There are many characteristics can be used to identify a document which cover characteristics of the documents, cited documents, and citing documents This research explored the inherent structure of a document collection as one of main components of information retrieval system. The characteristics examined are: descriptors, references (cited documents), and citations (citing documents). Three independent variables were studied: co-descriptor, bibliographic coupling, and co-citation. A test collection was constructed by searching on a single descriptor "information retrieval" in the CD-ROM version of Education Resource Information Clearinghouse (ERIC), covering the period 1981 through 1985. Descriptors were extracted from ERIC; cited and citing documents associated with the test collection were derived from Social Sciences Citation Index (SSCI), covering the period 1981 through 1990. Three hypothesis were tested in this study, that are: (1) the higher the frequency of co-descriptors between documents, the higher the frequencies of their bibliographic coupling and co-citation; (2) the higher the frequency of bibliographic coupling between documents, the higher the frequencies of their co-citation and co-descriptors; and (3) the higher the frequency of co-citation between documents, the higher the frequencies of their co-descriptors and bibliographic coupling. The results showed that all of three hypothesis are supported statistically and there is a significant linear relationship among the observed variables. It is mean that there is a significant relationship among descriptors, references, and citation, so that it can be used to construct the inherent structure of document collection in order to improve information retrieval system performance. Zainal A. Hasibuan* dan Mustangimah**; * Fakultas Ilmu Komputer, Universitas Indonesia ** Pusat Pengembangan Teknologi Informasi dan Komputasi, Badan Tenaga Nuklir Nasional Unive
