286 American Archivist / Vol. 52 / Summer 1989

Research Article Downloaded from http://meridian.allenpress.com/american-archivist/article-pdf/52/3/286/2747828/aarc_52_3_g562600um1063123.pdf by guest on 29 September 2021

Authority Control Issues and Prospects

DAVID BEARMAN

Abstract: Research on authority control reported in archival, library, and literature suggests that efforts to control topical subject terminology are inappro- priate and ineffective in an archival setting because researchers are unlikely to use the same terminology as that contained in the documents, and because most users value precision over recall (inclusiveness) in their searching. The author argues that archival retrieval will be enhanced by placing more emphasis on increasing the number of access points and less on achieving consistency in indexing. He describes various kinds of au- thority files and identifies several (occupation, time period, geographic coordinates, form- of-material, and function) that offer the most promise. He advocates the use of existing reference files and cooperative development of new ones, to be used not only in the traditional authority-control sense but also as valuable information resources in their own right.

About the author: David Bearman is the editor of Archives and Museum and Archives and Museum Informatics Technical Reports and is the founder of Archives & Museum Informatics, a consulting firm in Pittsburgh, Pennsylvania. He has written extensively on issues related to archival automation and archival information exchange. Authority Control 287

OVER THE PAST FEW years, authority con- on manual updating, break down and thereby trol has received increased attention in the fail to deliver on their promise.2 archival community.1 Experience in local Given the substantial intellectual, tech- automation applications and involvement nical, and administrative overhead in-

in national bibliographic has volved in authority control, it is not Downloaded from http://meridian.allenpress.com/american-archivist/article-pdf/52/3/286/2747828/aarc_52_3_g562600um1063123.pdf by guest on 29 September 2021 heightened archivists' awareness of both surprising that researchers have sought to system design and data quality issues. Au- determine whether vocabulary control thority control, as currently practiced in ar- works, how it could be made to work bet- chives and library systems, employs a ter, what alternatives exist, and if it is the controlled vocabulary to restrict what is en- best strategy to improve retrieval. Al- tered in the fields of a cataloging though the conclusions of the research lit- to values contained in external authority erature are far from uniform, two findings files. have emerged with great regularity. First, The decision to rely on authority files to vocabulary control works best when the impose consistency in the database leads to terminology of the documents and that of a series of demanding requirements. Some- the researchers are highly consistent, that where professionals must create and update is, when the collections and the use of them 3 the authority files. Indexers, catalogers, and are relatively homogeneous. Second, lim- others who assign terms for access points iting access points to authorized terms gen- must check look-up authority files to be erally results in lower recall (fewer finds) sure that their proposed terms are recog- and higher precision (fewer false finds). In nized, either as "preferred" terms or al- any case, authority control is less effective ternative values. Then the indexers may have overall than the introduction of richer lead- to do further research to determine whether in vocabularies, i.e., additional (uncon- or not the persons, places, or things in the trolled) terms that point to the appropriate 4 item or collection being cataloged are iden- controlled terms. tical to those of like name in the authority Taken alone, these conclusions should file. If valid lists are to be maintained, new discourage archivists from any further ef- terms must be confirmed by research in ex- forts to employ authority control for topical ternal sources and any conflicts must be subject access points, because neither of resolved and/or reflected in the database. the necessary situations exists for subject As a consequence of these requirements, access to archival information systems: use many systems that supposedly employ au- thority control, especially those dependent

2Joseph W. Palmer, "Subject Authority Control & Syndetic Structure: Myth & Realities: An Inquiry into 'Much of this literature is discussed, evaluated, and certain subject heading practices and some questions usefully added to by the authors published in Avra about the implications," Cataloging & Classification Michelson, ed., Archives and Authority Control, Ar- Quarterly, vol.7, no.2: 71-95; Catherine M. Thomas, chival Informatics Technical Report, vol.2, no.2 "Authority Control in Manual vs. Online Catalogs: (Summer 1988). Individual contributors are: Jackie An examination of 'See' references," Information Dooley, "An Introduction to Authority Control for Technology and Libraries 3 (December 1984): 393- Archivists," 5-18; Tom Garnett, "Development of an 8. Authority Control System for the Smithsonian Insti- 3Carol Tenopir, "Full Text database retrieval per- tution Libraries," 21-27; Marion Matters, "Authority formance," Online Review 9 (1985): 149-164; Ten- Files in an Archival Setting," 29-33; Avra Michel- opir, "Searching by Controlled Vocabulary or Free son, "Descriptive Standards and the Archival Profes- Text," Library Journal 112 (15 November 1987): 58- sion," 1-4; Richard Szary, "Technical Requirements 59. and prospectus for authority control in the SIBIS— "Katherine W. McCain, Howard D. White, and Archives Database," 41-44; Lisa Weber, "Develop- Belver C. Griffith, "Comparing retrieval performance ment of Authority Control Systems Within the Ar- in online databases," & Man- chival Profession," 35-40. agement 23 (1987): 539-553. 288 American Archivist / Summer 1989

of subject terminology, both within archi- vocabulary searches revealed differences in val collections and by researchers, is no- retrieval described only as differences, and toriously heterogeneous; and the archival recommended searching both controlled and literature has been fairly consistent in ar- uncontrolled terminology. The most influ-

guing that archival users value recall over ential of these studies reported that full text Downloaded from http://meridian.allenpress.com/american-archivist/article-pdf/52/3/286/2747828/aarc_52_3_g562600um1063123.pdf by guest on 29 September 2021 precision.5 This hypothesized failure of searches provide better recall but poorer topical subject-based authority control has precision.8 been empirically demonstrated by Avra The most comprehensive review of sub- Michelson.6 It would follow, then, that ar- ject authorities in library systems over the chivists should stop wasting their time on past twenty years, conducted by a partisan the effort to control topical subject termi- of vocabulary control, Elaine Svenonius, nology and instead should look for findings recently concluded that subject authority that can lead to more strategic approaches control isn't working.9 Svenonius assigns to vocabulary control. the blame for the failure of authority con- It is not necessary to dismiss all types of trol to a combination of what she calls "in- authority control for all archival purposes; trinsic variables," such as the quality of the research literature suggests that certain the vocabulary and the nature of the dis- kinds of authority control, if correctly im- cipline that it represents, and "external plemented, can benefit specific types of uses. variables," such as the skills of indexers Rather than emphasize the headings-man- and searchers and the criteria for retrieval agement aspects of name-authority control, evaluation. She particularly finds that the archivists should focus substantially greater problem lies in lack of distinction between efforts on building cooperative reference files types of controlled vocabularies and con- for occupation, function, geographic-co- cludes that "it would seem that the vocab- ordinate, time-period, and form-of-mate- ulary of a discipline should precede any rial terms; these will provide access to attempt to control that vocabulary for in- people, organizations, places, events, and formation retrieval."10 Her conclusions are records, respectively. not unlike the findings of Bhattacharyya, and the earlier Cranfield II experiments, When Does Authority Control Work? which demonstrated that the greater the ter- minological consistency in the field, the The first Cranfield experiments and sim- greater the benefit of vocabulary control.11 ilar large retrieval experiments of the early 1960s demonstrated that simple controlled Subject based authority control in ar- vocabularies with normalized word endings chives also fails in applications because (dropping suffixes such as ing and ed) and synonymy performed better than full vo- 7 "Karen Markey, Pauline Athcrton, and Claudia cabulary control. Subsequent studies com- Newton, "An analysis of controlled vocabulary and paring natural language and controlled free text search statements in online searches," On- line Review 4 (1980): 225-236. 9Elaine Svenonius, "Unanswered Questions in the Design of Controlled Vocabularies," Journal of the 5Mary Jo Pugh, "The Illusion of Omniscience: American Society for Information Science 37, no.5, Subject Access and the Reference Archivist," Amer- (September 1986): 331-340. ican Archivist 45 (Winter 1982): 33-44. "'Ibid., 336. 6Avra Michelson, "Description and Reference in UK. Bhattacharyya, "The Effectiveness of Natural the Age of Automation," American Archivist 50 (Spring Language in Science Indexing and Retrieval," Jour- 1987): 192-208. nal of Documentation 30 (1974): 235-254; Cyril W. 7Cyril W. Cleverdon, Report on the Testing and Cleverdon, Jack Mills, and Michael Keen, Factors Analysis of an Investigation into the Comparative Ef- Determining the Performance of Indexing Systems, 2 ficiency of Indexing Systems. (Cranfield, England: vols. (Cranfield, England: College of Aeronautics, College of Aerodynamics, October 1962). 1966). Authority Control 289

subject analysis of archival materials is ex- could dramatically improve access by user tremely problematic. Archival material does terminology.13 Together, these studies con- not have a subject per se. Archival material firm a long known information retrieval is of the activity that generates it, but sel- principle that, the more terms assigned to

dom is it consciously authored to be about a document, the better the chances of its Downloaded from http://meridian.allenpress.com/american-archivist/article-pdf/52/3/286/2747828/aarc_52_3_g562600um1063123.pdf by guest on 29 September 2021 something. For instance, the records of el- being retrieved. ementary school matriculation are often used It seems that when controlled vocabular- to answer demographic questions about im- ies augment retrieval precision, they do so migration, family size, or life-expectancy, because they increase the size of the lead- but this is not their "subject." Nor is their in vocabulary by providing more terms that subject "elementary education," although can match user queries and point to records this term is most often assigned. Their sub- in the database. Archivists should note that ject, insofar as they have one, is the reg- a vocabulary can play this role, whether or istration procedures of governmental not it is also used to "control" headings in agencies, but the data they contain will shed the database. They can capitalize on these little light on these procedures. Archival findings by emphasizing cross-references, materials are used to understand the con- broader and narrower terms, and synonymy texts of their creation, and may be ex- in the development of vocabularies and fo- ploited for the specific information they cus their attention on the construction of contain, but the perspectives brought by "front-end" systems, whether manual or users, both to the context of their creation automated, that assist users to expand their and to the data they may contain are too vocabulary and refine their search termi- diverse to support subject indexing. nology appropriately. Archivists should add If the research literature leads archivists terminology and enrich the links between away from control of subject terminology terms rather than attempt to validate terms, in favor of other access points, its lessons increase inter-indexer consistency, or en- do not end there. Archivists have typically force rules for choice of headings. But first, implemented authority control with the aim they need to determine which access points of increasing consistency in a database, be- to focus on, and what kinds of authority cause greater consistency would presum- lists are most appropriate to each. ably improve retrieval. Studies of library systems show, however, that when imple- Authority Control Issues in Archives mented, authority control doesn't improve retrieval overall as much as simply assign- User Queries and Access Points. Log- ing more terms. Ann Schabas found that ically, user queries should be the point of adding natural language from titles im- departure for defining a strategy to aug- proved recall by either PRECIS or Library ment access. Incredibly, archivists have no of Congress Subject Headings (LCSH) ac- published literature of user-query analysis cess; oddly, though, the two very different with which to begin. They do not even have approaches yielded otherwise similar re- a research literature that reports empirically sults for retrieval effectiveness.12 Anne on user expectations in use of archives. Piternick demonstrated that end-user the- Existing archival literature has been fairly sauri, rather than thesauri used by indexers, consistent in declaring that users of ar- chives value recall (inclusiveness) over

12Ann H. Schabas, "Postcoordinate retrieval: A Comparison of Two Indexing Languages," Journal 13Anne B. Piternick, "Searching vocabularies: A of the American Society for Information Science 33, developing category of online search tools," Online no.l (1982): 32-37. Review 8, no.S (1984): 441-449. 290 American Archivist / Summer 1989

precision, but no empirical evidence has thority control and then establish criteria by been offered.14 If this is true, authority which to decide whether they should, or control will not work in archives since should not, be controlled. studies regularly show that authority con- Potential Vocabularies. If we begin with trol either has no effect or decreases recall. the National Information Systems Task Downloaded from http://meridian.allenpress.com/american-archivist/article-pdf/52/3/286/2747828/aarc_52_3_g562600um1063123.pdf by guest on 29 September 2021 Until this assumption is shown empirically Force (NISTF) data dictionary and the fields to be wrong, however, personal experience of the MARC AMC format, we can readily would suggest that, with the exception of identify only fifteen fields other than topi- genealogists and biographers, archival vis- cal subjects that are: (1) likely to be searched itors are interested in seeing exemplary, by a researcher, or (2) likely to be used as rather than exhaustive, documentation. the basis of an administrative report.16 Furthermore, we know that archivists are These fields include: themselves the largest users of archives, followed closely by records creators,15 both action corporate name of whom almost certainly value precision creator event over recall, whether they are seeking a form function known item or a report of all instances of genre geographic name/ a specific kind of record. place language medium Therefore, authority control over some method of action occupation access points might well contribute to ar- personal name relator chival practice. To identify the most ap- propriate access points we should look more status closely at users' questions. What kinds of We know that vocabulary control performs terms appear in queries? How specific or best for retrievals in which high precision general are they? Do they invoke a place, is valued, as in the case of day-to-day ad- a time, a type of record, a person, or a ministrative retrievals to support activities function that generates documentation? Are such as records scheduling or appraisal. In our users seeking archival records or are this, as in most administrative retrievals, they more interested in information that finding a relevant precedent is more im- would be in our authority files rather than portant than an exhaustive search. One of in bibliographic records? How sophisti- the arguments advanced in favor of con- cated are they about the usages contem- trolling form-of-material and function vo- porary to the period they are researching? cabularies is that adopting controlled In the absence of appropriate research on vocabularies for form-of-material and func- user queries, we can only look at each po- tion will support access to records across tential archival access point and consider jurisdictions for which similar appraisal the extent to which authority control for each might improve a variety of different kinds of retrievals undertaken for different "National Information Systems Task Force, "Data purposes. Through a logical process, we Dictionary," comp. and ed. David Bearman, in Nancy Sahli, ed., MARC for Archives and Manuscripts: The need to determine what access points ap- AMC Format (Chicago: SAA, 1985). The list of can- pear to offer the greatest promise for au- didates for authority control does not differ greatly from that which would be proposed for library ma- terials, except that a librarian would immediately no- tice that it does not include title, or such absolute 14Pugh, "The Illusion of Omniscience;" Michel- identifiers as ISSN or ISBN and LC Card Number. son, "Description and Reference." Indeed, the absence of transcribed titles or unique ex- I5Paul Conway, "Research in Presidential Librar- tra-institution identifiers for unpublished materials is ies: A User Survey," The Midwestern Archivist 11 one strong argument for authority control in some other (1986): 35-56. access points. Authority Control 291

considerations will apply.17 Unfortunately, task forces of corporate entities, because the vocabularies we have for these access those names are only meaningful in their points are still underdeveloped. They need particular organizational context. Such to be enriched with substantially more lead- names cannot be used across organizations

in vocabulary and then tested. to locate similar functions or records, which Downloaded from http://meridian.allenpress.com/american-archivist/article-pdf/52/3/286/2747828/aarc_52_3_g562600um1063123.pdf by guest on 29 September 2021 Another area in which controlled vocab- correspond to the kinds of research ques- ularies might be powerful is the retrieval tions brought by users. Controlling such of names of entities—persons, corpora- terms fails, in the same way that efforts to tions, events, and geographical places—that assign concrete "levels" to organizational are unambiguously involved in the creation units fails, to support the intellectual move of records. Admittedly the names of people "from syndetic structure of records to syn- involved in creation of archival records are detic structure of entries."18 This is what less well known than the names of authors happened in the design of SPINDEX. There of publications, and lists such as the Li- is no inherent property of organizations that brary of Congress Name Authority file makes a third tier unit in one corporation barely begin to meet the needs of archi- equivalent to a third tier unit in another vists; still, a collocation of individuals who organization (or even another third tier unit are involved in the records creation process in the same organization). Nor are there is a worthy goal. We also know that re- properties in records systems that dictate searchers will often approach archives that a subdivision of one record system will seeking persons with specific biographical resemble that of another in its scope.19 characteristics, such as occupations, places Even though events, ranging from of birth, or political or religious affilia- groundbreakings to wars, are important in tions, rather than people with known names. the creation of records, events are known For this , archivists need to pay greater by many different names, and any signifi- attention to building biographical reference cant event is comprised of numerous lesser files than to aligning the form of name used events, demanding significant research. Lists for an individual through vocabulary con- of named events are not readily available. trol and headings management. Instead of using names as an access point, Corporate names are even more prob- archivists would be wise to exploit an at- lematic, because archival corporate entities tribute of events that can be represented come and go with astonishing frequency. uniformly by indexers and users: when they Collocating corporate entities is again a took place. Archivists should pay greater worthy aim, but archivists will be less well attention to chronological access, both to served by investments in authority control events authority files that use date ranges over the names of organizational sub-struc- and to records that use named time periods. tures, such as divisions, departments, and While users bring many geographical perspectives—geological, geocultural, and geolinguistic—to archives, geopolitical ter- "David Bearman, '"Who about what' or 'From minology is shared by users and records- Whence, Why and How': Intellectual Access Ap- proaches to Archives and the Implications for Na- tional Information Systems" in Archives, Automation & Access, ed. Peter Baskerville and Chad M. Gaf- field, (Victoria, BC: University of Victoria, 1986); 18Robert H. Burger, Authority Work: The Creation, David Bearman and Richard Szary, "Beyond Au- Use, Maintenance and Evaluation of Authority Rec- thorized Headings: Authorities as Reference Files in ords and Files (Littleton, CO: Libraries Unlimited, a Multi-Disciplinary Setting," in Authority Control 1985), 7. Symposium, ARLIS Occasional Papers #6, ed. Karen 19Max Evans, "Authority Control: An Alternative Muller (Tucson: Art Libraries Society of North Amer- to the Record Group Concept," American Archivist ica, 1987), 69-78. 49 (Summer 1986): 249-261. 292 American Archivist / Summer 1989

creating institutions. For terminology that structure, and presence or absence of scope differs depending on the disciplinary per- notes. The combination of these variables spective of the users, means are now being yields the following eight types of authority developed to provide graphical interfaces files: that dramatically increase the power of lead- Downloaded from http://meridian.allenpress.com/american-archivist/article-pdf/52/3/286/2747828/aarc_52_3_g562600um1063123.pdf by guest on 29 September 2021 • term lists, with and without scope notes in vocabularies to identify overlapping geo- • term lists, with and without syndetic graphical regions despite different user vo- structure cabularies.20 Coordinate-based searching, • reference files, with and without scope which lies at the heart of these graphical notes approaches, is based on defining the geo- • reference files, with and without syn- graphic coordinates of terms so that a va- detic structure riety of types of geo-terms from different disciplinary perspectives can be trans- Term lists are lists of single words or formed into coordinate expressions. For compound terms. In addition to the au- example, because the coordinates of a town thorized terms themselves, term lists may may fall within the coordinates of a valley, include data about the entities referred to researchers studying settlement of the val- when such additional information is re- ley should also retrieve items indexed by quired to uniquely identify the entity (birth the name of the town. Geographical coor- dates to prevent confusion between persons dinates also provide ways to solve prob- with the same name, for example). Library lems created by the movement of geo-names name authority files and subject headings over time.21 Overall, geographical access are term lists. and control over the coordinate expression Reference files are databases in which of geographical locations offers consider- data supplied for authorized terms goes be- able promise. yond what is required to distinguish be- tween like terms. Thus, a reference file for Types of Authorities & persons might include educational affilia- Implementations tions and degrees, honors and awards, and important life events. Reference files for Correct identification of fields that are events would name participants, discuss candidates for authority control in archives consequences, define the time of occur- is only the first step. Once fields that might rence, and identify related events. be controlled are selected, archivists need to identify the type of authority file most Scope notes are not about the entity named appropriate to each field and the most stra- by the term, but about how the term is used tegic implementation for each. by indexers. They define the conditions un- der which an authorized term may be as- Authority files are defined by three sa- signed as an access point within a given lient characteristics: type of file (term list application. or reference file), presence or absence of Lists in which the terms are arranged in an order other than one based on their 20James Ross, "Geographic Headings Online," meaning, such as alphabetical, or chrono- Cataloging & Classification Quarterly 5 (Winter 1984): logical, lack syndetic structure. Syndetic 27-43. 2IOreste Signore and Rigolctto Bartolli, "Control- structure refers to hierarchical and other as- ling Geographic Descriptions: A Case Study for His- sociative relationships between terms, re- torico-Geographical Authority," Proceedings of the flecting their linguistic significance. International Conference on Terminology Control for Museums, 22-24 September 1988, (Cambridge, Mu- Syndetically structured lists are sometimes seum Documentation Association, in press). called taxonomies or thesauri; they may or Authority Control 293

may not include scope notes and may or genre, function, and occupation, are attri- may not incorporate information that goes butes of authority files for records, corpo- beyond what is required to distinguish one rate entities, and persons. They have more authorized term from another. open-ended vocabularies than those subject

The fifteen MARC AMC fields that might to value-table control, and they do not lend Downloaded from http://meridian.allenpress.com/american-archivist/article-pdf/52/3/286/2747828/aarc_52_3_g562600um1063123.pdf by guest on 29 September 2021 provide access points do not all require the themselves to stylistic conventions. To de- same kind of authority control. Several, such cide whether users would benefit from the as action, language, medium, method of control of these elements, we need to dis- action, relator, and status, might be ade- tinguish among a variety of implementation quately controlled by relatively short term choices that determine how authority con- lists without scope notes or syndetic struc- trol will actually operate within a given ture; such lists are often called value ta- system. bles. The criterion for determining if a value Authority controls can be implemented table will suffice is whether users can in many ways, but three fundamental dis- achieve high precision searches when the tinctions may help archivists to appreciate system they are using shows them their the way implementation affects success, and choices. Generally, short lists of unambig- the why authority control is per- uous and discrete concepts are adequately ceived differently by users of different sys- controlled by value tables. tems: Other access points, such as events and • Some systems employ substitution of geographical places are best retrieved by terms and others do not. Term substi- dates and geographical coordinates which tution replaces values entered by users are controlled by conventions. Dating con- in data entry or in searching with au- ventions translate different ways of ex- thorized values from an authority file pressing the same date into a common so that only authorized values reside representation. If a searcher can translate a in the database and/or are searched. query into dating conventions, or the sys- tem can do so automatically, then the name • Some systems use pre-coordinated of the event need not be controlled in an terms while others assume post-coor- authority. Similarly, if a geographical or dination. geological term can be looked up in an au- • Some systems assume one consistent thority file that translates it into coordi- authority source; others permit mul- nates, and the database can be searched on tiple, inconsistent authorities to co- coordinates, then control over the terms used exist. to describe geographic places is exercised Term Substitution. Most archivists are by expanding the lead-in vocabulary in the best acquainted with the typical library au- authority file, rather than by restricting the thority control system, which is designed values in the database. Conventions can also for the management of pre-coordinated be useful to achieve greater consistency be- headings from a single authority source and tween indexers and users in formulation of is experienced as a substituting implemen- personal and organizational names, al- tation in data-entry mode only. In this type though name formulation conventions, such of system, the only terms permitted in au- as those dictated by AACR2, will not as- thority-controlled fields of bibliographic sure anywhere near the degree of common- records are the authorized versions. Alter- ality that dating conventions or geographical- native vocabulary is posted only in the au- coordinate-based location identifiers do. thority file. Users searching on an The remaining data elements, form, unauthorized term must know to look in the 294 American Archivist / Summer 1989

authority file to retrieve records that will on searching, the language actually in the match on authorized terms. In data entry, database is the natural language of the rec- on the other hand, the authorized term re- ord, for example, the functions of a polit- places any unauthorized term entered in the ical office as they were described at the bibliographic database. time the records were created, even if such Downloaded from http://meridian.allenpress.com/american-archivist/article-pdf/52/3/286/2747828/aarc_52_3_g562600um1063123.pdf by guest on 29 September 2021 This implementation has several disad- functions are now obsolete or subsumed by vantages. For staff describing archival rec- different functions in our contemporary ords, the same choice of term is imposed parlance. If a user employs any equivalent regardless of the term that was current at term in searching, the term used in the query the time the records were created. Thus dis- will look up its variants in the authority file tinctions that were significant to contem- and switch to any of them so as to collocate poraries are often lost. At the same time, records with like indexing concepts. Ar- researchers who are sensitive to such dif- chivists need to design search-term substi- ferences and employ terms from the period tuting authority systems in which the indexer of the records or the vernacular of the re- puts in what is in the record, the user puts gion will not automatically be shown rec- in what is in his or her head, and the system ords that satisfy their query; they must know finds all the appropriate matches and gives to look in the authority file for appropriate the user what he or she wants.23 When vo- related terms. Concepts with different cabulary control is implemented without meanings in a number of disciplines cannot data-entry term replacement, retrieval is be controlled by authority files unique to heavily dependent on the use of vocabular- specific domains because multiple and po- ies that are rich in lead-in terminology. tentially conflicting authorities are not rec- Fortunately, many designers of online li- ognized. brary catalogs are also prescribing such ro- bust front ends.24 Implementations that do not substitute terms at data entry are better suited to ar- Coordination. A similar failing of li- chives. Even though data-entry term re- brary catalogs for archives becomes appar- placement is the simplest implementation ent when we examine the pre-coordinated tactic for imposing commonality on lan- headings used in the Library of Congress guage, it violates subtleties of terminology Subject Headings (LCSH). Despite many by enforcing a "preferred" term, erasing failings exhaustively documented by li- differences in regional language and changes brarians,25 LCSH serves relatively well for in the meaning of terms over time. In the national library system, users are expected to search for books with a vocabulary from of Congress Response," Cataloging & Classification our own time. This strategy makes sense Quarterly 8: 7-19. when searching for contemporary non-fic- ^Michael Gorman in Mary W. Ghikas, ed., Au- thority Control: The Key to Tomorrow's Catalog. tion books and articles (fiction is not "sub- Proceedings of the 1979 Library Information Tech- ject" indexed). But in archives, where users nology Association Institute (Phoenix: Oryx Press, are seeking to locate evidence of precisely 1982). 24 the kinds of subtle social and cultural shifts Marcia Bates, "Subject Access in Online Cata- logs: A Design Model," Journal of the American So- reflected in the usages of the past, data- ciety for Information Science 37, no.6 (1987): 357- entry term substitution is not an acceptable 376; Charles R. Hildreth, Online Public Access Cat- strategy.22 In systems that "switch" terms alogs: The User Interface (Dublin, OH: OCLC, 1982); Walt Crawford, Patron Access: Issues for Online Cat- alogs (Boston: G. K. Hall, 1987). 25Pauline A. Cochrane, Redesign of Catalogs and Indexes for Improved Online Subject Access: Selected 22David Henige, "Library of Congress Subject Papers of Pauline A. Cochrane (Phoenix: Oryx Press, Headings: Is Euthanasia the Answer?" and "Library 1985), 475 pp.; Karen Markey and Francis Miksa, Authority Control 295

books because books are written with a the functionality to broaden and narrow consistent level of specificity and have a searches independently by facet. This ca- predominant subject matter. Thus, we may pability is essential in order to realize the find books about relatively specific topics, expectations of the architects of the newly

such as a biography of a person or a history adopted MARC field 654, designed for Downloaded from http://meridian.allenpress.com/american-archivist/article-pdf/52/3/286/2747828/aarc_52_3_g562600um1063123.pdf by guest on 29 September 2021 of a village or company, and we may find headings composed of Art and Architecture books about relatively global concepts like Thesaurus (AAT) and Medical Subject pollution, ideology, or authority control. In Headings (MeSH) terms. Such an imple- either case, a subject heading constructed mentation requires new systems functions, in the manner of Library of Congress head- however, because it is highly unlikely that ings, such as topical subject—place—time a user would otherwise invoke a term con- period, will place the book into a subject structed of precisely those facets selected category with similar materials. by an indexer without substantial assist- A single book does not contain materials ance.26 relating to the history of viticulture in Cal- Multiple, Independent, and Conflict- ifornia, damages caused by volcanoes in ing Authorities. It is possible to have sev- Turkey, and evidence of labor union pen- eral discrete authority files, reflecting sion frauds in the way that the records of different perspectives and disciplines, linked a civil court might. Also the context of the to a single field. Elsewhere, Richard Szary creation of the book and its form are not and I have advanced a number of theoret- relevant to the meaning of its text in the ical arguments in favor of such databases, way that contexts of creation and form-of- which we call Cultural Information Sys- material are relevant to understanding ar- tems, and Szary has since expanded on the chival records. In our hypothetical civil court requirements for an archival authority con- records series, one archival researcher may trol system, which supports such multiple be interested in depositions, another in the independent authorities.27 In subsequent acts of a particular judge, and a third in articles, I have discussed the problems of legal arguments. If archivists adopt pre-co- implementing multiple independent and ordinated strategies, they are likely to se- potentially conflicting authorities in a log- lect terms to describe records that will not ical design for an art historical database, match the unpredictable perspectives of and acknowledged the difficulties facing users. users of a system in which all knowledge, Implementations need to support searches and not just specialist knowledge, is sub- 28 for discrete facets of post-coordinated ject to conflict. Nevertheless, implemen- headings so that users can look for broader and narrower terminology on any dimen- sion, exploiting facilities provided by hi- 26For greater context to this debate, see the report erarchical thesauri. Users could then search of discussions following papers presented at the AAT independently on different facets in order session at the ARLIS/NA meeting in 1987 cited in to hone their search results. Even though Archival Informatics Newsletter 2, no.l (Spring 1988): 7; David Bearman and Toni Petersen, "Searching Da- some library catalogs provide for character- tabases Indexed Using the AAT," submitted to Art string searches or component-term search- Documentation. ing of pre-coordinated headings, they lack 27Bearman and Szary, "Beyond Authorized Head- ings;" Szary, "Technical Requirements and Prospec- tus." See also Richard Szary, "Design Requirements for Archival Authority Systems," (Paper delivered at the fifty-second annual meeting of the Society of "Subject Access Literature, 1986," Library Re- American Archivists, Atlanta, 2 October 1988). sources and Technical Services 31 (October 1987): 28David Bearman, "Buildings as Structures, as Art, 334-54. and as Dwellings: Data Exchange Issues in an Archi- 296 American Archivist / Summer 1989

tations supporting multiple independent By recasting such bibliocentric databases authorities are well suited to use in a setting using a relational data model, it is clear that in which users approach the records with records, persons, organizations, events, and such heterogeneous interests, and they are places should occupy separate files, with

attractive to archives because they open up defined relationships between the files, Downloaded from http://meridian.allenpress.com/american-archivist/article-pdf/52/3/286/2747828/aarc_52_3_g562600um1063123.pdf by guest on 29 September 2021 the possibility of importing authority files which support researchers whose informa- from a variety of external contexts.29 tion needs can be satisfied by data from any of these sources. Where Should We Go From Here? Because reference files contain infor- mation beyond that required for headings The shortcomings of library-system-based management, they can act in authority con- authority control for retrieval of archival trol roles without data-entry term substitu- materials should encourage us to look to tion. Each record includes all the variant other options. When we realize that we need terms that might be used in place of the to focus more on the benefits of expanded authorized name, thereby serving as the lo- lead-in vocabulary and less on achieving cus of lead-in vocabulary, and, at the same consistency of indexing, we will be at- time, each record contains numerous di- tracted to reference files instead of term mensions that can lead users out of the ref- lists and will place greater value on the erence file, to other reference files, and to benefits of authority control from the archival records. For example, a record searchers perspective as a means to identify about a person leads out to organizations alternative access points than on its value with which that individual was affiliated, for headings management. the place of the person's birth, others with Reference files enable us to extend the whom he was associated, and his publica- functions of authority control beyond head- tions and talks. By searching for records of ings management because they contain in- others with whom the subject of our query formation beyond that required to distinguish was affiliated, we may find information terms from each other. The concept of ex- about that subject that would not have been tending term lists into reference files grew indexed in even a detailed description of out of comparing the practices of catalogers the other records. involved in headings management and those The most common reference-authority of researchers compiling scholarly data- files in cultural repositories contain data bases. The sole function of the library cat- about persons or organizations. Extensive alogers' independent authority files for reference files are also created around rec- people, organizations, and geographical ords, whether accessioned as holdings, places is to control the terms in the head- scheduled for appraisal, or destroyed. Other ings of bibliographic records. Catalogers reference files include data about cultural, may build substantial value-added data- political, and intellectual events. Another bases, but users can only exploit them for contains information about spaces, whether the limited purpose of heading validation. these are buildings, archaeological sites, locales of collecting, or geological, geo- graphical, and geo-cultural places. It is tcctural Information Network," in Databases in the critical for archivists to recognize that many Humanities and Social Sciences-4, ed. Lawrence J. McCrank (Learned Information: Medford, NJ, 1989), of the terms that are used as access points 41-48. for records, such as roles, functions, and 29Carol A. Mandel, Multiple Thesauri in Online occupations, are attributes of entities other Library Bibliographic Systems. A Report prepared for than records, and that these must be linked the Library of Congress Processing Services (Wash- ington, DC: Library of Congress, 1987). to bibliographic files. Authority Control 297

The entities about which reference files Register.31 Because reference files can be are constructed are real-world things such acquired from other disciplines, especially as people and events, which are of consid- from the computerized databases of the erable interest to non-archivists as well as parent organization of the archives, em-

being useful ways to gain access to records. ploying a deep network of reference files Downloaded from http://meridian.allenpress.com/american-archivist/article-pdf/52/3/286/2747828/aarc_52_3_g562600um1063123.pdf by guest on 29 September 2021 Therefore, the reference files built by cul- for one application need not be prohibi- tural repositories are valuable databases to tively costly. It does, of course, require im- others. For example, organization history plementations that can support multiple, reference files built by archives contain in- independent, and conflicting authorities. formation about the authority, mission, It is time to implement a database of in- structure and function of organizational units dependent reference files supporting archi- that is essential to archival administration val description and information retrieval. but also represents a valuable resource both Several years ago, Richard Lytle and I de- for the organization itself and for outsiders. scribed the appropriate relationship among B.A.G. Fuller of the Mystic Seaport Mu- such files.32 Max Evans subsequently elab- seum reports: "the concept of separating orated how data about organizations could information into entity files and authority be converted into authority records that ex- files [is] immensely useful and versatile... ercise control over provenance data ele- We are using the format developed for our ments in the description record.33 For several vessel authority file as the basis for a union years now the participants in the Research list of all watercraft in American mu- Libraries Information Network have been 30 seums." In the process of providing ac- elaborating a vocabulary on functions of cess to the union list, Fuller is constructing organizations.34 It is time to put these to- a reference file on historical watercraft that gether with the appropriate retrieval sys- is a historical database in its own right. tem. Not all reference files used in archival The major reference file in this model information systems need to be created de that is still missing is that for "form-of- novo by archivists. Reference files in an material." Ever since the NISTF data dic- archival information system may be the pri- tionary introduced the concept, it has been mary databases of the discipline or organ- difficult to explain the concept of a cultural ization that created them. Archivists and abstraction corresponding to a kind of re- the builders of other cultural information cord without encountering skepticism about systems only need identify databases that whether such a property of records truly contain information which could be linked exists. Similar properties have been ad- to records, and then import such databases vanced by students of diplomatics, but until into their systems. Thus, geographical ref- recently these have not been carried for- erence files could come from a geological survey, corporate data from the SEC or the local Chamber of Commerce, data about 31David Bearman, "The National Archives and individuals from Who's Who, and data about Records Service: Policy Choices for the Next Five Years," For the Record, December 1981. the U.S. Government from the Federal 32David Bearman and Richard H. Lytle, "The Power of the Principle of Provenance," Archivaria 21 (1986): 14-27. 33Evans, "Authority Control: An Alternative." 34David Bearman, "Archives and Manuscript Con- 30B.A.G. Fuller, personal letter to author. See also trol Within Bibliographic Utilities: Challenges and "In Search of A System: A Report on Mystic Seaport Opportunities," American Archivist 52 (1989): 26- Museum's Experience as it Begins Computerization 39; Kathleen Roe and Alden Monroe, "The Role of of its Collections," Spectra 15, no.3 (Fall 1988): 12- Function in Archival Practice," (Fall 1988, unpub- 14. lished). 298 American Archivist / Summer 1989

ward into modern records.35 Recently, re- names of offices; they contain a field for search by information scientists studying SGML-like "fingerprints" and fields for electronic records has confirmed the exis- data elements typically found recorded in tence of document formalisms of the kind this type of record. Further work must be

suggested by the concept of form-of-ma- done before we can construct useful form- Downloaded from http://meridian.allenpress.com/american-archivist/article-pdf/52/3/286/2747828/aarc_52_3_g562600um1063123.pdf by guest on 29 September 2021 terial.36 of-material reference files, but what makes The underlying argument is that archi- this work exciting to archivists is that, if vists and historians employ rules for inter- we are to achieve reasonable quality re- preting the probable content of documents trievals from huge full-text databases, it is based on formal properties of records. For critical to be able to limit searches by form- instance, we expect that marriage certifi- of-material and internal structural compo- cates will provide information about the birth nents.38 place of parents, thus serving as a source Such fingerprints might also form the of information about migration patterns; and controlled vocabularies that link reference we expect military enlistment records to in- files of document types to databases of ar- clude health and mental profiles, serving chival records. Researchers knowing only as evidence of the population distribution that they are interested in smallpox vacci- of diseases and . We recognize nation might match their search on terms the form-of-material of a record quite in- in the description of school matriculation dependently of its content, thus knowing and military conscription records in the form- immediately when we see a memorandum, of-materials reference file and navigate an award, or an order form, without having across the SGML fingerprints to records in to read its contents. the archives matching the formal type. Research into the nature of documents is now demonstrating not only that "docu- Conclusions ment formalisms" (structural features of records that signal their contents to the cul- The research literature in library and in- turally attuned) exist, but also that ma- formation retrieval systems suggests what chines can be taught to distinguish between empirical research in archives has con- document types. Computers can parse doc- firmed, that authority control over topical uments for their internal components and subject terminology in archives is ineffec- "mark" them with such document-mark- tive. Other access points, including form ing languages as Standard Generalized

Markup Language (SGML), creating a sort 3s of electronic "fingerprint" of a form-of- David C. Blair, "Full-text Retrieval: Evaluation 37 and Implications," International Classification 13 material. Some of the elements of these (1986): 18-23; Jung Soon Roo, "An Evaluation of files are now becoming clear. They look a the Applicability of Ranking to Improve the Effectiveness of Full-Text Retrieval. I. On the bit like records schedules without dates or Effectiveness of Full-Text Retrieval," Journal of the American Society for Information Science 39, no.2 (March 1988): 73-78; "II. On the Effectiveness of Ranking Algorithms on Full-Text Retrieval," ibid., •"Luciana Duranti, "Diplomatics: New Uses for an 39, no. 3 (May 1988): 147-160; T. Saracevic, P. Kan- Old Science," Archivaria 28 (1989): 7-27. tor, A. Y. Chamis, and D. Trivison, "A Study of 36David M. Levy, Daniel C. Brotsky, and Kenneth Information Seeking and Retrieval: I. Background and R. Olson, "Formalizing the Figural: Aspects of a Methodology," ibid., 39, no. 3 (May 1988) 161-176; Foundation for Document Manipulation," (Systems T. Saracevic and P. Kantor, "A Study of Information Sciences Laboratory: Xerox Palo Alto Research Cen- Seeking and Retrieval: II. User, Questions, and Ef- ter, 31 August 1988, unpublished). fectiveness," ibid., 117-196; T. Saracevic and P. "Michael B. Spring, "Copymarks," (SLIS De- Kantor, "A Study of Information Seeking and Re- partment of Information Science: University of Pitts- trieval: III. Searches, Searches and Overlap," ibid., burgh, September 1988, unpublished). 197-216. Authority Control 299

and genre, function, and occupation, ap- entry-term switching, researchers would get pear to be more promising. These access the benefits of authority control without points are attributes of records, organiza- sacrificing the historical accuracy of de- tions, and people, and could be recorded scriptions. Some concrete suggestions can

in authority files for those entities if archi- be made for the structure of person, organ- Downloaded from http://meridian.allenpress.com/american-archivist/article-pdf/52/3/286/2747828/aarc_52_3_g562600um1063123.pdf by guest on 29 September 2021 vists expanded the concept of an authority ization, and records files in relational da- file from a term list to a reference file. If tabases in which each file serves as an such reference files were implemented with authority to the others. searcher-term switching rather than data-

The Papers of Andrew Johnson Volume 8, May-August 1865 Edited by Paul H. Bergeron These papers reflect Johnson's resolute attempt to begin the "restora- tion" of his native South during the first months of his presidency. 704 pages, illustrations, ISBN 0-87049-613-1, $45.00 Advice After Appomattox Letters to Andrew Johnson, 1865-1866 Special Volume No. 1 of the Papers of Andrew Johnson Edited by Brooks D. Simpson, LeRoy P. Graf, and John Muldowny 352 pages, illustrations, ISBN 0-87049-536-4, $29.95 cloth, ISBN 0-87049-549-6, $14.95 paper

The University of Tennessee Press Knoxville 37996-0325