Source-linked AI summary
Science overlay maps: a new tool for research policy and library management
Ismael Rafols, Alan L. Porter, Loet Leydesdorff
TL;DR
Science and technology institutions are undergoing major reforms while disciplinary structures are being bypassed without a clear alternative socio-cognitive structure. The paper presents science overlay maps to visually locate research and demonstrates their uses with institutions and an emergent research topic, while examining their application conditions and limitations.
Problem
Major reforms and developments that bypass disciplinary structures create a need to study research without a clear alternative socio-cognitive structure.
Method
The paper presents a science overlay technique for visually locating bodies of research and introduces possible uses for research analysis.
Results
The technique is illustrated through examples from universities, industries, funding agencies, and the emergent research topic of carbon nanotubes.
Takeaways & Limitations
Science overlay maps support research analysis at global levels and can animate diffusion patterns of new research fields.
Takeaways & Limitations
The paper examines the validity, reliability, and qualitative conditions of application of science overlay maps.
Abstract
from arXiv · showhide
We present a novel approach to visually locate bodies of research within the sciences, both at each moment of time and dynamically. This article describes how this approach fits with other efforts to locally and globally map scientific outputs. We then show how these science overlay maps help benchmark, explore collaborations, and track temporal changes, using examples of universities, corporations, funding agencies, and research topics. We address conditions of application, with their advantages, downsides and limitations. Overlay maps especially help investigate the increasing number of scientific developments and organisations that do not fit within traditional disciplinary categories. We make these tools accessible to help researchers explore the ongoing socio-cognitive transformation of science and technology systems.
3 Amsterdam School of Communication Research (ASCoR), University of Amsterdam,
The section is indexed with keywords covering science mapping, classification, interdisciplinarity, and research evaluation.
- The paper concerns science mapping and overlay representations.
- Classification and interdisciplinarity are central themes.
- Research evaluation is an intended application area.
1. Introduction
The introduction presents science overlay maps as a response to institutional change and increasingly fluid research structures. It introduces their uses while emphasizing validity, reliability, responsible interpretation, and practical accessibility.
- The introduction links the tool to reforms, changing intellectual environments, and societal demands for relevance.These pressures make strategic choices about institutional positions and directions increasingly necessary.
- Science overlay maps visually position an organisation or research field against a stable global representation of science.This supports readable comparisons for science policy and research or library management.
- The paper demonstrates overlay maps for assessing portfolios and animating research-field diffusion.Examples cover universities, industries, funding agencies, and carbon nanotubes.
- The study aims to make global-map methods accessible to prospective science-policy and research-management users.It avoids detailed technical bibliometric discussion while directing readers to further literature.
- The introduction examines map validity, reliability, application conditions, and potential misreadings.It proposes meaningful uses while flagging misunderstandings and possible sources of error.
- Maps are presented as more interpretatively flexible than rankings, supporting critical scrutiny and responsible use.The authors seek to avoid policy and management misuse associated with bibliometric indicators.
2. The dissonance between the epistemic and social structures of science
The section describes a dissonance between disciplinary representations and the changing social and cognitive organization of science. It frames overlay maps as a way to examine emergent fields and organizations in relation to established disciplines.
- Many scientific activities no longer align with traditional disciplinary boundaries.Research at the frontier is described as addressing problems rather than expanding a single discipline.
- Inter-, multi-, and transdisciplinary research characterize newer forms of knowledge production.The section also describes mission-oriented institutes and centres targeting societal problems across disciplines.
- New fields such as computer and cognitive sciences do not fit neatly into the traditional tree of knowledge.This illustrates the broader challenge posed by developments that cross established categories.
- Emergent organizations and fields must be positioned relative to traditional disciplines for evaluation.The section asks whether changes can be measured against a baseline and whether new developments are transient or relabeled older work.
- The paper asks whether overlay maps can illuminate cognitive and organizational dynamics in increasingly complex sciences.The stated applications include informing research-related choices such as funding.
3. Approaches to mapping the sciences
The section contrasts local and global approaches to mapping science and motivates overlay maps through the need for stable, shared coordinates. It also presents maps as useful but interpretively open tools with important limitations.
- Science maps represent fields or organisations by positioning related elements nearby and dissimilar elements farther apart.They use similarity measures derived from correlations among information items and project multidimensional matrices into two or three dimensions.
- Science maps have developed through earlier periods of innovation, dispute, and methodological refinement.The section describes their development from the 1970s and notes an emerging consensus on science’s core structure.
- Maps position units in networks rather than ranking them, helping non-experts read large and diverse bibliometric datasets.They can combine data types and support multiple views of an issue.
- Maps remain heuristic tools rather than closed answers to policy questions.Their interpretive flexibility can encourage reflexive awareness, but also leaves room for manipulating interpretations.
- Users may misunderstand map validity and utility, while ethnographic or sociological validation of bibliometric-tool use remains limited.The section emphasizes that map interpretation requires attention to analytical pitfalls.
- Local maps can accurately describe relations within a field but usually cover only a small area of science.Their study-specific units and coordinates make comparisons and cross-area positioning difficult.
- Proper comparisons require shared units of representation and positional coordinates across maps.A full science map is needed to obtain stable coordinates, requiring broad bibliographic coverage and robust classification.
4. Global maps of science: the emerging consensus
Global science maps show a robust coarse structure despite methodological differences, revealing clustered poles, bridges, and a torus-like organization. This consensus supports a stable basemap for comparative analysis.
- Independent global maps show strong agreement in science’s core structure despite different units, measures, classifications, and visualization techniques.
- Science forms a fragmentary, quasi-modular structure of clusters and gaps across multiple levels rather than a continuous body.
- The global map contains major biomedical and physical-science poles, with social sciences and humanities forming a third, smaller and more diffuse area.
- A torus-like arrangement connects disciplinary poles through bridges while preventing any single discipline from occupying science’s overall centre.
- Different representations can disagree substantially about positions such as mathematics, and dimensional reduction makes relative distances difficult to interpret.
- The emerging consensus warrants a stable global template that supports comparisons for researchers, science managers, and policymakers.
5. Science overlay maps: a novel tool for research analysis
Science overlay maps place an organization’s or theme’s research profile onto a common global map of science. Using Web of Science subject categories and citation-based similarities, the approach enables visual comparison across distinct research portfolios.
- Overlay maps address unstable local maps by displaying research profiles on fixed global units and positions.
- The method supports interactive exploration, but printed figures may omit labels needed to fully interpret individual nodes.
- The basemap uses 221 Web of Science subject categories, citation similarities normalized with Salton’s cosine, and Kamada-Kawai visualization.
- Factor analysis groups subject categories into 18 macro-disciplines, whose labels and colours organize the global map.
- Analysts retrieve documents, assign them to subject categories, and size map nodes according to the number of documents in each category.
- Overlay profiles distinguish universities’ research strengths, including Amsterdam’s diverse clinical portfolio, Georgia Tech’s engineering-related fields, and LSE’s social-science focus.
Appendix 1.
Overlay maps provide a common visual framework for comparing research profiles across organizations and scientific themes. Their cognitive units and accessible production process support interpretation beyond specialist scientometric analysis.
- Overlay maps enable immediate, visually rich comparisons of organizations’ and themes’ research profiles.
- They represent research with disciplines and specialties that are more stable and interpretable than analytical aggregates from local maps.
- Users of the Science Citation Index can produce overlay maps without being scientometrics experts.
- The displayed units can vary, allowing maps to serve different analytical purposes through publication, citation, reference, growth, or other indicators.
6. Conditions of application of the overlay maps
Useful overlay maps require transparent, traceable, parsimonious, and question-appropriate procedures. Their interpretations remain constrained by classification uncertainty, sample size, journal-based units, and visualization choices.
- Journal-based classification is imperfect because journals combine epistemic foci, researchers read across journals, and content may not match categories.
- Although classification errors may average out in large aggregates, different classifications can still produce meaningful disagreements and unstable positions.
- About 70 papers suffice to assign at least 40% of papers to a category with significance level 0.05 under a conservative accuracy estimate.
- Small publication sets should not support strong claims about research location; for individual researchers, references may serve as a disciplinary-profile proxy.
- Overlay maps should make classifications, similarity measures, visualization techniques, and procedures publicly reproducible and independently validated.
- Parsimony favors procedures simple enough for counter-expertise or non-expert scrutiny, even at some expense of detailed accuracy.
- Maps require a clearly formulated evaluation or foresight question and organizational context because they cannot otherwise provide a well-targeted answer.
- Overlay maps are especially useful for benchmarking collaborative activities and examining temporal change.
7. Use in science policy and research management
Science overlay maps address the difficulty of describing increasingly fluid research activities through fixed disciplinary labels. They support comparisons of positions, shifts, and dissonances across organisations and research topics, including benchmarking, collaboration exploration, and temporal analysis.
- Motivation: Disciplinary labels may misrepresent university and R&D activities because researchers publish and cite beyond their departmental fields.The problem is intensified by increasing cross-disciplinary publication and citation.
- Core use: Science overlay maps locate and compare disciplinary positions, shifts, and dissonances at institutional or thematic levels.They provide a common visual basis for examining activities that do not fit conventional categories.
- Related applications: Earlier overlay-map applications compared national disciplinary strengths and organisational or funding-agency publication profiles.Other studies used overlays to assess laboratory interdisciplinarity, research-topic diffusion, cross-disciplinary citations, and emerging-technology knowledge bases.
- Related applications: Overlay maps have been used to study research-topic diffusion, cross-disciplinary citations, and multidisciplinary knowledge bases in emerging technologies such as nanotechnology.These applications span both established research fields and specific emerging-technology specialties.
- Applications: Applications include benchmarking organisations, establishing collaborations, and capturing temporal change across universities, corporations, funding agencies, and emergent research topics.The examples include carbon-nanotube research as an emergent topic.
Benchmarking
The paper demonstrates overlay maps as readable, standardised tools for benchmarking portfolios, identifying collaboration opportunities, and following shifts in emerging research. Examples show both detailed disciplinary patterns and practical limitations in classification, visual interpretation, and local relational detail.
- Benchmarking: Overlay maps can benchmark organisations by comparing publication profiles on a common disciplinary map.Standardised mapping makes comparisons easier for science managers and policy-makers.
- Benchmarking: Pfizer and AstraZeneca share a biomedical profile, but Pfizer is stronger in nephrology while AstraZeneca is more active in gastroenterology and cardiovascular systems.Both profiles include pharmacology, biochemistry, toxicology, oncology, clinical medicine, and chemistry.
- Benchmarking: Overlay choices include displaying inputs, outputs, or outcomes and normalising by organisation size, category size, or both.The figures are normalised by organisation size but not category size; category normalisation highlights relative specialisation even when a category is a small production share.
- Exploring collaborations: Nestlé’s nutrition and dietetics activity lies nearer biomedical sciences than other food research, suggesting nutrition as common ground between food and pharmaceutical research.The example is linked to Nestlé’s strategic investment in health-enhancing functional foods.
- Exploring collaborations: Funding-agency profiles reveal substantial overlap, suggesting possible duplicated effort and opportunities for coordination or interdisciplinary collaboration.The maps can also help identify research communities for prospective workshops and evaluate prospective grantees.
- Capturing temporal change: Carbon-nanotube research has shifted from materials sciences and physical chemistry toward applied, biomedical, environmental, health, and safety-related areas.Growth is highest in applied fields such as textile materials and biomaterials, while biomedical and environmental activity also expand.
- Limitations: The main limitations are imperfect category attribution, possible visual errors after dimensional reduction, and lack of local relational structures.Map accuracy also depends on sufficient sample size and classification reliability.
- Advantages: Overlay maps contextualise disciplinary diversity and support reflexive, plural interpretation without reducing interdisciplinarity to a single index or ranking.Their readability and contextualisation can complement static indicators and tables.