Source-linked AI summary
Visualization of the Citation Impact Environments of Scientific Journals: An online mapping exercise
Loet Leydesdorff
TL;DR
The paper addresses how to analyze journal citation environments without imposing a single robust classification of specialties and disciplines. It provides journal-specific cited and citing maps with local-impact visualization, and concludes that the resulting perspectives can answer practical questions, including policy-oriented analysis of nanotechnology.
Problem
A single robust classification of specialties and disciplines is not viable because journal structures are fuzzy and perspective-dependent.
Method
The paper makes journal-specific cited and citing citation environments accessible through visualization, independently representing impact and activity dimensions.
Results
The approach can provide conclusive answers to practical questions, including access to policy priorities through both impact and activity dimensions.
Takeaways & Limitations
Nano Letters demonstrates how the two citation dimensions can be used to access an interdisciplinary policy-priority area.
Takeaways & Limitations
The analysis remains bounded by the conclusion that no single best classification is possible and by citation relationships that may be maintained chiefly by a single author.
Abstract
from arXiv · showhide
Aggregated journal-journal citation networks based on the Journal Citation Reports 2004 of the Science Citation Index (5968 journals) and the Social Science Citation Index (1712 journals) are made accessible from the perspective of any of these journals. The user is thus able to analyze the citation environment in terms of links and graphs. Furthermore, the local impact of a journal is defined as its share of the total citations in the specific journal's citation environments; the vertical size of the nodes is varied proportionally to this citation impact. The horizontal size of each node can be used to provide the same information after correction for within-journal (self)-citations. In the "citing" environment, the equivalents of this measure can be considered as a citation activity index which maps how the relevant journal environment is perceived by the collective of authors of a given journal. As a policy application, the mechanism of interdisciplinary developments among the sciences is elaborated for the case of nanotechnology journals.
1. Introduction
The introduction frames journal-citation mapping as a way to study the intellectual organization of science while avoiding an imposed, supposedly definitive classification. It motivates user-centered citation environments because disciplinary boundaries are fuzzy, perspective-dependent, and difficult to use as stable baselines for measuring change.
- Research motivation: Journal citation relations may reveal clusters that represent the intellectual organization of the sciences.This operationalizes knowledge through journals rather than institutions and people.
- Research motivation: Publication and citation frequencies vary among disciplines, making cross-disciplinary publication practices and global impact factors difficult to compare.The conventional impact factor does not account for the database’s intellectual structure.
- Methodological problem: Local impact measures require either robust journal classifications or an alternative unit of analysis such as aggregated journal-journal citation structures.Paper-level co-citation and co-word maps can show historical developments but do not themselves identify disciplinary organization.
- Methodological problem: A robust delineation of specialties and disciplines is not viable because journal-space boundaries are fuzzy and decomposition depends on analytical choices.Sensitivity arises from seed journals, thresholds, similarity criteria, clustering algorithms, and multidimensional structure.
- Proposed direction: Online journal maps let end-users choose the entrance journal, clustering, and perspective instead of relying on one objectified cluster-level representation.Cited and citing environments provide distinct perspectives on a journal’s position, with cited impact differing from citing behavior.
- Proposed direction: Cited and citing aggregates can also reveal hierarchies among journals, while visualization and quantification support practical analysis of citation impact.The two dimensions should be distinguished because impact and citing behavior can differ.
2. Methods
The methods construct journal-specific citation environments from asymmetric journal-citation data and visualize them using cosine similarities and Pajek. Node dimensions encode local citation contribution, with a second dimension correcting for within-journal citations.
- Normalization: The citation environment retains journals exceeding the relevant threshold, uses citation-matrix totals for normalization, and can apply the diagonal element as a correction.The seed journal determines the environment, while tail relations are discarded and values above one enter the matrix.
- Network construction: The study maps journals using cosine similarities between their citation-distribution vectors.Cosine similarity is chosen for sparse matrices and supplies positive values suitable for visualization.
- Network construction: The analysis creates separate cited and citing files for each journal and makes them available for visualization and further processing.The files can be read into Pajek and exported in formats such as DL.
- Environment definition: In 2004, cited and citing environments are normalized independently, so the two environments can differ substantially.This replaces the combined 2003 criterion with direction-specific environments.
- Impact encoding: Node size represents each journal’s percentage contribution to citations, while horizontal size applies a correction for within-journal citations.Ellipse shape indicates dependence on an inner circle of authors citing one another.
- Normalization: Citation counts were preferred to citation-per-publication ratios because publication counts and citation windows are ill-defined for this purpose.A validation study found total cites correlated more strongly with users’ appreciations than impact factor did.
3. Materials
The study builds journal citation matrices from Journal Citation Reports and organizes them in a relational system for journal-specific analysis. It compares citation environments with and without within-journal citations and reports pronounced sparsity and disciplinary differences in network density.
- Citation data were harvested from CD-ROM versions of the Journal Citation Reports for the Science Citation Index and Social Science Citation Index.
- A relational database stores the citation data so matrices can be extracted from perspectives based on selected seed journals or journal lists.
- The analysis is limited to individual journals, with a dedicated program generating input files for SPSS, UCINET, and Pajek.
- Citation environments are analyzed with and without within-journal citations, whose share is about 10% across the files but varies considerably among journals and specialties.
- Network density is more than twice as high in the sciences as in the social sciences, while all matrices are extremely sparse.
- The online visualizations use Pajek's Kamada–Kawai algorithm, while normalized cosine matrices and input files support alternative visualization and clustering techniques.
4. Results
The citation environments distinguish cited and citing perspectives, revealing information-science and policy-analysis groupings, strong self-citation effects, and journal-specific citation patterns. The visualizations quantify local impact, corrected impact, and citation activity while also exposing cases where the method breaks down.
- Cited environment: Scientometrics is embedded in information-science journals and shares an interface function with Research Policy and Administrative Science Quarterly.
- Citing environment: A three-factor solution explaining 73.3% of variance distinguishes information-science, Scientometrics–Research Evaluation, and science-policy-analysis citation patterns.
- Cited environment: In 2004, Scientometrics received 860 citations, including 366 within-journal citations (42.6%); nine environment journals supplied 222 citations (25.8%).
- Impact measures: After correcting for within-journal citations, Scientometrics accounted for 7.76% of citations in its environment, compared with 20.66% before correction.
- Impact measures: Node dimensions depend on the citation environment: vertical size represents local citation impact, while horizontal size represents the within-journal-citation-corrected value.
- Citing environment: The citing environment reaches broader literature but has a more focused core, with only five other journals cited above the one-percent threshold.
- Citing environment: JASIST and Research Policy are the most actively citing partners, but Research Policy's activity is dominated by within-journal citations at 473/514, or 92%.
- Limitation: For highly specific citation behavior, the visualization method can collapse to a single zero-width line, and the authors state that their methods break down.
5. Policy relevance: Nanotechnology
Nanotechnology is examined as a policy-relevant example of interdisciplinary development because its disciplinary boundaries remain difficult to delineate. Citation environments show that nanotechnology journals are visible across established fields but have not yet formed a strongly interconnected core.
- Nanotechnology is a policy priority, but its boundaries as an interdisciplinary field remain difficult to delineate.
- Nano Letters ranks as the leading core nanotechnology journal on both impact factor and total citations.
- Nano Letters connects nanotechnology with core nano journals, chemistry, applied physics, chemical physics, materials science, and polymer chemistry, with polymer relations more indirect.
- Nanotechnology journals do not yet actively cite one another as a core set, instead citing major general-science and neighboring-discipline journals.
- The pattern suggests that interdisciplinary developments first become visible in cited environments, while citing patterns remain weakly integrated because the specialty is not yet stabilized and codified.
6. Conclusions and discussion
The paper presents journal-based citation environments as a flexible way to analyze local impact, citation activity, interdisciplinary links, and journal structure. It also emphasizes that aggregation, cutoff levels, and database coverage constrain interpretation and should remain responsive to the user's analytical perspective.
- Methodological considerations: No single theoretically best classification exists, because relevant environments remain sensitive to journal inclusion, exclusion, and the chosen aggregation perspective.The study therefore leaves aggregation substantially to the user, while noting that substantive or methodological arguments are still needed.
- Citation environments: Journal-based environments let users inspect local citation patterns by selecting seed journals and examining their linked citation networks.The system provides Pajek input files and supports practical comparisons across different seed journals.
- Methodological considerations: Citation matrices differ across databases: the Science Citation Index is denser and more profiled than the Social Science Citation Index, making it easier to analyze in principle.The two indices also differ in their relevant distributions, so cutoff and aggregation choices cannot be assumed universal.
- Citation environments: The analysis distinguishes being-cited impact from citing activity and uses both dimensions to examine policy-relevant areas such as nanotechnology.Nano Letters is presented as an example accessible through both its impact and activity dimensions.
- Citation environments: Within-journal citations can be treated as additional information about journal structure rather than merely normalized away.The study reports uneven cited/citing ratios and sometimes up to 100% within-journal citation rates in the Chinese data.
- Methodological considerations: Reducing the Social Studies of Science cutoff from 1% to 0.1% made its extremely sparse citing environment visualizable.Cutoff choices affect whether interdisciplinary journal relations can be displayed in sparse networks.