---
title: "Citation bias"
type: "entity"
entity_kind: "concept"
status: "hub"
canonical_name: "Citation bias"
aliases: ["reference bias","selective citation"]
first_seen: "2026-11-03T00:00:00.000Z"
writer_model: "claude-sonnet-5"
connects_to: ["legitimation not origination","Matthew effect","evidence-based medicine","Peter C. Gøtzsche","Steven Greenberg"]
seek_code_commit: "d40ee39"
---


The practice of selecting or withholding a citation by how well the cited work's content fits a predetermined conclusion, rather than by the cited work's actual methodological bearing on the claim — "legitimation-selection," distinct from popularity-driven citation (the [[entity-matthew-effect|Matthew effect]]). Matters to this vault as a mechanism independently coined at least three times in three different research traditions that do not appear to cite one another on the point: in physics (Irvine & Martin, 1985), in clinical epidemiology as "reference bias" ([[entity-peter-gotzsche|Gøtzsche]], 1987, corroborated by Ravnskov 1992 and pooled by Duyx et al.'s 2017 meta-analysis of 52 studies), and in biomedical citation-network analysis (Greenberg, 2009) — the same convergent-coinage pattern the vault's [[moc-legitimation-not-origination]] cluster documents by case study (Song Jian's group citing Kwakernaak to buttress a predetermined population-policy argument).

A targeted 2026-11-03 search found no source that explicitly sets this mechanism and the Matthew effect side by side as two named poles of one distinguishing axis — the two literatures appear to be separate conversations; see [[observation-no-literature-located-crossreferences-citation-bias-and-matthew-effect-as-one-axis]] and the open [[question-does-an-outside-literature-name-the-legitimation-popularity-citation-distinction]].

## References
- [[claim-gotzsche-1987-first-demonstration-of-citation-bias-corroborated-by-ravnskov-and-duyx]]
- [[claim-greenberg-2009-citation-bias-names-selective-ignoring-of-contrary-data]]
- [[claim-greenberg-2009-citation-transmutation-names-hedge-erosion-mechanism]] — a related but distinct Greenberg mechanism (hedge erosion, not selective ignoring); kept apart in Greenberg's own vocabulary table.
- [[claim-bornmann-leibel-2025-cumulative-advantage-and-outcome-favoring-citation-named-as-citation-bias-instances]]
- Captures: 2026-11-03-does-citation-studies-or-sociology-of-science-literature, 2026-10-31-does-sociology-of-science-or-citation-analysis-literature

## Updates
- 2026-11-04: A follow-up search one layer up, in citation-*theory* overviews rather than citation-bias-measurement papers, found two outside sources that each place this mechanism and the Matthew effect inside one shared taxonomy under a common label — narrowing, but not closing, the 2026-11-03 negative finding above. Neither source uses the vault's exact two names side by side, and neither cites Gøtzsche, Greenberg, or Duyx. ([[observation-haustein-2016-and-bornmann-leibel-2025-narrow-but-do-not-resolve-citation-bias-matthew-effect-axis-question]], [[claim-haustein-bowman-costas-2016-matthew-effect-and-persuasion-hypothesis-named-as-sibling-poles]], [[claim-bornmann-leibel-2025-cumulative-advantage-and-outcome-favoring-citation-named-as-citation-bias-instances]])
- 2026-11-05: The vault's three best framings of how this mechanism relates to the Matthew effect disagree on structure, not just on vocabulary: [[observation-greenhalgh-kwakernaak-citation-petiska-chatgpt-citation-confirmed-legitimation-bridge|one vault synthesis]] nests the Matthew effect as a species under "legitimation" as genus (itself unsourced, see [[observation-legitimation-genus-matthew-effect-species-framing-is-unsourced-synthesis]]); Haustein et al. treat the two as coordinate under an unnamed fourfold "distortion" category; Bornmann & Leibel place them asymmetrically under "citation bias" itself — an outcome-favoring mechanism as its one highlighted worked example, cumulative advantage/preferential attachment a separate section away as a *generator* of bias rather than a co-equal instance. No located framing uses "legitimation" as the superordinate term. ([[observation-three-framings-of-legitimation-matthew-effect-relationship-disagree-on-structure]])
- 2026-11-06 (propagation-repair): the 2026-11-05 line above was corrected in place. WAS: "Bornmann & Leibel treat them as coordinate under 'citation bias' specifically ... and only one of the three nests at all." NOW: Bornmann & Leibel are asymmetric — outcome-favoring citation as the lead example of citation bias, cumulative advantage in a separate section as a mechanism that generates bias; so two of the three framings nest, and they nest differently. Propagates the 2026-11-06 correction in [[observation-three-framings-of-legitimation-matthew-effect-relationship-disagree-on-structure]] (itself following the 2026-11-05 audits of [[observation-haustein-2016-and-bornmann-leibel-2025-narrow-but-do-not-resolve-citation-bias-matthew-effect-axis-question]] and [[claim-bornmann-leibel-2025-cumulative-advantage-and-outcome-favoring-citation-named-as-citation-bias-instances]]).