Dynamic and Evolutionary Updates of Classificatory Schemes in Scientific Journal Structures Loet Leydesdorff Science & Technology Dynamics, Amsterdam School of Communications Research (ASCoR), Kloveniersburgwal 48, 1012 CX Amsterdam. E-mail: loet@leydesdorff.net; www.leydesdorff.net Can the inclusion of new journals in the Science Citation Index be used for the indication of structural change in the database, and how can this change be compared with reorganizations of relations among previously in- cluded journals? Change in the number of journals (n) is distinguished from change in the number of journal cat- egories (m). Although the number of journals can be considered as a given at each moment in time, the num- ber of journal categories is based on a reconstruction that is time-stamped ex post. The reflexive reconstruc- tion is in need of an update when new information be- comes available in a next year. Implications of this shift towards an evolutionary perspective are specified. Introduction Because the sciences develop dynamically, one expects to find change in trend lines of scientometric indicators. For example, scientific productivity changes over time, and it is also expected to differ among research groups. The varia- tion among research groups at each moment in time may interact with the processes of change over time. A policy analyst, therefore, may wish to ask “what do the results teach us?” Should policies nurture the “weak” units or rather “pick the winners” (Irvine & Martin, 1984)? Does a high score on an indicator predicate for further growth or rather predict relative stability or even decline? In other words: what is the strategic value of the measurement results using scientometric indicators? How do the indicated developments relate to a baseline for the comparison? The question of the construction of a baseline for the comparison (Studer & Chubin, 1980) has been prevailing in scientometric studies during the 1980s and 1990s without having been solved hitherto. Two important proposals for methodologies were made right at the beginning of the scientometric research program, notably (a) to make com- parisons at each moment only in terms of “like with like” (Martin & Irvine, 1983), and (b) to make comparisons over time only in terms of journal sets which are kept fixed ex ante during the period under study (Narin, 1976). The heuristics of comparing “like with like” can be considered as a definition of research groups in terms of institutional parameters (Collins, 1985), while the definition in terms of journal sets is expected to indicate the intellec- tual exchange among scholars in a field or specialty (Whit- ley, 1984). For example, an index of activity can be con- structed for the comparison among research groups or other units of analysis (Schubert, Gla¨nzel, & Braun, 1989). The units of analysis of knowledge production can be defined with reference to a relevant environment that one can mea- sure independently, for example, in terms of the journal sets used for the communication (Doreian & Fararo, 1985; Moed, Burger, Frankfort, & Van Raan, 1985; Leydesdorff, 1987). Can the changing positions of institutional units of knowledge production in changing intellectual environ- ments also be measured? Moed et al. (1985) proposed to normalize output performance measurement results in rela- tion to impact factors of journals used by the groups them- selves. In a similar vein, Schubert et al. (1989) developed the instrument of “expected” versus “observed” citation rates. Further questions can be raised here both methodolog- ically and theoretically. For example, the skewness of the distributions considerably complicates the issue of an ap- propriate normalization (Bonitz, 1997; Leydesdorff, 1995a). With other colleagues (e.g., Cozzens & Leydesdorff, 1993; Leydesdorff, Cozzens, & Van den Besselaar, 1994), I have been particularly interested in the measurement of structural change at the network level and how such change potentially redefines the universe (or, in other words, the paradigm) in which practicing scientists assess the rele- vance and the quality of the contributions of their col- leagues. In my opinion, the innovative dimension of the development of science and technology cannot be measured using ex ante fixed journal sets or institutional units; the institutions can be expected to aggregate both standardized routines and innovative activities. Received June 4, 2001; Revised January 3, 2002; accepted May 1, 2002 © 2002 Wiley Periodicals, Inc. ● Published online 7 August 2002 in Wiley InterScience (www.interscience.wiley.com). DOI: 10.1002/asi.10144 JOURNAL OF THE AMERICAN SOCIETY FOR INFORMATION SCIENCE AND TECHNOLOGY, 53(12):987–994, 2002