College and Research Libraries Research Notes Time Required for Shelf Reading-A Case Study James H. Sweetland While the subject is of importance to library ad- ministrators planning budgets, there is little information available in the literature on the time needed to maintain proper shelf order. A volunteer staff's recent experience with reading and reshelving an academic library's Z classifi- cation suggests such work can be done at a rate of 554 to 613 volumes per hour in a collection that includes serials. The value of the com- monly accepted statement that a library is effec- tively ''full'' at 86 percent of capacity is also verified. For academic libraries in the United States, the open stack library is the norm. 1 At the same time, there are very real costs associated with the open stack arrange- ment. 2 Perhaps most frustrating, how- ever, is the tendency of material to become disorganized-either by accident or inten- tion. The obvious solution to the misshelving problem is a regular program of shelf read- ing. Unfortunately, this is a very difficult program to implement-the general im- pression of potential workers (usually stu- dents) is that it is boring work, and it is. Of course, it is never-ending work as well. But perhaps the most important impedi- ment to regular shelf reading is the poten- tial cost. At least one library estimated that this task required about 10 percent of its student budget. 3 There is very little information on the ex- act costs. Bookstein, for example, pro- vides a useful formula for allocating a given amount of time among parts of the collection but other than assuming one person could read 600 volumes per hour, does not suggest what should be a reason- able amount of time. The literature is virtually silent on the is- sue, to the point that shelf-reading costs are not usually included in discussions of alternative shelf arrangements. Discus- sion of the advantages of open versus closed stacks tends to focus on the nature of the disadvantages of browsing5 or com- pare varieties of compact shelving with the traditional open stacks but to leave shelf reading out of the equations, includ- ing only comparative times for reshelving and paging. 6 Since formal inventories involve a form of shelf reading, however, some studies related to these do provide information. At Houston Public Library, using a 2-member team to check the shelves and James H. Sweetland is Assistant Professor at the School of Library and Information Science, University of Wis- consin, Milwaukee, Wisconsin 53201. 75 76 College & Research Libraries mark found items on the shelflist card, the team could handle an average of about 86 books per hour, or 43 books per person- hour.7 In a similar study at the University of Texas Health Sciences Library, the fig- ure was 40.6 books per person-hour using a two-member team. 8 Other libraries kept track of the shelflist cards rather than the volumes on the shelf. The University of Kansas, for exam- ple, reports that 1,300 cards can be checked against the shelf in 6 person- hours, for a rate of 217 cards per person- hour. 9 Results from the University of Michigan yield a rate of 300 cards on a first pass and 222 cards per person-hour on a second search (for items not found in the first). 10 Since one might assume at least 1 volume per shelflist card, the variation be- tween these and the first two reports is surprising. It is worth noting however, that Kansas did not check serials holdings and that Michigan participants felt their sample was deficient in serials. Searching for specific items is a task re- lated to straightening the shelves. Brigham Young University's Lee Library, searching for a group of 384lost books, re- ported a rate of 60 books per hour for the general stacks plus another hour for searching the Reserve Room for the re- maining 27 books, or an overall rate of 55 books per hour. 11 Purdue reports a full search of the entire library took 3 staff 3 weeks, a rate of 10 books per person- hour.12 Normally, however, shelf reading is not an inventory and is intended to prevent the obvious costs of searching. The only formal study of this process was per- formed at Princeton by a private consult- ing firm. Its final standard, based on a number of experiments, is "20 students working with three supervisors and one project coordinator for five hours per day for three days can read 1,500 standard sec- tions [7 three-foot shelves with 180-200 volumes per section]." This gives a range of 750 to 833 volumes per person-hour for the 24-person team. 13 Unfortunately, while the data on time are very explicit, it is apparent that the authors merely ac- cepted the usual library figures for the number of volumes shelved in a standard January 1988 section. And, when the teams found a misshelved item, they reshelved it only if it was close to its proper location, collect- ing the other items and leaving them for later work. Thus, the figure still does not give the full cost of what most library man- agers consider shelf reading. Recent experience at a larger academic library may help to provide a benchmark for the time involved in such work. In response to a perceived problem in the Library of Congress Z class (collection shifts resulting from construction of a ma- jor addition forced the library to reduce normal shelf reading), the Student Associ- ation of the School of Library and Informa- tion Science at the University of Wiscon- sin-Milwaukee volunteered to assist. After appropriate clearance from the li- brary administration, student volunteers spent 2 days reading the entire Z collec- tion (with the exception of seven sections at the end of the sequence that were inac- cessible due to construction). Given the lack of information on this type of work, each volunteer noted the beginning and end of time periods devoted to the project: it took 68.75 hours to complete the 235 sec- tions. In order to obtain more accurate data, a good estimate for the size of the collection was required. This was obtained by sam- pling every fifth shelf in the 235 sections. Measurements were obtained by com- pressing all volumes on the shelf to the left, measuring the material to the nearest inch, and then actually counting all vol- umes on the shelf (loose issues in Prince- ton files of similar boxes were also counted as a volume per box). The results give number of Z volumes in the classification, as well as capacity of the shelves and per- centage of capacity, within a 5 percent er- ror.14 At the time of the sample, the Z section was just under 85 percent full, giving an average shelf (35112 inches long) an expan- sion space of 5.3 inches. The volume count showed an average of .89 books per inch, or 10.7 volumes per foot in this collection, which includes bound journals and pam- p~lets. The range of items on a full shelf was between 9 and 70 volumes. Actual count of the 235 sections yielded 1,497 shelves, or an average of 6.37 shelves per standard (90 inches high) section. Apply- ing these results to the total shelf area gives an estimated capacity of 47,388 vol- umes, with 40,259 actual volumes present (±5 percent). This information is interesting in view of the usual assumption of 7 shelves per sec- tion. This classification actually has a ca- pacity of about 202 volumes per section. However, using the actual volumes per foot (10.7) times the nonnal assumption of 7 shelves per section, the theoretical ca- pacity would be 222 per section. Thus, while the area can only hold about 47,400 volumes in this classification, normal planning assumptions would put its ca- pacity at about 52,200. In any event, the project took about 69 hours. Thus, the shelf reading of the en- tire classification with highly motivated volunteers, many with some work experi- ence in libraries and all familiar with the Library of Congress classification, was done at an average rate of 583 volumes per hour (or between 554 and 613 volumes per hour). The ''reading'' included reshelving all Z-classification material in its proper place (including placement of all volumes of pe- riodicals in correct sequence); putting ma- terial outside Z on a nearby carrel for staff Research Notes 77 handling; and simple weeding (third and higher copies of all titles were placed on a nearby carrel for later deaccessioning con- sideration). The project provided an example of the validity of the ''common wisdom'' that a library is full at 86 percent of capacity, 15 since some reshelving required consider- able shifting of other material. The worst case was the replacement of 1 volume of a bound journal Uournal of Australian Librari- anship) on a full shelf. This 1 volume re- quired shifting 47 other shelves in order to gain space. While it is true that these shelves were all nearly full because they were bound periodicals, it is also true that the shifting was much faster and easier as a result. Thus, the current study suggests that Bookstein' s "about 600 per hour" is not far off as an estimate for shelf reading. A range of costs between regular shelf read- ing on the one hand and searching for items users can't find on the other should be considered against the costs of various closed-stack arrangements in future stud- ies of the value of shelf access. Ideally, other libraries will also make an effort to obtain similar data in their own shelf- reading projects, so that a de facto stan- dard will in time emerge. REFERENCES AND NOTES 1. Rutherford D. Rogers and David C. Weber, University Library Administration (New York: Wilson, 1971), p .238; Guy R. Lyle, The Administration of the College Library, 4th ed. (New York: Wilson, 1974), p.125 . 2. See Harold B. Shill, "Open Stacks and Library Performance," College & Research Libraries 41:220-26 (May 1980); Mathilde V. Rovelstad, "Open Shelves/Closed Shelves in Research Li- braries," College & Research Libraries 37:457-67 (Sept. 1976); F.W. Ratcliffe, "Problems of Open Access in Large Academic Libraries," Libri 18:95- 111 (1968). 3. William J. Hubbard, "Sources of Shelving Workload as a Cost Factor in Maintaining Open Stacks," Serials Librarian 8:75-82 (Winter 1983). 4. Abraham Bookstein, "Models for Shelf Reading," Library Quarterly 43:126-37 (April1973) . 5. See Richard J. Hyman, "Shelf Classification Research: Past, Present-Future?" Univ. of illinois Graduate School of Library Science, Occasional Papers no.146 (November 1980) for a useful sum- mary. 6. For example, Ralph E. Ellsworth, The Economics of Book Storage in College and University Libraries (Metuchen, N.J.: Assn. of Research Libraries and Scarecrow, 1969). 7. Jay B. Clark, "An Approach to Collection Inventory," College & Research Libraries 35:350-53 (Sept. 1974). 8. Virginia M . Bowden, "Inventory of a Monograph Collection,'' Bulletin of the Medical Library Associ- ation 65:445-46 (Oct. 1977). 78 College & Research Libraries January 1988 9. Clifford H. Haka and Nancy Ursery, "Inventory Costs: A Case Study," College & Research Libraries 46:169-72 (Mar. 1985). The authors indicate that they found "thousands of misshelved materials" but do not give details, "Letters," College & Research Libraries 47:83-84 (Jan. 1986). 10. R. E. Beck and J. R. McKinnon, "Development of Methods and Time Standards for a Large-Scale Library Inventory," in Case Studies in Systems Analysis in a University Library, ed. Barton R. Burkhal- ter (Metuchen, N.J.: Scarecrow, 1968), p.48-75. 11. Glenn R. Lowry, "A Heuristic Collection Loss Rate Determination Methodology: An Alternative to Shelf-Reading," Collection Management 4:73-83 (Spring/Summer 1982). 12. Barbara P. Pinzelik, Monitoring Book Losses in a Large Academic Library: Four Methods (Lafayette, Ind.: Purdue Univ. Libraries, 1979), p.ll-16 (ED 203 852). 13. Margaret Johnson Bennett, David T. Buxton, and Ella Caprioti, "Shelf Reading in a Large Open Stack Library," Journal of Academic Librarianship 5:4-8 (Mar. 1979). 14. A conservative sample from the assumed 1,645 shelves (7 x 235) to give a confidence interval of 5% at the .05level of significance is about 314. Herbert Arkin, Handbook of Sampling for Auditing and Accounting (New York: McGraw-Hill, 1963}, V.1, p.370. One of 5 shelves is 329. In the event, with 1,497 shelves (counted later}, the actual sample is quite larger than required. 15. Ralph M. Daehn, "The Measurement and Projection of Shelf Space," Collection Management 4:25-39 (Winter 1982); Keyes D. Metcalf, Philip D. Leighton, and David C. Weber, Planning Aca- demic and Research Library Buildings (Chicago: American Library Assn., 1986), p.155-56. Dissertations-An Online Dilemma Donald K. Hartman and Manuel D. Lopez There are few bibliographic aids available to the online searcher who has questions concerning dissertations and theses and the extent of their coverage by the various databases. A compari- son of two databases concerned with this "for- mat" material is not completely reassuring, while the survey of the individual databases cited did provide useful information for the on- line searcher and contributes to reducing the uncertainty of the situation. THE PROBLEM That uncomfortable feeling of uncer- tainty is back. As an academic librarian, information broker, or online searcher in a corporate structure you have just com- pleted a computerized search. The cita- tions are relevant but include several ref- erences to dissertations and/or theses. Does t~at mean you don't have to search Dissertation Abstracts Online? Questions nag. When did the database just searched start to include dissertations/theses? Were citations to dissertations added retrospec- tively? What about foreign dissertations? Are they included? If so, which countries? Just what percentage of the database is dissertations? The directory Computer- Readable Databases does indicate percent- ages but only in combinations of formats. What were the criteria used for selection of dissertations? Who assigned the subject headings/descriptors? The author, data- base personnel, or others? Was a thesau- rus used? If so, which one? THE SEARCH FOR ANSWERS A search by document type ( dt =theses? or dt=dissertation? or dt=doctoral?) of Donald K. Hartman and Manuel D. Lopez are reference librarians at the State University of New York, Buffalo, New York 14260.