Scientometrics

• The field originated from Garfield’s 1960s idea of an information-retrieval index, culminating in the Science Citation Index and quantitative science studies.The Science Citation Index was created as a novel instrument for empirically studying science, while Price’s work established foundations for quantitative science studies (Price, 1961, 1963, 1965, 1976; Garfield, 1979; Wouters, 1999).Source: Introduction · p. 2
• The launch of Scientometrics in 1978 and publication of Toward a Metric of Science institutionalized new quantitative perspectives involving historians, philosophers, and social scientists.The edited volume included Robert K. Merton among its contributors (Elkana et al., 1978).Source: Introduction · p. 2
• During the 1980s, sociology of science increasingly turned toward micro-analysis of scientists’ laboratory behavior, while scientometrics analyzed scientific literature at the macro level and developed at arm’s length from sociology of science.Source: Introduction · pp. 2–3
• Scientometric indicators became important for informing science policies and research management, while attention to evaluation and ranking increased after the first Academic Ranking of World Universities was published in 2004.The first Academic Ranking of World Universities was published by Shanghai Jiao Tong University in 2004.Source: Introduction · p. 3
• Large Internet-accessible datasets enabled network-science studies using coauthorship and citation data to model scientific network dynamics.These approaches often did not seek social-scientific explanation, but their visualization techniques were partly derived from and adopted by scientometricians.Source: Introduction · p. 4
• Scientometrics has experienced spectacular literature growth and increasingly functions as a research front through changing referencing patterns and interdisciplinary exchanges.Network-science researchers adopted and contributed visualization methods, while coauthorship and citation data supported statistical modeling of network dynamics.Source: Introduction · p. 4

• Scientometrics focuses on texts as empirical units, while authors, theories, and contents remain analytically distinct and require theorizing or hypotheses.Measures in one dimension can serve as proxies for another when supported by a research design.Source: The disciplinary organization of… · pp. 4–5
• Factor analysis can derive latent journal structures by reorganizing aggregated citation linkages among journals.Source: The disciplinary organization of… · p. 5
• The 2010 citation network of 32 journals revealed four communities, with Scientometrics mainly situated within the information sciences.The network included journals contributing more than 1% of references to Scientometrics, and the map reported modularity Q = 0.506.Source: The disciplinary organization of… · pp. 5–6, Figure 2 · p. 6
• JASIST and specialized outlets such as Journal of Informetrics became important publication venues, while Scientometrics remained the field’s core journal.Scientometrics interfaces with Social Studies of Science, general science journals, and technology and innovation studies, with more remote links to social network analysis and organization studies.Source: The disciplinary organization of… · p. 6
• Scientometric research also links to multidisciplinary and physics journals associated with network science, positioning the field among social sciences, information science, and advanced computing.These connections reflect the incorporation of physics, mathematical statistics, bibliographic databases, and big-data pattern analysis into scientometric research.Source: The disciplinary organization of… · p. 7

a. The measurement of impact

• Citation indexing and rankings have become increasingly central to research management and science policy, making debates about citation meaning and indicators a driver of scientometrics.Google Scholar and Scopus emerged as alternative citation indexes in 2004.Source: a. The measurement of impact · pp. 7–8
• Right-skewed citation distributions make arithmetic means ill advised, while publication-count denominators can distort impact across aggregation levels; raw scores therefore require normalization.A principal investigator may lose average impact when junior staff are included, even though the research group has greater cumulative impact.Source: a. The measurement of impact · p. 8
• Percentile and nonparametric statistics accommodate skewed distributions, with normalized article-level ranks aggregable to institutions, nations, and individual scholars’ œuvres.They also permit statistical significance testing and error estimates; the top-10% class is used as an excellence indicator in major university rankings.Source: a. The measurement of impact · pp. 8–9
• Nonparametric indicators attributed PNAS an impact significantly higher than Science or Nature, although the latter journals had an impact factor three times higher; they correlated more strongly with citations and publications than impact factors.Source: a. The measurement of impact · p. 9
• The h-index combines author productivity and impact but is crude, while Egghe’s g-index increases sensitivity to highly cited papers.These indexes can rank scholars within disciplines because publication and citation behavior varies across fields.Source: a. The measurement of impact · p. 9

b. The delineation of a reference set

• Reference sets are difficult to delineate because impact results are highly sensitive to assigning papers to disciplines, specialties, and interdisciplinary categories.This challenge is especially acute for universities containing multidisciplinary units.Source: b. The delineation of a reference set · p. 10
• Using citing papers as reference sets and fractionating citation credit by reference-list length accounts for papers addressing audiences with different citation practices.Following Moed (2010), Leydesdorff and Opthof (2010) proposed this procedure.Source: b. The delineation of a reference set · p. 10
• 19th to 2nd: Zhou & Leydesdorff (2011) found that fractional citation counting dramatically improved the Chinese Language and Literature department’s ranking at Tsinghua University.The example used all 2005 publications with Tsinghua University in Beijing as the institutional address.Source: b. The delineation of a reference set · p. 10
• Fractionation cannot assess uncited articles, and no single threshold eliminates the fuzziness caused by variable disciplinary groupings.Highly specific document sets may represent local environments carefully but remain difficult to aggregate into general conclusions, with exceptions varying across sciences.Source: b. The delineation of a reference set · pp. 10–11

Normative and constructivist theories

• Normative theories treat citations as manifestations of influence and a currency for paying tribute in science, whereas constructivist theories emphasize their rhetorical construction.This distinction follows Merton’s (1942) norms of communalism, universalism, disinterestedness, and organized skepticism.Source: c. Theories of citation · p. 11

Self-organization and layered networks

• Fujigaki (1998) interprets referencing as part of a self-organizing scientific system that selects knowledge elements for retention and further development.Citations help fields retain selected elements from the existing knowledge base for the next developmental round.Source: c. Theories of citation · p. 11
• Leydesdorff (1998) argues that texts and authors form layered networks in which citations can carry different meanings.Referencing can therefore connect the social and intellectual organization of science through mutual adjustments between selections in both dimensions.Source: c. Theories of citation · p. 11

Evolutionary model

• These interacting selection processes can produce trajectories that become meta-stabilized and globalized into regimes or paradigms, supporting an evolutionary model open to simulation and empirical research.The model links citation dynamics to the mutual adjustment of social and intellectual selections.Source: c. Theories of citation · pp. 11–12

d. Mapping science

• Mapping science visualizes relationships among scientific elements for information retrieval, understanding scientific dynamics, and informing resource and reward allocation.Its focus is the topology of relationships and its uses span retrieval, explanation, and policy decisions.Source: d. Mapping science · p. 12
• Scientometric mapping reveals the internal structure of disciplines, fields, and specialties by analyzing co-occurrences among subject terms, documents, authors’ œuvres, or journals.These maps are based on evidence from the literatures under study, with co-occurrence counts as the basic data.Source: d. Mapping science · p. 12
• Sciences develop as heterogeneous networks in which organized knowledge can be codified through mixtures of cited references and word usages.This heterogeneity extends across university, industry, and government domains; words are more volatile than citations. Callon et al. (1986)Source: d. Mapping science · p. 12
• Coauthorship relations alone do not adequately represent intellectual content, so social-network analysis must be combined with semantic maps.This combination generates richer representations of scientific communication networks.Source: d. Mapping science · p. 12
• A longer-term challenge is animating informed scientific networks over time while specifying statistical error, aided by advances in visualization and animation.Such advances can clarify the evolving roles of citations, title-words, keywords, and author names in scientific communication.Source: d. Mapping science · p. 12

e. Policy and management contexts

• Scientometric indicators are used in policy and management contexts, where utility is sometimes prioritized over validity.Source: e. Policy and management contexts · p. 13
• Impact-factor evaluations combine analyses across aggregation levels and may be applied to relatively small sets.Science maps and overlays can function heuristically at aggregate levels, but significance testing requires sufficiently large sets.Source: e. Policy and management contexts · p. 13
• Using indicators for low-level strategic choices, including “picking the winners”, is complicated by small sets, quality normalization, and pre-selected fundable proposals.Source: e. Policy and management contexts · p. 13
• Bornmann et al. (2010) found that matched “best rejected” proposers sometimes scored significantly higher than grantees on performance and impact indicators.Peer review may introduce unintended normative bias when selecting “excellent” proposals from “good” proposals; scientometric meta-evaluation may improve selection by highlighting that distinction is not always possible.Source: e. Policy and management contexts · p. 13
• Scientometric meta-evaluation responds to policy pressure by improving awareness that “excellent” and “good” research cannot always be distinguished.Source: e. Policy and management contexts · p. 13

Core contribution

• Scientometrics contributes quantitative models and measurements for studying science, technology, and innovation.The approach focuses on texts, communication, and the relations among authors, inventors, cognitions, institutions, and journals.Source: Conclusions and further perspectives · pp. 14–16

Textual mediation

• Scientific communication is textually mediated, enabling references to recombine local agency with cognitive organizers and allowing these processes to be traced through authenticated publications.Texts can travel across contexts, while manuscripts combine author names, institutional addresses, title words, and journal names before entering the published archive.Source: Conclusions and further perspectives · p. 14

Expanded scope

• The field has expanded across document types and domains, including patents whose citations, inventor names, and addresses reveal intellectual heritage and geographic technology diffusion.Patent analysis differs institutionally from public science because patents protect intellectual property, whereas public science also involves gift-giving.Source: Conclusions and further perspectives · pp. 14–15

Network dynamics

• Scientometric networks combine social and cognitive relations, whose forward information generation and backward meaning assignment can be distinguished as reconstructive incursion and historical recursion.Cognitive dynamics provide meaning retrospectively, whereas historical networks develop along the time arrow; anticipatory-systems theory distinguishes these operations.Source: Conclusions and further perspectives · p. 15

Knowledge modeling

• Modeling knowledge exchange cannot be reduced to co-authorship, co-word, or citation information because scientific models specify uncertainty and provide meaning to possible future states.Future-state construction can be modeled as hyper-incursion, while interacting reconstructions update and reinforce societies’ and economies’ knowledge bases.Source: Conclusions and further perspectives · pp. 15–16

Further perspectives

• Measuring communication and shared meaning across university–industry–government frames remains on the research agenda, but meaningful codification in scientific discourse remains unresolved.This marks a remaining step beyond measuring communication or sharing of meaning among reference frames.Source: Conclusions and further perspectives · p. 15