An experiment in mechanical indexing. The Scientific Citation Index

Abstract

The detailed analysis of the Science Citation Index (SCI) published by the Institute of Scientific Information of Philadelphia since 1963, shows the subject distribution of the periodicals processed in the 1961, 1964 and 1965 voluamst in 1965-66 medical sciences covered 42 per cent, agriculture 7 per cent, social sciences about 5 per cent. Natural sciences represented about 30 per cent, while technical sciences and technology hardly surpassed 15 per cent. Considering the vast field of technical sciences and technology this seems to provide an insufficient basis for proper literature searches. The claim of the SCI to rely successfully on the principle of interdisciplinary relations is exaggerated. The supposition of an "overall interdependency" cannot be regarded as a sound starting point for literature searches. The relative efficiency of the inverted snowball-system, being the basic principle for the preparation of the SCI, is insignificant as compared to the original snowball-system. The search from one author to an other is accompanied by a considerable amount of noise so that only about one third of the references found may be regarded as relevant to the subject. If the Chemical Abstracts as the basis of comparison is supposed to have a retrieval efficiency of 100, then the efficiency of Chemical Titles amounts to only 15, while the efficiency of the SCI in chemistry hardly reaches only 2,5. Due to the noise caused by the retrieval of irrelevant references, the latter percentage would be further reduced to about 0,8. Therefore, the SCI seems to provide by itself an insufficient tool for systematical literature searching. Besides the advantage of not requiring serious intellectual work for indexing, the outstanding advantage of the SCI lies in its new feature useful for authors as they are able to find out easily who have cited their papers in articles published in the perlodicals processed by the SCI. The SCI, thus, can be regarded as a list of the type Who from Whom.