The exchange of online public access catalogs for traditional card catalogs during the past decade has created a population of library users and librarians for whom this new technology provides the primary access to library collections. Most of these would acknowledge the contributions of the online catalog: locating an item when all or part of its title or its author's name is known is generally fast and easy. Topical access to the library's collection, however, is another matter entirely.
While subject searching is the most prevalent form of online catalog searching, it also causes the greatest problems. Studies of online catalog use and users have found that subject searches are the most frequent type of search, yet a surprising number of subject searches - 30 to 50 percent - fail to retrieve any catalog records whatsoever. The causes of search failure appear to be the users' lack of knowledge about Library of Congress Subject Headings (LCSH), misspelling and mis-typing of search terms, and problems in understanding Boolean logic and formulating Boolean queries.
On the other hand, when a subject search succeeds in retrieving catalog records, it frequently retrieves too much material for the user to effectively evaluate, producing information overload. The primary causes of information overload are the size of the catalog database (or the density of information in a given topical area) and the increasing numbers of items indexed using a given keyword or subject heading. The use of keyword searching and truncated headings in exact searching, and the tendency for users to specify broad general topics in searches, compound the information overload problem.
These twin problems of subject searching, search failure and information overload, are well known and have been documented in many studies of online catalogs (for a detailed review of this research, see ).
To combat these subject searching problems, the Cheshire II experimental online catalog system was developed. This paper will describe the Cheshire II system, its architecture, and the methods that are being used to evaluate the system and its use in a working library environment.