College and Research Libraries JOE A. HEWITT Sample Audit of Cards from a University Library Catalog This paper reviews results of studying a sample of cards (an audit) from a catalog, indicates uses, and suggests perspectives for other audit operations. ANY LARGE CARD CATALOG will contain a certain number of errors, conflicts, mutilated cards, and examples of obso- lete and misleading cataloging practices designed for the catalog when it was much smaller. The perfect catalog does not exist, and the nearly perfect catalog is very often counterbalanced by hoards of "problem" catalog cards stashed away in the catalog department. Al- though the state of maintenance of their catalogs is a universal concern of catalog librarians, this concern is not al- ways based on reliable data on the ac- tual condition of the catalog. One means of acquiring this information is to audit a sample of cards from the cat- alog for errors and inadequacies. In- formation provided by such an audit can be used for setting task priorities and manpower requirements for retro- active maintenance and, when per- formed periodically, for measuring the effectiveness of programs of current maintenance. The author-title section of the main public catalog at the University of Colo- rado Libraries has been audited, primar- At the time this paper was written, Mr. Hewitt was head, Catalog Maintenance De- partment, University of Colorado Libraries, Boulder. He is presently coordinator, Colo- rado Academic Libraries Book Processing Center (CALBPC), Boulder. 24/ ily for the purpose of setting priorities for projects of rehabilitation. Based on the information provided by the audit, the following projects were given top priority: (I) Updating of cross-refer- ence structure of the catalog; ( 2) sys- tematic search and replacement of muti- lated cards; and ( 3) replacement of the more misleading short cards, especially added entries in the form of see refer- ences. The audit allows us to concen- trate our limited resources for mainte- nance on the most pressing projects, without the nagging suspicion that there are other projects of more critical im- mediacy. It is noteworthy that the weak- ness of the cross-reference structure of the catalog was not suspected before the audit. The sample audit, though sub- stantially lacking in several ways, has proven well worth the time and trouble spent in conducting it. A search of the literature for reports of similar audits of university library catalogs produced only meager results. Two possible explanations for this lack of literature come to mind: (I) The sample catalog audit is a neglected or undiscovered tool in research libraries; or ( 2) the results of audits are consid- ered too localized in interest-or too em- barrassing-to be reported in the litera- ture. In either case, the unfortunate re- sult is that the research library perform- ing an audit does so in a vacuum. AI- though the primary usefulness of an audit is certainly not to compare cata- logs with other libraries, auditing with- out comparative data can be an acute disadvantage when reporting audit re- sults to the staff. Without comparative statistics from other libraries the staff may compare error and card deficiency rates with the only yardstick available- absolute perfection. The result may be that the catalog department may plunge, or be pushed, into a costly and unneces- sary project of editing and refiling the catalog. This report attempts to break the vac- uum's seal, regardless of its cause. This paper will demonstrate how audit in- formation has been useful in one uni- versity library, and it may lend perspec- tive to audits being performed else- where. ., The divided catalog at the University of Colorado's Nor lin Library consists of .approximately 1,000,000 cards in the author-title section and 600,000 in the subject section. Only the author-title section of the catalog was audited. (The audit required forty-four hours of professional time, not including the time required to write a final report and recommendations.) The sample totaled 2,500 cards, consisting of the first 100 cards in a random sample of twenty-five trays. The confidence level for the sam- ple is 95 percent, with precision inter- vals varying according to the propor- tions of each attribute tabulated. Ap- proximate precision levels were deter- mined by the use of a sampling table, and all were considered adequate for the purposes of the study.1 For example, the precision level for the filing error rate is approximately .005. With a filing error rate of 1.1 percent in the sample, we can be assured that in ninety-five times out of a hundred the actual filing error rate in the catalog will fall be- tween .6 percent and 1.6 percent. The composition of the sample by type of entry was as follows: Main en- Sample Audit of Cards I 25 tries, including primary entries for an- alytical sets-42.8 percent; title added entries-24.2 percent; added author, edi- tor, and translator entries-13 percent; added series entries-7.4 percent; con- tinuation cards (second cards to a given entry) -4.4 percent; references-5.3' percent; information cards-1 percent; and guide cards-1.1 percent. The rate of filing error in the sample was 1.1 percent. Errors resulting from a lack of knowledge of the filing rules made up 52 percent of the errors in the sample. The remaining 48 percent were mechanical or simple alphabetical er- rors. The filing of an added entry in a separate sequence after the same head- ing used as a main entry was the most frequent error involving violation of a filing rule. Since this reflects an earlier filing practice, it may be assumed that a number of these had not been refiled when the library's present rule was adopted. Other frequent rule errors were subfiling by main entry rather than - title, mistaking a subdivision of a cor- porate body for a title, and the filing of complete works in foreign languages al- phabetically rather than before the se- quence for the author's single works. In most cases errors involving violation of a filing rule did not result in loss of ac- cess through the misfiled entry, since the card was usually somewhere within the sequence for the heading. Alphabetical errors tended to misplace cards much more drastically. Initially, some staff members were alarmed at our filing error rate. After all, a rate of 1.1 percent means approxi- mately 11,600 filing errors in the author- title catalog alone. The staff member who found some mystical scale to con- vert this statistic into a patron incon- venience factor was truly staggered. For- tunately, figures were available from the 1953 .audit of the official catalog at the Library of Congress, which reported a filing error rate "in excess of 5 per- cent."2 This served to temper consider- 26 I College & Research Libraries • January 1972 ably the reaction to our own error rate. Of the cross-references in the sample 10.14 percent were blind, resulting in an estimate of 5,600 blind references in the catalog. The weakness of the cross- reference structure in the author-title section of the catalog was completely unexpected, and this statistic instigated a search for an offending procedure. The main cause has been identified as a gap in the cancellation process. In cancelling the last card under a heading, it had not been the practice to check systematically the authority file to see if references were made to that entry. The audit demonstrated that such a lapse, over a period of time, can have a con- siderable effect on the catalog. Several of the blind references had not been transferred into the subject section when the catalog was divided. Of the cards in the sample 1.48 per- cent were judged to be mutilated. Since mutilation was strictly defined to in- clude only cards with the call number or other essential information torn off or completely obscured by dirt and wear, this statistic may only be used to estimate the number of cards needing immediate replacement. There were a number of cards in the sample in vari- ous stages of deterioration which were not counted as mutilated. For that rea- son the study was of little help in esti- mating manpower requirements for re- typing over the next several years. The lack of more intensive analysis of card deterioration is a major deficiency of the study, and in future audits estimates will be made of the length of useful- ness of each card in the sample. Other card deficiencies recorded were obsolete cards (page analytics for peri- odicals, analytics to pamphlet collec- tion) -.8 per cent; short cards for added entries-7.3 percent of added entries; abbreviated cataloging (adding new edi- tions as added copies )-1.1 percent. Several card characteristics were re- corded for purposes other than deter- mining the state of maintenance of the catalog. The incidence of tracings on the verso rather than the face of the main entry was recorded for informa- tion useful in considering a proposal to reproduce our present catalog in book fonn by photographic means. Thirty- nine percent of the main entries had complete or partial tracings on the verso of the card, pointing out the consider- able cost of retaining a visible record of tracings in a book catalog. Two sta- tistics that will be useful in considering the question of reclassi:fication are that 35 percent of the cards in the sample for titles classed in Dewey were on LC printed cards with usable class numbers, and that. 90 percent of card deficiencies (not including filing errors ) were on Dewey cards. As stated earlier, the primary use of the audit has been to set task priorities for maintenance projects. However, a fairly reliable estimate of the rate of filing error in the catalog may well tum out to be the most valuable statistic pro- vided by the audit. Like many libraries, we are presently seeking relief from the considerable manpower drain of com- plete filing revision (student filers, cleri- cal revisers). Several alternatives to com- plete revision are being considered. Among these are nonrevision, sample re- vision, and partial or selective revision. If either of these alternatives is adopt- ed, our present filing error rate will be a base figure by which to measure the effect of the new method on the filing in the catalog. The auditing technique becomes more advanced in its application to the problem of filing error, from which a system of error control based on the pe- riodic sample audit and an established level of tolerable filing error may evolve. The difficult problem here will not be perfecting the auditing process, but rather establishing a realistic level of acceptable filing error. It is here that comparative statistics would be most useful in making this decision much less arbitrary. REFERENCES 1. Gene Brown and Lawrence L. Vance, Sam- pling Tables for Estimating Error Rates or Sample Audit of Cards I 21 Other Proportions (Berkeley: Institute of Business and Economic Research, University of California, 1961 ) . Table II, p.52. 2. Cited in Andrew D. Osborn and Susan M. Haskins, "Catalog Maintenance," Library Trends 2:280 (Oct. 1953).