The Extraction of Community Structures from Publication Networks to Support Ethnographic Observations of Field Differences in Scientific Communication

TL;DR

This paper presents a network analytic approach combining publication data and ethnographic insights to identify and compare overlapping scientific communities, aiding understanding of disciplinary differences and communication practices.

Contribution

It introduces novel methods for delineating research specialties and extracting community structures from publication networks, validated through a case study in physical and chemical sciences.

Findings

Overlapping sub-communities differ in disciplinary orientation.

Community structures correlate with research practices.

Methods improve delineation of research specialties.

Abstract

The scientific community of researchers in a research specialty is an important unit of analysis for understanding the field specific shaping of scientific communication practices. These scientific communities are, however, a challenging unit of analysis to capture and compare because they overlap, have fuzzy boundaries, and evolve over time. We describe a network analytic approach that reveals the complexities of these communities through examination of their publication networks in combination with insights from ethnographic field studies. We suggest that the structures revealed indicate overlapping sub- communities within a research specialty and we provide evidence that they differ in disciplinary orientation and research practices. By mapping the community structures of scientific fields we aim to increase confidence about the domain of validity of ethnographic observations as well as of collaborative patterns extracted from publication networks thereby enabling the systematic study of field differences. The network analytic methods presented include methods to optimize the delineation of a bibliographic data set in order to adequately represent a research specialty, and methods to extract community structures from this data. We demonstrate the application of these methods in a case study of two research specialties in the physical and chemical sciences.