Source-linked AI summary
Dynamic Animations of Journal Maps: Indicators of Structural Changes and Interdisciplinary Developments
Loet Leydesdorff, Thomas Schank
TL;DR
Distinguishing structural change from variation in the organization of science is difficult when observable changes may not reflect latent structural changes. The paper uses dynamic MDS to balance within-year and between-year stress, and its journal cases support a view of interdisciplinarity as an interface between existing specialties.
Problem
Observable variation has been difficult to distinguish from changes in latent structures, complicating analysis of structural change in journal-citation data.
Source: §1 · p. 2Method
A dynamic MDS algorithm lays out network time series by optimizing stress within each year and across consecutive years.
Source: §1 · p. 3Results
Nanotechnology emerged as an application domain in physics and chemistry focused on nanostructures, while advanced material sciences formed a citation cluster at their interface.
Source: §4 · p. 20Takeaways & Limitations
The conclusion characterizes interdisciplinarity as an interface emerging between two existing specialties, with potential stabilization beyond two specialties.
Source: §4 · pp. 19–20Substructures of citation matrices may evolve at different speeds.
Source: §1 · p. 3
Abstract
from arXiv · showhide
The dynamic analysis of structural change in the organization of the sciences requires methodologically the integration of multivariate and time-series analysis. Structural change--e.g., interdisciplinary development--is often an objective of government interventions. Recent developments in multi-dimensional scaling (MDS) enable us to distinguish the stress originating in each time-slice from the stress originating from the sequencing of time-slices, and thus to locally optimize the trade-offs between these two sources of variance in the animation. Furthermore, visualization programs like Pajek and Visone allow us to show not only the positions of the nodes, but also their relational attributes like betweenness centrality. Betweenness centrality in the vector space can be considered as an indicator of interdisciplinarity. Using this indicator, the dynamics of the citation impact environments of the journals Cognitive Science, Social Networks, and Nanotechnology are animated and assessed in terms of interdisciplinarity among the disciplines involved.
1. Introduction
The introduction frames a central challenge: separating structural change from variation in evolving journal-citation data. It motivates a dynamic MDS approach that balances fit within years and continuity across years.
1. Introduction
- Static structural models assume a fixed number of factors, although the number of relevant dimensions may vary over time.The passage notes that substructures of citation matrices may also evolve at different speeds.Source: §1 · pp. 2–3
- Changes in observable variation have been difficult to distinguish from changes in latent structures, while allowing factors and loadings to vary makes models unidentifiable without further assumptions.The earlier multivariate extension of entropy statistics also did not sufficiently separate variation from structural change in aggregated journal-citation indicators.Source: §1 · pp. 2–3
- ARIMA models had not been developed for large, evolving matrices such as networks, and different citation-matrix substructures may evolve at different speeds.The introduction also notes that autocorrelated measurements violate the independent-error assumption in regression models.Source: §1 · pp. 2–3
- The MDS-based algorithm optimizes stress within each year and across consecutive years to lay out network time series dynamically.The algorithm was implemented in Visone.Source: §1 · p. 3
- The study applies the algorithm to citation-impact environments of journals selected for expected interdisciplinary development.The introduction names Cognitive Science, Social Networks, and Nanotechnology as the journal cases.Source: §1 · p. 4
1. In the case of Cognitive Science, we elaborate on a study by Goldstone &
The section sets out prior observations and questions about interdisciplinarity in three journal citation environments. It highlights Cognitive Science and Social Networks' reported centrality in 2004 and frames Nanotechnology's available citation record as evidence of field evolution.
1. In the case of Cognitive Science, we elaborate on a study by Goldstone &
- Goldstone and Leydesdorff found Cognitive Science had the highest betweenness centrality in its citation-impact environment in 2004 and conjectured it would remain high.The journal's editors had expected it to play an interdisciplinary role in its field (Collins, 1977).Source: §1 · p. 4
- Social Networks had high betweenness centrality in its 2004 citation-impact environment, during a turbulent transition in network analysis associated with internet research.The cited discussion raises whether the journal's interdisciplinary position emerged and could be sustained.Source: §1 · p. 4
- Nanotechnology's citation data, available from 1996, provide an indicator of the field's evolution during 1996–2006.Nanotechnology was described as a government funding priority in the twenty-first century.Source: §1 · pp. 4–5
2. Data and Methods
The section describes a dynamic MDS layout for annual journal citation environments, extending stress optimization across individual years and their sequence. It also specifies the data and cosine thresholds used to build the vector spaces in which betweenness centrality is assessed.
2. Data and Methods
- The journal citation data come from the Journal Citation Reports and are available electronically by year from 1994 through 2006.The analysis uses cited structures to study citation impact environments.Source: §2 · p. 5
- Cosine normalization is used to visualize skewed distributions; thresholds are cosine ≥ 0.2 for the first two case studies and ≥ 0.5 for Nanotechnology.The higher Nanotechnology threshold reflects denser natural-science citation networks.Source: §2 · p. 6
- Betweenness centrality is calculated separately for each vector space, and the time-sliced layouts can be animated with node sizes encoding centrality measures.The cosine-normalized matrices are converted into a time-sliced Pajek project for further processing.Source: §2 · pp. 6, 10
- The authors extend a stress-minimizing algorithm to lay out dynamic networks, optimizing stress both within each year and across years.The dynamic stress function combines static stress with stress between successive years.Source: §2 · pp. 6, 8
- The dynamic weight controls temporal stability: zero reduces the method to static analysis, while higher values increase stability across years.The layout includes the previous and next year when optimizing each year's configuration, helping preserve the mental map between consecutive layouts.Source: §2 · pp. 8–9
3. Results
The results trace changing interdisciplinary roles and citation environments in the journal maps for Cognitive Science, Social Networks, and Nanotechnology. They also cover Science’s—and, to a lesser extent, Nature’s—role in shaping the interface from 2000 onward, the emergence of nanoscience journals, and Nanotechnology’s betweenness centrality from 1996 to 2006.
3.1 Cognitive Science
- 30.3% betweenness centrality placed Cognitive Science well ahead of Annual Review of Psychology’s 5.1% in the 2004 citation environment.Cognitive Science was cited by journals in otherwise poorly related fields.Source: §3.1 · pp. 10–11, Figure 1 · p. 11
- Cognitive Science maintained high betweenness across the years, while its citation environments changed except for a stable connection to education science.It functioned at the margins of cognitive psychology to connect specialties across disciplinary divides, while psychology and education remained dominant frames.Source: §3.1 · p. 12
- In 2004, Social Networks held a central position between sociology and management science and exact-science branches focused on network analysis; the Journal of Mathematical Sociology and Organization Science supported this interface.The account identifies 43 of the 54 journals citing Social Networks; Figure 2 shows betweenness centrality of 43 journals in the vector space of citation impact.Source: b. Social Networks · p. 13, Figure 2 · p. 14
- Social Networks remained primarily associated with sociology, gained management-journal interest in the late 1990s, and consolidated its sociology–management position during the last decade.Movement away from the sociology core occurred in some years but was not consolidated.Source: b. Social Networks · pp. 14–15
- Nanotechnology was initially embedded in applied-physics journals, then became a central focus as journal-level citations reorganized the applied-physics–physical-chemistry interface.Its animation covers 1996–2006, and the journal played an important role in the field-level interdisciplinary development.Source: c. Nanotechnology · pp. 15, 17
- Nanoscience journals had emerged in the meantime (Leydesdorff & Zhou, 2007; Leydesdorff, forthcoming).Source: c. Nanotechnology · p. 17
- Science, and to a lesser extent Nature, played a role in shaping the interface in 2000 and thereafter.Source: c. Nanotechnology · p. 17
- Figure 4 shows the development of Nanotechnology’s betweenness centrality in its citation impact environment during 1996–2006.Source: Figure 4 · p. 18, c. Nanotechnology · p. 18
4. Conclusions and discussion
The cases show that interdisciplinary development often forms an interface between two specialties, while broader or more stable bridges remain limited. Their trajectories vary: Cognitive Science stabilized an education interface, Social Networks remained largely within sociology, and Nanotechnology linked applied physics with physical chemistry.
4. Conclusions and discussion
- New developments can first appear in journals focused on the topic, which may attract citations from neighboring disciplines.The authors describe citation as a way to measure that neighboring-discipline attention.Source: §4 · pp. 18–19
- Cognitive Science stabilized its interface with education research and maintained high betweenness centrality through citations from outside its original discipline.Its relations with more cognitively remote specialties remained in flux.Source: §4 · p. 19
- Social Networks became a sociology methods journal, but its Internet-research connections did not make it an interdisciplinary node between natural and social sciences.Management links were established, while physics journals did not cite it regularly enough; most years its citation pattern remained within its mother discipline.Source: §4 · p. 19
- Nanotechnology helped connect applied physics and chemistry, transforming advanced materials into a citation cluster at their interface.The field reorganized, with the journal first drawing physicists' attention to nanotechnology and then forging a relationship with chemistry.Source: §4 · p. 20
- Interdisciplinarity tends to produce an interface between two existing specialties, with the potential for that interface to coevolve.The authors suggest that combining more than two disciplinary codes can cause confusion and impede intellectual development.Source: §4 · p. 20
- Reach beyond a single interface tends to remain incidental, while disciplinary differentiation and identification can counterbalance it.These competing tendencies may lead to changes over time.Source: §4 · p. 20
- No stable social–natural-science interfaces emerged in these cases, although such bridges might have been expected for Cognitive Science and Social Networks.Different disciplinary communication codes, including methodologies, may make forming such a bridge “a bridge too far.”Source: §4 · p. 21