Skip to main content

Were later-retracted papers accepted faster?

Yes, for some kinds of retraction. We compared 20,471 retracted research papers in PubMed with 150,703 papers from the same journal and year that were not retracted. The median retracted paper went from received to accepted 12.5 days faster than its journal-year norm. The gap comes from papers retracted for peer-review manipulation (27 days faster) and paper mills (18.5 days). Papers retracted for honest error were accepted at the usual pace. Most of the gap comes from the mass retractions of 2021 to 2024: counting each journal once, the typical difference is 1.5 days.

Published 7 October 2026 · Analysis plan written 2 October 2026 and published with the results; not publicly registered · All figures measured from PubMed · By Erik Jia. Spotted an error? erik@manusights.com

Retracted papers
20,471
Comparison papers
150,703
Journals
1,790
Published
2012 to 2024
Plan written
2 October 2026

By retraction reason

Peer-review manipulation

Retracted: 3,626

Median gap, days (95% CI): -27 (-41 to -18)

Faster (95% CI): 79.2% (77.9 to 80.5)

Call: Faster

Paper mill

Retracted: 5,921

Median gap, days (95% CI): -18.5 (-31 to -11.5)

Faster (95% CI): 70.5% (69.4 to 71.7)

Call: Faster

Other or unspecified

Retracted: 3,934

Median gap, days (95% CI): -4 (-8 to -1.5)

Faster (95% CI): 53.7% (52.2 to 55.3)

Call: Faster

Plagiarism or duplication

Retracted: 798

Median gap, days (95% CI): -1.5 (-5.5 to +2)

Faster (95% CI): 51.4% (47.9 to 54.8)

Call: No clear difference

Data or image problems

Retracted: 5,732

Median gap, days (95% CI): +1.5 (-0.5 to +3)

Faster (95% CI): 48.7% (47.4 to 50)

Call: No clear difference

Honest error

Retracted: 460

Median gap, days (95% CI): -1 (-7 to +4)

Faster (95% CI): 50.7% (46.1 to 55.2)

Call: No clear difference

A negative gap means the retracted paper was accepted faster than the median comparison paper in its journal and year. "Faster" is the share of retracted papers that beat that median. "Call" applies the decision rule fixed in the analysis plan. Reasons come from the Retraction Watch Database, matched to 17,230 of the 20,471 papers; "Other or unspecified" includes papers with no match. Reason classes were set in the analysis plan. A paper with several reasons takes the first match, in the order paper mill, peer-review manipulation, data or image problems, plagiarism or duplication, honest error.

What it means

  • 61.5% of retracted papers were accepted faster than the median paper in their journal that year (95% CI 60.9 to 62.2%). With no difference, that share would be 50%.
  • Papers retracted for peer-review manipulation were the fastest: 79.2% beat their journal-year median, by a median 27 days. For paper-mill retractions it was 70.5% and 18.5 days.
  • Papers retracted for honest error (50.7%), data or image problems (48.7%) and plagiarism or duplication (51.4%) showed no clear difference.
  • The gap appears in papers published from 2021. For papers published 2012 to 2019 taken together there was no clear difference (51.1% faster, median -1.5 days), although 2013 (6 days faster) reaches the "faster" call on its own.

In other words, we found no speed gap for honest error or data problems, though the honest-error group is small (460 papers, 95% CI -7 to +4 days) and cannot rule out a modest difference. The gap we did find tracks organised manipulation of the review process, the kind behind the mass retractions of 2022 and 2023.

Where the effect lives

The overall figure is pulled up by the journals and years hit hardest by mass retractions. Three checks set in the analysis plan show how concentrated it is.

Each journal counted once

Across 1,790 journals, the median journal-level difference was -1.5 days (95% CI -3 to 0). 925 journals leaned faster and 848 slower.

Without Hindawi journals

13,651 retracted papers remain. 54.3% were faster, by a median 5.5 days (95% CI -9 to -2.5).

Without the three largest publishers

7,420 retracted papers remain. 54.1% were faster, by a median 5 days (95% CI -11 to -0.5).

So the typical journal shows a difference of a day or two. The effect is real outside the largest mass retractions, but it is smaller there. A fast acceptance in an ordinary journal tells you very little about any one paper.

By publication year

Every year from 2021 to 2024 shows retracted papers accepted faster, by a median of 17 to 22.5 days. No year from 2014 to 2020 does; of the earlier years, only 2013 (6 days faster) does.

2012

Retracted: 398

Median gap, days (95% CI): -4 (-9 to +0.8)

Faster (95% CI): 53.8% (48.9 to 58.6)

Call: No clear difference

2013

Retracted: 574

Median gap, days (95% CI): -6 (-11 to -1.5)

Faster (95% CI): 55.4% (51.3 to 59.4)

Call: Faster

2014

Retracted: 716

Median gap, days (95% CI): -4.5 (-11.8 to +1.5)

Faster (95% CI): 54.7% (51.1 to 58.4)

Call: No clear difference

2015

Retracted: 823

Median gap, days (95% CI): -2 (-10 to +1)

Faster (95% CI): 52% (48.6 to 55.4)

Call: No clear difference

2016

Retracted: 775

Median gap, days (95% CI): -4 (-9.5 to +0.5)

Faster (95% CI): 54.2% (50.7 to 57.7)

Call: No clear difference

2017

Retracted: 980

Median gap, days (95% CI): 0 (-4 to +5)

Faster (95% CI): 49.4% (46.3 to 52.5)

Call: No clear difference

2018

Retracted: 1,213

Median gap, days (95% CI): 0 (-5 to +5)

Faster (95% CI): 49.6% (46.8 to 52.4)

Call: No clear difference

2019

Retracted: 1,520

Median gap, days (95% CI): +3.3 (-1 to +6)

Faster (95% CI): 47.1% (44.6 to 49.6)

Call: No clear difference

2020

Retracted: 1,791

Median gap, days (95% CI): -5 (-11.5 to +0.8)

Faster (95% CI): 53.5% (51.2 to 55.8)

Call: No clear difference

2021

Retracted: 2,791

Median gap, days (95% CI): -17 (-32.5 to -7.5)

Faster (95% CI): 64.5% (62.7 to 66.3)

Call: Faster

2022

Retracted: 6,062

Median gap, days (95% CI): -18 (-37 to -10.5)

Faster (95% CI): 72.8% (71.7 to 73.9)

Call: Faster

2023

Retracted: 1,393

Median gap, days (95% CI): -19.5 (-30 to -8.5)

Faster (95% CI): 64.6% (62.1 to 67.1)

Call: Faster

2024

Retracted: 1,435

Median gap, days (95% CI): -22.5 (-34 to -6)

Faster (95% CI): 66.3% (63.8 to 68.7)

Call: Faster

Recent papers have had less time to be retracted, and some comparison papers will be retracted later. Both push the result toward no difference, so recent years are, if anything, understated.

Robustness checks

The analysis plan declared seven checks. Six reach the same call. Restricting to 2012 to 2019 does not, which fits the year table, where 2013 (6 days faster) is the only earlier year with a "faster" call.

Without Hindawi journals

Retracted: 13,651

Median gap, days (95% CI): -5.5 (-9 to -2.5)

Faster (95% CI): 54.3% (53.5 to 55.1)

Call: Faster

Without the three publishers with the most retractions

Retracted: 7,420

Median gap, days (95% CI): -5 (-11 to -0.5)

Faster (95% CI): 54.1% (52.9 to 55.2)

Call: Faster

Only journal-years with 5 or more comparison papers

Retracted: 20,451

Median gap, days (95% CI): -12.5 (-17.5 to -8.5)

Faster (95% CI): 61.6% (60.9 to 62.2)

Call: Faster

Without labelled supplements and special issues

Retracted: 20,457

Median gap, days (95% CI): -12.5 (-17.5 to -8.5)

Faster (95% CI): 61.5% (60.9 to 62.2)

Call: Faster

Without journals whose received date looks like a revision date

Retracted: 20,409

Median gap, days (95% CI): -13 (-17.5 to -8.5)

Faster (95% CI): 61.6% (60.9 to 62.3)

Call: Faster

Only journals that also deposit a revised date

Retracted: 12,386

Median gap, days (95% CI): -16 (-24 to -11)

Faster (95% CI): 66.8% (66 to 67.6)

Call: Faster

Only papers published 2012 to 2019

Retracted: 6,999

Median gap, days (95% CI): -1.5 (-3 to 0)

Faster (95% CI): 51.1% (49.9 to 52.2)

Call: No clear difference

The check using only journals that also deposit a revised date matters most for interpretation. In those journals the received date is clearly the original submission, and the gap is larger (-16 days), not smaller.

Tougher checks (added 9 October 2026)

An independent review of this study raised four concerns: the confidence intervals, special issues, how some retractions were detected, and whether speed is a useful flag. We wrote four extra analyses into a plan before re-running the study from a fresh PubMed download. The re-run reproduced the result: 20,499 paired retracted papers, a median gap of -12.5 days, the same call.

Let a whole publisher move together in the confidence interval

The 95% CI was -17.5 to -8 days. With publishers clustered it is -17.5 to -2 days: still below zero, but much less precise.

Drop journal-years with 10 or more retractions

This removes 66.0% of retracted papers. The rest were accepted a median 2 days faster (95% CI -3.5 to -0.5), and 51.7% beat their journal-year median.

Peer-review manipulation, same journal-years dropped

335 papers, 32 days faster (95% CI -39.8 to -23.8).

Paper-mill papers that also name a peer-review problem

4,455 papers, 23 days faster; 79.0% beat their median.

Paper-mill papers with no peer-review reason

1,466 papers, 6 days slower than their journal-year median (95% CI +1 to +9.5).

So the timing signal belongs to papers whose peer review was compromised, not to paper mills as such. Outside the mass-retraction journal-years, the overall gap is about two days.

As a screening flag, speed is weak. Call a paper fast if its received-to-accepted interval is under half its journal-year median. That catches 23.0% of retracted papers and 9.2% of the others. In journal-years with at least one retraction, at most about 1 fast paper in 53 was later retracted (1.9%). Outside mass-retraction journal-years it is about 1 in 162.

Some compromised-review retractions were found partly because review was implausibly fast, so that group's gap is partly how the papers were identified. The screening figures are upper bounds for journals in general: every journal-year analysed had a retraction, and some comparison papers will be retracted later. Aggregates only: addendum results (JSON).

By publisher

Publisher is the Crossref member that registered each journal's DOIs. We report the 23 publishers with at least 30 retracted papers; 155 smaller groups are left out.

Hindawi Limited

Retracted: 6,822

Median gap, days (95% CI): -20 (-38 to -13)

Faster (95% CI): 76% (75 to 77)

Call: Faster

Elsevier BV

Retracted: 3,173

Median gap, days (95% CI): -3 (-7 to 0)

Faster (95% CI): 52.6% (50.9 to 54.3)

Call: No clear difference

Springer Science and Business Media LLC

Retracted: 3,056

Median gap, days (95% CI): -9 (-15 to -4.5)

Faster (95% CI): 56.6% (54.8 to 58.3)

Call: Faster

Wiley

Retracted: 1,775

Median gap, days (95% CI): -16 (-30.8 to +1)

Faster (95% CI): 60.5% (58.2 to 62.7)

Call: No clear difference

Public Library of Science (PLoS)

Retracted: 1,242

Median gap, days (95% CI): -30 (-51.5 to -10)

Faster (95% CI): 65.5% (62.8 to 68.1)

Call: Faster

Spandidos Publications

Retracted: 1,050

Median gap, days (95% CI): +3 (0 to +6)

Faster (95% CI): 46.6% (43.6 to 49.6)

Call: No clear difference

Frontiers Media SA

Retracted: 561

Median gap, days (95% CI): -3 (-12.5 to +7.5)

Faster (95% CI): 51.9% (47.7 to 56)

Call: No clear difference

Informa UK Limited

Retracted: 437

Median gap, days (95% CI): +5 (-5 to +9.5)

Faster (95% CI): 44.9% (40.3 to 49.5)

Call: No clear difference

MDPI AG

Retracted: 426

Median gap, days (95% CI): +0.5 (-1 to +2.5)

Faster (95% CI): 46.5% (41.8 to 51.2)

Call: No clear difference

Portland Press Ltd.

Retracted: 251

Median gap, days (95% CI): -2 (-10 to +10.5)

Faster (95% CI): 51.8% (45.6 to 57.9)

Call: No clear difference

Impact Journals, LLC

Retracted: 197

Median gap, days (95% CI): -6 (-16 to 0)

Faster (95% CI): 54.8% (47.8 to 61.6)

Call: No clear difference

Royal Society of Chemistry (RSC)

Retracted: 195

Median gap, days (95% CI): +3 (-3 to +7.5)

Faster (95% CI): 43.1% (36.3 to 50.1)

Call: No clear difference

S. Karger AG

Retracted: 179

Median gap, days (95% CI): +24 (-18 to +35)

Faster (95% CI): 37.4% (30.7 to 44.7)

Call: No clear difference

Publisher not identified

Retracted: 106

Median gap, days (95% CI): +3 (-3 to +9)

Faster (95% CI): 45.3% (36.1 to 54.8)

Call: No clear difference

Oxford University Press (OUP)

Retracted: 90

Median gap, days (95% CI): -5.8 (-14.5 to +15.8)

Faster (95% CI): 51.1% (41 to 61.2)

Call: No clear difference

American Chemical Society (ACS)

Retracted: 85

Median gap, days (95% CI): +1 (-6 to +10)

Faster (95% CI): 49.4% (39 to 59.8)

Call: No clear difference

SAGE Publications

Retracted: 65

Median gap, days (95% CI): +1 (-42.5 to +25)

Faster (95% CI): 49.2% (37.5 to 61.1)

Call: No clear difference

Bentham Science Publishers Ltd.

Retracted: 39

Median gap, days (95% CI): 0 (-16.5 to +8)

Faster (95% CI): 41% (27.1 to 56.6)

Call: No clear difference

FapUNIFESP (SciELO)

Retracted: 37

Median gap, days (95% CI): -1.5 (-27.8 to +7.5)

Faster (95% CI): 51.4% (35.9 to 66.6)

Call: No clear difference

Walter de Gruyter GmbH

Retracted: 36

Median gap, days (95% CI): -24.8 (-39.5 to +2.3)

Faster (95% CI): 63.9% (47.6 to 77.5)

Call: No clear difference

e-Century Publishing Corporation

Retracted: 36

Median gap, days (95% CI): +2 (-1 to +5)

Faster (95% CI): 44.4% (29.5 to 60.4)

Call: No clear difference

AME Publishing Company

Retracted: 34

Median gap, days (95% CI): -11 (-28.5 to +0.5)

Faster (95% CI): 70.6% (53.8 to 83.2)

Call: No clear difference

American Association for Cancer Research (AACR)

Retracted: 31

Median gap, days (95% CI): +1 (-28.8 to +36.5)

Faster (95% CI): 48.4% (32 to 65.2)

Call: No clear difference

A publisher's count reflects where retracted papers with deposited dates are concentrated, not how common misconduct is there. Hindawi's count reflects the mass retractions of its special-issue papers announced in 2022 and 2023. This table describes papers published 2012 to 2024. It is not an assessment of any publisher's current review process.

What this does not show

  • It does not show that any individual paper was reviewed improperly. The comparison is descriptive, and most quickly accepted papers are never retracted.
  • Received and accepted dates are deposited by publishers. Some journals reset the received date when a revision arrives. The checks above that exclude or isolate those journals give the same answer.
  • Only papers with both dates in PubMed are included. 4,451 retracted records had no usable received and accepted dates.
  • Retracted papers are the ones that were caught. A process that is good at hiding manipulation would not show up here.
  • Some papers retracted for compromised peer review may have been flagged partly because their review was unusually fast. If so, part of that group's gap reflects how the papers were found, not only how they were handled.
  • Much of the mass retraction came from special issues. Hindawi's special-issue papers were already known to be accepted faster than its regular papers: 56 days faster in 2022 (Petrou, 2025). Our special-issue check relies on PubMed labels and removed only 14 papers, so part of the overall gap may reflect the special-issue route rather than retraction as such. Dropping journal-years with 10 or more retractions, a rough stand-in, leaves a gap of about 2 days (see tougher checks).
  • Confidence intervals resample journal-years. Retractions also cluster by publisher and across years, so intervals that let a whole publisher move together would be wider. Clustering by publisher widens the interval to -17.5 to -2 days. The each-journal-once figure (-1.5 days) is the more conservative reading.
  • We searched for earlier paper-level studies of this question and found none, but that search was not exhaustive (see prior work).

Method

We enumerated every PubMed record tagged "retracted publication" and published 2011 to 2025 (30,172 records), kept research articles published 2012 to 2024 with one received date and one accepted date, and dropped front matter and intervals outside 1 to 1,825 days. That left 20,483 retracted papers.

For each journal and publication year containing a retracted paper, we drew a uniform random sample of non-retracted research articles from the same journal and year, up to 10 per retracted paper, checked against the same rules. We examined 182,006 candidate records and kept 150,703. 20,471 retracted papers had at least one comparison paper.

Each retracted paper's interval was compared with the median interval of its comparison papers. Confidence intervals for medians come from 10,000 bootstrap resamples of journal-years, so papers from the same journal and year move together. Shares use Wilson intervals. The analysis plan set the decision rule: "faster" only if the interval for the median difference lies entirely below zero and the lower bound of the share faster is above 50%.

We wrote the analysis plan on 2 October 2026, before fetching any dates. It was not lodged with a public registry, and it was first published together with the results on 7 October 2026, so that order cannot be checked independently. The later addendum plan was committed before its re-run.

A second script recomputed the headline and the Hindawi-excluded figures independently from the frozen inputs and matched them. One change was made during the run: the downloader was set to retry an intermittent PubMed error, which changed no analytic choice. No language model was used in the analysis.

Prior work

Earlier studies looked at parts of this question. None compared retracted and non-retracted papers within the same journal and year across PubMed.

  • Mascato Fontaíña and colleagues described 142 journals that retracted paper-mill papers, including their editorial times (Research Integrity and Peer Review, 2025). Their comparison is at the journal level.
  • Sevryugina and Jimenez measured peer-review time in 1,292 retracted chemistry papers (ACS Omega, 2023), and Abo-Elnour and colleagues did the same for 590 retracted pediatrics papers (European Journal of Pediatrics, 2025). Neither had non-retracted comparison papers.
  • Christos Petrou compared special-issue and regular papers at Hindawi, and publishing-speed trends at Frontiers and PLOS (The Scholarly Kitchen, 2025).

The widely cited studies of retracted papers ask a different question: why papers were retracted, and how long retraction took. This study looks at the other end of a retracted paper's life, its acceptance.

  1. Fang, F. C., Steen, R. G., & Casadevall, A. (2012). Misconduct accounts for the majority of retracted scientific publications. Proceedings of the National Academy of Sciences, 109(42), 17028-17033. https://doi.org/10.1073/pnas.1212247109

    Reviewed all 2,047 biomedical and life-science articles indexed as retracted in PubMed on 3 May 2012: 67.4% were retracted for misconduct and 21.3% for error.

  2. Steen, R. G., Casadevall, A., & Fang, F. C. (2013). Why has the number of scientific retractions increased? PLoS ONE, 8(7), e68397. https://doi.org/10.1371/journal.pone.0068397

    Measured the time from publication to retraction for the same 2,047 articles, 32.91 months on average, and found it shorter in high impact factor journals.

  3. Steen, R. G. (2011). Retractions in the scientific literature: is the incidence of research fraud increasing? Journal of Medical Ethics, 37(4), 249-253. https://doi.org/10.1136/jme.2010.040923

    Classified the reasons for 742 English-language papers retracted in PubMed between 2000 and 2010.

Data sources

PubMed (NCBI E-utilities)

Retracted-publication records, received and accepted dates, journal and year

Crossref REST API

Publisher (Crossref member) for each DOI prefix

The downloads carry every number on this page, with both kinds of confidence interval, under CC BY 4.0. They contain aggregates only. No paper, author, editor or journal is listed.

Questions

Were retracted papers accepted faster than other papers?

On average, yes. Across 20,471 retracted research papers in PubMed published 2012 to 2024, the median retracted paper was accepted 12.5 days faster than the median paper in the same journal and year, and 61.5% were faster than that median. The gap comes from papers retracted for peer-review manipulation and paper-mill activity.

Does a fast acceptance mean a paper is unreliable?

No. Papers retracted for honest error, data or image problems, or plagiarism were accepted at the usual pace, and across 1,790 journals the typical journal showed a difference of only 1.5 days. Most fast papers are never retracted. Speed is a pattern in organised manipulation, not a test for an individual paper.

Is a fast acceptance a good red flag for a bad paper?

Not on its own. Defining fast as under half the journal-year median, 23.0% of retracted papers were fast, but so were 9.2% of papers that were never retracted. In journal-years that had a retraction, at most about 1 fast paper in 53 was later retracted, and about 1 in 162 outside mass-retraction events.

Which retractions show the biggest speed gap?

Papers retracted for peer-review manipulation were accepted a median 27 days faster than their journal-year median, and 79.2% were faster. Papers retracted for paper-mill activity were 18.5 days faster, and 70.5% were faster.

Cite this resource

Citation formats

Free to use and cite, with attribution. See the reuse policy for what you may republish.

APA
Manusights. (2026). Were later-retracted papers accepted faster? A within-journal comparison of 20,471 retracted papers in PubMed (Version 2026.10.02-retraction-speed-v1) [Data set]. Zenodo. https://doi.org/10.5281/zenodo.23353186
MLA
Manusights. "Were Later-Retracted Papers Accepted Faster?" Version 2026.10.02-retraction-speed-v1, Manusights, 7 Oct. 2026, https://manusights.com/research/retraction-acceptance-speed. Data set, Zenodo, https://doi.org/10.5281/zenodo.23353186.
Vancouver
Manusights. Were later-retracted papers accepted faster? [Internet]. Version 2026.10.02-retraction-speed-v1. 2026 Oct 7. Available from: https://manusights.com/research/retraction-acceptance-speed. Data set: https://doi.org/10.5281/zenodo.23353186

Keep going

What do you need help with next?

Find another practical resource or follow a guided publishing path when you want deeper support across the process.