Proportion of authors citing a scholarly work who have read that work
- Measures
- Of all citations in scholarly publications, the share where at least one citing author has read the cited work itself (more than its abstract or its reference entry) before citing it.
- Population
- Citations in peer-reviewed scholarly literature across fields, counted per citation (each citing-paper to cited-work link)
- Conditions
- Contemporary practice (roughly 2000 to the present). "Read" means the citer consulted the original's text beyond the abstract; merely obtaining a copy or copying the reference from another bibliography does not count.
- Unit
- a share, shown as a percentage
Between 25 and 85% (90% interval), most likely about 55%.
Stated as an interval, the precision the evidence supports. broad, slightly left-skewed; no study observes reading directly
Assessment
The graph's estimate, all things considered: the figures and arguments it rests on, and how they were combined.
No study observes directly whether citing authors read what they cite, so the value is known only within wide bounds. The closest evidence is now survey data. In a survey of about 9,380 corresponding authors in 15 fields, Teplitskiy, Duede, Menietti and Lakhani (Research Policy, 2022) asked authors about two specific works from their own reference lists. About 40% of the cited works were known only "slightly well" or "not well", and a majority of citations (54%) had little or no influence on the citing author. Taking moderate or better familiarity as a sign of reading beyond the abstract suggests that roughly 60% of citations rest on reading. Self-report probably inflates this. Against that, the quantity counts a citation as read if any co-author read the work, which the single respondent may not know about. A smaller survey in the Journal of Documentation (2019) found that only about 23% of authors had read every source in a paper in full. That fits the same picture: almost everyone cites some works they have not fully read, but most individual citations are not of that kind. In surveys of psychologists, most agree that authors often cite after reading only the abstract, which is a perception and not a rate.
The widely repeated figure of about 20%, from Simkin and Roychowdhury's "Read Before You Cite!" (2003), comes from modeling how misprints spread through the reference lists of a few highly cited physics papers. It measures how often reference entries are copied, not how often works are read, and it comes from the setting where ritual citation is most likely. It works as a rough lower anchor and not as a general rate. The survey finding that highly cited works are read more carefully is in some tension with the idea that such works are the least read. That gap is one more reason not to treat 20% as typical.
Taking these together, about 55% of citations are backed by reading beyond the abstract, with a 90% interval of roughly 25% to 85%. The value very likely varies by field, by the age and fame of the cited work, and by whether the citation is substantive or perfunctory.
The work: which figures, how they were combined, and the arithmetic
Inputs: (1) Teplitskiy et al. 2022 survey (n about 9,380, 15 fields): about 60% of sampled references known moderately well or better. Reading: about 0.6 for one respondent's familiarity. Adjustments: self-report upward bias (perhaps -0.05 to -0.15); "at least one co-author read it" definition (perhaps +0.0 to +0.1); "slightly well" includes some works once read (+). Net about 0.55 to 0.6. The figure is taken from secondary reporting (Times Higher Education) and the arXiv preprint text, not from the table itself. (2) Journal of Documentation 2019 survey (n about 104): 23% of authors had read all cited sources entirely. This is per paper and requires full reading, so it is consistent with a per-citation rate well above 0.5 and constrains little. (3) Simkin and Roychowdhury 2003: point estimate of 0.2 from misprint propagation. This measures copying of reference entries in highly cited physics papers and is treated as near a floor. (4) Quotation-error literature (about a quarter of citations misrepresent their source) is consistent with imperfect engagement but does not fix a rate. Combined: central moved from 0.50 to 0.55, because the survey is the most direct per-citation evidence. Lower bound raised from 0.20 to 0.25, because the misprint figure is a floor for an unfavorable setting and the surveys point well above it. Upper bound kept at 0.85, since self-report could understate at the high end only slightly. Interval kept wide because no direct observation exists and the "read" threshold maps imprecisely onto the familiarity scales.
Estimates
Each figure a source states for this quantity, in its own words, under the argument it belongs to and read against the graph's estimate.
Grey band: the graph's estimate, with its central value as a line. Dark marks: each source's own figure. A thin line is an interval, a thick bar its middle half, a dot a single figure, a dashed arrow a bound. An open box is where an argument places the value.
Decomposition
The arguments the estimate rests on, each stated as it runs, with what it shows about the value and its premises linked inline. ↗︎ opens a premise; the map shows how they fit together.
Because identical misprints recurring across reference lists show that many citers copy references without consulting the original, fitting a stochastic copying model to how often each misprint appears gives the share of citers who consulted the original, about 20% for the highly cited physics papers studied. Treating consultation as reading, the argument concludes that about a fifth of citers read what they cite.
- Does it follow?
- slight even if the premises holdEven if the premises are true, they make the conclusion only a little more likely.constitution →
- Do the premises hold?
- 1 not separately assessed
Granting that shared misprints reflect copying without consulting the original, the model fixes how often reference entries were copied, not how often works were read. Someone can read a paper and still copy its reference entry from a database or reference manager. The sample is a few highly cited physics papers, where citing without engaging was assumed to be most common, although survey evidence suggests highly cited works are read more carefully than others. The 20% figure therefore works as a rough lower anchor for that setting, not as a general rate.
Provenance
Where this quantity's evidence comes from. The figures sources state are under Estimates above.
Readings
Figures from instruments taken to indicate this quantity. Each is evidence about the quantity, not its value: what was actually measured is stated with it, and how far it speaks to the quantity is the Steward's judgment.
- measured
- Self-reported familiarity of corresponding authors (about 9,380, 15 fields) with two specific references sampled from their own published papers, on a scale from "not well" to "extremely well"; also self-reported influence.
- bearing
- The closest direct, per-citation measure available. Knowing a work moderately well or better roughly corresponds to having read it beyond the abstract, so it suggests about 0.6 as a reading of the quantity; "slightly well" may include works once read. Self-report and a 2020s survey of corresponding authors only (not all co-authors) likely bias it upward relative to the truth, and the "at least one citing author" definition biases the true value upward relative to a single respondent's report. Figure taken from secondary reporting (Times Higher Education) of the survey.
- measured
- Survey of authors (n about 104 to 112, 27% response rate) asking whether they had read all the sources they cited in a given paper in their entirety.
- bearing
- Weak. It is per author and paper, not per citation, and "read entirely" is stricter than the quantity's "beyond the abstract". It shows most authors cite at least some works they have not fully read, but is compatible with a large majority of individual citations being read in part. Small sample, low response rate. Figures read from a search excerpt; the page itself could not be opened.
Our estimate is only about 20% of citers read the original.
Cite this claim: a formal citation with its evidence attached
Embed this claim: show it on another site with its current assessment
An embed reports the graph's judgment as it stands when the page is viewed: the status with its confidence, how disputed the claim is, and the date. It links back here.
<blockquote class="minerval-claim" data-claim-id="bd217086-caa8-4212-bc93-3f5fa6d068b2"><a href="https://minerval.ai/claims/bd217086-caa8-4212-bc93-3f5fa6d068b2">Proportion of authors citing a scholarly work who have read that work</a></blockquote> <script async src="https://minerval.ai/embed.js"></script>
<iframe src="https://minerval.ai/embed/claims/bd217086-caa8-4212-bc93-3f5fa6d068b2" title="Minerval: Proportion of authors citing a scholarly work who have read that work" width="480" height="250" style="border:0;max-width:100%" loading="lazy"></iframe>
[](https://minerval.ai/claims/bd217086-caa8-4212-bc93-3f5fa6d068b2)
The badge now: 
Contribution record · 1 exchange
Challenges and proposals from contributors, with the review each received. Substantive exchanges stay on the claim's public record.
Proposed as a quantity: of authors citing a scholarly work, the proportion who have read it, as opposed to copying the citation from another paper. Simkin and Roychowdhury (2003) estimated this indirectly from how misprints in reference lists propagate between citing papers, concluding that a minority of citers read the original. The estimate depends on a model of copying, so how well it measures reading, rather than copying references, is part of the assessment.
The proposal asks for the proportion of citing authors who have read a cited work, rather than copying the reference from another paper. This is a how-much question and is properly entered as a quantity. It concerns no private person and the wording is neutral and close to canonical form. The Simkin and Roychowdhury (2003) misprint-propagation estimate is a real, checkable, publicly discussed figure. The contributor correctly notes that the estimate depends on a copying model, so it is a reading from an instrument and not the quantity's value. How well it measures reading is for the Steward to assess. The quantity needs a population and a scope for "scholarly work", for example all works or a field, to be well defined (§2). The Matcher and Steward can complete that. Admitted on form and good faith; novelty is the Matcher's call.
Contribute
Every judgment on this page is open to challenge. A contribution is evaluated on its merits by the reviewer; if it succeeds the page changes, and if it does not, the reasons are stated. Either way the exchange becomes part of the claim’s public record.
Created by user · Nov 2, 2026. Every judgment on this page is accompanied by a reasoning trace.