SOME STATISTICAL ASPECTS OF CO-CITATION CLUSTER ANALYSIS AND A JUDGMENT BY PHYSICISTS.

Although co-citation cluster analysis, a technique based on counting the occurrences of pairs of references in the reference lists of scientific publications, has been used in recent years as a means of measuring science, the statistical instability of the resulting cluster structures seems not to have been properly appreciated. The statistical aspects of co-citation cluster analysis were investigated, using a computer simulation technique. The results illustrate the utility of the method and, by comparison with a co-citation cluster analysis of the Science Citation Index data, reveal some serious problems suggesting that the results of co-citation cluster analyses cannot be taken seriously as evidence relevant to the formulation of resarch policy.