WHAT HAPPENS AFTER SCIENCE SAYS: DO NOT TRUST THIS PAPER?
A scientific paper can be withdrawn. Its ideas can keep travelling.
A retraction is a journal telling readers that a published paper should no longer be relied on as it originally appeared. But the paper does not vanish. Later researchers can still find it, quote it, criticize it, copy a method from it, or keep using its claims.
We started with Retraction Watch's famous list of the ten most-cited retracted papers and asked a simple question: what are those later papers actually doing with the old work?
4,854later papers counted by Retraction Watch after the ten papers were retracted
6,458later papers we found independently in an open citation database
908later papers for which this site can show a short piece of the surrounding text
Why two different citation totals? Databases cover different journals and records. That disagreement is part of the story, not a mistake to hide.
01
The short version
You do not need to know anything about academic publishing to read the rest of this page.
1
“Retracted” does not always mean “everything in this paper was false.”
One paper on this list was pulled mainly because it repeated material published elsewhere. Two others were corrected and republished.
2
A citation is not automatically a vote of confidence.
A later paper may cite old research to criticize it, explain a scandal, discuss history, or warn readers about it.
3
Some later papers really do keep using the old claim.
Those are the cases worth finding. A raw citation count cannot separate them from criticism or history.
4
Sometimes the thing that survives is not a fact.
A measuring scale, method, category, research question or familiar story can keep moving through later papers even after the original paper is withdrawn.
5
The warning can arrive years after the idea has spread.
By then the paper may already have hundreds of descendants. Retraction changes the warning label; it does not rewind the network.
A POPULAR-SCIENCE GUIDE
Imagine recalling a book after thousands of other books have already quoted it.
That is roughly the problem. Scientific papers constantly point to earlier papers. Researchers call those pointers citations. They are how a new paper says, in effect, “this earlier work matters to what I am saying here.”
Years later, a journal may put a large warning on one of those earlier papers: retracted. The reason may involve bad data, manipulated images, a flawed analysis, a corrected replacement, or even duplicate publication.
The warning changes the paper's official status. It does not automatically update every later paper that already cited it. It also does not tell us what future authors will do.
That gives us a simple thing to check.
Take a famous retracted paper. Find the papers that mention it after the warning. Then read the words around each mention. Is the later author warning readers? Repeating the old claim? Using an old measurement tool? Pointing to a corrected version? Merely describing the controversy?
Once you do that, “this retracted paper received 1,000 more citations” breaks apart into many different stories.
02
Four examples show why the count alone is misleading
The same word, “retracted,” sits on very different stories.
A TOOL KEPT TRAVELLING
The Morisky medication scale
A 2008 paper helped establish a questionnaire used to measure whether people take their medicines. The paper was retracted in 2023 after the journal lost confidence in its conclusions. Later papers still use and describe the scale.
The thing being inherited may be the tool and its reputation, not one isolated result.THE PAPER WAS REPLACED
The Mediterranean-diet trial
The original 2013 PREDIMED paper was retracted and replaced in 2018 after problems with random assignment were found. Later authors sometimes still point to the old paper while discussing findings that survived the correction.
Here the problem may be an outdated reference rather than a dead scientific idea.THE CLAIM REALLY CAN PERSIST
Wakefield and vaccines
Some later writing still invokes the old vaccine-autism claim. Those cases matter. But other papers cite Wakefield precisely to explain why the claim was discredited.
Counting both as the same kind of citation hides the difference we actually care about.NOT EVERY RETRACTION MEANS FALSE DATA
The ZnO review
This highly cited review was retracted mainly because it duplicated previously published material.
“Cited a retracted paper” does not automatically mean “relied on a false result.”
03
Compare the ten famous papers
Click a paper to see what happened to it and how often later papers still mention it.
Same ten papers; only the order changes.
Choose a paper above to see what happened and whether a corrected version exists.
A retraction is a warning label, not a single diagnosis.
These ten papers include corrected studies, unreliable results, image problems, fabrication, ethical problems and duplicate publication. Before judging a later citation, first ask what happened to the original paper.
04
Why papers get pulled
Retraction Watch records many different reasons. A paper can have more than one. Larger boxes below mean that reason appears more often.
71,844published warnings and updates in this data snapshot
66,317of those are retractions
110different reason labels appear on retractions
Snapshot: 17 August 2026. One paper may have several reasons, so the boxes overlap.
05
Other researchers have already shown that this problem is real
We are not claiming to invent the study of retracted papers. Earlier work gives us useful checks and prevents us from pretending every later citation means the same thing.
READING THE SENTENCE MATTERS
Some later citations criticize the old paper.
Researchers have manually read citation passages and sorted them into supportive, neutral and critical uses. The surrounding words change the meaning of the citation.
THE PROBLEM CAN BE LARGE
Thousands of later papers fail to mention a retraction.
A large biomedical study examined more than 13,000 citation passages written after retraction. Only 5.4% explicitly noted that the earlier paper had been retracted.
ERROR DOES NOT ALWAYS SPREAD FOREVER
Second-generation effects differ from case to case.
One detailed study found no clear spread into papers that only cited the first generation. Another found 23 possible cases. That is a reason to inspect the words, not assume contamination.
THE WARNING SYSTEM IS IMPROVING
Publishers now have better rules for passing retraction notices along.
Crossref, Retraction Watch and NISO already provide much of the plumbing. This site focuses on the remaining human question: what did the next author actually keep using?
06
Wakefield gives us a useful reality check
The vaccine-autism paper is famous enough that researchers have already read and labelled hundreds of later citations by hand.
Not saying “retracted” is not the same as supporting the paper.
Among 337 later papers from 2011 to 2017, 209 did not explicitly say that Wakefield had been retracted. But 112 of those still spoke about the old paper negatively.
That is why this site never treats a missing warning word as proof of misuse. It only marks those cases as worth a closer look.
Based on the open Bologna Wakefield citation dataset.
07
One paper lets us watch the warning travel downstream
This is not proof that one bad paper automatically poisons everything after it. It is simply a rare, visible chain of events.
2002Jiang and colleagues publish a stem-cell paper in Nature.
→
2015A PLOS ONE paper cites Jiang while Jiang is still considered published research.
→
2024Nature retracts Jiang.
→
2026PLOS retracts the 2015 paper for several problems and explicitly notes that its Jiang reference has since been retracted.
The later warning names the earlier paper
This is a literal family-tree moment: a later paper's own retraction notice records that one of its sources changed status after publication.
Try it yourself: open a later citation and read what it says
Pick one of the ten retracted papers. Each dot is a later paper that cites it and for which we recovered nearby text. Click a dot to inspect the words around the citation.
You can help inspect the evidence. Open a dot and answer one question: what does the later paper appear to keep using from the older paper? Your choices stay on your own device unless you export them.
A line means only “this paper cited that paper.” It does not mean the later paper is wrong.
09
Read the examples as sentences, not just numbers
The table below shows a small set of the later papers and a short excerpt around each citation. Use the filters above to narrow it down.
mentions the problem / criticalold version when a replacement existsneeds a closer look
WHAT RETRACTION WATCH DOES WELL
It made a hidden problem visible.
The Top 10 list is memorable and useful. It shows that papers can keep receiving attention long after a journal has withdrawn them. Retraction Watch also maintains a much richer open database and has reported on studies that distinguish criticism from continued reliance.
WHY ITS SIMPLE PAGE STOPS SHORT
Counting references is easy. Reading thousands of papers is not.
A citation database can tell you that paper B points to paper A. The meaning lives in the surrounding sentence, often scattered across many publishers and access systems. So a simple leaderboard naturally stops at the count.
WHAT WE WOULD CHANGE
Do not let the number imply more than it knows.
The phrase “citations after retraction” mixes together criticism, history, accidental use of an old version, genuine reuse of a disputed claim and cases we simply cannot judge. A better leaderboard should make those differences visible from the start.
THE BIGGER IDEA
Science does not only pass facts from paper to paper.
It also passes tools, categories, methods, examples, habits, questions and familiar stories. A retraction is interesting because it gives us a dated event: the official status of one paper suddenly changes. We can then look downstream and ask what later work keeps carrying anyway.
That turns the scientific literature into something closer to a family tree. A descendant can inherit one feature while leaving another behind.
older paper→later paper cites it→what got carried over?→next generation
10
How we checked it
This part is here so you can repeat the work or disagree with us. It is written in the same plain language as the rest of the page.
STEP 1: GET THE WARNING DATES
We use Retraction Watch's open database, distributed by Crossref, to find when each paper was retracted and why.
STEP 2: FIND LATER PAPERS
We use OpenCitations, a free citation database, to find papers that point to one of the ten after its retraction date. Duplicate links are removed.
STEP 3: READ NEAR THE CITATION
Where Semantic Scholar has the surrounding text, we save a short passage around the citation. For Wakefield we also use a separate dataset that people already labelled by hand.
STEP 4: KEEP UNCERTAINTY VISIBLE
If the text does not clearly show criticism, support or use of an old version, we call it “needs a closer look.” We do not silently turn uncertainty into guilt.
One small date wrinkle:
Some citation records give only a year or a year and month, not an exact day. When that happens we keep borderline cases rather than pretending we know the exact order. The total of 6,458 may therefore include a small number of citations from the same month as a retraction.
Retraction Watch and Crossref provide the retraction records. OpenCitations provides the open citation links. Earlier studies by Bar-Ilan & Halevi, van der Vet & Nijveen, Schneider and colleagues, and Hsiao & Schneider already showed why citation context matters. NISO's CREC recommendations address how retraction warnings should travel through publishing systems.