Searchers of bibliographical databases are faced with a multitude of choices, but very few clues, in developing precise searchy strategies. this sutdy attemps to discover subject patterns and clusters among databases in order to provide clues about differences and similarities among them. Descriptors were taken at random from sev en subject categories in the ERIC thesaurus and used as search terms on Bibliographic Retrieval. Services' CROSS database. The resulting 1,830,312 postings from 54 databases were analyzed in two ways: (1) an Expectation Ratio was computed which allowed ranking databases in order of their relative responses to the subject searches; (2) a cluster analysis was conducted to discover possible subject relationships among databases. The results show five large cclusters: Technology, Life Sciences, Bibliography, Business and Industry, and Education; and 12 well-defined subclusters. It also shows several dimensions which cut across clusters, such as Research and Social and Economic Enterprise, and Indicates that there are nonobvious similarities and differences among databases which may provide clues to more effective search strategy development.