The Role of Journal-to-Journal Citation Maps in Science Mapping

Every scientific journal is part of a much larger conversation. When researchers cite earlier work, they leave behind a trail showing which ideas they built upon and which communities they belong to. When you collect millions of these citations and group them at the level of entire journals rather than individual papers, patterns begin to emerge. Some journals cite each other constantly, forming tight clusters. Others act as bridges between distant fields. A journal-to-journal citation map turns this invisible web of references into a picture you can actually study. It is one of the most powerful tools in scientometrics for understanding how scientific disciplines are organised and how knowledge moves between them.

What a journal-to-journal citation map actually shows 🔗

A journal-to-journal citation map is a visual network where each journal is a node, and the lines connecting them represent citation relationships. If articles in Journal A frequently cite articles in Journal B, a link is drawn between them. The more often this citation happens, the stronger the link. By plotting thousands of journals this way, researchers create a map that reveals the macro-level structure of science.

The core data behind these maps is the aggregated journal-journal citation matrix. Instead of counting how often one paper cites another, this matrix counts citations between whole journals. One axis lists the citing journals and the other lists the cited journals. Each cell records how many times one journal referenced another over a given period. This matrix is the raw fuel for mapping.

An important property of this matrix is that it is asymmetrical and mostly empty. Most journals never cite most other journals. Citations cluster into dense groups that represent specialties, while only a few interdisciplinary journals cite and are cited across many fields. This tendency means the matrix can be broken down into recognisable groupings, each one roughly corresponding to a research field or sub-field.

The data source behind the maps 🔗

Most journal-to-journal citation maps are built using data from Journal Citation Reports (JCR), a resource published originally by the Institute for Scientific Information and now associated with Clarivate. The JCR has compiled aggregated journal-journal citation data on a yearly basis since 1975, drawn from the Science Citation Index and later the Social Science Citation Index. This long, consistent record is what makes large-scale mapping possible.

For a single year, the citation data can be enormous. One widely cited study used the 2004 JCR data to connect more than 6,000 journals through over six million citations, then sorted them into dozens of modules representing fields. The scale gives you a sense of why visualisation is necessary. No one can read a six-million-entry table, but a well-drawn map makes the structure understandable at a glance.

Other databases now offer alternatives. Scopus also provides aggregated journal-journal citation data and can be used for mapping in much the same way. Researchers have compared maps built from Scopus against those built from JCR data and found that while the broad structure is similar, the specific journals selected and their positions can differ depending on the database. Google Scholar, by contrast, is harder to use for this purpose because it is not organised neatly around a controlled set of source journals.

Why aggregation matters 🔗

Working at the journal level rather than the article level has a clear advantage. Individual citation links between papers are noisy and sparse. Aggregating them up to the journal level smooths out this noise and reveals stable structures. These journal groupings are reproduced from year to year with considerable stability, which means they can serve as reliable indicators of how science is intellectually organised. At the same time, gradual shifts in these patterns can act as signals of evolutionary change in a field.

A short history of the method 🔗

The idea did not appear overnight. Derek de Solla Price, fascinated by the growth of scientific literature, suggested using aggregated journal-journal citations for analysis in the 1960s. Building on this, Narin and colleagues in 1972 used aggregated citations to distinguish groups of journals representing specialties, while Eugene Garfield developed citation analysis as a tool for evaluating journals. These were the founding moves of modern bibliometrics.

One early demonstration of citation analysis showed its investigative power. Garfield used citation data to disprove the common belief that Gregor Mendel’s foundational paper on genetics had been ignored by his contemporaries. The citation record revealed that Mendel’s work was in fact cited repeatedly, correcting a long-standing historical misconception. This showed that citation patterns carry real evidence about how ideas spread.

By the mid-1980s, several research teams were applying multidimensional scaling and related statistical techniques to JCR data to produce maps. Over the following decades, science mapping based on aggregated journal-journal citations grew into a substantial research field of its own, with many competing methods and indicators.

How the maps are built 🔗

Turning a citation matrix into a readable map involves a few key steps. Understanding them helps you read any map critically rather than taking its layout for granted.

Normalisation and similarity 🔗

Raw citation counts are biased by size. Large journals publishing many articles naturally generate and receive more citations. To compare journals fairly, the citation matrix is normalised, often using a vector-space model where each journal’s citation profile is treated as a vector and compared with others. Journals with similar citation profiles, meaning they cite and are cited by the same set of journals, are judged to be similar and placed close together on the map.

Positioning the nodes 🔗

Once similarity is measured, an algorithm decides where each journal sits. Early work used multidimensional scaling. Modern tools commonly use force-directed or spring-embedded algorithms, which treat strongly linked journals as if pulled together by springs and weakly linked ones as pushed apart. The result is a layout where physical distance roughly reflects intellectual distance.

Clustering into fields 🔗

Finally, journals are grouped into clusters. Some methods use information-flow approaches that model a “random surfer” moving through the citation network, partitioning thousands of journals into modules based on where this flow concentrates. Other methods use different clustering algorithms entirely. Each cluster typically corresponds to a discipline or specialty, and these clusters are what give the map its recognisable continents of science.

Popular tools for mapping 🔗

Several software packages have become standard for this work. Pajek is a long-established program for analysing and visualising large networks, and many early journal maps used it. VOSviewer, developed by Van Eck and Waltman, is now widely used and was built specifically for bibliometric mapping; one well-known map it produced shows the relations among 5,000 major scientific journals based on co-citation data. Gephi is another general network visualisation tool that researchers apply to citation data. These tools are largely free, which has helped make science mapping accessible to students and researchers worldwide.

Measuring influence, not just popularity 🔗

Citation maps do more than show structure. They also help measure the standing of journals, and here an important distinction appears. Simply counting citations tells you a journal is popular, but it does not tell you it is prestigious. A journal might collect many citations from low-ranked sources, while another collects fewer citations but from highly respected journals.

The breakthrough idea came from Gabriel Pinski and Francis Narin in 1976, who proposed that a journal should be considered influential if it is cited by other influential journals. This is a circular definition, but it can be solved mathematically through repeated iteration over the citation matrix until the values settle. The resulting Influence Weights rank journals by prestige rather than raw counts. This same circular logic later inspired the PageRank algorithm that powered early web search.

This network-based thinking also addresses a flaw in the traditional Journal Impact Factor. The Impact Factor is sensitive to self-citations because each self-citation adds to the numerator while the denominator stays fixed. Influence Weights, drawn from the eigenvectors of the aggregated citation matrix, are far less affected by such self-citation, making them a more robust measure of genuine standing. Modern indicators like the Eigenfactor and SCImago Journal Rank apply similar network principles.

What these maps let researchers do 🔗

The practical value of journal-to-journal citation maps lies in the questions they help answer about the scientific system.

Tracking the diffusion of ideas 🔗

Because citations flow in a direction, from citing journal to cited journal, maps can show how knowledge moves between fields over time. A striking example involves neuroeconomics. A map built from JCR data showed that in 1997 there were essentially no citations flowing between economics and neuroscience, yet within a few years a new bridge had formed between them. Maps make the birth of interdisciplinary fields visible as it happens.

Identifying influential and bridging journals 🔗

Maps reveal which journals sit at the centre of their clusters and which act as bridges connecting otherwise separate communities. A bridging journal that cites and is cited across multiple fields plays a special role in moving ideas around. For someone choosing where to publish or where to look for relevant literature, knowing a journal’s position in this larger structure is genuinely useful.

Delineating and managing disciplines 🔗

Decomposing the citation matrix into journal groupings has a practical payoff for research evaluation. Comparing a journal only against others in its true specialty is fairer than comparing across unrelated fields. These data-driven groupings can improve the consistency of journal collections and the baselines used in scientometric evaluation, which matters for libraries managing subscriptions and for institutions assessing research output.

Supporting science policy 🔗

At a higher level, maps help policymakers and funding bodies see the overall shape of research activity. They can highlight emerging areas of strategic importance, show where two fields are beginning to merge, and track how the structure of science evolves across years. Studies have used these maps to examine areas like nanotechnology, where new journals and citation links signal the rise of a fresh interdisciplinary domain.

Limitations to keep in mind 🔗

No map is the territory. Journal-level analysis hides variation within journals, since a single journal can publish work across several specialties. The choice of database changes the result, as JCR and Scopus do not contain identical sets of journals. The choice of normalisation, layout algorithm, and clustering method also shapes the final picture, meaning two analysts can produce different-looking maps from the same data. Citation practices differ between disciplines too, so fields that cite heavily can appear more central than they really are. A careful reader treats any single map as one informed view, not an absolute truth.

What do you think? If you were to map your own field using journal-to-journal citations, which journals would you expect to sit at the centre, and which might surprise you by acting as bridges to other disciplines? And given that the choice of database and method changes the map, how much trust should research evaluators place in any single citation map when making real decisions?

References

1. https://www.leydesdorff.net/cstpcd/
2. https://arxiv.org/pdf/1501.05462
3. http://www.eigenfactor.org/projects/mappingScience/
4. https://www.leydesdorff.net/scopusvsisi/
5. https://www.leydesdorff.net/jcr01/art/index.htm
6. https://arxiv.org/pdf/0909.3193
7. https://arxiv.org/pdf/1102.2891
8. https://arxiv.org/pdf/0707.0609
9. https://www.vosviewer.com/journal-map
10. https://arxiv.org/pdf/1002.2858
11. https://www.sciencedirect.com/science/article/abs/pii/S1751157719302950