Combined analysis of co-citation relations and words is explored to study time-dependent ("dynamical") aspects of scientific activities, as expressed in research publications. This approach, using words originating from publications citing documents in co-citation clusters, offers an additional and complementary possibility to identify and link specialty literature through time, compared to the exclusive use of citations. Analysis of co-citation relations is used to locate and link groups of publications that share a consensus concerning intellectual base literature. Analysis of word-profile similarity is used to identify and link publication groups that belong to the same subject-matter research specialty. Different types of "content-words" are analyzed, including indexing terms, classification codes, and words from title and abstract of publications. The developed methods and techniques are illustrated using data of a specialty in atomic and molecular physics. For this specialty, it is shown that, over a period of 10 years, continuity in intellectual base was at a lower level than continuity in topics of current research. This finding indicates that a series of interesting new contributions are made in course of time, without vast alteration in general topics of research. However, within this framework, a more detailed analysis based on timeplots of individual cited key-articles and of content-words reveals a change from more rapid succession of new empirical studies to more retrospective, and theoretically oriented studies in later years.
IntroductionCombined analysis of co-citation relations and wordprofile similarities is explored to improve the capability of quantitative techniques to depict structural and dynamical aspects of scientific research. In our foregoingThis study is part of a project financed by the Ministry of Education and Sciences, through the Netherlands Advisory Council for Science Policy (RAWB).Received February 24, 1989; revised October 4, 1989; accepted October 10, 1989. 0 1991 publication ("Mapping I," Braam et al., this issue) we emphasize the structural aspects ("local stability") of "science mapping," while in this article we focus on the analysis of dynamical aspects ("temporal stability") of scientific research.Starting with a clustering of documents that often co-occur in the reference lists of publications (cocitation clustering), publications in the dataset are grouped on the base of (one or more) citations to these clustered documents. This "classification" of publications is believed to partition the dataset according to participation of publications in research specialties Griffith et al., 1974;Small, 1977;Small & Crane, 1979). The prevalent idea that the "current research" publications of specialties are identified in this way, is based on theories of Price (1965) and Kuhn (1970). In particular the way researchers draw on earlier work, and their sharing of a set of "exemplars" (or "paradigm"), is considered to be reflected in the referencing practices of specialty members....