A longitudinal study of three discrete online public access catalog (OPAC) M ost previous studies of online public access catalog (OPAC) use and design have centered on transactionlog analysis and user survey results in the academic library environment. Measures of patron success or lack thereof have traditionally been expressed in the form of such concepts as "zerohit" analysis or the "branching" analysis of Kantor and, later, Ciliberti. 1 Missing from the majority of the literature on OPAC study, however, are the effects that use and design have had on public library patron borrowing practices.Major drawbacks to transactionlog analyses and user surveys as a measure of successful OPAC use include a lack of standardization and the inherent difficulties in interpreting resulting data. As Peters notes, "[s]urveys measure users' opinions about online catalogs and their perceptions of their successes or failures when using them, while transaction logs simply record the searches conducted by users. Surveys," he concludes, "mea sure attitudes, while transaction logs measure a specific form of behavior." 2 In both cases it is difficult, in many instances, to draw clear conclusions from either method.Circulation figures, on the other hand, measure a more narrowly defined level of patron success. Circulation is a discrete output that is the direct result of patrons' initiated interaction with one or many library collections, one or many levels of library technology. With the recent advent of such enhanced OPAC functionality as patronplaced holds on items from broader and broader catalogs, online catalogs now more than ever not only serve as search mechanisms but also as ways for patrons to directly obtain materials from multiple sources. It follows that an investigation of the possible effects such enhancements may have on general circulation trends is warranted.
■ Literature reviewDuring the midtolate 1980s, transactionlog analysis was introduced as an inexpensive and easy method of looking at OPAC use in primarily the academic library environment. Peters's transactionlog survey of more than thirteen thousand searches executed over a five month period at the University of MissouriKansas City remains particularly instructive today for its large sample and transferable design as well as its interpreta tion of results.
3Here analysis was broken into two phases. In phase one, usage patterns by search type and failure rates as measured by zero hits were examined as dependent vari ables with search type as the independent variable in a comparison study. Phase two took this one step further in the assigning of what Peters termed "probable cause" of zero hits. These probable causes fell into patterns that, in turn, resulted in the identification of fourteen discernable error types that included such things as typographical errors and searches for items not in the catalog. Once again, search type formed the independent variable while error type shaped the dependent variable in a simple study of error types as a percentage...