Groups of Highly Cited Publications: Stability in Content with Citation Window Length

Introduction The growing focus in research policy worldwide on top scientists makes it increasingly important to define adequate supporting measures to help identify excellent scientists. Highly cited publications have since long been associated to research excellence. At the same time, the analysis of the high-end of citation distributions still is a challenging topic in evaluative bibliometrics. Evaluations typically require indicators that generate sufficiently stable results when applied to recent publication records of limited size. Highly cited publications have been identified using two techniques in particular: pre-set percentiles, and the parameter free Characteristic Scores and Scales (CSS) (Glänzel & Schubert, 1988). The stability required in assessments of relatively small publication records, concerns size as well as content of groups of highly cited publications. Influencing factors include domain delineation and citation window length. Stability in size is evident for the pre-set percentiles, and has been demonstrated for the CSS-methodology beyond an initial citation period of about three years (Glänzel, 2007). Stability in content is less straightforward, considering for instance that more highly cited publications can have a later citation peak, as observed by Abt (1981) for astronomical papers. This paper investigates the stability in content of groups of highly cited publications, i.e. the extent to which individual publications enter and leave the group as the citation window is enlarged.