Facing Diversity of Science: A Challenge for Bibliometric Indicators

Bibliometrics has come of age and is generally accepted, together with peer review, as one way to describe the activity of players in science. Van Raan (this issue), with his team a prominent contributor to this specialty, describes in the focus article a variety of state-of-the-art indicators that show the concern of bibliometricians for improving measures used in sensitive contexts such as research assessment. I shall not try to comment on all aspects of this substantial contribution. The sections of the focus article (performance, interdisciplinarity, structure) deal in various ways with the central question of the diversity of science as either an object or a framework for performance indicators. I limit my discussion here to a few practical issues stemming from this diversity: How do you define or structure a scientific field? Is normalization straightforward? Which measure can be used for measuring knowledge flows and relations among fields? As the present challenges of science and technology indicators can hardly be disconnected from a historical perspective of scientometrics, a bit of hindsight is perhaps necessary. There is a lot of evidence since the work of Price (1963), Nalimov & Mulczenko (1969), and Garfield & Cawkell (1970) that many facets of scientific activity are amenable to measurement and modeling. Scientometrics handles as typical objects the networks generated by scientific activity, especially those implicitly carried by published outputs (networks of authors–institutions, of documents, of lexical contents, etc.). A typical one is the network of citation link-