribution to knowledge is well designed to surface.
The audit: four steps, about an hour
Step 1: Let your reference manager do the first pass. Zotero has checked libraries against retraction data since version 5.0.67, announced on 14 June 2019 — the feature “can now help you avoid relying on retracted publications in your research by automatically checking your database”, flagging retracted items in the list and warning when you try to cite one, using Retraction Watch data. If you already run Zotero, open it, let it sync, and look for the flags. If you use something else, this is a genuine point of difference worth weighing in our comparison of Zotero, Mendeley and EndNote at doctoral scale.
Step 2: Check the register directly for anything load-bearing. The Crossref and Retraction Watch data is open, so the ten or fifteen sources your argument genuinely depends on deserve an individual look rather than reliance on an automated flag. Search by DOI where you have one. Note that an automated check is only as current as its last sync, which is a reason to run this again immediately before submission rather than only once.
Step 3: Check the publisher’s page for your key sources. Retraction notices, expressions of concern and corrections are published by the journal, and an expression of concern — a formal signal that a paper is under question but not yet withdrawn — will not always appear in a retraction register at all. If a study carries your argument, open its current publisher page rather than the PDF you downloaded in year one, because the PDF on your disk will never update itself.
Step 4: Check upstream in your key reviews. For the two or three secondary sources doing the most work in your literature chapter, spend twenty minutes on what they rest on. Citation-context tools of the kind covered in our roundup of AI research assistants for doctoral work can help here, though none of them removes the need to look.
You found one. Now what?
Do not quietly delete it. Deletion is the worst available response, because it leaves a gap in the argument, it loses the reason you cited the work in the first place, and it forfeits an opportunity to demonstrate exactly the critical judgement your examiners are assessing. Work out which of four situations you are in.
The claim rested on it and nothing else supports the claim. This is the serious case and it needs your supervisor this week, not next month. Either find independent support for the claim or qualify it down to what your own evidence carries.
The claim rested on it but other work supports the claim independently. Recite the claim to the surviving sources, and remove the retracted one from that sentence.
You cited it as part of the field’s history. Keep it, and cite it as retracted. Referring to a withdrawn paper is legitimate when you are describing what a field believed and why; the requirement is that you say so, so that no reader mistakes it for live evidence.
The retraction is itself interesting. In some fields a prominent withdrawal changed the direction of research, and saying so is a contribution rather than an embarrassment. A candidate who can explain what was retracted, when, on what grounds, and what it did to the literature is displaying precisely the command of the field that a viva tests.
In every case, write two sentences about it into your notes for the examination. Being asked about a retracted citation you have already thought about is a prepared answer; being asked about one you have not noticed is the moment the examination changes tone. It sits in the same category as the other things worth knowing before you walk in, which our analysis of UK viva outcomes across 127 institutions sets out in full.
The habit that prevents it
Everything above is remedial. The preventive version costs nothing: store every source with its DOI, in a manager that checks retractions, from your first reading. A bibliography built that way audits itself continuously, and the check before submission takes minutes rather than an evening.
The same discipline pays off after the award, because the reference list travels with you when you publish a paper from the thesis — and journal editors and peer reviewers run exactly this check, with less patience than an examiner.
Keeping the bibliography clean and consistent across 80,000 words is the mechanical half of this problem. Tesify holds the chapter structure and formats your references as you cite them, so the reference list stays in step with the argument instead of drifting from it across four years. The reading, the judgement and every sentence remain 100% written by you. It is free to start.
Frequently asked questions
How many papers have been retracted?
Around 50,000 as of the 2023 consolidation, on Crossref’s own arithmetic: 14,000 Crossref retractions and 43,000 in the Retraction Watch database, with some overlap. The register grows continuously, so treat that as a floor rather than a current total.
Where can I check whether a paper has been retracted, for free?
The Retraction Watch database became an open Crossref resource in September 2023, released through Crossref’s Labs API as a downloadable CSV and later in a git repository, with a commitment to keep it “always open”. You can also check the publisher’s current page for the article, which is where notices and expressions of concern appear first.
Does Zotero tell me if I have cited a retracted paper?
Yes. Since Zotero 5.0.67, announced in June 2019, it checks your database automatically against Retraction Watch data, flags retracted items and warns you when you attempt to cite one. Let it sync before you rely on the result.
Is citing a retracted paper academic misconduct?
No, not in itself. Citing a withdrawn study unknowingly is an error of currency, not of integrity, and it is distinct from your institution’s academic misconduct regime. What would be a problem is knowing a paper is retracted and continuing to present it as live evidence.
Does a retraction always mean fraud?
No. Papers are retracted for honest error, unreproducible findings, duplicate publication, authorship disputes and publisher error as well as for misconduct. The grounds are stated in the retraction notice, and reading it is part of deciding what to do.
What is an expression of concern?
A formal notice from a journal that questions have been raised about a paper without a retraction having been issued. It may not appear in a retraction register, which is why the publisher’s own page is worth checking for sources your argument depends on.
What if a paper is retracted after I submit my thesis?
It is not held against you — you cited the record as it stood. If it happens before your viva, mention it yourself rather than waiting to be asked, because volunteering it demonstrates that you are still monitoring your field.
Should I remove a retracted reference from my bibliography?
Only if the claim it supported is supported elsewhere or has been withdrawn. Where the paper genuinely forms part of the story you are telling about the field, keep it and label it as retracted. Silent deletion loses the reasoning and leaves the argument with an unexplained gap.
How many UK theses cite retracted work?
Unknown. No UK body samples doctoral bibliographies for this, so no national figure exists. Any percentage you see quoted for UK theses specifically is not drawn from a UK collection.
Do UK universities publish retraction or misconduct statistics?
Not as a national series. UKRIO supports investigations and publishes guidance, and the Concordat to Support Research Integrity is the UK’s national policy statement, but neither produces an aggregate public count of retractions by institution.
Will my examiners check my references for retractions?
Not systematically, but they do not need to. An external examiner is chosen because they know the literature, and a withdrawn study in their own area is the kind of thing they recognise without looking. That asymmetry is the reason to run the check yourself.
How often should I run this check?
Twice at minimum: once when you begin assembling the final bibliography, and once in the fortnight before submission. Retraction lags publication by years, so the register moves underneath a long project in a way it never does under a short one.