Published papers get their references wrong more often than most readers assume. Studies since 1985 have found faulty bibliographic details in roughly a quarter to a third of journal references, and a cited source that does not say what the citing sentence claims in roughly one quotation in six. Neither figure has fallen over forty years of measurement.
EdCitation publishes this roundup. For each study we give the sample, the method, the finding and the limit, every figure read in the paper itself (or its abstract, where we say so) on 27 September 2026. The pre-submission routine built on this evidence is in how to check your references before journal submission. The first half of that routine, comparing every entry with the publisher's record, is what EdCitation's free Verify references does for a whole list at once.
What is the difference between a citation error and a quotation error?
A citation error is wrong bibliographic detail; a quotation error is a wrong account of what the source says. The literature has kept the two apart since De Lacey et al. (1985), and Baethge and Jergas (2025) restate the split: citation inaccuracy means technical faults such as a misspelt author or a wrong page, while quotation inaccuracy means the citing claim misrepresents the source.
Both kinds are graded. A major citation error usually stops a reader finding the source; a minor one, such as a dropped initial, does not. A major quotation error is a source that does not support the claim at all, or contradicts it; a minor one is a smaller slip that does not overturn the meaning. Many studies also count an indirect (secondary) quotation, citing a review for a finding that is really another paper's, as a minor error.
The studies, in one table
Oldest first. "Limit" is the main weakness the authors name, or the one their design imposes.
| Study | Field | What it looked at | What it found | Limit | Source |
|---|---|---|---|---|---|
| De Lacey et al. (1985) | Medicine | References in six medical journals, January 1984 | 15% misquoted; 24% with citation errors, 8% of those major | Abstract read | BMJ |
| Eichorn and Yankauer (1987) | Public health | 150 random references, three journals, May 1986 | 31% citation errors; 30% misused, half of them major | One issue per journal | AJPH |
| Simkin and Roychowdhury (2003) | Physics | 196 misprints in citations to one paper | 45 distinct; model puts readers at about 20% of citers | A model, one paper | Complex Systems |
| Lukić et al. (2004) | Anatomy | 199 random references, three anatomy journals, 2001 | 27% citation errors; 19% of 272 quotations wrong, 94% of those major | Abstract read | Clinical Anatomy |
| Wager and Middleton (2008) | Biomedicine | 66 surveys of reference accuracy | Median 38% citation errors, 20% quotation errors | Abstract read; to 2007 | Cochrane |
| Wright and Armstrong (2008) | Marketing | 50 studies citing one survey-methods paper | 49 of 50 reported its findings wrongly | One paper, by one of the authors | Interfaces |
| Greenberg (2009) | Neurology | 242 papers and 675 citations on one claim | Supportive data drew 94% of 214 citations | One claim | BMJ |
| Jergas and Baethge (2015) | Medicine | Meta-analysis, 28 studies, 7,321 references | 25.4% quotation errors, 11.9% major | Heterogeneity 97% | PeerJ |
| Mogull (2017) | Medicine | Recalculation, 15 studies, 5,535 quotations | 14.5% quotation errors; 4.1% to 34.2% by specialty | Some counts estimated | PLOS ONE |
| Smith and Cumberledge (2020) | General science | 250 random citations in high-impact journals | 25% error rate | Abstract read | Proc. R. Soc. A |
| Pavlovic et al. (2021) | Biomedicine | 4,912 citations of 13 highly cited papers | 15% of citing articles erred at least once | Abstract read | Clinical Science |
| Logan et al. (2024) | All fields | Scoping review, 105 reference-accuracy studies | 32.7% of references had an error | Punctuation rules unclear | JOLIS |
| Baethge and Jergas (2025) | Medicine | Meta-analysis, 46 studies, 32,074 quotations | 16.9% wrong, 8.0% major; no change over time | Heterogeneity high | RIPR |
How often are references in journals wrong?
About one reference in three carries at least one bibliographic error. Logan et al. (2024) gathered 105 primary studies that checked references in peer-reviewed articles, searched to November 2022, and put the overall error rate at 32.7%, with author names the commonest fault and similar rates across fields. Wager and Middleton (2008), in a Cochrane review we read in abstract, found a median citation error rate of 38% across 66 surveys, and found that giving authors instructions went with fewer reference errors in two studies.
The early counts in medicine
Single-journal studies give the same order of magnitude. Eichorn and Yankauer (1987) checked 50 randomly chosen references from the May 1986 issue of each of three public health journals: 31% of the 150 had citation errors, one in ten of them major (the reference could not be located). Lukić et al. (2004) took 199 references from the first 2001 issue of Annals of Anatomy, Clinical Anatomy and Surgical and Radiologic Anatomy. Of these, 27% had citation errors and 38% of those errors were major.
How often do papers misquote their sources?
In medicine, somewhere between one quotation in seven and one in four, depending on how it is counted; the most recent estimate is one in six. The method decides the number.
Jergas and Baethge (2015): the first meta-analysis
Jergas and Baethge (2015) screened 559 papers and pooled 28 studies published from 1985 to 2013, covering 7,321 references. Their estimates: 11.9% major errors, 11.5% minor and 25.4% in total. Heterogeneity was large (I² 97%), with single studies ranging from 6.7% to 83.0%. Taking out indirect references lowered the total to 21.5%, so indirect citation explained less than a sixth of the problem. In an average article of 50 references, they reckoned, about six are completely wrong.
Mogull (2017): a recount
Mogull (2017) argued that earlier reviews divided quotation errors by the number of references, although a reference is often quoted more than once, which inflated the rate. Recounting 15 representative studies, 3,337 references and 5,535 quotations, as errors per quotation gave 14.5% (95% CI 10.5 to 18.6), ranging from 4.1% to 34.2% between specialties. About two thirds of the errors were major. The quotation totals had to be estimated for 10 of the 15 studies, a limit Mogull states.
Baethge and Jergas (2025): the current estimate
Baethge and Jergas (2025) pooled 46 studies from 14 countries and 22 medical fields, 32,074 quotations made between 1984 and 2022, and found 16.9% wrong (95% CI 14.1 to 20.0): 8.0% major, 7.8% minor, and 5.3% secondary quotations. A meta-regression found no improvement over time. Higher journal impact factor went with fewer errors (Spearman's ρ = −0.253, 25 studies), a modest association.
Outside medicine
Smith and Cumberledge (2020), read in abstract, checked 250 random citations in high-impact general science journals against the cited material and found a total error rate of 25%. Wright and Armstrong (2008), also read in abstract, examined 50 studies that cited one paper on correcting for non-response bias in mail surveys: 49 of 50 reported its findings wrongly. Armstrong co-wrote the cited paper, which is worth knowing when weighing the verdict.
How do errors spread from paper to paper?
Mostly by copying: a reference or claim taken from another paper's list carries that paper's mistakes.
Simkin and Roychowdhury (2003) looked at misprints in citations to one physics paper, Kosterlitz and Thouless's 1973 article, with about 4,300 citations: 196 misprints, of which only 45 were distinct, and one page-number misprint repeated 78 times. Repeating someone else's misprint by accident is very unlikely, so the repeats were copied. Their model put the share of citers who had consulted the original, or a trusted database, at about 17% for this paper, and the abstract gives about 20% as its estimate. The title is the advice: read before you cite. It is a model fitted to one paper, and it measures copied references, not misquoted claims. The practical side of citing a source you have not read is in cited a source you didn't read?.
Greenberg (2009) traced every PubMed paper stating one claim about inclusion body myositis: 242 papers, 675 citations and 220,553 citation paths supporting it. Papers with supportive data received 94% of the 214 citations to primary data, and six papers that weakened or refuted the claim received 6%. Greenberg names three distortions: bias against critical papers, amplification by papers with no data of their own, and invention, such as a hypothesis turned into fact by citation alone.
Pavlovic et al. (2021), read in abstract, followed 4,912 citations in 2,995 articles citing 13 highly cited papers. At least one inaccurate citation appeared in 15% of the citing articles. The commonest fault was citing a finding the paper did not contain (38.4%), and a fifth of inaccurate citations came from chains of earlier inaccurate ones.
Where do the studies disagree?
On three points.
The quotation error rate
The 25.4%, 14.5% and 16.9% estimates are mostly one disagreement about denominators. Jergas and Baethge (2015) counted per reference; Mogull (2017) counted per quotation. Baethge and Jergas (2025) took the quotation-based approach and, re-running their 2015 set of 28 studies with the new imputed denominators, reached 17.0%. The 25% of Smith and Cumberledge (2020) comes from a different field and one sample.
How much is copying
Simkin and Roychowdhury (2003) suggest four citers in five had not consulted the original; Jergas and Baethge (2015) found indirect references behind less than a sixth of quotation problems. They measure different things: a copied reference can still carry a fair claim, and a misquotation can come from someone who read the paper.
Whether prestige protects
Baethge and Jergas (2025) found fewer quotation errors in higher-impact journals, yet Smith and Cumberledge (2020) found a quarter of citations wrong in the highest-impact general science journals. A strong journal lowers the odds a little. No field, era or journal studied so far was free of the problem.
What did Verify references say about our test list?
We pasted four real references into Verify references on 27 September 2026. One error was ours, put in on purpose: Mogull's surname typed as "Mogul", the kind of author-name slip Logan et al. (2024) found commonest.
| Our entry | Our error | Shown as | Reason, as returned |
|---|---|---|---|
| Jergas and Baethge (2015), with DOI | None | Verified | "The DOI resolves to this record and the title matches." |
| Mogull (2017), with DOI | Surname typed "Mogul" | Doubtful ("check this") | "The DOI and title are right, but the first author differs. Check the names." |
| Greenberg (2009), with DOI | None | Verified | "The DOI resolves to this record and the title matches." |
| Lukić et al. (2004), no DOI | None | Verified | "Title, year and first author match a published record." |
The counts came back as 3 verified, 1 doubtful, 0 not found, 0 retracted and 0 unchecked. The one-letter misprint was caught even though the DOI was right. The record is not perfect either: for Lukić et al. it spells three names with a breve (c̆, s̆) where the authors' names carry a háček (Marušić, not Marus̆ić), and the entry built from it repeated that. Read the names against the article.
Which EdCitation tool checks for citation errors?
For citation errors, the bibliographic kind that Logan et al. (2024) found in a third of references, Verify references is the best tool: every entry is compared with the record its publisher registered, nothing is ever written for you, and the reason names the part that disagrees, down to the single letter in our test. It is free with no account. Each reference comes back verified, doubtful ("check this") or not found; a retracted paper is flagged; and a reference it could not check is labelled that way, never as not found. Given the whole paper instead of a list, it matches every in-text citation to an entry and every entry back to the text.
Plainly: Verify references checks the reference, not whether your sentence represents the paper fairly. The quotation errors in the meta-analyses above are found only by reading the source beside the claim, and that stays with the author.
To rebuild a faulty entry, Cite a source takes the DOI, title, PubMed ID or ISBN and builds it in APA 7, MLA 9, Chicago 18 author-date, Harvard, IEEE or Vancouver. References from a file, in Pro at $8 a month, checks every reference in an uploaded paper and matches its citations to the list. EdCitation never writes any part of a paper, and that is its strength here: it finds, cites and verifies sources, and the reading and the argument stay yours.
How do I check my own citations before I submit?
Check the reference against its record, then the claim against its source.
- Paste the full list into Verify references. Fix every "check this" by opening the record: author names first, then year, title, volume and pages.
- Rebuild any entry you copied from another paper's list from the DOI with Cite a source, since copied entries carry copied misprints.
- Open every source your key claims rest on and find the passage, table or figure that supports each sentence. If you cannot point to it, change the sentence or the source.
- Replace secondary citations with the original. Search the original's title in Find sources; if you cannot get it, cite it as a secondary source.
- Cite from the results, not the abstract or the discussion. A finding the paper only mentions from others belongs to those others.
- Put each citation beside the claim it supports, not at the end of a paragraph that makes several claims, as Baethge and Jergas (2025) advise.
Related evidence: what studies say about students' referencing mistakes, retracted papers still being cited and how often AI tools invent references.
Quick questions
What is the citation error rate in published research?
A scoping review of 105 studies put it at 32.7% of references with at least one bibliographic error (Logan et al., 2024).
How common are misquoted citations in medical journals?
About one quotation in six. Baethge and Jergas (2025) found 16.9% of 32,074 quotations did not support the claim, half of them seriously.
What does "read before you cite" mean?
It is the title of Simkin and Roychowdhury's 2003 paper, which used repeated misprints to estimate that only about 20% of citers of one paper had consulted the original.
Can software check whether a source supports my claim?
Not reliably. EdCitation's Verify references checks that each reference matches the publisher's record and flags retractions; whether the source says what your sentence says needs a person reading it.
Have citation errors become rarer?
No. Both meta-analyses of quotation errors, Jergas and Baethge (2015) and Baethge and Jergas (2025), found no significant change over time.
References
- Baethge, C., & Jergas, H. (2025). Systematic review and meta-analysis of quotation inaccuracy in medicine. Research Integrity and Peer Review, 10, Article 13. https://doi.org/10.1186/s41073-025-00173-z
- De Lacey, G., Record, C., & Wade, J. (1985). How accurate are quotations and references in medical journals? British Medical Journal, 291(6499), 884–886. https://doi.org/10.1136/bmj.291.6499.884
- Eichorn, P., & Yankauer, A. (1987). Do authors check their references? A survey of accuracy of references in three public health journals. American Journal of Public Health, 77(8), 1011–1012. https://doi.org/10.2105/AJPH.77.8.1011
- Greenberg, S. A. (2009). How citation distortions create unfounded authority: Analysis of a citation network. BMJ, 339, Article b2680. https://doi.org/10.1136/bmj.b2680
- Jergas, H., & Baethge, C. (2015). Quotation accuracy in medical journal articles: A systematic review and meta-analysis. PeerJ, 3, Article e1364. https://doi.org/10.7717/peerj.1364
- Logan, S. W., Hussong-Christian, U., Case, L., & Noregaard, S. (2024). Reference accuracy of primary studies published in peer-reviewed scholarly journals: A scoping review. Journal of Librarianship and Information Science, 56(4), 896–949. https://doi.org/10.1177/09610006231177715
- Lukić, I. K., Lukić, A., Glunčić, V., Katavić, V., Vučenik, V., & Marušić, A. (2004). Citation and quotation accuracy in three anatomy journals. Clinical Anatomy, 17(7), 534–539. https://doi.org/10.1002/ca.10255
- Mogull, S. A. (2017). Accuracy of cited "facts" in medical research articles: A review of study methodology and recalculation of quotation error rate. PLOS ONE, 12(9), Article e0184727. https://doi.org/10.1371/journal.pone.0184727
- Pavlovic, V., Weissgerber, T., Stanisavljevic, D., Pekmezovic, T., Milicevic, O., Lazovic, J. M., Cirkovic, A., Savic, M., Rajovic, N., Piperac, P., Djuric, N., Madzarevic, P., Dimitrijevic, A., Randjelovic, S., Nestorovic, E., Akinyombo, R., Pavlovic, A., Ghamrawi, R., Garovic, V., & Milic, N. (2021). How accurate are citations of frequently cited papers in biomedical literature? Clinical Science, 135(5), 671–681. https://doi.org/10.1042/CS20201573
- Simkin, M. V., & Roychowdhury, V. P. (2003). Read before you cite! Complex Systems, 14(3), 269–274. https://doi.org/10.25088/ComplexSystems.14.3.269
- Smith, N., Jr., & Cumberledge, A. (2020). Quotation errors in general science journals. Proceedings of the Royal Society A, 476(2242), Article 20200538. https://doi.org/10.1098/rspa.2020.0538
- Wager, E., & Middleton, P. (2008). Technical editing of research reports in biomedical journals. Cochrane Database of Systematic Reviews, (4), Article MR000002. https://doi.org/10.1002/14651858.MR000002.pub3
- Wright, M., & Armstrong, J. S. (2008). The ombudsman: Verification of citations: Fawlty towers of knowledge? Interfaces, 38(2), 125–139. https://doi.org/10.1287/inte.1070.0317