Origin of the Serials Evaluator
As the SRP was being launched, a special project was undertaken within LSU Libraries to analyze the structure of the library market for ST serials. Faculty survey data obtained in the 1993 pilot project with the Department of Chemistry was used to determine whether cancellations could be made without seriously damaging the ST serials holdings at LSU Libraries. The results of the study have been published (Bensman 1996), and the main conclusion was that scientific value as measured by faculty ratings and total citations played no role in the pricing of scientific serials. Not only is this the case, the study also revealed that the library market for ST serials appeared to be severely bifurcated, with scientific value concentrating on the titles of the U.S. associations and costs concentrating on the serials of the commercial—largely foreign—publishers.
Because of this fact, major cancellations could be implemented without seriously affecting the ST value of the LSU Libraries' serials holdings. Moreover, during the course of the study, it became obvious that a computer program could be developed to take advantage of this situation. This software, ultimately called the Serials Evaluator, was created with the assistance of a programmer in the university's Administrative Information Systems. However, before one can understand how the Evaluator works, one must have an understanding of three principles that affect the structure of the ST journal system: (1) set definition, (2) skewed distributions, and (3) the social bases of ST value.
Set Definition and Its Statistical Consequences
Need for Set Definition
Any database in library and information science contains a witches' brew of variables. This is a result of the complex actions of social groups differing in size, subject interest, level of understanding and agreement, time framework, as well as purpose and intention. Due to these reasons, before library and information science data are analyzed, they should be broken down into well-defined sets. Otherwise, complex interactions among the variables will negate statistical relationships and lead to mistaken conclusions.
The usual way of sorting library and information science data into sets is to define the sets by subject matter. This is particularly important in serials analysis, where prices, citation rates, etc., differ vastly from subject group to subject group. For example, if one decides to cancel subscriptions by price alone, one runs the risk of canceling good science journals while leaving bad social science and humanities journals outside the range of analysis. However, set definition in library and information science is complicated by the interaction of processes that are best described by two eponymic bibliometric laws.
Bibliometric Laws That Affect Set Definition in Library and Information Science
The first of these is "Bradford's law of scattering" formulated by Bradford while he served as chief librarian from 1925 to 1938 at the National Science Library in South Kensington, England. In the formulation of his law, Bradford (1953, 148–59) started from the principle that "every scientific subject is related, more or less remotely, to every other scientific subject" and that therefore "the articles of an interest to a specialist must occur not only in the periodicals specialising on his subject, but also, from time to time, in other periodicals." He applied this principle in an analysis of two specific subjects, applied geophysics and lubrication, and the results of this study led him to state his famous law on the scattering of articles on a given subject among scientific journals in the following manner (p. 154):
Bensman (1982, 286–87) further analyzed Bradford's data to reveal (1) that in the applied geophysics set, 9.2% of journals accounted for 51.7% of the articles on that subject with the other 48.3% of these articles spread out over journals of other disciplines, and (2) that in the lubrication set, the same 9.2% of the journals accounted for 40.8% of the articles on this subject with the remaining 59.2% spread out over the journals of other disciplines.
The second bibliometric law that complicates set definition in library and information science is "Garfield's law of concentration." This law was formulated by Garfield, founder of the Institute of Scientific Information (ISI), which publishes the Science Citation Index (SCI), Social Sciences Citation Index (SSCI), and Arts & Humanities Citation Index (A&HCI). Along with its citation indexes, ISI produces two annual publications, the SCI Journal Citation Reports (SCI JCR) and the SSCI Journal Citation Reports (SSCI JCR), which give various citation measures for the serials covered by their respective indexes. A series of exploratory studies was conducted at ISI with a prototype of the SCI JCR containing citation data from one quarter of 1969, and these studies revealed that a multidisciplinary mix of 152 journals accounted for 50% of the citations processed for the SCI in 1969.
This finding caused Garfield to formulate his law of concentration, which he derived from Bradford's law of scattering by transposing the latter law from the level of a single discipline to that of science as a whole. Garfield devised a physical analogy to Bradford's law, one based upon a comet. In this analogy, the nucleus of the comet represents the core journals of a discipline's literature with the debris and gas molecules of the tail representing the additional journals that sometimes publish material relevant to the discipline. With this analogy in mind, Garfield described his law of concentration and its practical implications in the following manner (1979, 160):
[The bibliographic law of concentration] goes an important step beyond the Bradford law by stating that the tail of the literature of any one discipline consists, in large part, of the cores of the literature of all other disciplines, and that all the disciplines combined produce a multidisciplinary literature core for all of science that consists of no more than 1,000 journals. In fact, this multidisciplinary core might be as small as 500 journals. Though larger collections certainly can be justified in many cases, the single function of providing reasonably cost-effective coverage of the literature most used by research scientists requires no more than 500 to 1,000 journals.
The findings of the study with the 1969 SCI JCR data were replicated with an analysis of 1974 SCI JCR data (Garfield 1979, 21–23, 158–61).