In Search of Epistemic Networks

Abstract

The organization of bodies of scientific literature has hitherto been studied mainly in order to analyze the relations among documents — that is, from the perspective of information retrieval. The use of words and co-occurrences of words, however, makes it possible in principle to analyze the relations among words within documents, and to raise the question of the extent to which it is possible to reconstruct sciences by means of lexigraphical tools. In this study, I discuss various `network models' in science studies, using co-occurrences of words (terms, predicates) as indicators of organization within a scientific text. This approach is fully explored in the case of one research paper published in FEBS-LETTERS, at various levels of aggregation — sentences, paragraphs and sections. The major findings are: (1) the three-dimensional word structure at the section level cannot be found nor reconstructed at lower levels of aggregation; however, this structure can be most clearly identified as `theoretical', `observational' and `methodological'; and (2) word usage within sentences and within paragraphs is specific for the position of sentences in paragraphs, sentences in sections and paragraphs in sections. I argue that words and their relations can be used as an operationalization of epistemic networks in scientific literature as conceptualized in the pragmatic tradition in the philosophy of science, but that one has to add the notion of `nested structures' of networks at different levels of aggregations, as proposed by the French co-word analysts, in order to achieve a fuller understanding of the organization of texts.