Monday, July 27, 2026

Why Measuring Evolutionary Rate Is Harder Than It Sounds

At first, measuring evolutionary rate sounds easy. Choose a trait, measure it in old fossils and younger fossils, divide by time, and voilà: evolutionary speed.

But Simpson shows that the problem is much trickier.

The first complication is what to measure. Should we measure body size, tooth length, skull width, limb proportions, or some composite index of overall form? Different traits can evolve at different rates. A lineage may show rapid dental change because of dietary shifts, but little change in body size. Another may evolve new limb proportions while the skull remains conservative.

The second complication is scale. Evolutionary change can be measured within a population, between species, across genera, or across larger groups. A rate calculated for a short interval may look very different from a rate calculated across millions of years. Short-term rates may appear fast because they capture local fluctuations. Long-term rates may seem slower because they average out bursts, reversals, and pauses.

The third complication is variation. Organisms are not identical. A fossil sample contains individuals that differ in age, sex, environment, and genetic background. If we measure only one specimen, we may mistake individual variation for evolutionary transformation.

The fourth complication is time resolution. Fossils are usually dated in broad geological intervals. If a trait changed rapidly during a small part of that interval, the average rate may underestimate the true speed of change.

This means the evolutionary rate is not a single number waiting politely to be discovered. It is a constructed estimate, shaped by the trait under study, the available fossils, the timescale used, and the assumptions made.

Simpson’s deeper message is methodological: before asking “how fast did evolution happen?”, we must ask “fast in what, over what interval, in which lineage, and measured how?”

The speedometer of evolution works, but only when calibrated carefully.

Is Scientific Reports a Predatory Journal? A Criterion-by-Criterion Audit

The answer depends on what “predatory” actually means

Scientific Reports is one of the world’s largest open-access journals. It publishes across biology, medicine, physics, chemistry, Earth science, psychology and engineering. It charges authors an article processing fee, publishes a very large number of papers and regularly advertises guest-edited collections.

To some researchers, that combination triggers an immediate suspicion:

Is Scientific Reports a predatory journal wearing the Nature logo?

The evidence-based answer is no.

Scientific Reports is a legitimate, peer-reviewed, indexed open-access journal published by Nature Portfolio. It does, however, possess several features that are sometimes loosely associated with predatory publishing: a broad scope, high publication volume, author-paid fees and active solicitation of manuscripts.

Those similarities are superficial. Predatory publishing is not defined by size, open access or the existence of a publication charge. It is defined primarily by deception, absence of meaningful editorial services, sham peer review, hidden financial practices and failure to maintain the scholarly record.

A useful evaluation therefore requires more than asking whether a journal charges money. It requires opening the machinery and inspecting the gears.


What is a predatory journal?

The World Association of Medical Editors, or WAME, describes predatory or pseudo-journals as publications that collect article processing charges without meeting accepted standards of scholarly publishing. Such journals may claim to conduct peer review while publishing most submissions without genuine external scrutiny. They may also lack clear policies for conflicts of interest, corrections, retractions, archiving and fee disclosure.

WAME also warns against deciding that a journal is predatory on the basis of a single characteristic or blacklist. Some warning signs can occur in legitimate journals, particularly new journals and journals based in lower-income countries. The more reliable approach is to look for a cluster of serious red flags, especially deception and unverifiable claims.

For this audit, the journal is evaluated using criteria drawn primarily from:

  1. WAME’s framework for identifying predatory or pseudo-journals.
  2. The Think. Check. Submit. approach to journal evaluation.
  3. Independent information recorded by the Directory of Open Access Journals, or DOAJ.
  4. The journal’s own publicly available editorial, peer-review and publishing policies.

Official journal pages are useful for determining what a journal claims to do. Independent databases such as DOAJ are useful for checking whether at least some of those claims have been externally recorded or reviewed.


The audit

1. Is the publisher clearly identified?

What should be checked?

A legitimate journal should clearly identify:

  • the publisher;
  • the journal title and ISSN;
  • the people responsible for editorial management;
  • submission and support channels;
  • its relationship to any parent publishing organization.

Predatory operations often conceal their ownership, use an unverifiable address or adopt a name resembling that of an established journal.

What does Scientific Reports show?

The journal identifies itself as a Nature Portfolio journal and gives the online ISSN 2045-2322. Its contact page identifies the chief editor, deputy editors and senior editorial staff. Manuscripts are submitted through Springer Nature’s formal submission system, while author and reviewer queries are directed through identifiable support channels.

DOAJ independently records Nature Portfolio as the publisher and the United Kingdom as the publisher country. It also records the same ISSN.

Verdict: Pass

There is no meaningful evidence of concealed ownership, journal hijacking or a fictitious publisher.


2. Is the editorial board identifiable and academically credible?

What should be checked?

A suspicious journal may:

  • provide no named editor;
  • list only two or three unidentified board members;
  • give no affiliations or academic qualifications;
  • invent editorial-board members;
  • list scholars without their permission;
  • use the same board across unrelated journals.

Merely displaying a long list of names is not enough. The names should be connected to relevant disciplines and institutions.

What does Scientific Reports show?

The journal publicly lists its in-house editors, senior editorial board and subject-specific editorial-board members. The board is organized across fields including biology, medicine, physics, chemistry, materials science, Earth and environmental science, engineering and registered reports. Institutional affiliations are provided for board members.

The journal states that editorial-board members must hold a PhD or equivalent research degree and be active, independent researchers in a relevant field. Their responsibilities include determining whether manuscripts should be reviewed, selecting reviewers and making decisions based on reviewer reports.

That does not prove that every editor will handle every paper equally well. It does establish that the journal has a visible and professionally structured editorial system.

Verdict: Pass

The journal does not display the anonymous or evidently fabricated editorial structure commonly associated with predatory publishing.


3. Is the peer-review process clearly described?

What should be checked?

This is one of the most important criteria.

Warning signs include:

  • no mention of peer review;
  • guaranteed acceptance;
  • acceptance within a few days regardless of subject;
  • reviewer reports that are generic or irrelevant;
  • authors being asked to provide reviewers who are accepted without verification;
  • payment being treated as the condition for acceptance.

A journal’s peer-review policy should explain who screens manuscripts, who chooses reviewers, what reviewers evaluate and who makes the final decision.

What does Scientific Reports show?

The journal describes a multistage editorial process.

First, submitted manuscripts undergo initial checks covering authorship, competing interests, ethical approval and plagiarism. A manuscript that passes these checks is assigned to an editorial-board member who is described as an active researcher in the relevant field. The editor decides whether the paper is suitable for external review and normally selects two or three reviewers.

Reviewers are instructed to evaluate methodological and analytical soundness, statistical validity, support for the conclusions, data availability, engagement with the literature and ethical concerns. The handling editor can accept the manuscript, request minor or major revision, or reject it.

Authors may suggest reviewers, but the journal states that these suggestions are not necessarily followed. Authors may also request the exclusion of a limited number of researchers.

This is a conventional peer-review structure. It does not guarantee that every review is rigorous. No large journal can make that guarantee. The relevant question is whether the journal institutionally replaces peer review with payment. Its published process does not indicate that it does.

Verdict: Pass at the journal-policy level

There is no evidence that Scientific Reports systematically operates without peer review. Individual cases of weak review, if found, would demonstrate editorial failure in those cases, not automatically establish that the entire journal is predatory.


4. Does the journal promise implausibly fast publication?

What should be checked?

Predatory journals often promise publication within a few days. Such timelines may not allow enough time to identify reviewers, obtain reports, process revisions and check the final manuscript.

Speed alone is not suspicious. A rapid first decision may simply mean that the journal performs efficient administrative checks or rejects unsuitable manuscripts before external review. The more informative measure is the time from submission to final acceptance.

What does Scientific Reports report?

The journal reports a median of approximately 12 days to the first editorial decision and 138 days from submission to acceptance.

A 12-day first decision may sound fast, but it can include administrative checks and decisions not to send a paper for external review. A median acceptance time of 138 days is approximately four and a half months, which is compatible with reviewer selection, external assessment and revision.

The journal says that it aims to deliver an initial decision within 45 days once manuscripts proceed through its editorial process. It also describes minor and major revision pathways rather than promising automatic publication.

Verdict: Pass

The journal emphasizes efficiency, but its reported acceptance timeline is not an implausible “submit today, publish tomorrow” promise.


5. Are publication charges disclosed transparently?

What should be checked?

Charging an article processing charge, or APC, is not evidence of predatory publishing.

The relevant questions are:

  • Is the fee disclosed before submission?
  • Is the amount clearly stated?
  • Is it clear when the fee becomes payable?
  • Are taxes or additional costs explained?
  • Are waivers or discounts available?
  • Does editorial acceptance occur before payment?

A surprise invoice after acceptance is a significant red flag.

What does Scientific Reports show?

As of July 2026, the journal lists an APC of:

  • $2,850
  • £2,290
  • €2,490

The journal states that the applicable price is determined on the date of acceptance and that local taxes may apply. It also describes waiver and discount arrangements for eligible authors.

DOAJ independently records the same maximum APC amounts, notes that the journal has a waiver policy and describes the journal as conducting anonymous peer review and plagiarism checks.

The charge is substantial, but it is not hidden.

Verdict: Pass

Scientific Reports meets the superficial characteristic “charges authors to publish,” but it does not meet the more important predatory criterion of concealing or unexpectedly imposing those charges.


6. Is the journal’s scope implausibly broad?

What should be checked?

An extremely broad scope can be a warning sign when a small or obscure publisher claims to competently review almost every academic discipline. Some predatory journals use broad scopes to maximize the number of manuscripts from which fees can be collected.

However, broad-scope journals can be legitimate when they possess a sufficiently large editorial and reviewer network. PLOS ONE and several established multidisciplinary journals follow this model.

What does Scientific Reports show?

The journal publishes research across natural sciences, psychology, medicine and engineering. Its official scope includes biology, physics, chemistry, Earth science, biomedical and clinical science, psychology and multiple engineering disciplines.

Its editorial board is correspondingly divided into subject areas. The journal’s model is to assess papers primarily for scientific validity and technical soundness, rather than predicted importance or likely citation impact.

This broad, validity-based structure is characteristic of an open-access mega-journal. It creates a genuine quality-control challenge because a very large and distributed editorial system may produce variability between editors and disciplines. But breadth by itself does not demonstrate fraudulent publishing.

Verdict: Caution, but not a predatory finding

The broad scope is real and should make authors examine the quality of papers and editorial handling within their specific field. It does not establish that the journal is predatory.


7. Does the journal publish an unusually large number of papers?

What should be checked?

WAME identifies unusually large or highly variable publication volumes as a reason for closer examination. A journal that publishes thousands of papers must maintain enough editors, reviewers and integrity staff to process them credibly.

High volume is therefore a risk factor for inconsistency, not a diagnosis of predation.

What does Scientific Reports show?

Scientific Reports is explicitly structured as a high-throughput multidisciplinary journal. It does not reject manuscripts because editors believe the work is insufficiently fashionable, important or likely to attract attention. Instead, its stated threshold is whether the research is original, scientifically valid and technically sound.

This model naturally produces more published papers than a selective journal such as Nature, which applies an additional test of broad importance and exceptional significance.

The consequence is that readers should not interpret the Nature Portfolio branding as meaning that every Scientific Reports paper has passed the editorial selectivity used by Nature. The journals operate under different publication criteria.

Verdict: Caution, but not a predatory finding

The volume may contribute to uneven reviewing and variable article quality. It does not show that peer review is fictitious or that acceptance is purchased.


8. Are indexing and impact claims verifiable?

What should be checked?

Predatory journals frequently:

  • claim indexing in databases that do not include them;
  • describe Google Scholar as a selective indexing service;
  • display invented impact factors;
  • use metrics supplied by obscure companies with names resembling Clarivate or Scopus;
  • advertise misleading rankings without identifying the source or year.

The safest method is to verify indexing in the database itself rather than trusting a logo on the journal website.

What does Scientific Reports show?

The journal states that it is indexed in Web of Science, PubMed, PubMed Central, Scopus, Dimensions, Google Scholar, DOAJ and SAO/NASA ADS.

Its inclusion in DOAJ can be independently confirmed. DOAJ records that the journal has been included since 2011 and that its listing underwent a full review in March 2026.

The journal’s homepage currently reports a 2025 Journal Impact Factor of 4.9, attributing the figure to the 2026 release of Clarivate’s Journal Citation Reports.

An impact factor does not prove that every article is reliable. It does, however, differ fundamentally from an invented metric whose provider cannot be identified.

Verdict: Pass

The journal’s identity and major indexing claims are externally discoverable. There is no indication that it relies on counterfeit metrics.


9. Does it have meaningful research-integrity policies?

What should be checked?

A legitimate journal should have policies covering issues such as:

  • plagiarism and duplicate publication;
  • authorship;
  • conflicts of interest;
  • human and animal research;
  • clinical-trial registration;
  • data and code availability;
  • image manipulation;
  • confidentiality;
  • research with possible harmful applications.

Predatory journals may display generic ethics language copied from elsewhere, but provide no operational detail about how concerns are handled.

What does Scientific Reports show?

The journal publishes extensive policies on authorship, competing interests, plagiarism, duplicate publication, human and animal research, clinical trials, data availability, code availability, digital-image standards and research integrity.

Its initial quality checks are stated to include authorship, competing interests, ethical approvals and plagiarism screening.

For clinical trials, the journal requires registration in a suitable public registry and specifies reporting expectations. It also describes conflict-of-interest obligations for authors, reviewers, editorial-board members and staff.

The presence of policies does not prove flawless enforcement. Nevertheless, these are detailed and publicly inspectable policies, not merely a one-line declaration that “the journal follows international ethics.”

Verdict: Pass

The journal has a developed research-integrity framework inconsistent with the usual profile of a predatory operation.


10. Can the journal correct or retract the literature?

What should be checked?

A journal’s responsibility does not end when an article is published.

A credible journal needs mechanisms for:

  • corrections;
  • retractions;
  • expressions of concern;
  • editorial notes;
  • reader complaints;
  • post-publication scientific criticism;
  • linking updates to the original article.

Predatory journals frequently ignore complaints, remove papers without explanation or refuse to issue corrections because doing so would expose failures in their review process.

What does Scientific Reports show?

The journal has a formal correction and retraction policy. It can publish corrections, addenda, editor’s notes, editorial expressions of concern and retractions. The policy states that most formal updates are linked in both directions with the original article and indexed. Retraction decisions may be based on invalid results, research errors or violations of publication ethics.

The journal also provides a “Matters Arising” route for substantial scientific comments or clarifications concerning published papers. Some submissions may be sent to the original authors and then to independent reviewers.

The existence of retracted papers should not itself be treated as evidence that a journal is predatory. A zero-retraction journal may simply be a journal that never investigates anything. The more useful question is whether problems are publicly identified and connected to the scholarly record.

Verdict: Pass

The journal maintains visible post-publication mechanisms rather than treating publication as an irreversible cash transaction.


11. Is the content permanently archived?

What should be checked?

Predatory journals may disappear suddenly, taking published papers with them. Long-term archiving protects the literature if a journal ceases operation or changes platforms.

Authors should look for established preservation services such as CLOCKSS, Portico, PubMed Central or national-library deposit systems.

What does Scientific Reports show?

DOAJ records preservation arrangements through:

  • CLOCKSS;
  • Portico;
  • PubMed Central;
  • a national digital-preservation programme.

DOAJ also records the journal’s licensing options and deposit-policy information.

Verdict: Pass

The journal has identifiable preservation arrangements and is not dependent solely on the continued existence of one website.


12. Does it use spam-like solicitation or guest-edited collections?

What should be checked?

Unsolicited emails are common across scholarly publishing. They become concerning when they:

  • bear no relationship to the recipient’s research;
  • promise guaranteed or unusually rapid acceptance;
  • imitate a prestigious journal;
  • conceal fees;
  • use excessive flattery;
  • pressure authors to submit immediately.

Special issues and guest-edited collections also deserve scrutiny because weak oversight can turn them into volume engines.

What does Scientific Reports show?

The journal actively promotes calls for papers and guest-edited collections. Collection editors may solicit submissions and manage manuscripts, subject to conflict-of-interest rules.

However, the journal states that collection manuscripts are assessed under the standard Scientific Reports editorial criteria and Nature Portfolio policies. In-house editors reserve the right to take over management of a collection.

This reduces, but does not eliminate, the risk of inconsistent collection-level oversight. Researchers should still evaluate the qualifications of the guest editors, the relevance of the collection and the wording of any invitation.

Verdict: Caution

Active solicitation and numerous collections resemble practices used by some predatory publishers. In this case, however, they operate within an identifiable journal and policy framework. They are grounds for vigilance, not enough for a predatory label.


So, which “predatory” characteristics does Scientific Reports actually meet?

It meets four characteristics that are sometimes mistakenly treated as proof of predatory publishing:

CharacteristicDoes it apply?Is it diagnostic?
Authors pay an APCYesNo
The scope is extremely broadYesNo
The journal publishes at high volumeYesNo
It actively promotes calls for papers and collectionsYesNo
Fees are hiddenNoStrong red flag if present
Publisher identity is unclearNoStrong red flag if present
Editorial board is anonymous or unverifiableNoStrong red flag if present
Peer review is not describedNoStrong red flag if present
Publication is guaranteedNoStrong red flag if present
Acceptance occurs implausibly quicklyNo evidenceStrong red flag if present
Indexing claims are obviously falseNo evidenceStrong red flag if present
It uses invented impact factorsNo evidenceStrong red flag if present
Ethics policies are absentNoStrong red flag if present
Correction and retraction systems are absentNoStrong red flag if present
Long-term preservation is absentNoStrong red flag if present

The more accurate criticism of Scientific Reports

Calling Scientific Reports predatory is not supported by the available evidence. But that does not mean the journal should be treated as flawless or equivalent to Nature.

The strongest fair criticism is this:

Scientific Reports operates a very large, distributed editorial system that evaluates technical validity rather than exceptional significance. This creates the possibility of uneven review quality and substantial variation in the novelty, importance and reliability of individual articles.

That criticism is compatible with the journal being legitimate.

A journal can be:

  • legitimate but imperfect;
  • indexed but variably selective;
  • peer-reviewed but inconsistent;
  • reputable without being elite;
  • commercially successful without being predatory.

Scientific publishing is not divided into two magical kingdoms named “excellent” and “fraudulent.” Much of it occupies the foggy continent between them.


Final verdict

Based on publisher transparency, editorial-board structure, documented peer review, fee disclosure, recognized indexing, detailed ethics policies, long-term archiving and formal correction procedures, Scientific Reports should not be classified as a predatory journal.

It is better described as:

A legitimate, high-volume, open-access mega-journal with broad scope, moderate selectivity and potentially heterogeneous article-level quality.

Researchers considering submission should therefore ask two separate questions:

  1. Is this a legitimate scholarly journal?
    Yes.
  2. Is it the strongest and most appropriate journal for this particular paper, field and career objective?
    That depends on the article, the discipline and the alternatives available.

The first question is about predation. The second is about strategy. Confusing them turns a useful journal assessment into a branding argument.

Key references

  1. Laine, C. and Winker, M. A. Identifying Predatory or Pseudo-Journals. World Association of Medical Editors.
  2. Scientific Reports: Editorial Process.
  3. Scientific Reports: Guide to Referees.
  4. Scientific Reports: Editorial and Publishing Policies.
  5. Scientific Reports: Open Access Fees and Funding.
  6. Directory of Open Access Journals entry for Scientific Reports.
  7. Scientific Reports: Editors and Editorial Board.

The Authorship Fingerprint of Retractions: Do Single-Author and Many-Author Papers Fail Differently?

A retraction notice contains a strange little census. Alongside the title, journal, country, subject and reason, it also contains a byline. That byline is not just a list of names. It is a clue.

A single-author retraction often smells different from a twelve-author retraction. A two-author computer-science paper retracted after a peer-review investigation belongs to a different ecosystem than a fourteen-author biomedical paper retracted after image concerns. A 60-author clinical or COVID paper is different again.

Using the uploaded Retraction Watch CSV, I parsed author counts from the semicolon-separated Author field. I treated each semicolon-separated entry as one author entity, so consortium names or study groups listed as one entry count as one entity. This is a practical approximation, not a perfect bibliometric author disambiguation.

The dataset contained:

Dataset sliceRecords
Valid publication and notice dates70,589
Valid dates plus usable author counts70,473
Non-conference records with usable author counts56,428

For the main analysis, I focused on non-conference records because conference-proceedings batches strongly distort author-count and retraction-speed patterns.

The central result:

Author count appears to predict retraction timing in raw data, but the deeper story is that author count predicts what kind of problem the paper had.

More authors usually means more image/data/biomedical-style retractions. Fewer authors often means more peer-review, paper-mill, plagiarism, or proceedings-cleanup style retractions. The author count is not the disease. It is the footprint of the habitat. 🧪📊


1. The raw pattern: more authors, slower retraction, until the extreme tail

In non-conference records, the median author count was 4, and the mean was 4.58. Single-author records made up 14.7%, papers with more than 10 authors made up 5.25%, and papers with more than 20 authors were rare, only 0.34%.

The median retraction lag rises from 1.30 years for single-author papers to 2.52 years for papers with 11 to 20 authors. Then the very large-author group drops slightly, mostly because many very large bylines are linked to fast corrections, retract-and-replace events, COVID-era papers, or large collaborative notices.

Median time to retraction by author-count bucket

Non-conference records only. The 11-20 author bucket has the longest median publication-to-notice lag.

0years0.7years1.4years2.1years2.8years12-34-56-1011-20>20

Calculated from the uploaded Retraction Watch CSV.

Raw correlation supports the first impression. Across non-conference records, the Spearman correlation between author count and retraction lag was ρ = 0.151, with a very small p-value. Across all valid records, including conference material, it was ρ = 0.228.

But this is the tempting trapdoor.

When I fitted a simple adjusted model for log retraction lag, controlling for notice year, society status, subject categories, and broad reason tags, the independent author-count effect almost vanished: p = 0.684. In the same model, reason tags mattered much more. Image concerns, fraud/misconduct, and plagiarism/duplication were associated with longer lag; paper-mill/peer-review issues were associated with shorter lag.

So the correct interpretation is not:

“More authors cause slower retractions.”

It is:

“Many-author papers are more often found in subjects and reason categories that take longer to investigate.”

Authorship is a proxy. It is the smoke, not necessarily the fire.


2. Author count predicts the reason profile very strongly

The strongest pattern in the data is not timing. It is failure mode.

Single-author and low-author records are dominated by peer-review, paper-mill, and process-related tags. As author count rises, image concerns and fraud/misconduct tags rise sharply.

Retraction reason profiles change with author count

Non-conference records only. Reason categories overlap, so percentages do not sum to 100.

Paper mill / peer-review / AI
Image concerns
Fraud / misconduct
Plagiarism / duplication
0%15%30%45%60%12-34-56-1011-20>20

Calculated from the uploaded Retraction Watch CSV.

This is the big result.

Author bucketDominant signal
1 authorPaper-mill/peer-review/process issues, plagiarism/duplication
2-3 authorsStill strongly peer-review/paper-mill dominated
4-5 authorsMixed zone
6-10 authorsImage and data concerns rise
11-20 authorsHighest image and fraud/misconduct share
>20 authorsRare, heterogeneous, often large collaborations or clinical/public-health papers

A logistic model confirmed this pattern after controlling for society status, subject and notice year. For every e-fold increase in author count, meaning roughly multiplying the number of authors by 2.7:

Outcome tagDirection with higher author count
Paper-mill/peer-review/AI tagLower odds, OR = 0.637
Image concernHigher odds, OR = 1.95
Fraud/misconductHigher odds, OR = 1.44
Plagiarism/duplicationSlightly lower odds, OR = 0.937

That is a useful signature. Low-author retractions often look like publication-process failures. Many-author retractions more often look like data, image, and biomedical-forensic failures.


3. Time trend: the median author count is stable, but the composition changes

The median author count in non-conference retraction records is surprisingly stable, usually around 4. But underneath that calm median, the composition wriggles like a box of eels.

The share of single-author records spikes in years associated with batch corrections and low-author journal clusters. The share of >10-author records rises in some recent years, especially 2022, 2024, 2025 and partial 2026.

Single-author and many-author retraction records over time

Non-conference records only. The year 2026 is partial in the uploaded file.

0%7%14%21%28%20002002200420062008201020122014201620182020202220242026

Calculated from the uploaded Retraction Watch CSV.

The 2023 pattern is especially telling: the year had many retractions, but the >10-author share dropped to 2.98%, while single-author records rose to 18.49%. That fits the earlier observation that 2023 was heavily shaped by publisher-wide and paper-mill/special-issue cleanup.

The 2024 and 2025 pattern is different: fewer single-author records and more >10-author records. That suggests a shift toward biomedical, image/data, clinical, review, or multi-author investigations.

A useful era summary:

Notice eraMedian authorsSingle-author share>10-author share
≤2009413.5%3.9%
2010-2014411.8%4.8%
2015-2019417.9%5.3%
2020-2026414.2%5.4%

So the median hardly moves, but the tails do.


4. Subject matters: humanities and social sciences are low-author worlds, biology and medicine are many-author worlds

Author count is strongly shaped by subject.

Biology and medicine have higher author counts, while humanities and social sciences have many single-author records. This is expected from normal field culture, but the retraction patterns mirror it beautifully.

Author-count signatures by subject

Non-conference records only. Humanities and social sciences have many single-author records; biology and medicine have more large teams.

Single-author records
>10-author records
0%20%40%60%80%Biology / life sc...Health sciences /...Physical sciences...Social sciencesEnvironmental sci...Humanities

Calculated from the uploaded Retraction Watch CSV. Subject categories can overlap.

The subject-reason interaction is important:

SubjectMedian authorsDominant author-count interpretation
Biology/life sciences5More image/data/fraud signals, longer investigations
Health sciences/medicine5Multi-author clinical and biomedical work, mixed faster and slower pathways
Physical sciences/engineering4Mixed, includes paper-mill/special-issue and materials-image patterns
Social sciences2Many low-author records, peer-review/process and plagiarism signals
Humanities1Strong single-author culture, plagiarism/duplication and process issues
Environmental sciences3Many low-author batch-like records, but with some multi-author ecological/climate studies

The adjusted correlations also show this. Author count correlated with lag in biology, medicine, physical sciences and environmental sciences. But it did not meaningfully correlate with lag in social sciences or humanities. In those fields, the author-count range is too compressed, and the retraction machinery is often driven by plagiarism, policy, or peer-review concerns rather than image/data forensics.


5. Country-specific patterns: author count is a collaboration fingerprint

Country analysis needs caution. I used an exploded-country approach: if a record listed China and the United States, it counted once for China and once for the United States. This does not assign responsibility. It maps affiliation presence.

The country patterns are striking.

Country author-count signatures among retracted records

Top countries by non-conference country-paper occurrences. Multi-country papers are counted once for each country listed.

Single-author records
>10-author records
0%15%30%45%60%ChinaUnited StatesIndiaRussiaSaudi ArabiaIranUnited KingdomJapanPakistanGermanySouth KoreaEgyptItalyFranceCanada

Calculated from the uploaded Retraction Watch CSV.

Several country signatures emerge.

Russia: the single-author and plagiarism/duplication signature

Russia has a median author count of 2, with 43.4% single-author records. Its plagiarism/duplication/copyright theme is very high, about 72.1% in this non-conference country slice. This is a very different profile from image-heavy biomedical retractions.

Italy, France, Germany, Canada and the United States: many-author long-tail worlds

Italy has the highest >10-author share among the large country groups shown: 23.7%. France, Germany, Canada and the United States also have high >10-author shares.

These countries also have substantial biomedical, clinical, institutional, and long-tail correction profiles. Many-author papers here are often not paper-mill style records. They are more likely to be clinical, biomedical, collaboration-heavy, or image/data investigation records.

Saudi Arabia and Pakistan: many-author, multinational collaboration signatures

Saudi Arabia and Pakistan have high median author counts, both around 6, and high >10-author shares. They also have very high multinational shares in the earlier country analysis. In other words, their author-count pattern is partly a collaboration-network pattern.

China and India: huge counts, moderate author numbers, strong process signals

China has the largest record count by far, median 4 authors, with 14.8% single-author records and only 3.9% >10-author records. Its non-conference records are heavily associated with peer-review/paper-mill/publisher-investigation themes.

India has median 4 authors, low single-author share (4.6%), and modest >10-author share (4.3%). Its profile is also strongly shaped by peer-review and publisher-investigation clusters.

So again, author count is not national character. It is publication ecology.


6. Publishers and journals: the author-count worlds are completely different

The strongest author-count patterns appear when we look at publisher and journal ecosystems.

Some publishers have many low-author records associated with paper-mill or peer-review cleanup. Others have higher-author biomedical, clinical, or image-heavy portfolios.

Publisher ecosystems differ by author-count profile

Non-conference records only. Selected publishers with large record counts are shown.

Single-author records
>10-author records
0%6%12%18%24%HindawiElsevierSpringerWileySpringer NatureTaylor & FrancisIOS PressSAGEPLoSSpandidosOUPFrontiersCell PressBMCRSCASBMB/JBCACS

Calculated from the uploaded Retraction Watch CSV.

Low-author, fast, process-heavy journal clusters

These include many paper-mill or peer-review-heavy journals:

JournalMedian authorsSingle-author shareMedian lagMain signal
Arabian Journal of Geosciences152.9%0.30 yearsPeer-review/process
Journal of Environmental and Public Health248.6%0.98 yearsPeer-review/process
Wireless Communications and Mobile Computing238.9%1.16 yearsPeer-review/process
Security and Communication Networks236.0%1.38 yearsPeer-review/process
Computational Intelligence and Neuroscience227.6%1.25 yearsPeer-review/process

These are not typical “old lab-data investigation” retractions. They look like special-issue, publisher-audit, paper-mill, or peer-review pipeline failures.

Many-author, slower, biomedical/data-heavy clusters

These include:

JournalMedian authors>10-author shareMedian lagMain signal
PLoS One616.4%4.31 yearsMixed, image/data, peer-review
Scientific Reports615.0%1.96 yearsImage/data and mixed integrity issues
Journal of Biological Chemistry68.5%7.32 yearsImage-heavy, misconduct-heavy
Bioscience Reports54.6%3.13 yearsImage and batch signals
Journal of Crohn’s and Colitis717.4%11.98 yearsLong-lag clinical/review-like correction
Cochrane Database of Systematic Reviews41.2%8.32 yearsReview lifecycle corrections

This contrast is one of the strongest in the analysis. A two-author paper in a paper-mill-heavy journal and a twelve-author biomedical paper in an image-heavy journal are both “retracted,” but they belong to different weather systems.


7. Society vs non-society journals: society retractions have larger teams

Using the conservative society-linked publisher classification from the previous analysis, society-linked non-conference records had:

GroupRecordsMedian authorsMean authorsSingle-author share>10-author shareMedian lag
Society-linked4,89655.864.7%9.7%3.00 years
Non-society / unclassified51,53244.4515.6%4.8%1.67 years

Society-linked records have larger author teams and longer retraction lags. They are also more image-heavy and fraud/misconduct-heavy, while non-society/unclassified records are more peer-review/paper-mill-heavy.

Society-linked records have larger author teams

Non-conference records only. Society-linked records are more concentrated in 6-10 and 11-20 author buckets.

Society-linked
Non-society / unclassified
0%9%18%27%36%12-34-56-1011-20>20

Calculated from the uploaded Retraction Watch CSV.

This explains much of the society-journal pattern from the previous post. Society-linked retractions are not necessarily more numerous, but they are more likely to sit in biomedical, biochemical, chemistry, society-proceedings, and higher-team-size journals. That produces a different retraction clock.

Within society-linked records:

Author bucketMedian lagImage concernsFraud/misconduct
1 author1.97 years6.5%13.9%
2-3 authors2.38 years27.2%23.4%
4-5 authors2.85 years38.8%27.8%
6-10 authors3.51 years53.5%26.0%
11-20 authors3.79 years52.6%30.4%

Within non-society/unclassified records, the same direction exists, but the paper-mill/peer-review signal is much stronger in low-author buckets.

So society status modifies the author-count interpretation:

In society journals, more authors usually means more image/data/forensic correction.
In non-society/unclassified journals, low-author retractions are heavily shaped by peer-review and paper-mill correction.


8. The >20-author exception: giant bylines are rare and heterogeneous

The largest author-count records are fascinating exceptions.

The maximum parsed author count in the dataset was 88, from a Science paper on the emergence and spread of the SARS-CoV-2 Omicron variant in Africa. It was retracted very quickly, with a lag of about 0.05 years, and the reason involved contamination/materials and unreliable conclusions.

Other very large bylines include:

Approx. authorsType of recordTypical pattern
80+COVID/genomics/public-health collaborationsFast correction possible
80+Mendelian randomization/large consortium studyData concerns, retract-and-replace or updated notice
60+Clinical trial/COVID ICU papersRemoval, retract-and-replace, date/notice complexity
50+Nature/biomedical consortium papersInstitutional/data/manipulation concerns

This is why the >20-author bucket does not simply continue the lag increase. Very large bylines are a special species. They include consortia, public-health surveillance, clinical collaborations, and multi-country computational studies. Their corrections can be rapid if the problem is centralized, obvious, or administrative.

The author-count curve therefore has a bend:

Retraction lag increases from 1 author to 11-20 authors, but the extreme mega-author bucket is too rare and too heterogeneous to behave like a simple continuation.


9. Hypothesis-by-hypothesis evaluation

HypothesisResultEvidence
More authors mean slower retractionRaw yes, adjusted noSpearman ρ = 0.151 in non-conference records, but adjusted model author effect p = 0.684
Author count predicts reason typeStrongly supportedLow-author records are peer/paper-mill-heavy; many-author records are image/data/fraud-heavy
Single-author records are mostly in humanities/social sciencesPartly true, but not enoughHumanities and social sciences are single-author-heavy, but many low-author records also come from publisher batch corrections
Many-author records are mostly biomedical/clinicalBroadly supportedBiology and medicine have median 5 authors and the largest >10-author shares
Country patterns differ by author countSupportedRussia has a single-author/plagiarism signature; Italy, France, Germany, US, Saudi Arabia and Pakistan have stronger many-author signatures
Publisher/journal ecosystems differStrongly supportedHindawi/Springer/IOS low-author process-heavy clusters vs PLoS/BMC/Cell Press/JBC many-author data/image clusters
Society journals have different author-count behaviorSupportedSociety-linked records have higher median authors, fewer single-author records, more >10-author records and longer lag
Mega-author papers behave like ordinary many-author papersNot supported>20-author records are rare, mixed and often corrected faster than 11-20 author papers

10. What the author count really tells us

A byline is not just a list of contributors. In this dataset, it behaves like a weak but useful diagnostic.

Author-count patternLikely retraction ecology
1 authorPlagiarism, peer-review/process, humanities/social sciences, batch cleanup
2-3 authorsPaper-mill/peer-review-heavy, computing/engineering and special-issue clusters
4-5 authorsTransition zone, mixed problems
6-10 authorsBiomedical, image/data, society-journal and lab-science records rise
11-20 authorsStrong image/data/fraud/institutional-investigation signal
>20 authorsConsortia, clinical/public-health, large collaborations, heterogeneous and rare

The most important conclusion is that author count is a context marker. It points toward field, journal, publisher, collaboration structure and reason category. It does not by itself tell us whether a paper is fraudulent, careless, or unlucky.

A one-author retraction may be plagiarism.
A three-author retraction may be a paper-mill node.
A seven-author retraction may be duplicated western blots.
A fifteen-author retraction may be a clinical or biomedical investigation.
An eighty-author retraction may be a fast correction in a consortium study.

The byline is a map legend, not the map.


Data cautions

Several caveats matter:

  1. Author count was parsed from the Retraction Watch author field, using semicolon-separated entries. Group authors may be counted as one entity.
  2. Records are not always unique scientific articles, because some entries are updated notices, expressions of concern, corrections, or retract-and-replace events.
  3. Conference records were excluded from the main analysis, because conference-proceedings batches strongly distort author-count and timing.
  4. Country fields were exploded, so multinational papers count once for each country listed.
  5. No publication denominator is available, so this analysis describes retraction records, not retraction rates per published paper.
  6. Reason categories overlap, so percentages do not sum to 100.

Final thought: the byline is the paper’s seismograph

The number of authors on a retracted paper does not tell us guilt. It tells us terrain.

Small bylines often sit in fast-moving process failures: peer review, paper mills, special issues, plagiarism, metadata cleanup. Medium-to-large bylines sit more often in slow-moving forensic failures: images, data, misconduct investigations, biomedical records, society journals, and clinical ecosystems. Very large bylines are rare exceptions, often shaped by consortia and centralized corrections.

So the authorship pattern is not a morality score. It is a seismograph.

It tells us whether the tremor came from a paper-mill factory floor, a humanities desk, a computational special issue, a biochemical blot archive, a clinical collaboration, or a giant pandemic consortium.

The byline, quiet little row of names that it is, carries the crackle of the whole publishing ecosystem. 🔬📉