Map the roots and networks: co-citation, coupling and collaboration

Keywords show what a field talks about. References show what it builds on, and co-authorship shows who works with whom. Together they complete the picture.

Co-citation analysis: the intellectual base

Two references are co-cited when a later document cites both (Small, 1973). References that are frequently cited together tend to belong to the same school of thought or methodological tradition. Clusters in a co-citation network therefore reveal the intellectual foundations of the field: for a digital health collection you would expect clusters around technology acceptance theory (Davis, 1989; Venkatesh et al., 2003), clinical informatics and machine learning methods. Co-citation can also be run on cited first authors or cited sources (journal co-citation), which is coarser but more robust to reference formatting.

Because co-citation depends on later citing behaviour, it works best for established literatures; papers from the last two or three years have not yet been co-cited often. Start with the most locally cited references table, the list of works your collection cites most, which is often the first thing reviewers look at.

Bibliographic coupling: the research fronts

Two documents are coupled when they cite the same references (Kessler, 1963). Unlike co-citation, coupling is fixed at publication, so it works for recent papers and maps the current research fronts. The clustering by coupling map plots coupling clusters by centrality and impact (their documents’ normalised local citations) and labels each with its most frequent keywords. Coupling can be aggregated to authors, sources or countries: two countries are coupled when their papers cite the same work.

Historiograph and reference spectroscopy

The historiograph (Garfield, 2004) places the most locally cited documents on a timeline and draws arrows for direct citations between them. It shows how the key ideas of the field developed, from early foundational papers to recent ones that build on them. Reference publication year spectroscopy (RPYS; Marx et al., 2014) counts cited references by their publication year; years that stand out above the five-year median point to seminal works. Look up the most cited references from those peak years to identify the field’s historical roots.

Collaboration networks: the social structure

Co-authorship networks link authors, institutions or countries that publish together. Clusters show research groups or regional blocs; nodes with high betweenness centrality are brokers connecting otherwise separate groups; isolated large nodes are productive but inward-looking. The countries’ collaboration world map draws international co-authorships as arcs. Combine it with the corresponding author’s countries table: a country with many single-country publications (SCP) and few multiple-country publications (MCP) is productive but nationally focused. In the demo data, 28.7% of documents have international co-authors.

The three-field plot

The three-field (Sankey) plot links three fields at once, for example countries, authors and keywords, or cited references, authors and sources. It shows, in one figure, which countries’ authors work on which topics, and is a compact way to connect the social and conceptual structure.

Interpreting networks responsibly

• Name clusters by reading their key documents, not only their labels.
• Report the settings: the number of items, the minimum link weight, normalisation and clustering resolution. Different settings give different maps.
• Do not over-read positions: the distance between two nodes on a map is an approximation of relatedness, not an exact measure.
• Remember that self-citation and citation cartels can inflate co-citation, and that hyper-authored papers (with dozens of authors) can dominate collaboration networks; consider excluding them.