Research Integrity Violations
The ACE-RIV taxonomy, version 1.0
The ACE Research Integrity Violations (ACE-RIV)
The ACE-RIV taxonomy is the classification system ACE uses to describe the errors and integrity problems it documents in audited research. It draws on the standards published by the Committee on Publication Ethics (COPE), the Council of Science Editors (CSE), the International Committee of Medical Journal Editors (ICMJE), and the European Code of Conduct for Research Integrity (ALLEA). The RIVs are then organized into four domains: (1) Authorship and Publication, (2) Statistical and Data Integrity, (3) Ethical Oversight and Transparency, and (4) Indicators and Minor Issues.
Quick reference
Table 1 lists every category with its severity tier and the unit in which instances are counted. Each number links to the full entry below.
| RIV | Name | Severity | Counted per |
|---|---|---|---|
| Section 1 | Authorship and Publication | ||
| 1.1 | Guest, Honorary, or Gift Authorship | Major | credited author, per article |
| 1.2 | Ghost Authorship | Major | omitted contributor (article if unidentifiable) |
| 1.3 | Authorship for Sale and Paper Mill Activity | Critical | article |
| 1.4 | Plagiarism | Major | passage, figure, or table |
| 1.5 | Duplicate Publication and Salami Slicing | Critical | redundant article |
| 1.6 | Citation Manipulation | Major | article |
| 1.7 | Peer Review Subversion | Critical | article |
| 1.8 | Irrelevant Citations and Reference Padding | Major | article |
| 1.9 | Trial Registration Failures and Outcome Switching | Major | unregistered trial, or switched outcome |
| 1.10 | Incompatible Authorship or Contribution Statements | Major | article |
| 1.11 | Misrepresentation of Cited Sources | Major | citation |
| 1.12 | Undisclosed Generated Text | Major | article |
| Section 2 | Statistical and Data Integrity | ||
| 2.1 | Anomalies and Extreme Values | Major | anomalous value |
| 2.2 | Statistical Inconsistencies and Impossible Data Sets | Critical | impossible statistic |
| 2.3 | Selective Reporting, Suppressed Data, and Omitted Methodology | Major | omitted outcome, analysis, or method element |
| 2.4 | Data Miscalculation or Erroneous Reporting | Major | incorrect value |
| 2.5 | Image and Figure Manipulation | Critical | image or panel |
| 2.6 | Data Set Anomalies | Critical | distinct anomaly |
| 2.7 | Image and Figure Errors | Major | figure or panel |
| Section 3 | Ethical Oversight and Transparency | ||
| 3.1 | Failure to Disclose Conflicts of Interest | Major | undisclosed interest, per article |
| 3.2 | Bypass of Ethical Oversight | Critical | article |
| 3.3 | Ethics Approval Anomalies | Critical | article |
| 3.4 | Implausible Study Timelines | Major | article |
| 3.5 | False or Unfulfilled Data Availability Statements | Major | article |
| Section 4 | Indicators and Minor Issues | ||
| 4.1 | Minor Typographical Error | Minor | error |
| 4.2 | Indicators of Paper Mill Involvement | Indicator | article |
| 4.3 | Tortured Phrases and Text Spinners | Indicator | article |
How to read a RIV
RIV labels are meant to act as categories of integrity issues. When a report records an instance of a violation, it means that the public record contains evidence matching that RIV as defined below; it does not assert that any author, institution, or publisher acted deliberately, which ACE is not in a position to determine. Several RIVs (for example, RIV 1.3 and RIV 1.7) describe practices that are usually intentional when they occur, and readers should understand a finding under those RIVs as evidence consistent with the practice rather than a determination that it took place.
The Criteria of a RIV are the conditions that must be met for a single instance of that RIV to be recorded. Typical evidence is the kind of material ACE relies on to show them.
Each RIV has an associated severity tier. The tiers are:
Critical – On its own, undermines the reliability of the article’s findings.
Major – Materially affects the accuracy, completeness, or trustworthiness of the article.
Minor – Does not change the data, results, or conclusions. Noted for the record and for the benefit of secondary analyses.
Indicator – Not a violation in itself but a pattern that raises the probability of one. Recorded only alongside other findings.
What is not a RIV
The taxonomy describes the integrity of the published record: whether what an article reports is internally consistent, consistent with the public record around it, and produced by the process it describes. It does not grade the quality of the science. A small sample, an underpowered design, a debatable choice of statistical model, a weak control condition, a conclusion that reads more into the results than they support, or a disagreement about interpretation is not a violation and is not recorded as one, unless the article omits something it committed to report (RIV 2.3) or misstates what a source says (RIV 1.11). ACE reports may comment on such matters where they bear on how the findings should be used, and say so when they do, but those comments carry no RIV number and no count. Conduct outside the audited article, such as an author’s record on other papers, is likewise not a RIV in the audited article, although it may be cited as context under RIV 1.3 or RIV 4.2.
Threshold of evidence and status of an instance
An instance is recorded only when every criterion of the RIV is met, as shown by material a reader can examine: the article and its supplements, the registration or protocol, a public database, the raw data where released, the journal’s own metadata, or a reproducible computation on the article’s numbers. Suspicion, pattern alone, or the reputation of the authors is not enough; an audit that cannot show that every criterion is met records nothing under that RIV, however strong the impression.
Each instance carries one of two statuses, stated where the instance is recorded. An instance is recorded when every criterion is met and no reasonable reading of the record avoids it. An instance is provisional when the criteria are met on the most plausible reading of an ambiguous record, but information held by the authors or the journal, such as raw data or a committee record, could resolve it; the report says which reading it adopts, why, and what information would settle the question. Unless a report says otherwise, an instance is recorded.
When a response or new evidence resolves a provisional instance, or shows a recorded one to be wrong, a new version of the report marks it withdrawn, or reclassified under another RIV, and gives the reason. The summary table changes accordingly and the report’s version history says so. Nothing is removed silently.
Counting instances
ACE reports count violations in instances, so the size of a problem is visible and comparable across reports. An instance is the smallest piece of evidence that meets every criterion of a RIV by itself and can be checked on its own. Each RIV states what counts as one instance. RIVs about an article as a whole (for example RIV 1.3 or RIV 3.3) count once per article; RIVs about the data (for example RIV 2.2 or RIV 2.4) count once per affected value, statistic, image, or outcome.
Three rules keep the counts honest. Each instance belongs to one article and one RIV; evidence that fits two RIVs is recorded under the more specific one. Different evidence in the same article can support different RIVs, and counts are never added across RIVs. Severity belongs to the RIV, not to the count, although many instances of one RIV do weigh on the recommendation.
A report states each finding as a count, for example “three instances of Peer Review Subversion (RIV 1.7)”. The summary table at the end of the report is generated from those statements, so it always matches the text: instances per RIV, a link to the section that documents them, and the total. Where space is short, the compact form “RIV 1.7 (×3)” is used.
From findings to a recommendation
Every ACE report ends with a recommendation to the publishing journal. It follows from the instances recorded, by the rules below, applied article by article. Any Critical instance leads to a recommendation of retraction, because a Critical RIV by definition undermines the article’s findings. Major instances lead to a recommendation of correction when they are isolated and the corrected article would still stand; when they leave the findings without evidentiary value, the recommendation is retraction. Minor instances on their own call for no editorial action, but an accumulation of them can support a correction or, in combination with other findings, a retraction. Indicators never change the recommendation on their own, but they may point to further investigation. Where every instance that would drive the recommendation is provisional, the recommendation is that the journal obtain the underlying data or records, together with the action that should follow if they are not produced.
The recommendation is the authors’ professional opinion, offered in good faith on the evidence set out in the report. The decision belongs to the journal.
Boundaries between neighboring categories
Some categories sit next to each other, and the same evidence must always land in the same place. Impossibility separates RIV 2.2 from RIV 2.4: if the reported values cannot all be true, it is RIV 2.2, however small the discrepancy; if they could be true but are wrong on recomputation, it is RIV 2.4. Evidence of alteration separates RIV 2.5 from RIV 2.7: a figure with a wrong label or value is RIV 2.7 until duplication, splicing, or generation is shown. Attribution separates RIV 1.4 from RIV 1.5: reuse of other people’s work is plagiarism, reuse of the authors’ own is duplicate publication. Purpose separates RIV 1.6 from RIV 1.8: citations that benefit an identifiable party are manipulation, citations that serve no purpose at all are padding. Evidence about the article separates RIV 1.3 from RIV 4.2: a documented link between the article and a sale or a paper mill is RIV 1.3, patterns that merely resemble such products are RIV 4.2. An error that alters no data, result, or conclusion is RIV 4.1, whatever it looks like; one that does is RIV 2.4 or RIV 2.7. When an item would satisfy two RIVs, it is recorded once, under the more specific one, and the other is mentioned in the text without a count.
Relation to other frameworks
The ACE-RIV is a classification for audits of the published record, not a definition of misconduct. In the terms used by the United States Office of Research Integrity, fabrication and falsification correspond most closely to RIV 2.2, RIV 2.5, and RIV 2.6, and plagiarism to RIV 1.4; but a finding under those RIVs is a finding about the record, and whether conduct meets a legal definition of misconduct, which requires intent and a formal process, is for institutions and regulators to determine. The authorship RIVs (1.1, 1.2, and 1.10) follow the ICMJE criteria for authorship and the CRediT contributor roles. The publication RIVs (1.5, 1.6, 1.7, and 1.8) correspond to the situations addressed by COPE’s guidance and flowcharts on redundant publication, citation manipulation, and peer review manipulation, and RIV 4.2 draws on the COPE and STM guidance on paper mills. The registration and reporting RIVs (1.9 and 2.3) follow the ICMJE registration policy and the CONSORT statement. A journal applying COPE guidance will find that each RIV identifies the guidance that applies and the evidence the journal will need.
Numbering, versions, and citing the taxonomy
Numbers are stable from version 1.0 onward. A RIV keeps its number for as long as the taxonomy exists, so that a RIV cited in a published report always resolves to the same definition, and new RIVs are appended to the end of a section. Version numbers have two parts. A minor version (1.1, 1.2, and so on) adds RIVs, adds evidence or boundary notes, or clarifies wording without changing what an existing number means. A major version (2.0) is reserved for any change that alters the meaning, criteria, unit of counting, or severity of an existing RIV. Every report states the version it used, and its findings are read against that version; every version of this page is archived, and the current version is always this page.
To cite the taxonomy:
Jané, M. B. (2026). Research Integrity Violations: The ACE-RIV taxonomy, version 1.0. Audit Clinical Evidence. https://auditclinicalevidence.org/riv.html
@techreport{ACERIV2026,
author = {Jané, Matthew B.},
title = {Research Integrity Violations: The ACE-RIV taxonomy, version 1.0},
institution = {Audit Clinical Evidence},
type = {Taxonomy},
year = {2026},
url = {https://auditclinicalevidence.org/riv.html}
}Proposing a change
Anyone may propose a new category, a change to an existing one, or a correction. Open an issue in the ACE repository describing the RIV, the criteria that would have to be met, the evidence that would show them, and the unit of counting, or write to contact@auditclinicalevidence.org. Accepted changes appear in the version history at the end of this page.
Section 2: Statistical and Data Integrity
RIV 2.1: Anomalies and Extreme Values Major
Reported data, statistics, or results display extreme or unrealistic values, or distributions that are exceedingly improbable under natural variation, without an explanation in the article.
RIV 2.2: Statistical Inconsistencies and Impossible Data Sets Critical
Reported statistics are internally incompatible with one another, or could not have arisen from a real data set of the stated size and structure. The distinguishing feature is impossibility: the reported values cannot all be true at once.
RIV 2.3: Selective Reporting, Suppressed Data, and Omitted Methodology Major
Published outcomes exclude relevant data, negative results, or conflicting findings that bear on the study’s conclusions, or the article omits methodological details that are necessary for independent verification and reproducibility.
RIV 2.4: Data Miscalculation or Erroneous Reporting Major
Quantitative results are calculated, transcribed, or reported incorrectly. Unlike RIV 2.2, the reported values are possible; they are simply wrong, as shown by recomputation from the article’s own numbers.
RIV 2.5: Image and Figure Manipulation Critical
Images, figures, or graphical data contain duplicated, spliced, rotated, cloned, or artificially generated elements, or images taken from other sources are presented as original findings without attribution and authorization.
RIV 2.6: Data Set Anomalies Critical
The raw data set, where available, shows structural irregularities or improper manual alteration that compromises its integrity. This RIV applies only when data at the participant or observation level can be examined.
RIV 2.7: Image and Figure Errors Major
Images, figures, or graphical data contain incorrect values, mislabeled axes or groups, or results that contradict the text or tables, without evidence of manipulation (which is RIV 2.5).
Section 3: Ethical Oversight and Transparency
RIV 3.1: Failure to Disclose Conflicts of Interest Major
Financial, personal, or professional interests that could reasonably be perceived as compromising the objectivity of the work are not disclosed.
RIV 3.2: Bypass of Ethical Oversight Critical
Required ethical review (for example, by an institutional review board or animal care committee) was not obtained, or participants were not treated in accordance with applicable legal and ethical requirements, including informed consent.
RIV 3.3: Ethics Approval Anomalies Critical
The ethics approval or IRB reference given in the article cannot be reconciled with the study: it belongs to a different study, does not match the approving committee’s format, refers to a committee that does not exist at the stated institution, or differs across articles that report the same study.
RIV 3.4: Implausible Study Timelines Major
The dates reported for approval, registration, recruitment, intervention, follow-up, submission, or publication cannot all be true, or imply a study that could not have been conducted as described.
RIV 3.5: False or Unfulfilled Data Availability Statements Major
The article states that data or materials are available when they are not, or the authors do not provide data that the statement commits them to share.
Section 4: Indicators and Minor Issues
RIV 4.1: Minor Typographical Error Minor
Spelling, grammatical, formatting, or transcription errors that do not alter the underlying data, the statistical results, or the conclusions. Recorded because such errors can still propagate into meta-analyses and secondary analyses.
RIV 4.2: Indicators of Paper Mill Involvement Indicator
Patterns within the article or its authorship that are characteristic of industrialized publishing. These are indicators, not violations: they are recorded only alongside substantive findings and never on their own.
RIV 4.3: Tortured Phrases and Text Spinners Indicator
The text contains unnatural synonyms produced by automated paraphrasing (“tortured phrases”), such as “flag to clamor” for “signal to noise” or “bosom malignancy” for “breast cancer”, which indicate text spinning to evade plagiarism detection or generated text.
Version history
Version 1.0 (September 2026). Initial release.