Images are popular to digitize because they are attractive to users with varying levels of sophistication, and they may be manipulated in more ways in their digital form than in their print form, for instance by zooming in. They frequently belong to “hidden” collections of primary sources that were not easily accessible before digitization, and they give users valuable views of specific structures and places unencumbered by someone else’s interpretation, as users would find in a book. In the words of Coyle and Hillman, “these collections [of primary sources] … are not the product of the scholarly enterprise, but instead the precursor.”¹⁸ To the cataloger performing subject analysis, however, these images can be much more challenging than monographs or other library materials, precisely because of their form. Free from the interpretation a book gives its topic, an image often lacks the context that tells the cataloger what is important about it, how users may wish to search for it, or under what circumstances they may wish to find it. The problem of ambiguity is made more acute by the fact that, for an image, the cataloger is responsible for providing the only words by which this item may be recalled, which would not be the case for a full-text searchable digitized text. While this problem is somewhat mitigated in the case of postcards, which typically include a caption that provides some information about the focal point of the image, postcards remain much closer to contextless images than to interpretive monographs.

This lack of context gives rise to the question of how specific (and how general) the subject analysis should be. For instance, for a postcard of St. Peter’s Episcopal Church in Washington, N.C., should the cataloger assign a subject heading for the specific name of the church? LC practice is to assign subject headings at a level of specificity matching the content of the work being cataloged.¹⁹ But would exclusively assigning a name heading impede access for those interested in Washington churches but not knowing the name of this particular church? Should more generic headings be assigned, such as “Anglican church buildings—North Carolina—Washington,” even though such headings are normally assigned to works that discuss such churches collectively? What about users interested in Washington architecture? Should the heading “Buildings—North Carolina—Washington” also be assigned? In that case, where should one stop? On the question of specificity, Cataloging Cultural Objects (CCO) advises, “The greater the level of specificity … in catalog records, the more valuable the records will be for researchers.”²⁰ In keeping with this advice, two headings were assigned: a specific name heading and a heading for a generic category of entity: “Anglican church buildings—North Carolina—Washington.” This practice of dual (specific and generic) heading assignment is an attempt to make up for the lack of a syndetic structure of broader and narrower terms contained within LC subject authority records. That this deviation from LC practice creates a more useful repository is again supported by CCO: “If it is not possible to link to hierarchical authorities, it may be necessary for catalogers to enter both specific and generic terms in each record to allow access, which may differ from traditional bibliographic practice.”²¹

It could be argued that it is unnecessary to create specific subject headings for named entities such as St. Peter’s Church, since keyword access to the titles, or captions, of the postcards should render the desired entity findable. Many similar projects rely on such keyword access and do not include subject analysis at all, or assign only generic subject headings. It is true that many users may discover the images in Joyner Library’s repository through an external search engine such as Google, which operates exclusively on a keyword basis. The value of specific subject headings, however, is increased by their linking potential. If the metadata contains a specific heading, once users view an image of St. Peter’s Episcopal Church within the repository, they are able to click on that heading to find all the pictures of that particular church in the repository, and not of any other church. If future links are to be made between the digital repository and the library catalog, the data will be consistent, and the user could find books about the church and pictures of the church in one search. Taylor emphasizes this superiority of access points in establishing explicit relationships between materials: “When relationships are merely described (e.g., mentioned in a note [or a title]) … the user is left with chance as the means for discerning related information packages that may be useful. Access points can make relationships explicit.”²² Any institution committing the time and effort to digitization projects such as this one should be concerned with providing quality metadata to make the resulting collection navigable by users. Minimization of the necessity of relying on chance for information discovery is the mark of quality metadata.

While some complexities arise in determining the correct level of specificity to which to adhere in subject analysis, the real challenge comes with the name authority work involved in assigning specific headings for a project such as this one. These images are about a particular location, such as a building, street, beach, or waterfall. The local nature of these named entities means that few of them will be found already established in the LC authority files. This requires that the cataloger either determine the correct form of the heading according to complex rules, or else forgo assigning a specific heading for a named entity. Thus the very quality that makes these images valuable to the user—their unique representation of named entities not a focus of published works—increases their complexity from the perspective of the cataloger.

Choosing the correct form of name for these entities is made more complicated because many of the headings for buildings, structures, and various types of corporate bodies fall into an acknowledged group of ambiguous entities. According to the guidelines contained in the LC Subject Cataloging Manual instruction sheet H405, some of these named entities (e.g., Banks, Cemeteries, Churches) are established in the LCNAF following name heading conventions and others (e.g., Bridges, Courthouses, Dwellings) in the Subject Authority File (LCSAF) following subject heading conventions.²³ Following one set of conventions rather than the other could result in a different heading. It has been argued that “the separation of controlled names and terms into a Name Authority File and a Subject Authority File is artificial” because it adds no value for the user and is ignored by many library systems.²⁴ The policy of separation nevertheless remains codified, and debating the pros and cons of these rules is beyond the scope of this paper.

The process of creating headings was sometimes relatively straightforward. For example, the postcard with the caption “A. C. Monk Tobacco Company, Farmville, North Carolina” (figure 2) required the creation of a corporate name heading with no representation in the LCNAF. AACR2 directs the cataloger to enter a corporate body under the name by which it is commonly identified in items issued by the body, or lacking those, from reference sources.²⁵ A search of OCLC’s WorldCat revealed no items issued by the body, necessitating consultation of reference sources. Following the NACO practice of using the item itself as a reference source when no conflict is found in WorldCat, the cataloger created the name as found on the item itself: A. C. Monk Tobacco Company.

Sometimes significantly more work was necessary to determine the correct form of the name. Take, for example, the three postcards entitled St. Peter’s Church, Washington, N.C.; St. Peter’s Protestant Episcopal Church, Washington, N.C.; and St. Peter’s Episcopal Church, Washington, N.C. No heading exists in the LCNAF for this church, so the cataloger needed to make a decision about the correct form of the name to use. A search of WorldCat revealed three books issued by the church, with varying usage: Saint Peters Parish, St. Peter’s Episcopal Church, and St. Peter’s Church, the last with a variant usage on the cover, St. Peter’s Episcopal Church. Using the predominant form of the name found in works issued by the body, St. Peter’s Episcopal Church (Washington, N.C.) was chosen.

Frequently, variant names are used for the same structure on different postcards in the collection. For example, a particular pavilion in Wrightsville Beach called variously Lumina, Greater Lumina, or Lumina Dancing Pavilion, is the subject of a number of postcards (figure 3). No appropriate heading exists in the authority file. According to the Subject Cataloging Manual H405, pavilions are established in the LCSAF following instruction sheet H 1334, “Buildings and Other Structures.” This instruction sheet states, “Enter the heading for a particular building or structure directly under its own name, in uninverted form, and qualify it by the name of the geographic entity in which the structure is located.”²⁶ But what is this structure’s “own name?” The conflict of the various names used was settled by a book about Wrightsville Beach that contained a chapter on Lumina.²⁷ There the author states that Lumina was an entertainment center that opened in Wrightsville Beach in 1905. A new center containing a movie theater opened in 1909 and was called Greater Lumina.²⁸ Following the instruction in the Subject Cataloging Manual H1334, the cataloger created the heading “Lumina (Pavilion : Wrightsville Beach, N.C.).”

The preceding examples give a sense of the complexity involved in assigning specific subject headings for named entities in a project such as this one. It could be argued that the library brought this challenge on itself by the choice of LCSH and LCNAF as the controlled vocabulary for the repository, saddling itself with requirements to follow complex rules of questionable applicability outside the context of NACO contribution. The choice of LCSH and LCNAF for items in the repository was made prior to undertaking the Eastern North Carolina Postcard Collection project and had to do with the fact that many of the first materials in the repository were digitized books. Perpetuating this choice came from a desire to make all metadata added to the repository compatible with each other in terms of controlled vocabulary. The choice of LCSH and LCNAF also makes the repository more compatible with the library catalog, increasing consistency for users and furthering the library’s goal to integrate digital collections more fully with the rest of the library collections. The use of such a widely used standard also increases potential for interoperability not only within the library’s collections but with other collections from other institutions, and makes possible the future sharing of authority data should such a step be deemed viable or useful. In fact, it is probable that, apart from a simplification of the rules for heading construction, the only solution to the challenge of complexity is found in sharing as much data as possible. If more institutions continue to undertake the digitization of material with a local focus, it is reasonable to expect a proliferation of geographic-based NACO and SACO funnels mirroring the proliferation of locally focused digitization projects.

The large number of named entities pictured in this collection presents an additional challenge. As noted above, the local nature of the named entities indicates that few of them will be found already established in the authority file. Cuccurullo warns of the impending flood of names to be established as a result of cataloging digitized collections at the image level: “The number of such headings [for buildings and structures] is likely to increase exponentially as libraries focus attention on cataloging of digital collections.”²⁹ The Eastern North Carolina Postcard Collection of 404 postcards depicts approximately 429 named entities, with some entities appearing on more than one postcard and some single postcards depicting more than one entity. Of these, 113 were associated with authority records in the LC authority file; the remaining 316 were not.

Most institutions have only a limited amount of personnel time dedicated to doing subject analysis and authority work for digital projects. As noted previously, the work for the Eastern North Carolina Postcard Collection was limited to a portion of one staff member’s time. When time and personnel are limited, it is prudent to invoke the law of diminishing returns on that time investment, a concept Younger calls “utility” in authority control.³⁰ By “utility,” Younger means that authority control efforts should be directed to those areas in which it has the most potential positive effect on user retrieval. “Categories,” she writes, “can be defined for names requiring more or less control.”³¹ Though Younger’s article is concerned mainly with personal names of authors, her thesis is applicable to named entities as subjects as well.

Despite generally espousing the principle of specificity, cataloging staff did not create all the specific headings possible for the postcard project. In an attempt to meet the challenge of high volume, a selection process was applied to pare down the number of specific headings created. This selection process was informed by Younger’s concept of utility. Those images not receiving a specific subject heading received only a generic category subject heading. In deciding whether to create a specific heading for a named entity, the cataloger tried to be as consistent as possible across a generic subject category. The decision about how to treat subject categories was guided mainly by considerations of the effect on user retrieval. User retrieval is greatly improved by specific headings when that heading adds unique information to the subject file, when there are many similar names to keep organized, or when there are many potential links to other materials about the entity through specific subject analysis. When these conditions were met, the cataloger was more likely to create specific headings for that category of entity. Other considerations included practical concerns regarding the prohibitive amount of research involved in specific heading creation. By the end of the project, 126 unique new specific headings for proper named entities had been created. The table of generic category treatment (see appendix) shows the breakdown of how generic subject categories were treated.

To make a decision about how to treat each subject category, it was necessary to create a checklist of three questions. The first question on the checklist was whether assigning the specific names in the subject category added unique information to the subject file. A generic term plus geographic qualifier has a high likelihood of resulting in exactly the same words as a generic heading with geographic subdivision. For instance, specific headings were created for hotels because most of the hotels had nongeneric names (e.g., Hotel Kennon adds the unique word “Kennon” to the subject file). Courthouses, which have generic names like Pitt County Court House (figure 4), did not receive specific headings, rather only generic heading with the nearly identical words “Courthouses—North Carolina—Pitt County.” Assigning exclusively generic headings only for courthouses actually results in more consistency because then “courthouse” is always entered as a single compound word in the subject file rather than as it appears on the postcard, where it is sometimes “courthouse” and sometimes “court house.” If it was judged that there was a positive effect for user searching through unique and more consistent information in the subject file, the cataloger was more likely to assign specific headings to a given subject category.

The second question on the checklist was whether this category was a focus of the postcard collection and of the larger library collection. That is, are there a significant number of images in this category, perhaps several images of each entity in the category, to keep organized? If so, this was a point in favor of assigning specific subject headings in that category. For example, the largest category of named entities is street names. Thirty-seven distinct named streets are depicted in the postcards, several streets featuring in as many as eleven different postcards. Specific street name headings were assigned because of the benefit perceived in linking directly to other images of the same street. Having only one image of a pier, however, the collection does not gain much in navigability by a specific name heading in this subject category. If the library owned other materials by or about the entities in that category, it was also a point in favor of specific heading assignment. For instance, because the library collection includes many histories of local churches, it was useful to assign specific headings for churches to integrate the postcard collection with the existing library collection.

Finally, were the names in this subject category actual proper names? Some names are clearly proper names (for example, St. Peter’s Episcopal Church), but what about Old Tar River Bridge? Is that the name of a bridge, or simply a description of a bridge that goes over the Tar River? The “name” is capitalized on the postcard, but that does not necessarily signify a proper name. Subject Cataloging Manual H1334 instructs catalogers, “Do not formulate a heading for a named structure that consists solely of a generic term with a geographic qualifier unless there is evidence that this is also the proper name of the structure.”³² Being certain it was a proper name would probably require prohibitively extensive research into city records, and even then it may be difficult to be positive. CCO warns, “Catalogers should never use a specific term unless they have the research, documentation, or expertise to support that use.”³³ This last item on the checklist shows how, in addition to considerations of effect on user retrieval, decisions about the treatment of particular subject categories must be guided by practical concerns regarding the amount of research involved in specific heading creation as well as a desire to avoid introducing misleading data into the repository.

When in doubt, the cataloger tended not to create specific headings because, should circumstances change (for instance, if the library should digitize a whole collection of images of university buildings), specific subject analysis could then be done on categories initially left generic. A vital complement to Younger’s concept of utility is her assertion that cataloging of monographs should be viewed as more of a dynamic, iterative process than it traditionally has been. “Authority control,” Younger writes, “requires continual evaluation of how a name fits into the larger context of the catalog.”³⁴ This iterative process is perhaps more compatible with digital repository metadata creation, in which standards are more flexible and policies more subject to continual revision than they have been in traditional monograph cataloging.

As an attempt to meet the challenge of the high volume of authority work associated with this digital project, a checklist of considerations was used to select various subject categories for which to create specific name subject headings. Nevertheless, the volume of authority work produced by such a relatively small project was significant. The exact approach would probably not be scalable to larger projects, at least not without substantial increases in committed personnel. While the use made of the checklist may differ from project to project, the process of separating the images into generic subject categories and the checklist itself remain valuable tools in meeting the challenge of the high ratio of new headings to image descriptions that is typical for these types of projects.

Though many new headings were created for named entities, no actual authority records were created because there was no structure to accommodate them in the repository database. While initially a time saver, this omission ultimately results in inefficiencies that constitute an additional challenge for the cataloger. A lack of structure for authority data is not atypical for metadata projects, where most of the emphasis is on resource description. In the non–MARC metadata context, “doing authority control” is usually used to mean merely “using controlled vocabulary.” However, this is akin to doing authority work without the authority record. In traditional bibliographic databases with authority control, the data in an authority record includes not only the authorized form of the heading but also cross-references to variants of the name—in the case of corporate bodies to earlier and later names and in the case of headings established under subject conventions—to broader or narrower terms. It also includes additional information about the entity represented by the heading, useful when determining the appropriateness of a heading to a given bibliographic record, and references to the sources of information used when establishing the heading. Lacking authority records, the library’s repository lacks these data.

According to FRAD, the conceptual model developed to accompany the Functional Requirements for Bibliographic Records (FRBR), authority data have five functions, some aiding the cataloger and some the user.³⁵ Authority data should

1. document decisions for the cataloger;
2. serve as a reference tool for the cataloger creating more descriptions;
3. control the form of access points so that a user may find all relevant items under a certain access point;
4. support access to the bibliographic file by the user, e.g., through cross-references and additional information; and
5. link bibliographic and authority records for automatic bibliographic file maintenance.³⁶

In the metadata implementation of the Eastern North Carolina Postcard Collection, only the third function listed above is fully executed. Though some workarounds partially fulfill the other four functions and mitigate the effect of the lack of an authority file, inefficiencies remain that could be eliminated by the introduction of an actual authority file. These inefficiencies are a major source of challenges for the user as well as the cataloger.

One example of a workaround addresses the functions of documenting cataloging decisions and serving as a reference tool for future cataloging. In the absence of MARC authority field 670, records in the repository include an additional note field that cites outside sources consulted in creating the description and access points for each image. For instance, in the case of the image of the Lumina Pavilion, this note field contains a citation to the history of Wrightsville Beach consulted to confirm the name of the pavilion. This note is only used for citations of outside reference sources, not simply to justify the form of the name, as would be the case for MARC field 670. Though not visible to users, these data, which could be useful to a future cataloger adding records to the repository, will not be lost. Tying this information to the image description rather than to the named entity itself, as it would be in an authority record, is not ideal. The future cataloger would have to locate that particular image and its description to use the information in the note unless the note were repeated in each description with an access point for that named entity.

FRAD states that authority data also support access to the bibliographic file by the user through cross-references and additional information.³⁷ The lack of these data in digital repositories is a clear disadvantage to the user, who may not know the exact name of an entity for which to search. A portion of this problem has been addressed by adding headings for generic subject categories in the subject analysis for each image, hopefully helping more users find what they are looking for. The inability to navigate through cross-reference structures and broader and narrower trees is a detriment, as is the inability to view additional identifying information to help the user determine whether a named entity is actually the intent of their search. These shortcomings, however, are frequently found even in databases with authority files if they are implemented imperfectly.

Automatic maintenance of headings in the bibliographic file, the final FRAD function of authority data, is impossible without linked authority records, of course. Without the ability to effect a global change by changing one authority record, the cataloger must manually update each affected bibliographic record if it is determined that a heading needs to change. There is no workaround for this. The change process can be made as easy as possible with the inclusion of a good search function that identifies all the records that need to be changed and with the ability to cut and paste the updated headings into each record, but nevertheless the process is inefficient and prone to human error.

There is no doubt that workarounds, when they are even a possibility, are responses to the limitations of the current system. From the cataloger’s and presumably the user’s perspectives, the addition of authority file functionality to the repository would be a welcome development, aiding information retrieval and making maintenance more efficient. Such functionality could be accomplished in a variety of ways, whether by the addition of an actual authority file to the database or through links to an external authority file. The latter may be preferable, since it would eliminate the need to maintain a separate authority file just for the repository. Regardless of how it is done, programming complexities would ensue. Whether the investment of time and effort would result in overall savings depends on projections for the future growth of digitization efforts at a particular institution. For a small project such as the Eastern North Carolina Postcard Collection, success in maintaining a certain level of consistency in subject analysis despite the lack of an authority file was due in major part to the concentration of subject heading assignment responsibility in one person. Success may not be scalable to larger projects, however. If digitization of local materials is a growth area for an individual library, new solutions will need to be found.

Fortunately, progress toward these new solutions is being made. While authority control has frequently been an afterthought to metadata creation, some evidence exists that this is changing. One sees increasing interest in authority data in the metadata environment. Metadata Authority Description Schema (MADS) was developed as a counterpart to MODS by the LC.³⁸ Derived from the MARC 21 authority format, MADS records could satisfy the five functions of authority data listed in FRAD.³⁹ Although it has been available since 2005, MADS has not been widely implemented nor experience with it broadly addressed in the professional literature. Practical discussions of MADS have started to appear in blogs, however.⁴⁰ Other metadata communities besides the library cataloging community also have begun to recognize the importance of authority data. The archival community, for example, has developed Encoded Archival Context (EAC) to complement the more established Encoded Archival Description (EAD) by housing authoritative data about the creators of archival collections separately from but linked to the descriptions of the collections.⁴¹ The implementation of EAC, according to Pitti, would not only enable easier collocation of materials with a particular provenance, but also facilitate the sharing of archival authority data across institutions, enable the expression of relationships between different creators, and take advantage of the unique source material held in archives to make this contextual information available.⁴² Whether use of schemas such as MADS and EAC will become widespread remains to be seen, although it seems clear that their use has the potential to address the challenge of inefficiencies created by the lack of authority data in digitization project repositories.