Does Conceptual Centrality Pay Off in Citations Everywhere? Topic-Dependent Evidence from Library and Information Science

Abstract Citation counts are widely used as indicators of scholarly recognition, but a paper’s citation performance need not correspond closely to its position within the conceptual structure of its research field. This study explores that distinction in Library and Information Science by introducing a topic-relative measure of conceptual centrality, defined as the leave-one-out cosine similarity between a paper’s embedding and the centroid of the other papers in its topic. Using a Scopus corpus of 14,720 papers organized into 52 empirically derived research topics, centrality shows a positive but modest aggregate association with citation rate (Spearman’s rs = .107, p < .001). The association becomes stronger after accounting for the journal citation environment (standardized β = .095 for centrality versus .312 for journal environment), consistent with suppression rather than confounding. More notably, its strength is not uniform across topics. Among 45 topics with at least 40 regression observations, 26 show a statistically distinguishable positive slope, 19 are inconclusive, and none shows a reliable negative slope. This heterogeneity is not an artifact of topic sample size (p = .844). The pattern persists when centrality is constructed using only prior publications (F = 4.79, p < .001), and a residual analysis shows that review articles are modestly overrepresented among papers cited more than expected (7.7% versus 5.6%, p = .008). The findings support a multidimensional view of scholarly recognition in which conceptual embeddedness and citation attention are related but not interchangeable, and their relationship depends on the topical context in which a paper sits.