College and Research Libraries Research Notes Library Collection Deterioration: A Study at the University of Illinois at Urbana-Champaign Tina Chrzastowski, David Cobb, Nancy Davis, Jean Geil, and Betsy Kruger A suroey of bound items in the books tacks of the University of Illinois library at Urbana- Champaign was conducted following the meth- odology used in the 1979 suroey of the Green Library stacks at Stanford University. A reli- able random sampling technique was used. The suroey found that 37.0% of the items at Illinois are seriously deteriorated (paper is embrittled), 33.6% are moderately deteriorated (paper is be- coming brittle), and 29.4% are in good condi- tion (paper shows no signs of deterioration). The total cost of the suroey was $1,845.45 (ex- cluding permanent staff salaries). The method- ology can be adapted by other libraries for col- lection condition suroeys. In 1959, publication of the results of W .J. Barrow's research on paper deterioration included the estimation that most twentieth-century printed books may have a shelf life of only fifty years or less. 1 Now, thirty years later, the reality of this sobering prediction is becoming painfully obvious to libraries and other cultural in- stitutions worldwide. Recently a renewed emphasis on preservation has been appar- ent in the library literature, as well as at professional conferences. The collections of all institutions, from the largest research libraries to small mu- seums, historical societies, and public li- braries, are vulnerable to deterioration. The problem of progressive decay crosses all media to affect virtually all types of li- brary material, including photographs, microprint publications, sound record- ings, and computer tapes, among others; but paper deterioration is of particular sig- nificance to any collection condition study, and has been the focus of several surveys in recent years. The results of a study conducted in 1979 in Stanford's Green Library stack collec- tion appeared in 1982, 2 followed by the publication in 1985 of a massive survey of book deterioration at Yale University, 3 and in 1987 the results of a preservation study conducted in 1985 at the Syracuse University libraries. 4 Sample sizes of these three studies were 400, 36,500, and 2,548, respectively; they yielded results indicat- ing levels of embrittlement ranging from 12% (Syracuse) to 37.1% (Yale). Armed with this information and aware Tina Chrzastowski is Chemistry Librarian; David Cobb is Map and Geography Librarian; Nancy Davis is As- sistant Agriculture Librarian; Jean Geil is Special Collections Coordinator, Music Library; and Betsy Kruger is Assistant Acquisitions Librarian and Binding Librarian at the University of Illinois, Urbana, Illinois 61801 . 577 578 College & Research Libraries of growing concerns regarding its own collections, the preservation committee of the University of Illinois Library at Urbana-Champaign recommended in 1987 that a modest survey be performed similar to that previously conducted at Stanford University. As in the Stanford study of its Green Library stacks, the pres- ervation committee at lliinois chose to sur- vey the collections in the books tacks of the university library. With holdings of 5.3 million volumes, this section of the library is the most representative of the total col- lections. Although the bookstacks also house the government documents collec- tion and the Asian library, these special- ized holdings were not included in the study. As in the Stanford survey, un- bound items were excluded from the sta- tistical sampl~. The survey objective was to gain insight on the deterioration of the library collec- tion by providing specific data as to the condition of three distinct elements: pa- per, bindings, and boards and covers. A paper fold test was also performed. While not as comprehensive in scope as the Yale or Syracuse surveys, this survey can be re- produced by other libraries at minimal cost. METHODOLOGY The survey's sample size was derived from a table of sample sizes for selected confidence levels and tolerances pre- sented in M. Carl Drott' s article, "Ran- dom Sampling: A Tool for Library Re- search. " 5 Sample sizes shown in Drott' s table are valid for surveys of over 30 items but less than 10% of the population. The populationofbooksin the Universityofll- linois at Urbana-Champaign bookstacks is estimated at 5.3 million volumes. Using Drott' s table, the sample size was set at 384 items. This would allow a 95% confi- dence level and a 5% tolerance level. Tolerance and confidence are two types of error measures. Drott defines tolerance as "a measure of the accuracy of our result'' and confidence as II a measure of how certain one is that the true answer lies within the limits stated in this tolerance.'' To state at a 95% confidence level that 37% (plus or minus 4%) of books surveyed at September 1989 lliinois are in poor condition means that there is a 1 chance in 20 (5%) that the actual percentage of surveyed books that are in poor condition is greater than 41% or less than 33%. Preparing the Sample The validity of the survey depended on random selection of items to be surveyed. Sets of random numbers for selecting each item by floor, range, column, shelf, and book were computer generated. For each item, 1 random number was assigned for floor, 6 for range, column and book, and 2 for shelf. These multiple options pre- vented a large number of rejects due to disparate shelving situations. Criteria and Grading This study closely paralleled the meth- odology of a similar study ferformed at Stanford University in 1979. Books were evaluated according to three separate cri- teria: condition of paper, binding condi- tion, and the condition o{ the boards and covers. Paper condition was given twice the weight of binding or cover and boards in the final scoring. Paper is weighted by two since it represents the intellectual por- tion of the book and its deterioration is of prime concern to the library. Each book was evaluated for paper dis- coloration, tears, missing pieces, and rough edges. In addition, each item was given a paper fold test, which was used along with the evaluation score to deter- mine the paper condition in the study. The last numbered page was selected to insure that actual text pages were tested rather than fly leaves. A maximum of six folds was used for this survey. The sever- ity of these characteristics determined the score each book was given for paper con- dition. Binding condition was scored by the quality of the stitching at the spine and how well the pages adhered to it. The boards and covers were scored by examin- ing the outer portion of the book and the inner hinges where the boards are at- tached to the text block. The combined grading of these catego- ries was used to determine the overall score for each volume. Detailed grading procedures are found in the Stanford study. 7 WORKSHOP Eleven students from the university's graduate school of library and information science were hired for the study. A work- shop was held to train these surveyors to conduct the study and complete the sur- vey forms. Emphasis was placed on the method for locating the materials to be evaluated using maps and random num- bers, criteria for evaluating a book, and how to apply the grading system and com- pute the overall score. A tour of the library stacks was included; a pretest to grade sample books completed the training. Survey supervisors attended the pretest session to meet the students and observe their training. THE UIUC SURVEY Due to well-trained surveyors and a tested, well-documented methodology, the survey was completed as planned and on schedule. The Survey Form The survey form (Appendix A) was de- signed to collect all the required data in an easy, progressive way. At the head of the survey form were the random numbers used to locate the book for evaluation; a line identifying the call number allowed supervisors to review the surveyors' work. Rejects from the survey were few, num- bering only 22; the survey form required a reason for rejection and these items were later reviewed by supervisors. All rejects were the result of stack areas that did not meet the requirements of the random numbers; that is, there were too few ranges, columns, shelves, or books. The evaluation portion of the survey form required surveyors to simply circle the values given to each category. A final, overall score was the last item to be com- pleted, and concluded the form. Survey Day Survey forms were pre-sorted by floor in order to reduce the distance and time Research Notes 579 between each item. Each surveyor was given 50 survey forms but was instructed to stop once 40 forms had been completed. In addition, each student had full written instructions and floor maps. All questions were to be referred to the supervisor on duty for that floor. Only a few questions were asked of the supervisors; to reduce labor costs for future surveys, we recom- mend only one supervisor be scheduled to handle all surveyor questions. Over 90% of the surveyors finished the required 40 forms the first day (5.5 survey hours). All forms were completed by noon the second day. RESULTS OF THE SURVEY Table 1 summarizes the data obtained in the survey. The data indicate (at a 95% confidence level and a 5% tolerance level) that of the random sample of 384 books in the stacks: 29.4% are in good condition (weighted average = 0); 33.6% are in moderate condition (weighted average = 1); 37.0% are in poor condition (weighted average = 2). As in the Stanford survey, the overall weighted averages at Illinois (0 = 29.4%, 1 = 33.6%, 2 = 37.0%) correspond closely with the paper condition scores (0 = 32.0%, 1 = 31.0%, 2 = 37.0%; see ta- ble 2). This is because, at both institutions, the condition of the paper was given twice the weight of binding or cover and boards in the final scoring. Binding showed ex- tensive deterioration in 4.7% of books sampled, and boards and covers showed extensive deterioration in 8.3% of books sampled. Table 2 compares the results of the sur- vey of the University of Illinois library bookstacks with those of the Stanford sur- vey of their Green Library stacks. Illinois' methodology closely duplicated Stan- ford's, making comparisons generally valid: 29.4% of Illinois' sample and 32.8% of Stanford's were in good condition; 33.6% of Illinois' and 40.8% of Stanford's were in moderate condition; and 37.0% of Illinois' and 26.5% of Stanford's were in poor condition. The survey showed that, at the Univer- Number of titles Percent of titles Weighted average Condition of paper Condition of binding 0 Good 123 32.0% Condition of boards and covers Condition of Paper 1 Moderate 119 31.0% TABLE 1 LEVELS OF DETERIORATION OF TOTAL SAMPLE (N = 384) Condition of Condition of Binding Boards and Covers 2 0 1 2 0 1 2 Poor Good Moderate Poor Good Moderate Poor 142 272 94 18 191 161 32 37.0% 70.8% 24.5% 4.7% 49 .7% 41.9% 8.3% TABLE 2 UNIVERSITY OF ILLINOIS (1988) AND STANFORD (1979): COMPARATIVE LEVELS OF DETERIORATION University of lllinois (n = 384) Stanford University (n = 400) 0 1 2 0 1 2 Good Moderate Poor Good Moderate Poor 29.4% 33.6% 37.0% 32.8% 40.8% 26.5% (113) (129) (142) (131) (163) (106) 32.0% 31.0% 37.0% 33.3% 40.3% 26 .5% (123) (119) (142) (133) (161) (106) 70.8% 24.5% 4.7% 70.8% 25.5% 3.8% (272) (94) (18) (283) (102) (15) 49 .7% 41 .9% 8.3% 56.3% 36.3% 7.5% (191) (161) (32) (225) (145) (30) 0 Good 113 29.4% Weighted Average 1 Moderate 129 33.6% 2 Poor 142 37.0% Percent Change U of I 0 1 2 Good Moderate Poor -3.4 -7.2 +10 .5 -1.3 -9.3 +10 .5 0 -1 +.09 -6.6 +5 .6 +0.8 U1 00 0 sity of Illinois, paper conditions are poor in 37% of the collection, as contrasted with 26.5% at Stanford. Because paper is given twice the weight of the other criteria, the condition of paper has a greater impact on the overall level of deterioration; as a result, the proportion of books with a weighted average score of "poor" is iden- tical to that for poor paper conditions at both Illinois and Stanford. Environmental factors probably account for most of this difference. Central Illinois is subject to high heat and humidity in summer and to frequent temperature fluctuations throughout the year. In contrast, Stanford enjoys a moderate coastal climate with only mild temperature fluctuations throughout the year. The stacks of both li- braries are partially air-conditioned- Stanford in 1980 and Illinois in 1982. Less immediately obvious factors, such as use and age of materials, may also play a role in the higher rate of deterioration at Illi- nois. The impact of the nine-year time span between the two surveys is probably marginal. Weighted averages equalling 2 for bind- ing were 4.7% and 3.8% for Illinois and Stanford respectively, and weighted aver- ages equalling 2 for boards and covers were, respectively, 8.3% and 7.5%. Con- ditions of binding at both institutions were strikingly paralleL The percentage of books with boards and covers in good con- dition was higher at Stanford (56.3% vs. 49.7%). Table 3 shows the distribution of weighted averages by date group. For date groups with fewer than 100 samples (1850-99 and pre-1850) statistically valid conclusions cannot be drawn. Of books published after 1950, 9.4% are in poor con- Research Notes 581 dition. This figure jumps to 67.7% for books published between 1900 and 1949, supporting the Barrow estimate that books published in the twentieth century have only a 35- to 50-year life span. These findings roughly support Stanford's, which found 6.2% of books published be- tween 1950 and 1979 and 44.6% of those published between 1900 and 1949 to be in poor condition. Table 4 displays these comparative findings. In conclusion, the survey results are so- bering: 70.6% of the books sampled evi- denced some degree of deterioration. Moderate deterioration was evident in 33.6% and extensive deterioration in 37.0%. Of books judged to be in poor con- dition, paper condition was the overriding factor. In general, the University of Illi- nois' survey results parallel the findings at Stanford, although paper deterioration is more extensive at Illinois. It is possible to submit these results to standard statistical testing using the chi square formula. 8 Specifically, we tested ta- ble 2 for its summary findings. We defined our null hypothesis as: the state of deterio- ration of the sampled books is indepen- dent of whether they are located at the University of Illinois or Stanford Univer- sity libraries. Applying the chi square for- mula x?- = _E (fofefei and a degree of freedom of 2 ( df = 2) we reached the following results: Weighted average Condition of paper Condition of binding Condition of boards and covers x2 = 10.18 x?- = 11.59 x?- = .49 x2 = 3.35 If the variables are independent we TABLE 3 Date Group 1950- 1900-49 1850-99 Pre-1850 LEVELS OF DETERIORATION OF TOTAL SAMPLE (N = 384) BY DATE GROUP Total 191 141 41 11 0 Good 103 (53.9%) 10 (7.1%) 0 (0.0%) 0 (0.0%) 1 Moderate 70 (36.6%) 44 (31.2%) 10 (24.4%) 5 (45.5%) 2 Poor 18 (9.4%) 87 (67 .7%) 31 (75.6%) 6 (54.5%) 582 College & Research Libraries §' II 0\ 0\ ~ 0\ t-.. ~ ~ 0\ 0\ 00 ~ ~ ~ ~ I I I I J!!S ~ 0 0 0 0 Lt') 0 "' .... 0\ 0\ 00 00 Ol? ~ ..... ~ ~ September 1989 would expect the chi square statistics to be lower than 5. 99, the critical value for chi square with df = 2 and a 95% confidence level. Therefore, we cannot reject the null hypothesis for binding and boards and covers. The higher values for weighted av- erage and condition of paper suggest the null hypothesis to be void. One explana- tion, as we have mentioned, is the envi- ronmental differences which may have led to higher deterioration percentages at the University of lliinois. COSTS OF THE SURVEY Costs for this survey were kept to a min- imum due to administrative constraints and through the use of permanent staff. Most of the preparation for the survey was performed by permanent members of the library's staff or by a graduate assistant as- signed to the preservation committee. The seven permanent faculty members on the committee spent an estimated 191.5 hours on the project, an average of 27.6 hours each. This estimated time includes a sur- vey pre-test, 42.5 hours of survey supervi- sion, report writing, and committee meet- ings. The graduate assistant spent an estimated 100 hours on the project. Eleven graduate library school students (includ- ing one alternate) were hired to perform the survey at a rate of $6.00 per hour; stu- dent and graduate assistant labor costs to- taled $1,573.00. Programming and computer services were donated by a professional member of the library's staff, while computer analy- sis was completed by a free-lance pro- grammer for $150. Miscellaneous supplies were provided by the university library, and copying costs totaled $122.45. The to- tal cost of the survey was $1,845.45, ex- cluding permanent staff salaries. Including the cost of permanent profes- sional salaries (median hourly salary at il- linois is $13. 75) increased the actual cost of the survey by $2,633.13, to $4,478.58. La- bor costs could have been reduced with greater use of students or non-academic staff. Our experience showed that profes- sional time on the survey could have been reduced significantly by having fewer sur- · vey supervisors. For example, profes- sionallabor costs could have been reduced by 18% by using only one supervisor on the day of the survey. CONCLUSION This survey of the condition of materials in the bookstacks of the University of Illi- nois library has yielded data for the pur- pose of making informed decisions for an- nual budgeting and other aspects of preservation planning. The information will also be essential in documenting the case for increased administrative support for library preservation. At the same time, it would be a mistake to assume that this study necessarily mir- rors the condition of the university library collections in their entirety. In the interest of obtaining data on binding, as well as covers and boards, unbound materials were excluded; among the latter are some of the most severely deteriorated items in the bookstacks. Some heavily-used gen- res (such as printed music and maps) do not fall within the scope of the survey. On the other hand, the percentage of older materials is considerable higher in the general bookstacks than is typical in most departmental libraries. It has been demonstrated that useful results can be gained quickly and at mini- mal cost using trained student assistants. As in similar studies elsewhere, there may be small but unavoidable variations in the Research Notes 583 manner in which different surveyors ap- ply the measurement criteria. This risk can be considerably minimized by running a pilot study to bring unforeseen difficulties to light, and by conducting a well-planned instructional workshop to offer surveyors hands-on experience in evaluating vol- umes under the close supervision of proj- ect planners. Having completed this collection sur- vey, the preservation committee has es- tablished baseline data for future studies. These could involve the utilization of a new sample, or rechecking the same titles. The study could also be repeated in the undergraduate library, other departmen- tal libraries, or with categories of special materials excluded from the original sur- vey (such as maps, printed music, or some collections of foreign imprints). Following the lead of Buchanan and Coleman's 1979 study at Stanford, 9 condi- tion data has now been gathered for sev- eral major university library collections. In order for valid statistical comparisons to be drawn, it is recommended that other institutions conduct condition surveys utilizing the methodology reported here. As more precise information becomes available regarding the extent of library collection deterioration, it is hoped that the preservation issue will be duly recog- nized as an impending national crisis. REFERENCES 1. W. J. Barrow, DeteriorationofBook Stock, Causes and Remedies, ed. Randolph W. Church, (Richmond, Va.: The Virginia State Library, 1959), p.15. 2. Sara Buchanan and Sandra Coleman, "Deterioration Survey of the Stanford University Libraries Green Library Stack Collection," in Preseroation Planning Program Resource Notebook, comp. Pamela W. Darling (Washington, D. C.: Association of Research Libraries, Office of Management Studies, 1987), p .159-230. 3. Gay Walker and others, "The Yale Survey: A Large-Scale Study of Book Deterioration in the Yale University Library," College & Research Libraries 46:111-32 (Mar. 1985). 4. Randolph Bond and others, ''Preservation Study at the Syracuse University Libraries,'' College & Research Libraries 48:133-47 (Mar. 1987). 5. M. Carl Drott, "Random Sampling: A Tool for Library Research," College & Research Libraries 30:119-125 (Mar. 1969). 6. Buchanan and Coleman, "Deterioration Survey," p .159-230. 7. Ibid. 8. Ray L. Carpenter and Ellen Storey Vasu, Statistical Methods for Librarians (Chicago: American Li- brary Assn., 1978), p.77-86 . 9. Buchanan and Coleman, "Deterioration Survey," p .159-230. 584 College & Research Libraries September 1989 APPENDIX A: SURVEY FORM (Circle which random numbers were used to locate the book.) REJECT reason: FOLD TEST number of folds: 1 2 3 4 5 6 score: 0 1 2 CALL#: SHORT TITLE: DATE OF PUBLICATION: PLACE OF PUBLICATION: 1 2 3 4 5 6 7 write place of publ. here EVALUATION (Circle one in each category.) A. PAPER 0 1 2 B. BINDINGS 0 1 2 C. BOARDS/COVERS 0 1 2 SCORE 0 1 2 COMPACT Concentrated... Condensed... Compact ... No matter how you say it, .smaller is better. People want more information in a smaller package. And that's what you get with Biological Abstracts® in microform as your life sciences reference tool. Delivering the scope of coverage you've come to expect from BIOSIS®, BA in microform provides you with rapid access and saves a lot of shelf space. In fact, you can store 18 years' worth of BA in microform in the same amount of space taken up by just one year of the printed publication! See for yourself. Contact BIOSIS, Marketing Section, 2100 Arch Street, Dept. CRL989CT, Philadel- phia, PA 19103-1399 USA. Telephone (215) 587-4800 worldwide; toll free 1-800-523-4806 (USA except PA); Telex 831739; Fax (215) 587-2016. BIOSIS is a not-for-profit organization serving the biological community since 1926. Biologics/ Abstracts is a registered trademark of BIOSIS. BIOSIS · BIOSIS is a registered trademark of Biological Abstracts Inc. The world just got smaller. Small enough, in fact, to fit in the palm of your hand! Introducing PAIS ON CD-ROM, the enhanced compilation of PAIS' acclaimed international public/ social policy indexes now available on convenient, durable compact disc. With this powerful, innovative information resource at your fingertips: • You get 24-hour-a-day, on-site access to the most authoritative indexes covering national and international political, economic, and social issues. · • Skilled searchers and novices alike can access nearly 300,000 stringently edited records-indexes to peri- odicals, directories, monographs, government docu- ments, and more-published worldwide in English, French, German, Italian, Portuguese, and Spanish. • You can search this rich, eclectic range of material with greater speed and ease than ever before-and come away with references to hard-to-find statistics, demographics, business trends, background infor- mation, and fresh perspectives. • You can search as long as you like, as often as you like-and actually pay less per retrieval the more you use this remarkable new tool. Find out how the world just got smaller-and your job just got easier-with PAIS ON CD-ROM. Call toll-free today for complete details-and ask about our 30-day money-back guarantee! Public Affairs Information Service, Inc. 621 West 43rd Street • New York, NY 10036-4396 • 800-288-PAIS 212-736-6629 (outside the U.S.) · l11 Prillt: PAIS BULLETIN • PAIS FOREIGN LANGUAGE INDEX • PAIS SUBJECT HEADINGS OnliM: PAIS INTERNATIONAL 011 Compad Diu: PAIS ON CD-ROM MARCIVE GPO CAT/PAC IS SIMPLY BETTER the U.S. Government Printing Office database of government publications since 1976 on CD-ROM BETTER BECAUSE: the RECORDS are better, as the result of an ongoing clean-up project with Texas A & M University, Rice University, and Louisiana State University libraries. BETTER BECAUSE: the ACCESS is better. Keyword access to all Titles/Series, Subjects, Agencies/ Authors. See and see also references. Browse by SuDoc Number, Technical Report Number, Stock Number, Item Number, OCLC Number, Monthly Catalog Entry Number. BETTER BECAUSE: the VALUE is better. Introductory annual subscription price: $995. P .0. Box 47508 San Antonio, TX 78265 1-800-531-7678