College and Research Libraries Exploring the Intellectual Organization of an Interdisciplinary Research Institute Bryce L. Allen and Brett Sutton Planning and implementing library services for interdisciplinary research communities pose special challenges for academic librarians. Data were col- lected on journal reading patterns in an interdisciplinary research institute (the Beckman Institute at the University of Illinois at Urbana-Champaign). Analy- sis of these data produced a number of maps of the intellectual structure of this user community. This understanding of the structure of the academic commu- nity and how it changes over time provides a basis for developing library services that will meet the special needs of this community. cademic libraries are some- times in the position· of pro- viding information services to parts of the academic commu- nity that are not organized along tradi- tional departmental lines. Specialized research institutes focusing on particu- lar scientific problems are one example. Many campuses now have centers for the study- of specialized topics ranging from cognitive science to the breeding of Chinese pigs. Interdisciplinary groups concerned with topics such as Latin American Studies or Women's Studies constitute another example of special- ized institutes. Providing services to such groups can present challenges for academic libraries that are organized according to tradi- tional areas of subject specialization. Subject specialization, institutionalized in departmental libraries (such as a physics library or a mathematics library) or in the work of subject bibliographers, seems to be based on an assumption of a homogeneous user population. Librari- ans establish a specialized library or specialized services because we think there is a population of users who have similar information needs. Physics li- braries are created, or physics bibliog- raphers hired, to serve physicists and physics students. If these users did not have similar needs, the departmental li- brary or subject specialist bibliographer would not be the best approach to serv- ice. Sometimes academic libraries are af- fected by institutional inertia and remain organized along traditional disciplinary boundaries even when these boundaries no longer reflect the academic communi- ties the libraries serve. The debate con- Bryce L. Allen and Brett Sutton are Assistant Professors in the Graduate School of Library and Information Science, University of fllinois at Urbana-Champaign, Urbana, Illinois 61801. Research for this article was done under the aegis of the Project on Scholar Communication and Information Transfer, uigh Estabrook, Project Coordinator, and was funded in part by grants to the project from the Research Board of the University of Illinois. The authors wish to thank the federated researchers in the project, Theodore Brown, Director of the Beckman Institute, and the librarians and staff at the Beckman Institute Library for their helpful suggestions and comments. 499 500 College & Research Libraries cerning the role of academic branch li- braries is summarized admirably by Leon Shkolnik.1 Librarians, even those in specialized libraries, recognize the necessity of pro- viding flexible service to a hetero- geneous user community and know from experience that the needs of users, even within a particular discipline, are not homogeneous. At some universities, when local resources are available and interest is high, special collections are established to meet the needs of emerg- ing user groups. For example, computer science collections emerge from mathe- matics and science libraries, or area stu- dies libraries are created to meet the needs of special programs. If special col- lections cannot be created, some depart- ments may establish their own informal reading rooms, although these services may be less than ideal. More often, librarians find ways to meet the needs of a heterogeneous user community by developing channels of communication within existing library structures. Sub- ject bibliographers who develop good working relationships with faculty and students can respond to many different user needs. Collaboration among subject specialists can help to assure an appro- priate balance of materials in library collec- tions. But these labor-intensive solutions cannot always keep pace with changes in the academic community. Sometimes li- brary users working in interdisciplinary areas may have to adapt to the tradi- tional organization of collections and services by visiting each of the appro- priate service areas in tum. In the case of specialized research in- stitutes or interdisciplinary working groups such as area studies depart- ments, to assume that there is a homo- geneous user population is particularly erroneous. A research institute on cogni- tive science, for example, may involve linguists, psychologists, and computer scientists. A Latin American Studies de- partment may have political scientists, sociologists, and literary scholars. In sit- uations where the population of users is not homogeneous, librarians charged with providing information service should November 1993 seek systematic information about the intellectual organization of the user community. If librarians can learn more about an intellectual organization, they are in a better position to ensure that users and librarians can cooperate to provide the information people need. The library staff is able to meet the users halfway, adapting to the organization of the user community even as users adapt to the organization of the library. Librarians establish a specialized library or specialized services because we think there is a population of users who have similar information needs. Recent studies, including the work of Julie Hurd and Paul Metz, have demon- strated that interdisciplinary work is now widespread in academic communi- ties.2.3 Although our work focused on an interdisciplinary research institute, we believe that the methods outlined below may have broad application as academic librarians consider the ways in which they will organize services to meet the needs of their user communities. STRUCTURE OF USER COMMUNITIES AND SERVICE PATTERNS The problem addressed in this · re- search, generally stated, is: How can li- brary service be structured for user communities that do not fit the typical department-centered or discipline-based structure? Our approach to resolving this problem was to develop ways of ex- amining the intellectual organization of the user community. In other words, it is not enough to acknowledge that interdisciplinary research institutes seem to be structured differently from other user groups on campus. They can, theoretically, have a variety of internal organizing structures, each of which might suggest a different approach to providing library service. The challenge is to figure out how these institutes are structured. It is possible that an interdisciplinary research institute might promote inter- disciplinary work and examine topics that can be viewed from a variety of perspectives and yet reflect, in micro- cosm, the usual academic organization of a university. For example, a Latin American studies unit might consist of groups of economists, political scien- tists, and historians who interact within their disciplinary groups but do not cross disciplinary boundaries in their re- search. If this were true, then the user community would be organized into a number of stable independent com- ponents, each component having its own particular information needs. In such a circumstance the library could organize its service by providing reference tools, selective dissemination of information (SOl), or collections that are tailored to the specific information needs of each component of its user community. This would be accomplished by developing a number of small" departmental" service units to meet the information needs of each of the components of the user com- munity. But it is also possible for the user community to have a more dynamic in- tellectual organization in which library users do not limit themselves to materials from their own discipline. Users may cross disciplinary boundaries, working with · one set of colleagues at one time, and another set of colleagues at another time. Their information needs may vary signifi- cantly from day to day. In such a dynamic situation, discretely structured information services would be counter-productive. More flexible, integrated approaches to reference, SDI, and collection building would be needed. Ideally, the organization of library services for such a user commu- nity would need toechothecomplexstruc- ture of the user community. A centralized, integrated information service offering a variety of reference tools, information re- trieval, and SDI would probably provide the most appropriate service to such a dy- namic community. Members of the user community could select from these varied services the tools that might best meet their information needs at a partic- ular time. Interdisciplinary Research Institute 501 Other ways in which user communi- ties could organize themselves require other service strategies. In some user communities, there may be hierarchical organizations of users in which informa- tion needs depend at least in part on a user's level in the hierarchy. In this type of community, services could be de- signed to meet the needs of each of the hierarchical levels, with new services being offered to people as: they change their levels in the hierarchy. Librarians who serve businesses or work in manage- ment information systems often encounter hierarchically organized situations. An in- terdisciplinary research unit could be or- ganized hierarchically, but this is certainly not traditional for academic organiza- tions. Similarly, there may be user com- munities in which there is a majority group of users and one or more minority user groups. In such cases, information needs may depend on whether the user is part of the majority group or takes a minority approach to scholarship. For such a community, it might be appro- priate to design information services for each of the "approaches"' to the research topic. Finally, it is' not uncommon for user communities to be organized around research projects. In this type of research institute, library services could be designed for each project, as sug- gested by Harry Llull:' These examples illustrate the impor- tance of understanding the intellectual structure of a user community so that librarians may offer appropriate infor- mation services to that community. Aca- demic librarians, usually have the opportunity to gain an understanding of the intellectual organization of their user communities. They see and talk to users from all parts of that community on a regular basis, participate- in meetings and colloquia, and take, part in the gover- nance of their academic community. But busy librarians may not be able to partic- ipate extensively in academic affairs, and theix: unfamiliarity with a user com- mu:nity and its intellectual organization may prevent optimal planning and eval- uation of information services. Our ob- jective in this researdt was to explore a 502 College & Research Libraries more formal means of identifying the ways in which user communities are or- ganized, and so to provide a set of tools that might be of assistance to librarians who are faced with the task of providing information services to an interdiscipli- nary research institute. INDICATORS OF USER COMMUNITY STRUCTURE In approaching this task, we con- sidered many of the characteristics of a user community that might serve as in- dicators of the intellectual organization of that community. For example, infor- mal communication patterns are one such indicator, as are patterns of re- search collaboration. It is quite possible to assess who communicates with whom, and who works with whom, and to map the structure of the user community from these indicators. However, both of these indicators are relatively narrow in applica- bility. Identification of infonnal communica- tion patterns through social network analysis is helpful, but possible only in relatively small user communities. In a large research institute or interdiscipli- nary studies group, particularly one in which the composition of the user commu- nity changes rapidly over time, such methods are costly and difficult to man- age. Research collaboration (usually iden- tified through coauthorship of ~pers) focuses on only one kind of interaction. We thought that using this indicator would provide an incomplete under- standing of the nature of the user com- munity we were studying. From the possible indicators that we considered, we selected journal use (viewed from several perspectives) as the indicator that seemed most appro- priate to our task of identifying the intel- lectual organization of the particular user community we studied. We as- sumed that if people used the same jour- nals, they were likely to share academic interests, speak the same technical lan- guage, and to share some interest with the authors who published in the jour- nals. In other words, people who use the same journals are closer together in the intellectual organization of a community November 1993 than people who seldom or never use the same journals. This assumption is shared by Hurd and by many of the stu- dies of interdisciplinarity she cites in the thorough literature review contained in her article. 5 THE RESEARCH SITE The library user group we studied came into existence as the result of the establishment of the Beckman Institute at the University of Illinois in 1989. De- signed as an interdisciplinary research institute to investigate aspects of human and machine intelligence, the Beckman Institute brought together scholars from a variety of disciplines ranging from physics to psychology and from philos- ophy to computer science. Faculty and students were organized into research groups, each with designated office and laboratory spa-ce and support staff. Be- cause these groups did not map directly onto existing academic departments, and were instead organized around general research areas, they were useful in providing a first approximation of in- tellectual organization. There were, however, reasons to ques- tion the adequacy of this organizational structure as an indication of the intellec- tual structure of the Beckman Institute. Some of these research groups existed before the creation of the institute as nat- ural communities of scholars sharing re- search interests, but others appeared to have no specific research agenda and served rather as umbrella organizations under which individual faculty pursued particular research agendas. In other cases, topics of scholarly investigation and study appeared to be shared by several groups. By focusing on journal use, an indicator of intellectual activity, we were able to identify the high-level intellectual organization of this user community, as opposed to its institu- tional structure. We attempted to show how the groups initially established in the institute combined into broader groups or clusters that shared research interests and perspectives. The scholars and administrators who planned the Beckman Institute saw the overall collection of the University of Illinois Libraries as supportive of the re- search and instruction of the institute. A special academic library was included within the institute, designed primarily as a service point rather than as a collec- tion. The Beckman Institute Library es- tablished a small collection of highly used journals, but the greatest emphasis in this library was on the provision of elec- tronic access to information through lo- cally generated and mounted databases, CD-ROM bibliographic services, and on- line searching. Another important service was a heavily used article copy service, in which photocopies of articles from journals were requested by Beckman In- stitute faculty and students, and pro- vided free of charge by the library from the University Library's collections, or through interlibrary loan. Our objective in this research was to explore a more formal means of identifying the ways in which user communities are organized, and so to provide a set of tools that might be of assistance to librarians who are faced with the task of providing information services to an interdisciplinary research institute. The objectives of this research were to explore means of data collection and analysis that reveal the intellectual struc- ture of a user community. We were inter- ested in investigating a variety of ways to obtain insights about the high-level organization of scholars (faculty and students) within an interdisciplinary re- search institute. As a secondary objec- tive, we wanted to identify methods of analysis that could be used by pro- fessionallibrarians. We used a variety of microcomputer hardware and software that would be available to many pro- fessionals in academic libraries. In other words, we wanted not only to explore the intellectual organization of one re- search community, but to do so in a way that would benefit professionals work- ing with similar research communities. Interdisciplinary Research Institute 503 DATA COLLECfiON Data on the journal reading patterns of members of the user community of the Beckman Institute Library were gathered in three ways. As part of a larger project examining scholarly communication and information transfer within the institute, all full-time faculty members appointed to the institute were interviewed using a structured interview schedule. One set of questions asked faculty members to identify the journals they read regularly in connection with their research. Re- sponses to these questions were aggre- gated by research group for analysis, so that the results could be presented anonymously. The frequency with which journal titles were mentioned by all of the faculty within a research group was tabulated into a separate journal list for each research group. Although this mechanism suppressed differences be- tween scholars within groups, it pro- vided a data set that was ideal for establishing the higher-level intellectual organization of the institute. The second data collection was accom- plished with the cooperation of the Beck- man Institute Library. Records of the article copy service were summarized in such a way that individual users' read- ing could not be identified. These data were aggregated into a similar set of journal lists for each research group for each of the first four academic terms in which the service was offered. Each list also contained the frequency with which articles from a particular journal were re- quested by users associated with a partic- ular research group. These lists were further aggregated into a single list of the first fourteen months of the service for each department. The data from the ar- ticle copy service were less complete than the interview data in certain ways. Some of the members of the user community who were interviewed made no use of the library's service, and so were not included in the article copy service data. Similarly, we assumed that scholars do not request from the library photocopies of articles from journals to which they subscribe. As a result, these data did not include 504 College & Research Libraries some of the central reading habits of members of the user community. However, the data from the library ar- ticle photocopy service were also more complete than the interview data in some ways. Articles requested by all members of the user community were recorded, so reading patterns of the stu- dents of the institute as well as of its faculty were included. Also, articles were requested from many journals that would not have been considered as "read regularly." As a result, a broader picture of reading patterns was obtained from the library data. If librarians can learn more about an intellectual organization, they are in a better position to ensure that users and librarians can cooperate to provide the information people need. November 1993 advantages in this type of data. Because the data are publicly available, this ap- proach to data collection is unobtrusive. Because authors usually cite articles directly relevant to their scholarly work, citations provide a reliable measure of the information that the scholars think is important. There are also some dis- advantages associated with using cita- tion data to explore the intellectual structure of a user community. Citations may reflect previous reading, rather than current research interests, and so may be less relevant to the current intel- lectual structure of the community than the other measures used in this research. In the case of articles coauthored by scholars from outside the user commu- nity, we do not know whether the cita- tions reflect the reading of .members of the user community or that of the coauthors. In addition, this method of data collection is constrained by the selection of source articles by Institute The third method of data collection for Scientific Information, the producer involved scanning three current years of of the citation indexes. It is possible that citation indexes (both Science Citation citations from articles not covered by the Index and Social Science Citation Index) for citation indexes would have added to articles written by faculty from the Beck- our understanding of the intellectual man Institute. The citation records from structure of the community. Finally, each citation index entry were copied these data require considerable database into a database that allowed us to com- expertise to convert the citation index pile a list of all of the journals cited by entries into lists of journals consulted by scholars from each research group. In groups of scholars. this case, citations from published arti- Identifying the same variable (in this des were taken as evidence of the read- case, journal use patterns) through ing patterns of the members of the user different types of data provides many community. There are a number of advantages. In this study, there were rea- TABLEt COMPARISON OF THREE SOURCES OF JOURNAL USE DATA Source of Data Advantages Survey Direct Provides contextual information Stresses most used titles Copy Service Reflects current interests Unobtrusive Provides frequency of use Citations Broader base Unobtrusive Multidisciplinary Provides frequency of use Disadvantages Possibly incomplete Subject to selective recall Obtrusive Provides title, not frequency of use Does not include some scholars May not reflect all journal reading May reflect previous reading May overemphasize peripheral subjects Limited by lSI coverage May reflect reading of co-author sonably high correlations (r =.52 to .62) between the similarity matrices pro- duced from the three types of data, so we were confident that these different ap- proaches were reliable measures of jour- nal use. It would have been possible to combine the results into a single index of journal use, but we thought that separate analyses of the data would reflect their richness, and would allow the advan- tages of individual measures of journal use to compensate for the disadvantages of other measures. Table 1 outlines our perception of the advantages and dis- advantages of each of the three ap- proaches to data collection that were used in this research. DATA ANALYSIS The lists of journals consulted by members of research groups were ana- lyzed to produce a cosine similarity measure for each pair of groups. The cosine measure is a value between 0 and 1 reflecting the similarity between pairs of groups, based on the frequency with which the same journals were men- tioned or requested by members of those groups. The result was a matrix of simi- larities that showed how similar each group was to every other group. The similarity matrices formed input to two additional kinds of analysis. The first was multidimensional scaling, which produces a two-dimensional configura- tion. It creates a map of the intellectual community, with groups shown to be sim- ilar to each other appearing closer together on the map and with dissimilar groups farther apart. The second analysis was hierarchical cluster analysis. This analy- sis placed research groups with similar journal use patterns into higher-level clusters. These clusters were then located on the two-dimensional maps produced by multidimensional scaling. Frequently this type of analysis has been helpful in understanding the intellectual structure of disciplines; technical details are pro- vided in articles by Katherine N. McCain and by Henry Small and E. Sweeney.6.7.s Lists of journals consulted by groups of scholars can also be used to generate similarity measurements between jour- Interdisciplinary Research Institute 505 nals. If two journals are consistently read by the same people, it is assumed that their content is similar. In the case of the interview data, similarities between journals were identified using cosine measures. This allowed a map of journal titles, analogous to the maps of the intel- lectual community, to be created. In addi- tion, simultaneous clustering of journals and groups showed which parts of the user community were most closely as- sociated with the use of particular clusters of journal titles. FINDINGS Comparison of Data from Three Sources Interview Data. The faculty ap- pointed to the Beckman Institute were interviewed to assess a variety of issues related to information gathering and scholarly communication. As part of these interviews, scholars were asked to identify the journals they regularly consulted in connection with their research. Figure 1 presents the clusters of groups identified on the basis of the responses of the re- searchers during these interviews. This map presents the result of the analysis of the interview data described above. Before moving to a discussion of the differences between this map and those produced from other data, it may be helpful to outline the main features of all the maps. Each letter represents a research group. For example, in the Psy- chology cluster, the letters represent groups with names like "Cognitive Neu- roscience" and ''Visual Processing." The groups whose reading patterns were similar are closer together on the map. Normal maps have an orientation (for example, north at the top) that helps users read them. Regarding the type of map created in this project, the orienta- tion of the map is a matter for interpreta- tion. It seemed clear to us that the top of the map contained groups who adopted approaches that might be called "scien- tific" while those toward the bottom seem to have adopted engineering ap- proaches. Toward the top, for instance, are groups investigating neuronal pat- terns and other aspects of cognitive science, while toward the bottom are 506 College & Research Libraries November 1993 Scientific Approach M a c r 0 L e v e 1 f- r- I Psychology F X G 0 l E'\Stneermp 1 D Techr A H R I I N I c E s K L IP ology J u w I Science v I - v - M 0 1 e c u 1 a r L e v e 1 Bnoineerino Approach FIGUREl Ousters from Interview Data groups investigating organometallic materials or decision processes in com- puterized networks. The other "direc- tion" on the map was more difficult to interpret. At first, we thought that social science was on the left and pure science on the right. Since a number of engineer- ing groups were found on the left, this interpretation was inappropriate. Our final interpretation of the horizontal di- mension on this map was that groups toward the right of the map are con- cerned with phenomena at the molecu- lar level (or even atomic level) while those to the left of the map are concerned with what we have called the "macro" level. Toward the right of the map are groups doing scanning tunneling micro- scopy and working on materials chemis- try, while to the left are groups investigating much larger phenomena, such as the architecture of intelligent systems (including the brain). The cluster boundaries superimposed on the letters represent the results of cluster analysis, in which similar groups were placed into high-level clusters. The labels assigned to these clusters are based on our interpretation of the com- mon interests in the research groups rep- resented in each cluster. For example, the Science cluster contains groups like "Molecular Biophysics" and "Prokaryote Genome Analysis." A comparison of the different views of the user community presented in this article will show that there appear to be six of these high-level clusters in the Beckman Institute (Ap- plied Science, Artificial Intelligence, En- gineering, Psychology, Science and Technology), but that the boundaries be- tween some of the clusters are not well defined in some of the data. Library Article Photocopy Service Data. Over fourteen months, careful records were made of all journal titles Interdisciplinary Research Institute 507 Scientific Approach Science L M K 0 M 1 a e c c r c u 0 1 a M r L e s v E L rhn~:r v e y e 1 App1ied Science v e R 1 Bngineering Approach FIGURE2 Clusters from Article Copy Service Data requested by faculty and students of the Beckman Institute from the Beckman Li- brary. From these lists, a map of intellec- tual structure of the user community based on actual reading patterns rather than reported reading patterns was derived. Figure 2 shows the clusters of groups identified on the basis of four- teen months of data from the article pho- tocopy service. The similarities between this map of the intellectual structure of the conuriu- nity and the one produced from inter- view data are clear. For example, the Psychology and Science clusters are roughly the same. There are also a num- ber of important differences between the two data sets that are illustrated by these two figures. The first difference is that the · Engineering cluster identified through interviews is absent from the library data. It would appe~r that re- searchers in this cluster made little or no use ·of the library's article copy service. In other words, the interview data iden- tified a group of potential library users in the intellectual community who were not making use of one of the important services of the library. On the other hand, the library data showed a clear difference between the Applied Science and Technology clusters that had been grouped together as a Tech- nology cluster based on the interview data. This seems to indicate that there are a few m~in journals shared by these clus- ters and all of the scholars mentioned these journals in the interviews. However, when detailed evidence of reading was examined, it showed that there was sufficient difference in reading patterns to create two separate groups. Because of the additional detail provided by the li- brary data, a clearer picture of the con- 508 College & Research Libraries November 1993 Scientific Appro~ch M a c r 0 L e v e 1 1- - I Psychology ll F X N G D IJ A K I y ~ppl Technology u ~cie J I p w ~z s c ed "ce E Q I Science K L a T R y M ' v I - - M 0 1 e ·C \1 1 a r L e v e 1 Engineering Approach FIGURE3 Clusters from Citation Data figuration of the intellectual community emerged. Although it is difficult to see differ- ences in detail by examining the maps, the analysis of these two types of data showed that some of the research groups were included in one duster based on the interview data, and in a different duster based on the library data. For example, group N appears in the Psy- chology duster in the maps based on interview and citation data, in the Science duster in the map based on the library data, and even in the Applied Science Cluster in one of the longitudinal maps discussed below. These groups are of great potential interest, because they seem to be located at the boundary be- tween dusters. Such boundruy-spanning groups can perform an essential role in the development of an intellectual com- munity. From the perspective of library service, they are likely to require a broader range of information than groups con- sistently found in the same cluster. In other words, boundary-spanning groups defy traditional approaches to library serv- ice, and require innovative approaches by library professionals to meet their wide- ranging information needs. Citation Data. Figure 3 is a map of the intellectual community of the Beckman Institute based on citation data. Clearly, this map resulted from a very different kind of data, and it is not sur- prising that there are a number of differences between this map ana those presented above in the relative location of the clusters. In addition, more groups ap- pear on this map than on the map generated from library data, because all " publishing faculty were included, whether or not they made use of the library photo- copy service. Thus there is more detail Interdisciplinary Research Institute 509 Scientific Approach M a c r 0 L e v e 1 r- r- I - Jj PaychCl logy Q A I c Ev ADDiied Science I y D I N Science z L K M R s I I I - - M 0 1 e c u 1 a r L e v e 1 Engineering Approach FIGURE4 Clusters from Article Copy Service, June/December 1990 (2,380 Uses) on this map. However, the similarities among the versions of the intellectual structure of this user community pro- duced by the three different sources of data on reading patterns are striking. Longitudinal Analysis of Library Data Figures 4, 5, 6 and 7 show how the intellectual organization of the user community changed over a period of eighteen months. We believed that ana- lyzing the library data would provide an appropriate longitudinal view of these changes. Repeated interviews would have provided comparable data, but they would have required considerable intrusion into the user community. In addition, the data from the library article photocopy service were more sensitive to small-scale changes in reading pat- terns than the citation data. One final consideration in this choice was the fact that library users were likely to be re- questing materials to which they did not routinely have access (through personal subscriptions, for example). We thought that this made the library data more in- dicative of emerging interests, and par- ticularly of cross-disciplinary interests. It is possible to examine the evolution of the user community by considering each of the main clusters in turn. The Psychology cluster initially had two components, one more applied than the other. These two clusters eventually merged. The Science cluster has main- tained a relatively consistent core of re- search groups, but there has been some migration back and forth between this cluster and the Applied Science clusters. Like the Psychology cluster, the Applied Science cluster has two components, which separated and combined at differ- ent points in time. It appears to be one of 510 College & Research Libraries November 1993 Scientific Approach M a c r 0 L e v e 1 r- - u D I ,..- N G Fl Psychology I A X I ~ z c M I Science K L Applied Science s E Q v y R I - - M 0 1 e c u 1 a r L e v e 1 Engineering Approach FIGURES Clusters from Article Copy Service, January /May 1991 (2,410 Uses) the more dynamic clusters, with groups shifting between the Applied Science and the Science cluster, although there is a relatively consistent core of groups. In summary, this user community exhibits a great deal of dynamism, which may be important in designing information services for the community. Reading Patterns as Journal Clusters Most libraries in North America use broad, general-purpose classification schemes to organize their collections and to assist in the identification and re- trieval of information. Although such schemes have the advantage of being applicable to a large variety of library situations, sometimes they may be too inflexible to be used successfully in a specialized research institute. Analysis of the kind discussed here can provide an alternative for classification of library materials and for information retrieval in a specialized environment. In essence, the alternative classification emerges from an analysis of the extent to which scholars view titles as meeting the same information need. For example, the lists of journals developed for each research group from interview data were ana- lyzed to produce similarity measures be- tween journals. If scholars frequently cited the same two journals as relevant to their research interest, this indicated that the journals were similar in subject coverage. After similarity measures were calculated for all journals, multidi- mensional scaling and cluster analysis were used to produce a map of the jour- nal literature. Figure 8 shows the organization of journal titles revealed by the responses in interviews. Each dot represents a jour- nal title, and journals that appear close Interdisciplinary Research Institute 511 Scientific Approach M a c r 0 L ~ v e 1 1- 1- I Psychology G F X L.- N J. v M y E AI plied s ier Q s R I Science A u ce Ap I L ~1 c a ~lied S I - jcience - M 0 1 e c u 1 a r L e v e 1 Engineering Approach FIGURE6 Clusters from Article Copy Service, May I August 1991 (2,248 Uses) to each other were more likely to have been cited as being read frequently by the same scholars. This journal map is similar in many respects to the maps of the intellectual structure of the user community. This is hardly surprising, since it is based on similar analysis of the data used to create the intellectual structure maps. Simultaneous clustering of scholars and journals can provide a direct asso- ciation between clusters of users and clusters of journals. This was done for the interview data, and table 2 provides an illustrative output. It shows (as one would expect), that Psychology journals are primarily associated with the Psy- chology cluster. More revealing is the nature of these journals. Rather than em- phasizing the full range of journals in Psychology, this cluster of library users appears to be focusing upon physiologi- TABLE2 JOURNALS ASSOCIATED WITH THE PSYCHOLOGY CLUSTER Behavioral and Brain Sciences Cognition Cognitive Psychology Cognitive $cience Electronencephalography and Clinical Neurophysiology Journal of Experimental Psychology: General Journal of Experimental Psychology: Human Perception and Performance Journal of Experimental Psychology: Learning, Memory and Cognition Journal of Neuroscience Neuropsychologia · Neuroscience Psychological Review Psychophysiology 512 College & Research Libraries November 1993 Scientific Approach M a c r 0 L e v e 1 r- ATiic~l 1- I PayL:j G N '"~ruence v R D I I LScience a K z Applied Science EM Q s y c I - - M 0 1 e 0 u 1 a r L e v e 1 Engineering Approach FIGURE7 Clusters from Article Copy Service, September /December 1991 (3,273 Uses) cal psychology and cognitive psychol- ogy. Details of this sort may be most useful in planning services such as SDI for segments of user communities. SUMMARY OF THE MAPPING TECHNIQUE The three methods of data collection used in this research identified an or- ganizational pattern among the library user community associated with the Beckman Institute. Data from interviews with scholars, citation analysis, and pat- terns of use of one specialized library service contributed to this understanding of the user community. Each type of data collection had advantages and disadvan- tages, and differences among the data sug- gest that any librarian who wants to . explore a user community should make use of more than one data source. The data were collected, aggregated, and analyzed using a variety of micro- computer hardware and software that would be available to many academic librarians. The interview data were tran- scribed and entered into text files on IBM-PC clones. A utility program was written to scan each text file in turn and to aggregate answers to the same ques- tions into a separate file. Data from this file were entered into a standard spread- sheet (in this case, MS Excel) to create the lists of journals identified by groups of scholars as being regularly used. Cosine similarity measures were calculated by a small program written in BASIC, al- though SPSS-PC+ could have been used. Multidimensional scaling and cluster · analysis were done on the Macintosh version of SYSTAT, although SPSS-PC+ would have produced the same output. Interdisciplinary Research Institute 513 Scientific Approach M a c r 0 L e v e I - ~ I Psyc~olog,r \ • • • • • • • • • • Engineer Art flclal lntt lllg:nce ~: • Technology • • • I • • ng • • • • • • • • • I Science • • • • • • • • • Applied Science • • • • I - - M 0 I e c u I a r L e v e I Engineering Approach FIGURES Journal Clusters from Interview Data The citation data were downloaded from the online (DIALOC) version of Social Science Citation Index and the CD- ROM version of Science Citation Index. They were imported into an INFORMIX database, using a program written inC to convert the text files into structured files for entry into appropriate fields in the database. Because of available re- sources in the School where we were working, we used a UNIX version of INFORMIX, but this version is very sim- ilar to that available for the M5-DOS en- vironment. Again, any programmable relational database on a microcomputer would be able to manage this data and produce the necessary output. Because off-the-shelf hardware and software were used in this project, and because only moderate computer, database, and statistical expertise were employed, the methods of data collection and analysis illustrated here could be adopted by many academic librarians who wish to explore the intellectual organization of their user communities. CONCLUSIONS In this section, we summarize the re- sults of this investigation of the intellec- tual structure of the library user community from one interdisciplinary research institute, and suggest how these results might be taken into account 514 College & Research Libraries in designing information services for that user community. First, by using a number of methods to assess journal reading, we were able to identify parts of this user community that were not making use of the library. These clusters appear on maps of the community that were generated from interview data, and from citation data, but not on the maps that were generated from library use statistics. This "black hole" in the library data points out the importance of using a number of methods to assess user com- munities. If the library data had been used by itself, this absence would not have been as noticeable. This also sug- gests that current information services provided by the library are not con- Data from interviews with scholars, citation analysis, and patterns of use of one specialized library service contributed to this understanding of the user community. sidered useful (or perhaps usable) by part of the user community. Redesigning information services to meet the needs of these clusters is a challenge that the li- brary staff may wish to address. Another insight emerges from these maps. Some research groups were ex- cluded from the high-level clusters that made up the intellectual structure of the community. Providing library service to outlying members of a user community can be another challenge for a library. Frequently, we establish services to meet the needs of the majority, and so tend to disregard the needs of the minority. One of the great values of this kind of analysis of intellectual structure of a user com- munity is that it identifies minorities: small clusters and clusters of one. Equipped with an understanding of this intellectual structure, librarians are in a position to explore the special informa- tion needs of minority users, and toes- tablish programs that meet those needs. In some cases, services that meet the needs of the main clusters in the commu- November1993 nity can be adapted to meet the needs of outlying members. In other cases, it may be necessary to consider specialized services to meet the needs of those minority groups. This analysis also showed that reading patterns in this user community changed from one semester to the next. These changes were not always dra- matic, and perhaps indicated changes in perspective rather than in information needs. On the other hand, a library would be ill-advised to ignore such changes, particularly when the short- term changes are shown to be a part of long-term trends. For example, it seems clear that the combination and separa- tion of clusters are phenomena to which the library should respond. As noted above, some of this activity results from the movement of boundary-spanning groups from one cluster to another. Plan- ning for library service to a rapidly changing user community seems to re- quire new and flexible approaches. It would be possible, for example, to create a highly selective SDI service for boun- dary-spanning scholars. This SDI service would be designed to alert scholars to new initiatives and developments in fields other than their primary area of research interest. This kind of wide- ranging SDI iS quite different from usual library services that are keyed to a nar- row interest profile. A case can be made for a service that would identify impor- tant developments in all areas except the narrow area of subject interest of a user. Some of the semester-to-semester changes in reading patterns result from the addition of new individuals and groups to the Beckman Institute, and from the departure of some individuals and groups from the user community. Thus this user community is dynamic in this sense as well as in the ways dis- cussed above. The rate at which groups have come into the institute, and have begun to make full use of the facilities of the institute (including the library) has varied considerably. Therefore, library service must be flexible enough to meet the needs of a rapidly changing user community. Finally, the classification of journal titles that emerged from this research could be used in conjunction with the maps of the high-level structure of the user community to provide specific serv- ices. For example, tables of contents from journals associated with user clusters could be routed to those users. Another use of this emergent classification might be in providing access to online searching. The databases that cover a cluster of journals could be made directly available to users from the associated user cluster, while databases covering other journal clusters might be made available primarily in an intermediated mode to those users. We do not mean to suggest that this kind of classification is useful for organizing the library's collection. Because it emerges from a dynamic intellectual organiza- tion of the user community, the classifi- cation is likely to change too quickly to allow its use for an essentially static function such as the organization of the library collection. Rather, it is of greatest use in the organization of services such as database searching or SDI, whether or not such services make direct use of a Interdisciplinary Research Institute 515 particular collection. In addition, emer- gent classifications of this sort may serve to point out to scholars relationships be- tween research topics of which they were not initially aware. Equipped with an understanding of this intellectual structure, librarians are in a position to explore the special information needs of minority users, and to establish programs that meet those needs. In summary, analyzing the intellectual structure of the user community can pro- vide important insights about how that community functions, and accordingly about how its information needs might be met. We believe that this type of map- ping is within the technical capabilities of many academic librarians, and that it can be useful to them as they consider the difficult issues associated with pro- viding information services to an inter- disciplinary research institute. REFERENCES AND NOTES 1. Leon Shkolnik, "The Continuing Debate over Academic Branch Libraries," College & Research Libraries 52(July 1991): 343-51. 2. Julie M. Hurd, "Interdisciplinary Research in the Sciences: Implications for Library Organization," College & Research Libraries 53(July 1992): 283-97. 3. Paul Metz, The Landscape of Literatures: Use of Subject Collections in a University Library, ACRL Publications in Librarianship, no. 43 (Chicago: ALA, 1983). 4. Harry Llull, "Meeting the Academic and Research Information Needs of Scientists and Engineers in the University Environment," Science & Technology Libraries 11(Spring 1990): 83-90. 5. Hurd, "Interdisciplinary Research in the Sciences." 6. Katherine W. McCain, "Cocited Author Mapping as a Valid Representation of Intellec- tual Structure," Journal of the American Society for Information Science 37(1986):111-22. 7. Katherine W. McCain, "Mapping Authors in Intellectual Space: A Technical Overview," Journal of the American Society for Information Science 41(1990): 360-68. 8. Henry Small and E. Sweeney, "Clustering the Science Citation Index Using Co-Cita- tions 1: A Comparison of Methods," Scientometrics 7(1985):393-409. .I You get one chance with authority control, so it's important to get it done right. LTI guarantees that its affordable, machine-only authority control will link 95% or more of your library's controlled headings to an LC or LTI authority record. No exceptions! No excuses! When manual review is requested, only professional librarians are used as editors and link rates approach 100%. LTI maintains the complete LC MARC authority ftles (updated weekly), supplemented with over 410,000 LTI authority records and 350,000 proprietary "cross links." Contact LTI for more information on authority record link results. ~~uthority Control for the 21st Century,, • LIBRARY TECHNOLOGIES, INC. 1142E Bradfield Road Abington, PA 19001 (215) 576-6983 Fax: (215) 576-0137 (BOO) 795-9504·