Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. Using GIS to Measure In-Library Book-Use Behavior Xia, Jingfeng Information Technology and Libraries; Dec 2004; 23, 4; ProQuest pg. 184 that development rests with the local site. For the NWDA consortium, this development , using the code base, ha s been manageable. The current stat e of interface dev elop ment for the NWDA project can be reviewed at http :// nwda. wsulibs .wsu.edu / project_info /. Conclusion In se lecting an EAD searc h- and- retrieval system, one important qu es tion for th e consortium was, Which software solution had the best prosp ects for migration in the futur e? Because of the inherent strength s of nativ e XML technology in comp ari - son to the other product categories list ed in tabl e 1, a nati ve XML data- bas e appeared to be the best appro ach, and TextML provided the best combination of licensi ng costs, software capabilities, and support. It is important to note that the dis- tinctions betw een nativ e XML data- bas es and databases that support XML throu gh extensions (XML- enabled databa ses) ma y b eco me more difficult to dis cern over time, in part du e to the existi ng exper tise and in vestme nts in RDBMS technolo- gies. 16 Nevertheless, capabilities cen- tral to native XML, such as the us e of an XML-based query language, are integral to th e success of such hybrid syst ems . References and Notes 1. Daniel Pitti , "Encoded Archival De scriptio n: The Development of a n Encoding Standard for Archival Finding Aids ," The American Archivist 60, no. 3 (Summ er 1997): 269. 2. Daniel Pitti, "Encod ed Archival Des cription: An Introducti on and Over- vi ew," 0-Lib Magazine 5, no. 11 (Nov. 1999). Accessed Nov. 2, 2004, www.dlib. org / dlib / november99 / 11 pitti.html. 3. Daniel V. Pitti and Wendy M. Duff (ed s.), "Introduction," in Encoded Archival Description on the Internet (Binghamton, N.Y.: Haworth, 2001), 3. 4. James M . Roth, "Serv ing Up EAD: An Exp lorat o ry Study on the Deploy- ment and Utilization of Encod ed Arch- ival Description Finding Aids," The American Archivist 64, no. 2 (Fall /Win ter 2001): 226. 5. Sarah L. Shreeves et al., "Har ves t- ing Cultural Heritage Metadata Using the OAI Protocol," Library Hi Tech 21, no. 2 (2003): 161. 6. Nan cy Fleck and Michael Sead le, "EAD Harvesting for the National Ga llery of th e Spoken Word" (pap er pre- sent ed at th e Coa liti on for Netw orke d Information fall 2002 Task Force meeting, San Anton io, Tex., Dec. 2002). Accessed Nov. 2, 2004, www.cni .org/tfms/2002b. fall/handouts/H-EAD-FleckSeadle.doc. 7. Anne J. Gilliland -Swe tland , "Po pu- larizi ng th e Finding Aid : Exploiting EAD to Enhance Online Discover y and Retrie- val," in Encoded Archival Description on the Internet (Bing h a mton , N.Y.: Haworth, 2001), 207. 8. Ibid, 210-14. 9. Charlott e B. Brown and Brian E. C. Schottlaender, "The Onlin e Arch ive of Cal- ifornia: A Consortia! Approach to Encode d Archival Descrip tion, " in Encoded Archival Description on the Internet (Bingham ton, N .Y.: Haworth, 2001), 99. 10. Ibid , 103-5. OAC ava ilable at: www . o ac.c dlib.org / . Accessed Nov. 2, 2004. 11. Christopher J. Prom and Thomas Habing, "Using the Open Archiv es Initia- tive Protocols w ith EAD," in Proceed ings of th e Second ACM/IEEE-CS Joint Confe r- ence on Digit al Librari es (Portland, Ore., July 2002). Accessed Nov . 2, 2004, http:// dli .grai ng er.uiu c.edu / publications / jcdl20 02/ p14prom.pdf. 12. Marc Cyrenne, "Go ing N at ive: Wh en Should You Use a Native XML Database?" AIM E-DOC Magazine 16, no . 6 (Nov./Dec. 2002), 16. Accessed Nov. 2, 2004, www. edo cmaga zine.com/ article_ n ew.as p?ID=2 5421. 13. Product categor y decisions ba sed upon definiti ons and classifications avail- able from: Ronald Bourret, "XML Database Products." Accessed Nov. 2, 2004, www. rpbourret .com/ x ml / XMLDa ta base Prods .htm. 14. Cyrenn e, "Going Native," 18. 15. Bill Stockting, "EAD in A2A," Microsoft Power Point pres entation. Accessed N ov. 2, 2004, www.agad.a rchiwa . gov.pl/ ead /s tocking.pp t. 16. Uwe Hohenstein, "Supp orting XML in Oracle9i ," in Akm a l B. Chaudhri , 184 INFORMATION TECHNOLOGY AND LIBRARIES I DECEMBER 2004 Awais Rashid, and Rob erto Zic ar i (eds.), XML Data Management: Native XML and XML-Enabled Database Systems (Boston: Addison-Wesley, 2003), 123-4. Using GIS to Measure In-Library Book-Use Behavior Jingfeng Xia This article is an attempt to develop Geographic Information Systems (GIS) technology into an analytical tool for exam- ining the relationships between the height of the bookshelves and the behavior of library readers in utilizing books within a library. The tool would contain a database to store book-use information and some GIS maps to represent bookshelves. Upon ana- lyzing the data stored in the database, dif- feren t frequencies of book use across bookshelf layers are displayed on the maps. The tool would provide a wonderful means of visualization through which analysts can quickly realize the spatial distribution of books used in a library. This article reveals that readers tend to pull books out of the bookshelf layers that are easily reachable by human eyes and hands, and thus opens some issues for librarians to reconsider the management of library collections. Several years ago, when working as a library assistant reshelving books in a university librar y, the author noted that the majority of books used inside the library were from the mid-range laye rs of bookshelves. That is, b y pro- portion , few books pulled out by library readers were from the top or bottom layers. Books on the layers that were easily reachable by readers were frequentl y utilized . Such a book-us e distribution patt ern made the job of reshelving books easy, but created some inquiries: how could book locati ons influ ence th e choices of readers in selecting books? If this was not an isolated observation, it must have exposed an inter es ting Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. phenomenon that librarians needed to pay attention to . Then , by finding out the reasons , librarians might becom e capable of guiding, to some extent , us ers' selectiv eness on library books by deliberately arranging col- lections at design ated heights on book sh elves. A research study was designed to develop Geographical Information Systems (GIS) into an analytical tool to examine former casual observa- tions by the author. The study was conducted in the MacKimmie Library at the University of Calgary. Thi s paper highlights th e results of the study that aimed at assessing th e behavior of library readers in pulling out books from bookshelves . Thes e book s, when not checked out, are cat- egoriz ed as "pickup books" becau se they are usually discarded inside a library after use and then picked up by library assistants for reshelving. Like many other libraries , the MacKimmie Library does not encourage reasd ers to reshelve books th emse lves. ArcView, a GIS software, was selected to develop th e tool for this study because GIS ha s the functions of dynamicall y analyzing and di s- playin g spatial data. The research on library readers pullin g out books involv es the measur emen ts of book- shelf heights, and thu s deals with spatial coordinates. With the capabil- ity of presenting book shelves in dif- ferent views on map s, GIS is able to provide readers with an easy und er- standing of the anal ytical results in visual forms, which make any textu al description s wordy . At the same time, some GIS products are avail- able now in most academic libraries, thus giving develop ers convenient access to use. Hypothesis When library users decide to check books out of a library, the se books are what the y think of as useful. Peopl e are usually hesitant to carry home books that are of little or uncertain use, not only because of the limit on the numb er of check-out books , but also bec ause of the physical work required for carrying them. Moreover, some items, such as periodicals and multimedia materials, are either des- ignated as "refe rence only" or have a very short loan period . It is reasonable to beli eve that user s carefully select what they want from library collec- tions and keep these book s for handy use outside the library. By contrast , in-library book use repre sents a different category of library readers' behavior . There are two general categories of in-library book us e: readers bringin g their own books into a library for use, and readers pulling out book s from book- shelves inside a librar y. The former is commonly seen when students study textbook s for examinations (not the topic of this study), whil e the latter is a little more complex. 1 As library users approach book- shelves to extract book s, th ey may or may not hav e a definit e target. When coming with call numb ers, peopl e will deliberately draw the books they want for reading, photoc opyi ng , or referencing. Ho wever, there are time s when user s on ly wander in bookshelf aisles of desired collections, uncer- tain about singling out specific books . Th ey may simply shelf-shop to randomly select whatever is inter- esting to them, or they may locate a subject of need and go to the storage position(s) to look for whatever books are there. No matter what these readers' intention s are, they roam among collections, pick book s for quick u se, and leave them inside the library after use, although some materials may also be checked out. Because of such arbitrary selec- tions from library collections , physi- cal con venie nce sometimes influence s library users in takin g books from booksh elves-they ma y look around for books on bookshelf layers that are at a reach able height. The standard library bookshelf is hi gher than the average person's height and is struc- tured to have five to eigh t layers. In aca demic libraries, "wood shelving is available in three heights: 82 in. (2050 mm), with a bottom shelf and six ad justabl e shelves; 60 in. (1500 mm), with a b ottom shelf and four adjustable she lves; and 42 in. (1050 mm), with a bottom shelf and two adjustable shel ves ." 2 For regular col- lections in mo st academic libraries, bookshelve s are usually about eighty- two inches high and hav e seven lay- ers. Books on the top lay er are out of reach for many reader s, requiring them to use a ladder to draw a book from it. Many users are hesitant to use ladders. Even worse, a reader will have to bend over or squ at down to view the contents of books on the bot- tom layer of a bookshelf . Hence , the hypothe sis is that books used inside a library are prima- rily distribut ed among the mid-ranged layers of bookshelves. Specifically, if a bookshelf ha s seven lay ers, books placed on layers two through six are most frequently consulted. This is the subject of this research paper . Background A considerable number of studies have investigated the utilization of books that are checked out of a library. An esti mate made in 1967 pointed out that over seven hundred research results pertained to this topic. ' How ever, the situation of books used inside a library has not been given enough attention. One of the reasons for this seeming neglect comes from the belief that the records of library book s in circulation provide similar info rma tion as those of books used within libraries." Thi s misunder- standing wa s lately criticized by other researchers who discov ere d the dif- ferences in use behavior between Jingfeng Xia (jxia@email.arizona.edu) is a student at the School of Information Resources and Library Science at the University of Arizona, Tucson. USING GIS TO MEASURE IN-LIBRARY BOOK-USE BEHAVIOR I XIA 185 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. libr ary readers takin g books h ome and those using books inside librar- ies. 5 Research ers hav e now recog - ni zed that correlations between the two sets of data are n ot as strong as they seemed to be. Such reco gnition, unfortunately, ha s not resulted in mor e consequ ent work to explor e the issu e of in-lib rary book use. This is probabl y due to th e difficulties of co llecting data or the la ck of appropriate research meth- ods .6 Also, the majority of rele va nt surv eys w ere conducted several de cades ago and focu sed primaril y on exp loring a go od method of sam - plin g in-library book us e.7 Am ong the se studies , Fu ssler and Simon pre- ferr ed to carry out researc h by dis- tributing questionnaires am ong library reader s; Drott u sed random- sampling m et hods to statisti ca lly examine th e importance of librar y- book use; and Jain, as well as Salv erso n, emphasized dividing th e survey time s into differ ent investi ga- tion units when conducting res earch. Simil a rly, M orse point ed out the compl ex ity of measurin g lib rary- book u se a t wo rk , advocating an involv ement of computerized opera- tion s in librar y-book man ag ement. The sampling strategies and ana- ly tical methods implemented in pa st studie s are still applicable to curr ent res earc h. Non etheless, because many new technol ogie s ha ve come into view since th en, it is quite likel y tha t som e new ways of obtaining and analy zing th e d ata of in-library book use can now be developed. Th e n ew app roac hes must have the capability of providing not only accurate m eas- urem ent of the data but also the me ans for easy manipulation . Th eir result s must be able to enhance th e und ers tandin g of us er behavio r in expl ori ng th e reso urc es of existing collection inv entorie s . One of th e solutions is an analytic al tool. An analytical tool can control data collection and anal ysis by computeri- zati on . If the system is ab le to accu- mul ate const antly upd ated records ov er time, it will remedy the probl em of poor sampling th at man y resear- chers hav e encount ered, be cause an alysis will then b e done on all the data rather th an w ith certain isolated samples. The development of m odern technologi es makes such data collec- tion and storage po ssible and easier than ever before. On e exampl e of the technologi es is the radio freque ncy identification (RFID) tag system that ha s been adopted b y some public and acade mic librar ies recently.8 Thi s sys- tem stores a tag in each librar y item with the item's biblio gra phic informa- tion, and uses an antenna to keep tr ack of th e tag. By automatically com- municating with dat a stored in the tags, the system can collect dat a on all librar y collections in a timel y manner and export them into pred esigned d atabases for easy man ag ement. Data an a lys is and pres enta tion comprise ano ther p ar t of the an aly ti- cal mechani sm. Researc hers h ave to carefully evaluate existing technolo- gies in order to select prop er prod- u cts or de ve lop parti cular pro gra ms to integrate with RFID (if used) and th e databases. It is fortunate th at GIS techno log y is available with numer- ous functi ons for analyzing and demonstrating data , especiall y spa- tial data. Da ta visuali za tion through GIS produ cts has been very good, which giv es them advantages over other analytic al, stati stical, or repor t- in g produ cts. Combining RFID and GIS into one system would seem to be th e per- fect solution-the former can effec- tive ly carry out dat a collection and th e latter can efficiently perfo rm data analysis and presentation. H ow ever, while GIS products h ave been u sed in libraries in the Unit ed States for more th an a dec ade, mo st academic lib- raries are hesitant to invest in RFID because of its high costs . GIS technol- ogy alone, however , can still provide sufficient functions to be dev eloped into such an analytical tool. Up to n ow, tho se librarie s that have provid ed GIS serv ices only use the software that assists in the uti- lization of geospatial data and map- 186 INFORMATION TECHNOLOGY AND LIBRARIES I DECEMBER 2004 ping te chnologie s for users .9 GIS is not expl oited enou gh to aid the man- age ment of librari es them selves and the res earch of librar y collections. Some commercial GIS software, such as "Lib rary Decision" by Civic-Tech- nologi es, has be en recently marketed to support the analysis of library- user d a ta for public libraries. 10 Ho wev er, it only wor ks w ell on the data of conventional geographical nature, that is, th e distributi on and location of librari es and th eir users with the mapping of city bl ocks and streets . It does not app ly to a librar y an d its books, and especiall y not to the distribution of books us ed insid e th e librar y. Such products are also not ap plicabl e to acad emic librari es that do not always concentrate on the ana lysis of geog rap hical area s of their us ers. Even so, GIS h as all the function s that such a propos ed analytical tool demands. It is suit able for assisting in the research of in-library book us e where library floor layout s or other facilities can be d raw n into maps on multiple-dimensional views. At the same tim e, bookshelves wi th individ- ual lay ers can be treated as an innova- tive form of map by GIS technology (see figur e 1), makin g visible the rela- tionship of book u se to the height of the book sh elf. As soon as th e presen- tation mechanism is linked to data- bases, any updat es on book use will be mirror ed visuall y. Method This proj ect is one of a serie s of proj- ects for deve lopin g GIS into a tool to manag e and anal yze the u sage char- acteristi cs of library books . The other projects include u sing GIS to measure book u sability for the de velopm ent of collection inventorie s; to assist in the managem ent of libr ary physic al space an d facilities; and to locat e library items . 11 In order to make GIS workable for the subject of this paper , the focus was placed only on the exploration of corr ela tions b e tween b ooks helf Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. Figure 1. The front view of one bookshelf rack on the fifth floor of the University of Calgary MacKimmie Library. Eight bookshelves assemb le the range. Here, different shades of color represents the numbers of books used on each individual layer. The display is only for demonstration and not to actual scale. height s and book-use frequenci es in an aca demic library environm ent. Th ere are two major step s to con- ductin g this research : collectin g data and d ev eloping a GIS anal y tical tool. Since MacKimmie Librar y did not in ves t in RFID at th e tim e thi s resea rch was undertak en , p ersonal ob servations were mad e to record b ook-use data. 12 The dev elopm ent of th e GIS tool involves creatin g a sm all d a tabase to store data and facilit ate d ata analysis. It also requir es creatin g seve r al bookshelf and sh elf-r an ge m ap s to pre sent anal ytical result s in visualized forms. Arc View-the mo st p opul ar GIS produ ct in th e w orld- was ut ilize d for the de ve lopment. This paper presents only a p or tion of co llection areas at MacKimmie Library. Part of the fifth floor, wh ere som e collections of humaniti es and social sciences are stored, w as selected becau se this floor is amon g the busi est of th e floors used by read ers. It is filled with sixty-eight ran ges of b ook- shelves containin g book s from call numbers B to DU. The terms used in this paper includ e bookshelf, referring to one unit of furnitur e fitted with horizontal sh elves to h old book s; rack, which includ es more than one bookshelf standin g tog eth er in a line ; and range, comp osed of two racks standing b ack-t o-back. Bookshelves on the fifth floor are arr anged to sur- round a group of facility rooms in the central area. Stud y corridors are set between booksh elves and the wall. Each booksh elf ran ge consists of two bookshelf rack s, each of which in turn has eight individual book- shel ves . All of the book shel ves are about eight y-two in ches high and have seven laye rs. Th e laye rs, except for the top on es th a t are open, are equal in height , w idth , and length. Data Collection Personal surv eys wer e taken by the author to not e d own each call number of books that w ere n ot in their origi- n a l p os ition s on the sh elv es, but in stead were found discard ed on the floo r, tables, chairs, sofas , or on top or in front of other stocked book s . Boo ks on th e sh elving carts ar e also account ed for. The surveys we re sep- ar ately con ducted three times a d ay - mo rnin g, afternoon, and ev en in g- in ord er to cat ch as m any book s u sed in a day as p oss ible. To avoid reco rdin g the sa me boo k mor e than on ce, n o duplicat e call numbers w ere acce pt ed for any single da y even thou gh th e sam e book wa s found in diff erent locations on that day. On the oth er hand , the sam e call number coul d be ent ere d int o the records on th e second day alth ough it was recorded th e d ay befo re a nd remained in th e sa m e pla ce w ith out b eing pick ed up by librar y ass is tants . (Thi s dupli ca te reco rdin g was ve ry rare beca use of th e routin e work of book pi ckup by libra ry ass istants.) A period of two w eeks w as d esignated for the sur vey in th e first h alf of December 2002. Th e final exam in ation week was pl ann ed becau se it represents a week of h ea vy book u se, although previous resea rch found th at readers in this w ee k tend ed to u se library collection s less th an their own stud y mat erials." A suppl em ent a ry surve y th a t a lso las ted two w eeks, includin g a final exam ina tion wee k, wa s condu cted in th e lib ra ry in late spring 2004. To simplify the rese arch , some excepti ons w ere established for d a ta collection. Pe riodicals were exclud ed beca use th ey have a very short loa n p er iod (gen erall y one day) . Libra ry u sers m ay pr efer to read articl es in journ als w ithin the library and thu s w ill h av e a clear idea as to wh a t m aterials to read. '' Books belon ging to oth er floo rs of the librar y, o r b oo ks b elon g ing to th e fifth floor but found out sid e th e area were not includ ed in th e an alysis. Furthermore, du e to the n atur e and time limit of thes e ob ser - v ation s, b ooks pulled out of tar geted bo okshelves were not distingui sh ed from b oo ks taken from book sh elves at rand om . Thi s information can onl y become ava ilable throu gh int erv iew s USING GIS TO MEASURE IN-LIBRARY BOOK-USE BEHAVIOR I XIA 187 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. with library users, which can be another rese arch project. Each book shelf laye r wa s recorded with and signified by two call num- bers: the start and end numbers of books. For example, the call numbers "BF1999 .K54" to "BH21 .B35 1965," representing books stored on a partic- ular layer, were record ed to identify that layer . Because book shifting can happen from time to time, such recording of start and end call num- bers for individual book shelf layers only reflects the condition s when this research wa s undertaken and may need updates whenever changes occur. Data Manipulation and Visualization Using a bookshelf lay er as the recording unit is essenti al for the analysi s of the relationship between book use and bookshelf height. Each book used can be classifi ed to fit in one unit according to the call num - ber of the book. Therefore , building a databas e with a table for lay ers will be an important part in the develop- ment of such an analytic al tool. The LAYERS tabl e includes a data field as an identifi e r to stand for the sequenc e of e ach layer-1 for the top layer, 2 for th e next layer down , and so on , in addition to storing the start and end call numbers of books for each lay er. If more than one book- shelf in th e library has seven layers, layer identifiers will it erate from booksh elf to bookshelf . Therefore, this tabl e will also need an identifier for each individual book shelf with which lay ers are associated. The dat abase will also contain such information as bookshelf ranges, bookshelf racks, and books , all of which are individual database tables that are joined with each other by relational keys. Among them, the RANGES table is simply character- ized by its id entifier, and is designed to repre se nt two rack s of book - shelves that stand back to back. The BOOKSHELVES table is identified by the call numbers of the start and end books stor ed across individual bookshelves rather than on individ- ual layers. Furthermore, th e BOOKS table is primarily filled with the data of individual book call numbers as well as book pickup time s and book discard locations . GIS h as lim ited ability for orga n- izing da tab ase struc tu re. If n ecessa ry, oth er da tab ase managemen t sys tem s, su ch as Microsof t Access, can b e incor p ora ted . Qu ery codes are built to ge t su mmarize d infor m ation for speci fic p ur poses, and th e agg re- ga ted da ta are exp or ted int o GIS data bases for fur the r sp a tial an alysis or con venie nt vis u al prese nt ati on . Da ta vis u aliza tion can be show n at differe nt leve ls- by layer, books helf, rack, and range . Th e firs t attempt at ma king a vis u al dem on stra tion of this researc h is for th e area of in di - vi du al b ooks helves at layer leve l (see figur e 1). Th e follow in g qu ery w ill return necessary summ arize d infor- ma tion: SELECT sum(b.call_no) AS total_num, l.layer_id, l.shelf_id FROM (BOOKS b INNER JOIN LAYERS l ON b .some_id = l.some_id) WHERE b.call_no > l.start_no and b.call_no < l.end_no GROUP BY l.layer_id, l.shelf_id ORDER BY l.shelf_id, l.layer_id. At the same time, another attempt is made to d emonstrate book num- bers per layer, at bookshelf level, across multipl e bookshelf ranges. This demonstration provides a better visualization in the GIS di splay so that an ov erall view of the height distributions of book usage over cer- tain collection areas can be presented (see figures 2 and 3). To achi eve such visualization, data must be com- pared in order to get information about which layer of a bookshelf 188 INFORMATION TECHNOLOGY AND LIBRARIES I DECEMBER 2004 contains the most frequently used books and which holds thos e that are rarely visit ed . This demonstration indicates that any alternative selec- tion of analytical-display units can be easily performed by making mod- ifications on the query that works on aggregating data . Technically, data visualization can be presented by using an y GIS soft- ware, although ArcView is used here because it has been availabl e in the systems of many academic libraries. Bookshelf ranges in MacKimmie Library 's fifth floor were drawn into map features . In order to show them with a three-dim ensional view, each of the seven layers was given a sequential number as its height value , and all book shelves were treated as having the same height. These height values are tre a ted as the z values in any three-dimensional analysis. Then, by associating the numbers of books from the database with the heights of layers on the map, ArcView is able to sketch the hei ght distributions of in- library book us e in new perspectives, dramatically improving the under- standing of book use. In order to implement the visual- iza tion of all layers across a book- shelf range, lay ers were drawn as map features (see figure 1). Layer heights and widths are in appropri- at e proportion . (Individual book s on each layer are for demonstration only, and thus are not in the exact shape and number.) Figure 1 shows how a bookshelf rack has been pre- sented as a GIS map, which is a totally new idea in the applications of GIS visualization . The databas e and visualization mechanism constitute what is referred to in this paper as the analyt- ical tool. One will find that th e devel- opment is relatively easy and the tool is incredibly simple. However, it is a dynamic device. If expanded into other parts of the library collections, this tool will become an integrated system that is able to assi st in the management of library book use and Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. ••••m•==== ===----::"'-:-=-=-=-=-=-=-=-=-=-=-===:-::::-::-":"".:-:-~.,Jgl4 file Edit 7:J Scene Iheme .S1.liace 6t~ s ~ I:!~ IJID ~ ~ liiffl~[i] !HJ~ ~~~I§]~ [QI (ill ---- ..................... _ -~ ¥.l ,, Ill Figure 2. A three-dimensional view of bookshelf ranges on the fifth floor at the MacKimmie Library. The height of each bookshelf represents the corresponding height of the layer from which most books were removed. This display is not to actual scale. I -'! st a,t IIJ gJ.1V