192 American Archivist / Vol. 50 / Spring 1987 Description and Reference in the Age of Automation AVRA MICHELSON Abstract: This article is a report of the results of a survey conducted in 1986 to deter- mine the effects of descriptive practice on retrieval capabilities of archive and manu- script materials described in bibliographic utilities. Forty repositories inputting into RLIN AMC were surveyed on (1) standardization in the choice and construction of headings, (2) levels of authority control, and (3) treatment of out-of-scope materials. Massive inconsistency in descriptive practice was found. The author makes five recom- mendations to correct inconsistency in the file and the retrieval difficulties that result. About the author: Avra Michelson assists in the development of the archives catalog of the Smith- sonian Institution Bibliographic Information System (SIBIS). She received an M.A. in American Studies from the State University of New York at Buffalo (1976) and a M.L.S. with a specializa- tion in archives administration from the University of Michigan (1983). This article was written as a product of the author's participation in the 1986 Research Fellow- ship Program for Study of Modern Archives administered by the Bentley Historical Library, Uni- versity of Michigan, and funded by the Andrew W. Mellon Foundation and the Research Division of the National Endowment for the Humanities, a federal agency. The author thanks Francis X. Blouin and William Wallachfor assistance beyond the call of duty; Paul Conway, Lisa Weber, and members of the RLG Task Force on Archives and Special Collections for comments on meth- odology and research design; David Blair and Frank Boles for assistance in research analysis; and David Bearman for patience, support, and careful readings of countless drafts. D ow nloaded from http://m eridian.allenpress.com /doi/pdf/10.17723/aarc.50.2.j889j531t25g1401 by C arnegie M ellon U niversity user on 06 A pril 2021 Description and Reference 193 DURING THE PAST DECADE, the desire to share information on archival holdings in an automated environment emerged near the top of the archival profession's agen- da. Although there were a variety of op- tions, many repositories determined par- ticipation in bibliographic utilities to be the most viable method for information exchange. David Bearman's 1981 report on the work of NISTF (National Infor- mation Systems Task Force) outlined more than a dozen possible prototypes for information sharing.1 The choice to automate through bibliographic utilities represented merely one option that of- fered the archival community distinct benefits. First, it provided a way to inte- grate archival and manuscript collections with other types of material in library cat- alogs, offering repositories whose mother institutions were RLIN or OCLC mem- bers both a vehicle for greater local visi- bility and the promise of wider use. Se- cond, participation in bibliographic utili- ties offered improved local and network- wide subject access to holdings. Third, and not incidentally, bibliographic utili- ties rapidly introduced many archives to automation. Information sharing through bibliographic networks, thus, al- lowed archivists to mainstream materials into library catalogs, exchange informa- tion on holdings with other archives, and begin the automation process within re- positories. I n f o r m a t i o n sharing through bibliographic utilities involved tradeoffs, however; joining these systems meant adopting system standards, which led to some fairly predictable problems for ar- chival repositories. For example, because there were few archival descriptive stan- dards suitable for an automated environ- ment, repositories, in order to automate quickly, had little recourse but to accept the library conventions used by biblio- graphic utilities—such as Anglo- American Cataloging Rules, second edi- tion (AACR2), Library of Congress Sub- ject Headings (LCSH), and the Library of Congress Name Authority File (LCNAF). As a result, library cataloging largely supplanted customary archival de- scription in preparing the automated re- cord at this early stage of archivists' work. Resolving tensions between exist- ing local practice, the needs of a system, and the requirements of a group of users is a complicated matter. Because a num- ber of archives have participated in bibli- ographic utilities for several years, a body of data exists that permits evaluation of the status of information sharing within the profession, the implications of the use of library standards for archival retrieval, and the steps that can be taken to im- prove network access to descriptions of archival collections. The actual performance of national networks in providing access to archival and manuscript holdings has never been tested. Various elements influence searching and system-wide retrieval capa- bilities in automated bibliographic net- works, including format design, system architecture, the application of stan- dards, and the compatibility of standards with particular formats of material. The research discussed below examined the variant descriptive practices of the forty repositories inputting into the Research Library Group's (RLG) Research Library Information Network (RLIN) Archival and Manuscript Control (AMC) file in 1986. By surveying these repositories, it was possible to evaluate the extent to which lack of uniform descriptive prac- tices complicates accessing information 'David Bearman, "Towards National Information Systems for Archives and Manuscript Repositories: Al- ternative Models; First Working Paper on Scenarios for Multi-Institutional Exchange and their Implications for the Profession," unpublished, 20 August 1981. D ow nloaded from http://m eridian.allenpress.com /doi/pdf/10.17723/aarc.50.2.j889j531t25g1401 by C arnegie M ellon U niversity user on 06 A pril 2021 194 American Archivist / Spring 1987 within a data base. The survey included questions on three types of practices: (1) standardization in the choice and con- struction of headings, (2) levels of au- thority control, and (3) treatment of out- of-scope materials (i.e., holdings that are outside a repository's topical, geographi- cal, or chronological collecting policy guidelines). Thirty-six of the forty reposi- tories actively inputting into RLIN AMC responded to the survey, representing 88 percent of the records in the data base. Network environments rely on the use of standard conventions to facilitate in- formation retrieval. Because computers are unforgivingly literal, the degree to which cooperating repositories can agree or "converge" on what they put into the system directly affects what they can get out of it. Although standard conventions guide their work, archivists are far from achieving standard practice. Extreme in- consistency in describing materials is the key problem facing archival reference in the age of automation. Archivists are in- consistent in both how they describe and in what they describe. Resolving these discrepancies will profoundly affect the ability of the reference archivist to re- spond to user queries. This article will re- port inconsistencies found in the RLIN AMC data base, examine the implica- tions of these inconsistencies, and suggest methods for improving retrieval in auto- mated bibliographic networks. The first area of inconsistency appears in the assignment of topical index terms. Previous information retrieval studies have shown a positive relationship be- tween interindexer consistency and retrieval effectiveness; that is, the more convergence in cooperating repositories' choices of index terms, the greater the likelihood of retrieving relevant materials from the data base.2 In this study, inter- indexer consistency among those input- ting into RLIN AMC was measured by asking all repositories to assign topical in- dex terms to the same three descriptions of collections, using their own descriptive procedures (see Appendix A). Consisten- cy was calculated by determining the total number of different terms selected to de- scribe each collection in ratio to the num- ber of terms selected by all repositories. An unrealistically high level of conver- gence might be expected, because survey respondents performed this exercise with the equivalent of an identical card catalog description in hand, preventing many of the opportunities for divergence that arise in drafting descriptions from the be- ginning. The findings, however, contradicted initial expectations. In the first of three descriptions, the Carter family of Indi- ana, 21 indexing repositories assigned 162 different access points to this collection (see Appendix A).3 No term was assigned by all indexers, resulting in an indexing consistency rate of zero. Even when terms were reduced through truncation to their most global representation, such as WOM#N or INDIANA^ (which would result in an unwieldly number of hits if searched as such), interindexer consisten- cy still equaled zero; there was not one term, collapsed to its most generic root, that was chosen by all repositories to de- scribe this collection. The other two collections included an even more extreme bias toward interin- dexer convergence. Each repository was asked to describe one collection—either !William S. Cooper, "Is Interindexer Consistency a Hobgoblin?" American Documentation, no. 20 (July 1969): 276 'Although thirty-six repositories responded to the survey, not all repositories answered every survey ques- tion. When less than thirty-six repositories responded to a particular question, the number of responses has been supplied in the text. D ow nloaded from http://m eridian.allenpress.com /doi/pdf/10.17723/aarc.50.2.j889j531t25g1401 by C arnegie M ellon U niversity user on 06 A pril 2021 Description and Reference 195 record group or a manuscript collection —that represented its predominant hold- ings (see Appendix A). This self-selection should have resulted in overrepresented consistency, because convergence de- pended on the agreement of a few reposi- tories indexing familiar materials. Again, however, there was no consistency. The extreme lack of consistency in the assign- ment of topical index terms constitutes the major finding of this study. Survey analysis reveals that reposito- ries avoid certain kinds of topical terms, further complicating network-wide ac- cess. Terms identifying "occupations" and "conferences or meetings" tend to be used significantly less often than other points of entry. Respondents indicate that 19 percent of records are not as- signed occupation index terms even when appropriate. Twenty-two percent of re- cords are not assigned conference or meetings index terms. Although the per- centage is slightly higher for the latter, the absence of occupation terms poses the more serious problem because occupa- tion terms permit searches related to a particular activity (e.g., teaching, nurs- ing, publishing). The ability to search by activity-related terms is central to archi- val retrieval, yet survey reports indicate that search queries using occupation terms will result in a significant number of misses. Considerable nonconformity also ap- pears in the preferred level of specificity in choosing topical terms. For example, some respondents described a trip by rail to Florida as "Voyages and travel," while others termed it "Railroad travel— United States." Records documenting the sale of public lands in Indiana might be "Indiana—Government property" or "Land titles—Registration and transfer —Indiana." The eighteen repositories that entered 61 percent of the records surveyed favored narrow terms. Seven re- spondents, representing 13 percent of the records surveyed, favored broad terms. The other seven repositories, representing 25 percent of the records surveyed, used both types of terms. Although survey re- sponses indicate that repositories are much more apt to assign specific terms rather than broad terms when creating records, irregular practice in this area cre- ates further retrieval difficulties. Lack of uniformity in the choice and construction of topical index terms is compounded by a second inconsistency, inadequacies in RLIN's syndetic struc- ture. Syndetic structure is the linking mechanisms used with groups of words or phrases in an information system.4 An authority file that uses " s e e " and "see also" references is the most commonly found syndetic structure in library cata- logs. Authority control files are the key mechanism for ensuring consistency within bibliographic catalogs by distin- guishing names, showing relationships (among variant forms of names, parent bodies, and earlier to later names), and documenting decisions. Such files there- by promote consistency in the subsequent determination of relationships and identi- fication of headings.5 RLIN AMC users use the Library of Congress Name Authority File (LCNAF) for system-wide authority control for 'Ritvars Bregzis, "The Syndetic Structure of the Catalog," Authority Control: The Key to Tomorrow's Catalog, Proceedings of the 1979 Library and Information Technology Association Institutes, ed. Mary W. Ghikas (Phoenix, Az: Oryx Press, 1982), 26. 'David R. McDonald, "Data Dictionaries, Authority Control, and Online Catalogs: A New Perspective," - Journal of Academic Librarianship 11, no. 4 (1985): 219. An authority file distinguishes two persons with the same name who were alive at approximately the same time by indicating that one was a biologist and one a molecular scientist, or by providing place of birth, names of publications, or other differentiating informa- tion. Similarly, an authority file leads one from Samuel Clemens to Mark Twain or indicates that the Atomic Bomb Casualty Commission has been known by at least five other names. D ow nloaded from http://m eridian.allenpress.com /doi/pdf/10.17723/aarc.50.2.j889j531t25g1401 by C arnegie M ellon U niversity user on 06 A pril 2021 196 American Archivist / Spring 1987 personal names and corporate entries. The respondents reported searching from 0 to 100 percent of their personal and cor- porate name terms in the LCNAF, with an overall average of 90 percent. Al- though most RLIN AMC users regularly search the LCNAF, their terms are not apt to appear in the file because it con- tains names associated with published works. Only 36 percent of the index terms entered into RLIN AMC have been searched and found in the LCNAF; as a result, approximately two-thirds of re- spondents' personal and corporate name index terms have been entered into the system with no network-wide authority control. This lack of rigorous system- wide authority control significantly com- promises access capabilities. The survey results likewise indicate an alarming lack of authority control on the local level. Nearly one-half of the records input by surveyed repositories relied ex- clusively on the LCNAF and/or the RLIN AMC file for authority control of personal and corporate names. These re- positories maintained no separate local authority file, but instead used RLIN AMC in that capacity. For many reposi- tories, using RLIN AMC in this way ap- pears a reasonable accommodation to the costs of automation. Maintaining an au- thority file is extremely labor intensive.6 RLIN AMC, however, is a catalog, not an authority file. It is not designed nor in- tended to alleviate inconsistency and pro- vides no comprehensive mechanism for linking variant forms of headings, distin- guishing names, and documenting deci- sions. Reliance on a catalog for authority control ultimately harms retrieval. If two-thirds of the names entered are not screened for system-wide consistency and there is inadequate control of one-half of the respondents' RLIN AMC records, in- consistency is surely epidemic. The third area of inconsistency con- cerns the descriptive treatment of out-of- scope material. The extent to which such materials are assigned access points is sig- nificant because the ability to share infor- mation on these holdings constitutes a key benefit of national networks. Net- work members, however, stand divided in their willingness to make out-of-scope materials accessible through the AMC file. Nine repositories, representing one- half of the records surveyed, reported that their indexing of out-of-scope mate- rials was inferior to their indexing of core holdings. In many cases, these materials are insufficiently described in finding aids and thus must be reprocessed to permit better access, tremendously taxing a re- pository's available resources. Determin- ing the appropriate balance between local priorities and network demands is auto- mation's key challenge for the profes- sion. With respect to out-of-scope mate- rials, survey results indicate that RLIN AMC repositories currently tend to favor local priorities. Providing network-wide access to in- formation in bibliographic utilities is complicated, because the use of standard conventions has not yet produced a stan- dard practice. But to achieve effective re- trieval, data bases require convergence on what is entered. Determining the most cost-effective use of a file characterized by massive inconsistency in subject head- ings, authority control, and description of holdings presents formidable challeng- es for the reference archivist. There is, however, reason to persevere, as biblio- graphic automation offers archivists a 'One Library of Congress cataloger estimates that 50 percent of LC's descriptive catalogers' staff time is spent exclusively on name authorities. See Lucia J. Rather, "Authority Systems at the Library of Congress," Authority Control: The Key to Tomorrow's Catalog, 158. D ow nloaded from http://m eridian.allenpress.com /doi/pdf/10.17723/aarc.50.2.j889j531t25g1401 by C arnegie M ellon U niversity user on 06 A pril 2021 Description and Reference 197 significant tool and an extraordinary cat- alyst for professional growth and devel- opment. The strategies adopted to im- prove data input and retrieval must be well researched and wisely selected. Based on this study, five corrective ap- proaches are recommended, some quick- ly accomplished and others long-term in nature. First, measures should be taken to up- grade indexing. This is not to suggest that all archivists require basic instruction in subject cataloging. Indeed, repositories usually chose very appropriate, although different, terms in completing the survey. Although the zero consistency among this survey's respondents is extreme, all stud- ies on retrieval have found some inconsis- tency.7 Retrieval experiments report con- siderable disparity even when the same indexer performs an identical exercise at two different times.8 Thus, information retrieval scientists have concluded that substantial interindexer inconsistency forms the rule rather than the exception.9 Information scientists hesitate to gen- eralize about rates of consistency that can be reasonably attained. They do, how- ever, point to factors such as the degree of vocabulary control, the subject of ma- terials indexed, and the conditions under which indexing is performed, as elements that affect indexing quality.10 A consis- tency rate of zero is unacceptable, how- ever, regardless of limitations. Yet con- centrating the profession's efforts exclu- sively on upgrading indexing is no solu- tion. Archivists can best address this pro- blem by creating network users' groups whose purpose is to determine common use, promote adherence to conventions, provide needed training, and monitor participants' practice. Agreeing to agree is the prerequisite to achieving consisten- cy. As a second step, the inconsistency re- lated to the use of LCSH, an orderless controlled vocabulary, must be reduced. The library community has criticized LC subject headings for many years, citing problems with terminology, form and structure, complexity and size, currency and prejudices, and the use of Cutter's rule of specific entry.'' Despite its adop- tion of rules to guide subject term selec- tion, Library of Congress practice has wavered seriously on adherence to estab- lished principles.12 LCSH were adopted for use with automated systems in spite of a lack of consensus within the library community. As one observer remarked, using LC subject headings for computer access with MARC makes MARC "rath- er like a modern jet plane powered by a late nineteenth-century model of a steam engine; the thing might possibly move or even fly, but it will soon be prone to acci- dents, unreliable, and above all, the 'David C. Blair, "Indeterminacy in the Subject Access to Documentation," Information Processing and Management 22, no. 2 (1986): 230. 'Frances I. Hurwitz, " A Study of Indexer Consistency," American Documentation, no. 20 (January 1969): 92-93. 'See for instance William S. Cooper, "Is Interindexer Consistency a Hobgoblin?" 268; Pranas Zunde and Margaret E. Dexter, "Indexing Consistency and Quality," American Documentation, no. 20 (July 1969): 259; Blair, "Indeterminacy in the Subject Access to Documentation," 220; and Michael R. Middleton, " A Com- parison of Indexing Consistency and Coverage in the AEI, ERIC and APAIS Data Bases," Behavioral and Social Sciences Librarian 3 (Summer 1984): 140. '"Middleton, "Comparison of Indexing Consistency and Coverage," 140. 1' Pauline A Cochrane, Critical Views of Library of Congress Subject Headings: A Bibliographic and Biblio- metric Essay; and an Analysis of Vocabulary Control in the Library of Congress List of Subject Headings (Syracuse, N.Y.: Eric Clearinghouse on Information Resources, Syracuse University, 1981). 12See Richard S. Angell, "Library of Congress Subject Headings—Review and Forecast," Subject Retrieval in the Seventies: New Directions, Proceedings of an International Symposium held at the Center of Adult Education, University of Maryland, College Park, 14-15 May 1971, ed. Hans Weelisch and Thomas D. Wilson (Westport, Ct: Greenwood Publishing, 1972), 148-49 and 153; and Cochrane, Critical Views. D ow nloaded from http://m eridian.allenpress.com /doi/pdf/10.17723/aarc.50.2.j889j531t25g1401 by C arnegie M ellon U niversity user on 06 A pril 2021 198 American Archivist / Spring 1987 streamlined features of the fuselage will be wasted because of the slow speed at- tained."13 Both economic and political realities, however, make serious consideration of revamping LC subject headings unfeasi- ble. Thesaurus construction is prohibi- tively expensive. Development time for a small specialized thesaurus can require up to three years. Studies estimate that a small-scale thesaurus of only 2,500 terms would require two full-time staff six months to complete.14 The creation of a subject-controlled vocabulary or thesau- rus designed for use with bibliographic utilities and appropriate to archival and manuscript materials would entail thou- sands of terms and call for the input of countless subject specialists as well as the development of a sophisticated mecha- nism for continual administration, up- date, and change. Further, nearly all na- tional, academic, and public libraries in the United States, and many such librar- ies abroad use LCSH.15 The overhaul of LCSH would require either complete ar- chival detachment from the library com- munity or a commitment from innumera- ble archivists, librarians, administrators, resource allocators, and the bibliographic utilities to transform LCSH. Archivists are not in a position to inaugurate either change. Topical term selection, while currently inconsistent, is nevertheless not random; some terms are chosen more than others. Providing on-line access to LC subject headings with a running count of each heading's use within AMC would pro- mote greater consistency in term selec- tion. Archivists could use this file when creating index terms and, where choices exist, select the heading most often used in the data base. Because archival term selection is original and seldom deriva- tive, archivists have a greater need than librarians to know the extent to which headings are used within a system. Incor- porating information on heading-use pat- terns into the term selection process ulti- mately should lead to greater conver- gence. RLG plans to provide online ac- cess to LCSH in RLIN, but current plans do not include provision for reporting on the use of terms.16 Augmenting the syndetic support avail- able on RLIN AMC offers a third way to improve access. The authority file pro- vides the foundation of the automated li- brary system.17 It is the primary tool used throughout the data processing industry to maintain consistency within data bases.18 The need for authority control when automating in a cooperative net- work has been firmly documented. While archivists might argue about the high costs of implementing authority control, they cannot ignore the greater costs asso- ciated with excessive searching or failed retrieval. Archival participation in the Library of Congress's Name Authority Cooperative (NACO) offers a beginning. Qualified RLIN AMC users will soon be able to contribute to the LC Name Authority File through NACO, which permits li- braries throughout the country " t o pro- vide their own local name authority data to be included in the LC automated name authority file and made available as a 13Hans Wellisch, "Subject Retrieval in the Seventies—Methods, Problems, Prospects," Subject Retrieval in the Seventies: New Directions, 15. '"State Historical Society of Wisconsin Archives Division MSAGP Subject Access Position paper, unpub- lished, ca. 1980, 5. "Wellisch, "Subject Retrieval in the Seventies," 16. "Conversation with Ed Glazier of RLG, 28 July 1986. "Initial Considerations for a Nationwide Database, prepared by Edwin J. Buchinski, ed. and revd. Henriette D. Avram and Sally McCallum, Network Planning Paper (Washington, D.C.: Library of Congress, 1978), 1. "McDonald, "Data Dictionaries," 222. D ow nloaded from http://m eridian.allenpress.com /doi/pdf/10.17723/aarc.50.2.j889j531t25g1401 by C arnegie M ellon U niversity user on 06 A pril 2021 Description and Reference 199 whole."19 Attention also must be devoted to developing local authority files. In the age of automation, providing access to materials necessarily includes the cost of developing substantial syndetic systems.20 As a fourth route to greater access, ar- chivists should direct resources to devel- oping an archival science of searching. Some may consider the term "science of searching" an overstatement. Results of this survey suggest, however, that suc- cessful retrieval of primary source mate- rials from bibliographic utilities requires systematic data gathering, analysis, and testing. Additional research especially is needed because the process of providing access in RLIN AMC to primary materi- als differs in three ways from that of pro- viding access to books. First, archival records describe hetero- geneous collections that require many more index terms than those used to de- scribe monographs. The average number of index terms assigned to records by the survey respondents was thirteen; the av- erage number assigned to books by the Library of Congress is 3.21 Retrieval specialists have discovered that "infor- mation systems do not scale up. That is, retrieval strategies that work well on small systems do not necessarily work well on larger systems."22 The greater number of access points created for ar- chival and manuscript collections, there- fore, significantly affects retrieval. Sec- ond, archival retrieval is complicated by less adequate authority control and the tendency toward less convergence of terms because archival cataloging is pri- marily original and seldom derivative. Consequently, greater inconsistency characterizes the file. Third, the expectations of library and archival users differ, which creates con- flicting demands on the system. The re- trieval of some relevant citations, either books or journal articles, normally satis- fies most library patrons. Scholars using primary source materials, however, are more likely to expect an exhaustive listing of the relevant collections. Complete in- formation on their topic then allows them to develop a research strategy. Two dif- ferent kinds of retrieval are involved: the library patron needs precision retrieval; the scholar, recall retrieval. Precision and recall are the most widely used measures of retrieval effectiveness. Precision asses- ses how well a system retrieves only rele- vant documents, or the probability that a retrieved document will prove relevant. Users who want a few relevant citations are served best by high-precision retriev- al. Recall, on the other hand, measures how well a system retrieves all relevant documents, or the probability that a doc- ument relevant in any degree will be re- trieved. An inverse relationship exists be- tween a system's ability to be precise and its ability to be exhaustive. This only "Suzanne L. Liggett, "The Name Authority Co-op Project at the Library of Congress," Crossroads, Pro- ceedings of the First National Conference of the Library and Information Science Association, 17-21 Septem- ber 1983, Baltimore, Md., ed. Michael Gorman (Chicago: American Library Association, ca. 1984), 121. !0Little has been written on archival authority systems. The work of David Bearman, Max Evans, and Richard Szary comprises current thinking and deserves wide reading within the profession. See Richard V. Szary, "Expanding the Role of Authority Files in the Archival Context," paper presented at the annual meet- ing of the Society of American Archivists, Austin, Texas, 1 November 1985; Max J. Evans, "Authority Con- trol: An Alternative to the Record Group Concept," American Archivist 49 (Summer 1986): 249-61; David Bearman and Richard Szary, "Beyond Authorized Headings: Authorities as Reference Files in a Multi- Disciplinary Setting," paper delivered at ARLIS/NA Conference on Authority Control, 10 February 1986. An online archival authority system is currently under development for the archives catalog of the Smithsoni- an Institution Bibliographic Information System. 2'Sally McCallum, "Evolution of Authority Control for a National Network," Authority Control: The Key to Tomorrow's Catalog, 56. "David C. Blair and M. E. Maron, "An Evaluation of Retrieval Effectiveness for a Full-Test Document- Retrieval System," Computing Practices 28 (March 1985): 298. D ow nloaded from http://m eridian.allenpress.com /doi/pdf/10.17723/aarc.50.2.j889j531t25g1401 by C arnegie M ellon U niversity user on 06 A pril 2021 200 American Archivist / Spring 1987 known " l a w " of information retrieval performance presents the key stumbling block archivists will encounter when transferring library retrieval methods to an archival setting.23 Searching strategies devised for libraries cannot necessarily serve as models for archives. Instead, concentrated research leading to the de- velopment of search strategies for pri- mary source materials is needed. As a beginning, RLIN AMC users might maintain a record of research ques- tions searched in AMC, the nature of each query (precision or recall), the search strategies employed, the search query statements, and the search results. The goal would be to collect sufficient data in order to identify successful pat- terns that might guide the searching of ar- chival data bases. Precision requests probably will be quite perfunctory; the system can readily provide some relevant items for most searches. Recall retrieval, however, is apt to be more complicated. To retrieve all relevant network material, archivists must conduct repeated search- es, using all related and synonymous terms in countless combinations until rel- evant records are no longer retrieved. Re- call searches, therefore, inevitably entail considerable computer time and retrieve many irrelevant items.24 Consequently, archivists should not hold unrealistic ex- pectations of the system. They must com- pare the most effective types of search strategies with different types of inquiries to provide cost-effective, efficient service to users. Retrieval research should result in a set of model searches or prototypes that are particularly effective for archival materials. By identifying searching as es- sential to enhancing retrieval, archivists will build on the existing strength of the reference archivist as expert intermediary between users and materials. Fifth, to improve access to archival and manuscript collections in biblio- graphic utilities, archivists must come to terms with the treatment of out-of-scope materials. Research has shown that RLIN AMC members usually did not adequate- ly describe these holdings. The benefits of participation in a national network will only increase through cooperation in this area. Network participants should agree to share full information on out-of-scope materials processed in the future, and they should seek outside funding to up- grade access to those out-of-scope mate- rials already entered into the data base in order to correct the existing file. In conclusion, the research discussed above has shown (1) that the use of stan- dard conventions has not yet produced a standard practice among archivists; (2) that the lack of consistency in archival practice impedes the ability to access in- formation; (3) that attaining consistency within bibliographic utilities will be diffi- cult to achieve; and (4) that the resolution of these problems requires research, allo- cation of resources, and a willingness to balance local priorities with those of the network. Automating through bibliographic utilities entails numerous compromises for archival and manuscript repositories. An awareness of the intrinsic limitations should temper unrealistic expectations of these systems. Adopting library stan- dards used with bibliographic utilities cre- ates problems in providing access to ar- chival holdings. But automating through this route also offers advantages. In addi- tion to mainstreaming primary source materials into the library research com- "See Karen Sparck Jones, ed., Information Retrieval Experiment (London: Butterworth and Co., 1981), 2; M.H. Heine, "The Inverse Relationship of Recall and Precision in Terras of the Swets' Model," Journal of Documentation, no. 29 (1973): 81; and Cyril W. Cleverdon, "On the Inverse Relationship of Recall and Pre- cision," Journal of Documentation, no. 28 (1972): 199. "Elizabeth D. Barraclough, "Opportunities for Testing with Online Systems," Information Retrieval Ex- periment, 129. D ow nloaded from http://m eridian.allenpress.com /doi/pdf/10.17723/aarc.50.2.j889j531t25g1401 by C arnegie M ellon U niversity user on 06 A pril 2021 Description and Reference 201 munity and enhancing access to archival tion. The knowledge gained from partici- holdings, this process can provide an ef- pation in this process will prepare ar- fective way to gain needed education in chivists to undertake the tasks needed to information systems, challenge archivists create a new generation of information to begin transforming internal practices, systems more authentic to archival re- and encourage the development of a trieval. body of users expert in archival automa- Appendix A: Questionnaire and terms selected by surveyed repositories to describe collections. Below you will find three hypothetical descriptions of collections. Assuming the infor- mation is complete, please create RLIN AMC records, on paper, for 2 of these collec- tions, following the internal conventions of your repository. Please send me descrip- tions formatted on your own worksheets or whatever your repository normally uses to create a record. For the purposes of this exercise, complete only the bibliographic fixed and variable fields, not the archival control segment, and do not conduct any authority work on personal or corporate names (the information supplied is assumed to be cor- rect). Be sure to create index terms (6XX & 7XX fields) in accordance with standard practice within your repository. Assume, however, that your repository is located in the state of Indiana, and has state history as its collecting theme. Question: To be Completed by all Repositories 1. Carter family of Indiana 3 linear feet Papers, 1815-1950 and 1967, of the Carter family of Muncie and South Bend, Indiana. Contain papers of Mark Carter, Muncie businessman and postmaster, concerning banking, milling, and railroads, and includes letters from his son Leonard, concerning his studies in the 1840s at Indiana University and Brown University, travels in Europe, and his participation in the Dred Scott slavery case; papers of his wife, Rose Vaill Carter, local teacher and woman's rights advocate, relating in part to the Civil War, her interest in the cause of coeducation and suffrage for women and her involvement in the First Presbyterian Church of Muncie; papers of their son, Leonard, attorney and regent of Indiana University, 1883-1884, concerning business matters, family affairs, post- Civil War politics in Virginia, and his campaign for Indiana state supreme court justice in 1885, and his work on the board of regents, particularly as relates to the School of Medicine at Indiana University; papers of Leonard's son, Mark Carter, South Bend at- torney and Grand Master of the Knights of Templar in the United States, largely con- cerning freemasonry activities, but also including Indiana University student notebook, 1877, of course taught by Charles K. Mathews, and scrapbook, 1875-1876, of universi- ty life; papers, 1916-1932 of Leonard's daughter, Maria Carter Murray, South Bend physician, concerning her work as a settlement house reformer and on behalf of the re- productive rights of women; papers of Mark's son, Abbott Carter, concerning his inter- est in political issues, 1936-1946, as reflected in correspondence with the state's con- gressional delegation, and letters, 1967, from his grandson, Allan, concerning opposi- tion to the Vietnam war and the draft at Indiana University; and related papers of other family members, notably the Halsey family of Missouri. D ow nloaded from http://m eridian.allenpress.com /doi/pdf/10.17723/aarc.50.2.j889j531t25g1401 by C arnegie M ellon U niversity user on 06 A pril 2021 202 American Archivist / Spring 1987 Charles K. Mathews, 1835-1902 Abbott Carter, 1887-1968 Leonard Carter, 1823-1894 Maria Carter Murray, 1856-1938 Mark Carter, 1796-1882 Mark Carter, 1857-1943 Rose Vaill Carter, 1797-1876 Allan Carter, 1949- Terms assigned by repositories for the Carter family of Indiana: Abortion—Moral and ethical aspects (1) Banks and banking (3) Banks and banking—Indiana (1) Banks and banking—Indiana — Muncie (2) Banks and banking—Muncie (Ind.) (1) Banks and banking—19th century (1) Birth control (2) Birth control—Indiana — South Bend (1) Birth control—Moral and ethical aspects (1) Birth control—Law and legislation (1) Business (1) Business records (2) Businessmen (3) Businessmen—Indiana (2) Businessmen—Indiana—Muncie (1) Churches—Indiana—Muncie (1) Coeducation (4) Coeducation—Indiana (1) Coeducation—19th century (1) College students (1) College students—Indiana (1) College students—Indiana—Political activity (1) Correspondence (2) Education (1) Education, Higher—Providence (R.I.) (1) Electioneering—Indiana (1) Elections — Indiana—1885 (1) Elections and election campaigns—Indiana (1) Essays (1) Europe (2) Europe—Description and travel (5) Europe—Description and travel—1800-1918 (2) Family—Indiana (1) Family—Missouri (1) Family papers (1) Family records (1) Feminists (1) D ow nloaded from http://m eridian.allenpress.com /doi/pdf/10.17723/aarc.50.2.j889j531t25g1401 by C arnegie M ellon U niversity user on 06 A pril 2021 Description and Reference 203 Flour and feed trade (1) Freemasonry (3) Freemasonry—United States (1) Freemasons (2) Freemasons—United States (1) Indiana (2) Indiana—Commerce (1) Indiana—History (4) Indiana—History—Civil War, 1861-1865 (3) Indiana—History, local (1) Indiana—Industries (1) Indiana—Muncie (2) Indiana—Politics and government (7) Indiana—Politics and government—19th century (1) Indiana—Politics and government—1865-1950 (1) Indiana—Politics and government—1929-1938 (1) Indiana—Politics and government—1939-1945 (1) Indiana—Social conditions (1) Indiana—South Bend (2) Lawyers (4) Lawyers—Indiana (3) Letters (2) Letters—19th century—Indiana (1) Letters—20th century—Indiana (1) Medicine (1) Medicine—Indiana—South Bend (1) Medicine—Study and teaching (1) Military service, compulsory (1) Military service, compulsory—Draft resisters (1) Military service, compulsory—Public opinion (1) Military service, compulsory—United States—Draft resisters (1) Milling—19th century (1) Mills and mill-work (1) Mills and millwork—Indiana—Muncie (1) Mills and mill-work—Muncie (Ind.) (1) Missouri (2) Missouri—History—Sources (1) Muncie (Ind.) (8) Muncie (Ind.)—Churches (1) Muncie (Ind.)—Commerce (1) Muncie (Ind.)—History (2) Muncie (Ind.)—Industries (1) Muncie (Ind.)—Manufactures (1) Muncie (Ind.)—Social life and customs (1) Notebooks (4) Notebooks—19th century (1) Physicians (4) Physicians—Indiana—South Bend (1) D ow nloaded from http://m eridian.allenpress.com /doi/pdf/10.17723/aarc.50.2.j889j531t25g1401 by C arnegie M ellon U niversity user on 06 A pril 2021 204 American Archivist / Spring 1987 Political letter writing—20th century—Indiana (1) Politicians (1) Politics, Practical (1) Postal service—Indiana—Postmasters (1) Postal service—Muncie (Ind.) (1) Railroads (1) Railroads—Indiana (5) Railroads—Muncie (Ind.) (1) Railroads—19th century (1) Reconstruction—Virginia (1) Reformers (1) Scrapbooks (8) Scrapbooks—19th century (1) Settlement houses—Reform (1) Settlements, social (1) Slavery (2) Slavery—Anti-slavery movements (1) Slavery—Law and legislation—United States (1) Slavery—Legal status, laws, etc. (1) Slavery—United States (3) Slavery—United States—Law and legislation (1) Slavery—United States—Legal Status of slaves in free states (3) Slavery in the United States—Indiana (1) Slavery in the United States—Law and legislation (1) Slavery in the United States—Legal Status of slaves in free states (1) Social reformers (1) Social reformers—History—Indiana (1) Social reformers—Indiana—Muncie (1) Social science (1) Social settlements (4) Social settlements—Indiana—South Bend (2) South Bend (Ind.)—Social life and customs (1) South Bend—Social conditions (1) South Bend (Ind.)—History (2) South Bend (Ind.)—Benevolent and moral institutions and societies (1) South Bend (Ind.) (6) Student movements (1) Students—Indiana—Political activity (1) Teachers (2) Teachers—Indiana—Muncie (1) Travel (1) Travel—Europe (1) United States (1) United States—History—1849-1877 (1) United States—History—Civil War, 1861-1865 (9) United States—History—Civil War, 1861-1865—Women's work (1) United States—History—Vietnamese conflict—1961-1975—Public opinion (1) Universities and colleges—Indiana (2) D ow nloaded from http://m eridian.allenpress.com /doi/pdf/10.17723/aarc.50.2.j889j531t25g1401 by C arnegie M ellon U niversity user on 06 A pril 2021 Description and Reference 205 Universities and colleges—19th century (1) Universities and colleges—Rhode Island (1) Vietnamese conflict—1961-1975 (4) Vietnamese conflict, 1961-1975—Draft resisters—Indiana (2) Vietnamese conflict, 1961-1975—Protest movements (7) Vietnamese conflict, 1961-1975—Public opinion (1) Virginia (1) Virginia—History—1865-1950 (1) Virginia—Politics and government (4) Virginia—Politics and government—1865-1950 (3) Voyages and travel—Europe—19th century (1) Woman—Rights of women (1) Woman—Suffrage (2) Women—Education (2) Women—Suffrage (8) Women—Suffrage—19th century (1) Women—Suffrage—United States (1) Women in church work (1) Women physicians (3) Women physicians—Indiana (2) Women social reformers (1) Women teachers (1) Women's rights (10) Women's rights—19th century (2) Women's rights—20th century (1) Women's rights—Indiana (3) Women's rights—United States (1) Question: Create a Record for Either #2 or #3 Below (use whichever description most closely resembles the collections found in your repository). 2. Indiana. State Land Office 207 volumes and 3 linear feet. Records, ca. 1818-1924 and 1944-1946, of the Indiana State Land Office; contain plat and tract books containing the record of the survey and sale of public lands in Indiana, 1818-1920; surveys of lumber on state-owned land, 1890-1919; records of lands owned by the Indiana Harbor Belt Railroad, including taxes paid on these lands, 1879-1915; and record of plats and notes on the surveying of the state road between Muncie and South Bend; also, records pertaining to the policy concerning suburban development and the sale of state lands. Terms assigned by repositories for the Indiana State Land Office: Administrative agencies—Indiana (1) Cities and towns—Indiana—Growth (1) Forests and forestry—Indiana—Mensuration (1) D ow nloaded from http://m eridian.allenpress.com /doi/pdf/10.17723/aarc.50.2.j889j531t25g1401 by C arnegie M ellon U niversity user on 06 A pril 2021 206 American Archivist / Spring 1987 Indiana (1) Indiana—Forest policy (1) Indiana—Government property (1) Indiana—History (1) Indiana—Public lands (4) Indiana—Surveys (1) Land—Indiana—Taxation (1) Land grants—Indiana (2) Land subdivision—Indiana (1) Land titles—Registration and transfer—Indiana (1) Land use (1) Land use—Indiana (1) Land use—Planning (1) Maps (1) Notes (1) Patents (1) Plats (3) Plats—19th century (1) Plats—20th century (1) Public lands (1) Public records—Indiana—State Land Office (1) Real estate development (1) Real property (1) Real property, exchange of (1) Real property—Indiana (1) Real property—Indiana—Maps (1) Real property—Maps (1) Real property tax—Indiana (1) Roads—Indiana—Surveying (2) Surveying—Public lands (1) Surveying—Public lands—Indiana (1) Surveys (Land) (2) Surveys—19th century (1) Surveys—20th century (1) Tracts (1) 3. Philip Slater diaries, 1840-1847, 1854-1858, 1885, and 1887-1888. 5 volumes Farmer in Vernal Township, Monroe County, Indiana. Description of farm life in New York and his settlement in Monroe County, Indiana; also diary of his son Edwin, 1887-1888, recording farm and church activities, local affairs, the gubernatorial elec- tion of 1888, and weather; and diary of his daughter Grace describing a trip by rail to Florida. Philip Slater, 1828-1902 Edwin Slater, 1852-1935 Grace Slater, 1877-1956 D ow nloaded from http://m eridian.allenpress.com /doi/pdf/10.17723/aarc.50.2.j889j531t25g1401 by C arnegie M ellon U niversity user on 06 A pril 2021 Description and Reference 207 Terms assigned by repositories to the Slater diaries: Agriculture (1) Agriculture—Indiana (1) Agriculture—Indiana — Monroe County (1) Agriculture—New York (State) (2) Agriculture—Social aspects—Indiana—Monroe County (1) Agriculture—Social aspects—New York (State) (1) Churches—Indiana—Vernal (1) Churches—Monroe County (Ind.) (1) Diaries (6) Diaries—19th century (1) Elections—Indiana (1) Elections—Indiana—1888 (1) Family records (1) Farm life (1) Farm life—Indiana (6) Farm life—Indiana—Monroe County (1) Farm life—Monroe County—Indiana (1) Farm life—New York (State) (7) Farmers (3) Farmers—Indiana—Monroe County (2) Farmers—New York (1) Farms—Indiana—Monroe County (1) Farms—New York (State) (1) Florida—Description and travel (4) Florida—Description and travel—1865-1950 (1) Governors—Indiana (1) Indiana (1) Indiana—Governors—Election (1) Indiana—Governors—Election, 1888 (1) Indiana—History (2) Indiana—Monroe County (1) Indiana—Politics and government (3) Indiana—Politics and government—19th century (1) Indiana—Politics and government—1865-1950 (1) Indiana—Religious life and customs (1) Indiana—Social life and customs (1) Journals (1) Monroe County (Ind.) (3) Monroe County (Ind.)—Climate (1) Monroe County (Ind.)—History (3) Monroe County (Ind.)—Social conditions (1) Monroe County (Ind.)—Social life and customs (2) Monroe County (Ind.)—Vernal Township—History (1) New York (1) New York (State) (1) New York (State)—History—1865—Indiana—History (1) D ow nloaded from http://m eridian.allenpress.com /doi/pdf/10.17723/aarc.50.2.j889j531t25g1401 by C arnegie M ellon U niversity user on 06 A pril 2021 208 American Archivist / Spring 1987 Railroad travel (1) Railroad travel—United States (5) Rural families (1) United States (1) United States—Description and travel (1) United States—Description and travel—1865-1900 (1) Vernal (Ind.)—Social life and customs (2) Voyages and travel (1) Weather (1) Weather—Monroe County (Ind.) (1) Women—Diaries (1) D ow nloaded from http://m eridian.allenpress.com /doi/pdf/10.17723/aarc.50.2.j889j531t25g1401 by C arnegie M ellon U niversity user on 06 A pril 2021