Lawrence (2001)found computer science articles that were openly accessible (OA) on the Web were cited more. We replicated this in physics. We tested 1,307,038 articles published across 12 years (1992-2003) in 10 disciplines (Biology, Psychology, Sociology, Health, Political Science, Economics, Education, Law, Business, Management). A robot trawls the Web for full-texts using reference metadata ISI citation data (signal detectability d'=2.45; bias = 0.52). Percentage OA (relative to total OA + NOA) articles varies from 5%-16% (depending on discipline, year and country) and is slowly climbing annually (correlation r=.76, sample size N=12, probability p < 0.005). Comparing OA and NOA articles in the same journal/year, OA articles have consistently more citations, the advantage varying from 36%-172% by discipline and year. Comparing articles within six citation ranges (0, 1, 2-3, 4-7, 8-15, 16+ citations), the annual percentage of OA articles is growing significantly faster than NOA within every citation range (r > .90, N=12, p < .0005) and the effect is greater with the more highly cited articles (r = .98, N=6, p < .005). Causality cannot be determined from these data, but our prior finding of a similar pattern in physics, where percent OA is much higher (and even approaches 100% in some subfields), makes it unlikely that the OA citation advantage is merely or mostly a self-selection bias (for making only one's better articles OA). Further research will analyze the effect's timing, causal components and relation to other variables.
Comparing OA and NOA articles in the same journal/year, OA articles have consistently more citations, the advantage varying from 36%-172% by discipline and year.
large within-journal/year comparisons consistently identify the reported association, though they do not establish causation
The yearly ratio OAc/NOAc … is growing annually in every citation range.
Table 3 reports consistently strong positive annual correlations across all six citation ranges
The overall percentage of OA articles is slowly climbing annually (correlation r=.76, sample size N=12, probability p < 0.005).
the stated descriptive correlation matches Table 1 and the twelve-year pattern in Figure 2c
This is unlikely to be the sole or even the primary cause of the OA advantage.
external cross-field comparisons only indirectly bound article-level self-selection in this observational sample
The effect is greater with the more highly cited articles (r = .98, N=6, p < .005).
Figure 4 supports the six-bin correlation, but the small number of citation ranges limits certainty
Derived from the full evaluation — not a separate score.
Strengths
The analysis covers 1,307,038 articles published from 1992 through 2003 in ten biological and social-science disciplines, substantially broadening the settings in which the OA–citation association is examined.
↳ Abstract; Method §2; Results, Figure 2
A 200-article Biology sample was manually checked, producing reported misses, false alarms, signal detectability, and bias estimates. The paper also explains the expected direction of classification bias.
↳ Signal detection §3, Figure 1
The Conclusion identifies researchers, institutions, and funders as stakeholders and presents self-archiving as the practical pathway through which the findings could affect access policy.
↳ Conclusion §5
Limitations
Table 2 reports r=.10 and r=-.36 as p<.05 with N=12, which is incompatible with ordinary two-sided Pearson tests. This undermines confidence in that table’s inferential claims, although it does not directly invalidate the principal Figure 2 association.
↳ Results §4, Table 2
Within-journal comparison does not control article quality, author characteristics, selective self-archiving, or other article-level causes of citation differences. The external physics and astronomy comparisons bound self-selection only indirectly.
↳ Method §2; Results §4, self-selection discussion
The paper says its correlations cannot determine causality and reserves timing analysis for future research, yet the title and Conclusion present self-archiving as a way to maximize citation impact.
↳ Title; Abstract; Results §4; Conclusion §5
The principal contribution is the expansion of an OA–citation comparison to ten disciplines, twelve years, and more than 1.3 million articles. The within-journal design and crawler validation support the descriptive association, but they do not resolve article-level self-selection or establish whether OA availability preceded citation accumulation. That distinction matters because the title and final self-archiving recommendation are causal and actionable despite the paper’s explicit caveat in §4. Table 2’s incompatible correlation and significance pairs add a separate reporting concern, although the table addresses an intermediate pattern rather than the headline Figure 2 result.
Nabu’s assessment, alongside the field’s view.
Are you an author of this paper?
Limited2.7
Confidence highThe study meaningfully extends prior computer-science and physics findings to 1,307,038 articles across ten disciplines and twelve publication years. Its advance lies in scale and cross-disciplinary generality rather than a new identification strategy.
“Comparing OA and NOA articles in the same journal/year, OA articles have consistently more citations”
Within-journal comparisons and a hand-checked crawler validation provide useful design controls, but article-level self-selection and the timing of OA availability relative to citation accumulation remain unresolved. Table 2 also contains correlation and p-value pairs that raise a reliability concern, while causal framing extends beyond the observational design.
“In and of themselves, these correlations and temporal patterns cannot determine causality.”
The progression from collection and classifier validation to results is followable, and the body explicitly distinguishes association from causation. The causal title and recommendation to self-archive to maximize impact materially exceed that qualification.
“publish and self-archive” so as to maximize research access and impact
The paper connects its analysis to prior computer-science, physics, astronomy, download, and citation research and explicitly discusses self-selection. Its causal and timing limitations are not consistently traced into the final policy recommendation, constraining the contextual interpretation.
“Further research will analyze the effect’s timing, causal components and relation to other variables”
Lower confidence on Methodological Rigour, Positioning — domain match limited.
Concerns4 of 4 checks
Table 2 contains apparently incompatible correlation and p-value pairs for N=12. The inconsistency affects an intermediate analysis rather than the principal cross-disciplinary association, but it calls the reported significance pattern into question.
The bibliometric analysis uses published citation metadata and public web-accessibility records, so no human-participant ethics requirement is indicated. Missing conflict, preregistration, data, and code declarations are not treated as conduct concerns for this study type and period.
Flags: 0 declared / 5 total
13 of 13 checkable references verified
21 references in manuscript 8 have no canonical index record — counted, but not index-checkable 7 references confirmed by manual review
References not verified automatically are resolved manually within 24 hours.
No retraction or expression of concern found in Crossref (includes Retraction Watch).
Sources: Crossref (incl. Retraction Watch) ✓
Where this paper’s evidence sits on the path from initial observation to real-world use.
The analysis uses real publication and citation records, giving it practical relevance beyond a laboratory setting. It does not test an access mandate, implementation strategy, or intervention, so additional causal and operational work is required.
“Further research will analyze the effect’s timing, causal components”
AI-generated, human-governed. Something look off? Contact us to request a review.
Keep reading
Other papers Nabu has evaluated blind, starting with the same field.
Baptiste Scancar, Jennifer Anne Byrne, David Causeur, et al. · 2026 · Media and communications
These findings suggest that paper mills inflate the citation metrics of supported publications and affected journals at scale.
TentativeChao Ma, Yiwei Li, Feng Guo, et al. · 2019 · Economics
papers published between October and December on average get as much as 18.5% fewer citations than those published in the other months in the year
TentativeMasashi Kawai, Koichi Ojiro, Makoto Asaoka, et al. · 2022 · Other social sciences
日本の機関リポジトリによるオープンアクセス論文の数は限られるものの，その被引用数と様々なタイプのオープンアクセス論文の被引用数の比較からは，日本の機関リポジトリによるオープンアクセスのメリットは確認できなかった．
Tentative