Measuring Multidisciplinarity Using the Circle of Science Kevin W. Boyack 1 and Richard Klavans 2 1 kboyack@mapofscience.com SciTech Strategies, Inc., Albuquerque, NM 87122 (USA) 2 rklavans@mapofscience.com SciTech Strategies, Inc., Berwyn, PA 19312 (USA) Abstract In this paper we introduce a method for measuring multidisciplinary research that is based on the assumption that the underlying structure of science is stable and has extremely low dimensionality. Introduction Popular measures of multidisciplinary research (MDR) are based on the assumption that the underlying structure of research has very high dimensionality. For example, the most common approach is to characterize research into tens or hundreds of weakly related disciplines where each discipline is composed of journals and relatedness is based on the citation data from the articles in these journals. One can then use simple measures of dispersion to characterize MDR, such as publications or references across disciplinary categories (Leydesdorff, 2007; Porter, Cohen, Roessner, & Perreault, 2007; Porter, Roessner, & Heberger, 2008; Rinia, van Leeuwen, Bruins, van Vuren, & van Raan, 2002). Studies using network centrality measures between journals assume an even higher level of dimensionality to science (Leydesdorff, 2007; Rafols & Meyer, 2009). This paper is based on an alternative assumption – that the underlying structure of research has extremely low dimensionality. We have recently shown that one can represent the vast majority of relatedness between disciplines using one dimension (Klavans & Boyack, 2009a). Disciplines can be placed around the perimeter circle of science in such a way that disciplines that are related tend to be close to each other. Millions of articles can correspondingly be placed on the edge of the same circle using co-citation analysis, where co-citation clusters (of papers) are assigned to their dominant disciplines. The location of a department or a researcher can then be plotted as the average position of all of their corresponding publications. Research that is closer to the center of the circle is more multidisciplinary. Research that is not multidisciplinary is located at the edge of the circle. In this paper we describe this approach along with some advantages and disadvantages associated with its use. Data and Models This paper builds on three data sources: a meta-analysis of 20 maps of science, a disciplinary analysis of 16,000 journals, and a co-citation analysis of 2.08 million reference papers and the corresponding 5.68 million articles that cite these references. Following is a short description of each data source and how they relate to each other. The meta-analysis of 20 maps of science focused only on maps that were relatively comprehensive (covering the majority of the natural and social sciences). The earliest maps (from 1939 and 1948) were done by hand; experts drew major areas of science on a piece of paper with the assumption that proximity (between areas) corresponded to relatedness. Subsequent maps used computer algorithms to locate clusters of reference papers, journals, journal categories and, in one case, college course prerequisites. Various facets of these 20