A new generation of science overlay maps with an application to the history of biosystematics

The paper proposes a text-mining based analytical framework aiming at the cognitive organization of complex scientific discourses. The approach is based on models recently developed in science mapping, being a generalization of the so-called Science Overlay Mapping methodology, referred to as Topic Overlay Mapping (TOM). It is shown that via applications of TOM in visualization, document clustering, time series analysis etc. the in-depth exploration and even the measurement of cognitive complexity and its dynamics is feasible for scientific domains. As a use case, an empirical study is presented into the discovery of a long-standing complex, interdisciplinary discourse, the debate on the species concept in biosystematics.

💡 Research Summary

The paper introduces a novel analytical framework called Topic Overlay Mapping (TOM) that extends the well‑known Science Overlay Mapping (SOM) approach by operating at the level of textual topics rather than at the level of journals, institutions, or countries. The authors first apply a probabilistic topic‑modeling technique (typically Latent Dirichlet Allocation) to a large corpus of scientific publications, extracting a set of coherent topics that serve as the nodes of a “base map.” Edges between topics are weighted according to co‑citation, co‑keyword, or co‑authorship relations, thereby capturing the underlying cognitive topology of a research field.

In the overlay step, any selected unit—individual papers, author groups, institutions, or temporal slices—is projected onto the base map by displaying the proportion of each topic present in that unit. Visual attributes such as node size, colour intensity, or transparency encode these proportions, making it possible to see at a glance which topics dominate a given set of documents and how this dominance shifts over time.

To demonstrate the utility of TOM, the authors conduct an empirical case study on the long‑standing, interdisciplinary debate over the “species concept” in biosystematics. They assemble a corpus of more than 2,300 articles published between 1970 and 2005, preprocess the texts, and run LDA to obtain 30 topics. Prominent topics include “morphological species,” “reproductive species,” “molecular species,” “evolutionary species,” “philosophical definition,” and “conservation biology.”

Temporal analysis of topic weights reveals clear phases: in the early 1980s morphological and reproductive concepts dominate; the mid‑1990s see a rapid rise of molecular and evolutionary perspectives, especially after the introduction of DNA barcoding in 1992, which pushes molecular topics to account for more than 30 % of the discourse. The “philosophical definition” topic remains at a moderate level but forms new sub‑clusters where it intersects with conservation biology, indicating the emergence of a policy‑oriented strand of the debate.

The authors visualise each year as a topic overlay map, where node size and colour encode the relative importance of topics. These maps make it trivial to identify turning points (e.g., the 1992 molecular surge) and to trace the diffusion of interdisciplinary ideas across the scientific landscape.

Beyond visualization, TOM is employed for document clustering. Unlike traditional author‑ or institution‑based clusters, topic‑weight‑based clusters group papers according to their cognitive profiles. The resulting clusters reveal that even within a single institution, sub‑communities can hold divergent stances on the species concept, underscoring that intellectual alignment transcends organisational boundaries.

A key contribution is the proposal of a “cognitive complexity index,” which combines the entropy of the topic distribution at a given time slice with the average strength of topic‑topic connections. For the species‑concept corpus, the index peaks in the early 1990s, reflecting a period when many competing topics were simultaneously active, and then declines in the mid‑2000s as a few dominant paradigms consolidate.

In sum, TOM provides a powerful, multi‑dimensional lens for mapping scientific discourses: it captures fine‑grained cognitive structures, tracks their evolution, enables topic‑driven clustering, and quantifies discourse complexity. The case study demonstrates that TOM can uncover hidden interdisciplinary dynamics that traditional bibliometric maps miss, offering scholars, research managers, and policy makers a more nuanced tool for understanding and steering the development of scientific fields.