CCO in Context:
bibliographical & archival standards
CCO Boot Camp
ALA, New Orleans, June 2006
(http://pages.nyu.edu/~sc3/CCOinContextNOLA.ppt)

In the next hour or so, I would like to provide some context for CCO from the viewpoint of a library cataloger who has been involved with the development of VRA Core and who has had numerous conversations with visual resources catalogers about cataloging images. At the same time, my actual practical cataloging experience is mostly with books, especially art books. Since our setting here is within ALA, I’m assuming that most of you are more familiar with bibliographic records for books and other materials collected by libraries than for cultural objects.

[slide: conceptual differences] The cardinal principle in the Anglo-American cataloguing rules tells us to “bring out all aspects of the item being described, including its content, its carrier, its type of publication, its bibliographic relationships, and whether it is published or unpublished.” That is, we catalog the work in hand. We then have general rules and specific rules based on these attributes. In North America, we have mostly followed this principle except, notably, for reproductions and when using the single-record approach wherein we use one record for a multipart or serial run in various formats or editions or for copies of closely related editions such as museum and trade editions of an exhibition publication.

Somewhat similarly, for cultural objects and their surrogates, the first question is to determine what you are cataloging. For original objects, the work is usually determined by accession. That is, if an altarpiece is considered one accession by the repository that cares for it, the “work” to be cataloged will be the altarpiece rather than one of the panels. If a repository only has one panel, the “work” will probably be the panel even if other panels from the altarpiece are known in other museums. Surrogates of these panels will probably follow the museum path. The first category in the Categories for the description of works of art is “Object/Work” and states that “the level of cataloging represented in the record should be recorded in CATALOG LEVEL.” While this sounds like the definition circles around the term being defined, it is a critical notion in cataloging not only the item in hand but also in providing the context for surrogates and parts and views of the work, as well as for other related works.

[slide: work] Though I have started using the term “work,” this is not equivalent to the “work” that we find in the Functional requirements for bibliographic records. There, the “work” is the peak of the hierarchy of Group 1 entities: work, expression, manifestation, item. We’ll get back to FRBR but I wanted now to be sure that we all understand that “work” is not the same in the cultural objects arena as in AACR and MARC. And there is also a significant difference here between the cataloging of original objects and of surrogates. A visual resources cataloger will speak of the work but rarely has the original artwork to be cataloged. Rather, the work is being cataloged in order to provide the parent record for a record for the image, often called a surrogate. The important item being cataloged is not the item in hand, but rather the item being represented. In this, VR cataloging is like our exception to AACR in the cataloging of reproductions; we know when we are cataloging a microfilm of an incunable that the important descriptive elements relate more to the pages of the book than to the measure of the microfilm though we do need that information in order to use the reproduction. With digital reproductions where we can hide that access information behind a click, the details become less important to the user. This is similar for users of image collections: just give me the slide or show me where to click so that I can see the artwork.

“Work” is also used in VR cataloging as part of “work type” to provide a broad breakdown of materials. There has been considerable discussion on VRA-L about controlled vocabulary for “work type” and the need to use a fairly limited number of terms. For example, “furniture” rather than “chair” or “Windsor chair” or “wooden chair.” Other speakers in this preconference will talk in more detail about “Work type” but the equivalent in book cataloging is not generally explicit except perhaps in the MARC leader byte that determines format, for example, books, maps, scores, sound recordings, videorecordings. These leader bytes may be amplified by data in 006 or 007 that refine the format. In AACR descriptions, the physical description may include terms that can implicitly provide for work type. These are, however, not as controlled as you probably want to do in a visual resources database.

[slide: self-describing] Before I move onto a discussion of AACR to set a context for CCO, I need to talk about one more concept that is fundamentally different in the VR cataloging universe. This is “self describing” which may need no defining but does need to be accepted before we go further. The rules in AACR talk a great deal about transcription from the title page or other source of information. Even maps have captions that may form the basis of the description. There are of course those items which have elusive information and the cataloger must construct a title. Such a constructed title might be based on the content of a book or map. Once transcribed or constructed, this information is relatively stable. Electronic resources and serials are not as stable but we do have rules for handling variations over time, and the item usually provides the impetus for the title change. For example, when we are aware that the title on an electronic resource has changed, we see that change and adjust our record.

Now imagine you have a cultural object in front of you to be cataloged. Think of a landscape painting, a Dürer print, JFK’s rocking chair, a Pennsylvania Dutch crock, or a photo of an historic train station. While you yourself could describe this object, the object does not describe itself. This is partly because we are putting into words something which is non-verbal. Even if there is an inscription on the piece, it may not be considered the title. [slide: Moran] An example is this Thomas Moran drawing. The pencil inscription in the upper left is “From the top of the Great Falls, Yellowstone, 1871, TM” and this is actually used as the title on the page from Gerald Peters Gallery where I found this illustration. On the other hand, when I was at the Amon Carter Museum, one of our Thomas Moran drawings had a similar note on the drawing but it was not used for the title on the museum label. For many objects, the repository that holds the item may have all of the information that you need to describe the item. It may provide the answers to the critical questions: who, what, where, when. It may not. The critical point is that the item’s record cannot be constructed from the item itself. Cultural objects must be placed in their cultural context.

[slide: AACR] AACR represents a stage in a long development of cataloging rules for library materials. The immediate sequence of stages comes from mid-19th-century guidelines laid down by Panizzi for the British Museum but lists of library holdings go back to ancient and medieval times. AACR is morphing into a new edition to be called Resource description and access, or RDA, but the tradition basically continues. This is not the place, nor the time, to talk about RDA in depth since it is still evolving. But there is considerable effort to make those rules generalizable to non-library communities. That is, of course, the reason we are all gathered here: there is a new standard being published for another community, partly because resources are fundamentally different and need some specific guidelines. My fear is that rules which are too generalized will be useful to no one.

RDA at this point is separating rules for punctuation and abbreviation into an appendix. Readers of CCO will find many instances prescribing particular punctuation, some in variance from library cataloging rules. [slide: capitalization] For those of us accustomed to library capitalization, these examples jar a bit. But I promise this is mutual in that VR and art history colleagues aren’t especially fond of my library capitalization prejudices.

The terms “description” and “access” have been part of the name of the ALA cataloging committee that looks after AACR for more than 20 years and I think it is interesting that the title of the new edition of the rules will include those words. Before the first edition of AACR, “description” and “access” were covered by different rule books. With AACR, the rules were brought together but are in different sections of the rules. That will continue with RDA.

[slide: history of rules] Using a rule book or data content standard allows us to build similar records for the same item. As automation has changed our cataloging and the resources we catalog, the sharing of cataloging becomes more important. Anyone who has been cataloging for more than a few years knows how much impetus there is to standardize local practice so that copy cataloging can be expedited. Databases for VR collections and museum collection systems are somewhat younger than library opacs. The history is roughly the same however in that the first generation of automation generally tried to recreate the traditional records in electronic form. The museum collection system had the same basic information that had been in the registrar’s books. The slide collection may have expanded the information beyond the abbreviated information of a slide label. Nonetheless, the label continues to be important in some collections but the story is much like library catalogs. There aren’t many libraries that are still producing cards for a card catalog. There are fewer and fewer image collections producing abbreviated labels that contain the bulk of the available information about the image.

The tradition of sharing cataloging in libraries goes back at least to the late 19th century when the Library of Congress began to distribute cards. We used to edit the cards for variant editions or reuse the information for similar titles. Somewhat similarly, slide room curators would use the caption information for slides taken from books and journals, or used the data provided by a vendor. Such caption or vendor information was probably transformed differently in each collection to fit on the slide label according to local guidelines.

[slide: classification] Another role of the slide label was to provide guidance for placing the slide in a particular drawer in the collection. VR catalogers use the term “classification” for this function. While the role is similar to the call number, the literature of VR cataloging often uses the term “classification” to mean much more than mere call number. Particularly in physical slide collections, the word “classification” substitutes for cataloging, since the placement of the slide is so fundamental to its use where at least the faculty or curators are more likely to browse than use some catalog to find items.

As long as the data needed to be shortened to fit on a label, local abbreviations evolved over time according to user dictates and other local needs. Libraries have long enjoyed more separation of our cataloging rules from the dictates of a specific scholarly community. Discussions of slide classification have often dwelled on the prejudices of a particular scholar, of reclassifying the slides in a particular area to the arrangement preferred by a new faculty member.

[slide: organization of rule book] For those of us who grew up as catalogers with AACR, the arrangement of the rules greatly ease our use of the rules. The basic descriptive chapter is followed by chapters for specific formats. Some of the chapters have not been heavily used by specialist catalogers, and a number of manuals were written after the first publication of AACR2 to cover the special needs of those format catalogers. For example, there is the manual for Graphic materials edited by Elizabeth Betz Parker and Archives, personal papers and manuscripts edited by Steven Hensen. These manuals combined rules for description and access.

[slide: aacr vs cco] AACR is not as all-encompassing as CCO. RDA follows in this path. Neither talks in depth about data structure. Because cultural objects are more inherently contextual, there is also much more coverage of the relationship of the object to its context in CCO than in AACR or RDA. Nonetheless, CCO like AACR and RDA builds on the tradition and legacy of cataloging works of art.

[slide: rda] RDA is arranged somewhat like AACR, divided into sections addressing description, relationships, and access point control. CCO’s basic arrangement is in more parts, principally because it has to address more questions. Consider how much bigger RDA would have to be if it also included MARC to say nothing of the subject cataloging manuals. Any data content standard, whether AACR, RDA or CCO, needs a structure in order to be used.

[slide: marc] Library catalogs have shown that most any records can be dumped into MARC. MARC and AACR are however deeply intertwined. Trying to fit non-AACR data into a MARC environment will involve separating data that CCO keeps together, just as trying to fit non-MARC data into a MARC opac can be problematic. Since some image collections have used MARC-based opacs for access, the mapping of descriptive and access elements has been done. Because MARC is not a data content standard, this interdependence of MARC and AACR makes for difficulties when data constructed by other data content rules doesn’t fit well into MARC. There are however several visual resource collections that do successfully use MARC opacs. Two notable examples are the National Gallery of Art and the University of California, San Diego. When you get home or to the internet cafe, search on “Ghent Altarpiece” at the National Gallery or the UCSD catalog to see how numerous views can be accommodated in a MARC environment. The National Gallery uses holdings records to describe the views, with a “view title” on each holding record. UCSD uses layered titles to accommodate different views of a single work. The Morgan Library, in conjunction with the Index of Christian Art, has also done extensive work with MARC records to build browsable manuscript surrogates but I think I’d describe the page records as dependent on the parent record for the manuscript. Maria Oldal will be speaking later in this preconference about some of the Morgan’s use of CCO.

[slide: aacr & marc] So just like AACR needed MARC, the VRA Core set of elements needed data content and data structure standards before they could really assist in the sharing of cataloging across collections. CCO is a giant step forward in providing a content standard for the Core. [slide: CCO picture] And the Core evolved as CCO was being edited. The new version of the VRA Core, called Core 4, has also evolved. In addition to changes because of CCO, attempts to provide an XML schema for Core 4 meant that elements, qualifiers, and attributes had to be clarified.

[slide: headings & attributes] In AACR, we build headings that include various qualifiers and attributes. Though MARC has been mapped to XML, attributes are often embedded in the heading. For example, a subject heading that includes chronological and geographic subdivisions absorbs the attributes and the subdivision coding tells you the nature of those subdivisions. The library cataloging environment works together, based on a set of data content, data value, and data structure standards and guidelines. CCO similarly is one part of another environment.

We are now seeing a lot of studies and projects which are working to blend access to resources from different environments. As library opacs become part of the picture, such mapping from one structure to another may be done less and replaced by indexing that can look for data values taken from different structures. That is, rather than putting image collections into the opac, I look forward to the day when my search interface will look at both the opac in MARC and the image collection in, let’s say, CCO and VRA Core.

There is much promise in some of these new interfaces to resources. We look with anticipation to such experiments as North Carolina State’s implementation of Endeca or Amazon’s A9. The Getty has published CDWA Lite which will expedite OAI data harvesting. I have concerns about the fundamental differences between data built according to different standards, but I also have dreams of access to scholarly resources being as “easy” as Google is for certain kinds of information.

CCO recommends the use of the Getty vocabularies: Art & architecture thesaurus, Union list of artist names, and the Thesaurus of geographic names, known by their initialisms AAT, ULAN and TGN. AAT and TGN were built with full hierarchies. A dream system would be able to inherit those hierarchies so that the user would have access to broader, narrower and related terms or places. Place name headings built according to AACR often have a bit of hierarchy because of a larger-place qualifier. VR catalogers adding place name access often include other parts of the hierarchy but I don’t know of anyone who has built in much more hierarchy than an AACR heading. Most of the access you’d provide using AAT or TGN however would mix pretty well with AACR/MARC headings.

[slide: motherwell] A record from the Union list of artist names or ULAN is considerably richer than one from LC/NAF built according to AACR. This is partly because of the separation of bibliographic and authority records. If one could build an opac that used all of the data in the LC/NAF authority record, the difference would not in many cases be so great. For example, here are the headings for Robert Motherwell from LC/NAF and ULAN. The LC/NAF heading doesn’t include dates though they are included in note fields on the authority record. This points out another difference between book and cultural object cataloging. We have not generally thought it necessary to explicitly give a person’s context on book records, but rather to simply provide a unique entry for a person. Slide collections might very well have used either “American” or “painting” as the primary classification for something by Motherwell.

[slide: motherwell in schema] As an example of the richness of such a heading as this in the XML schema for VRA Core 4, here is the Motherwell example exploded into the display heading and the coded version with attributes, subelements, and qualifiers.

[slide: frar] When I first heard about the development of the Functional requirements for authority records, I was hopeful that it would resolve how I could use the Motherwell heading from LC/NAF or ULAN in either a library or image collection environment and still consider myself using the specific vocabulary. FRAR is very interesting but it doesn’t go much beyond the heading that we are accustomed to in the library cataloging world.

[slide: authority info] Using the Getty vocabularies as envisioned by CCO would allow for much richer access, especially if integrated with the indexing and retrieval. LCSH does include some broader terms and some of us load LCSH into our opacs. Imagine however if the search interface could show you hits up, down and across the subject terms that are available in the fully-developed AAT hierarchies. [slide: houses by form] For example, what if you searched split-level houses but your database only had ranch houses and the retrieval could indicate that you had the related term. This might very well satisfy your immediate need.

Similarly, the hierarchies of the Thesaurus of Geographic Names could make for richer searching if records inherited the places in the hierarchy. This would probably be especially useful for nonwestern art where the national borders may not match cultural borders. [slide: Timbuktu] For example, if one were looking for objects from Timbuktu, it could be helpful to find other things from Mali. As I looked at this example, I was somewhat saddened to find that there is not a sub-continental level such as West Africa which might be helpful. Hmm, perhaps this is where there’s a place for social tagging which I’ll talk about in a little while.

When we move from the LC/Name Authority File to the LC/Subject Authority File, more context is provided, still less than the Getty vocabularies but more in SAF than NAF. [slide: naf to saf] The style of AACR is used in LCSH. For example, place qualifiers are added according to AACR2. Not all headings are placed in a hierarchy but many are, with broader-term references. For particular instances of an object such as a building or monument, there is a reference from the object type subdivided geographically. The geographic subdivision is however not to the local place name.

[slide: buildings] Buildings are an important part of the cultural objects universe. I hope you’ll forgive a bit of a digression on headings for named buildings. Churches have long been considered as corporate bodies and are established according to AACR. Many other buildings are not considered corporate bodies and are established according to the Subject cataloging manual. The subject rules call for using the latest name and this is often a significant issue for historical collections as the modern name for a building may not be what is represented in an historic photograph.

As we move beyond the world of mainstream book cataloging, we find that the words of LC/NAF and LCSH are no longer fully satisfactory. [slide: other vocabs] Some years ago, a set of guidelines were developed for genre terms for fiction and other works of literature (Guidelines on subject access to individual works of fiction, drama, etc., 2nd ed., 2000). Most of these terms are constructed much like LCSH; they may even duplicate the terminology of LCSH. Similarly, subject terminology for works of art and architecture were not fine enough for detailed access to such objects. Toni Petersen and Pat Moholt started working on the Art & Architecture Thesaurus around 1980 and it became one of the Getty vocabularies some years later.

AAT was focused on the subject and genre terminology needed for cultural materials. Scholars were consulted along with information specialists. Because it was a new language, it didn’t have a legacy of outdated terminology. I am very glad to report that LCSH is much more responsive to changes in terminology these days, and many of us contribute new terms and suggest revisions to existing terms.

More important than the lack of terms, however, was the practice of cataloging at the book level. I remember an early comparison of LCSH and AAT wherein the AAT terms were applied at a chapter level and compared to the LCSH applied to the book as a whole. While this was an unfair comparison, it showed how LCSH’s vocabulary was intended for cataloging books.

AAT is constructed in facets such as Materials, Objects, Agents, and Styles and Periods. The houses example a bit ago was from the Objects facet. Within each facet, terms are arranged in hierarchies. There are guide terms which are not meant to be used for indexing but for arranging the terms. [slide: AAT example] Most terms have both a noun form and an alternate form which is usually an adjective. AAT does not have extensive guidelines for putting the terms into subject phrases, or modified descriptors. Most applications that I know of have either adopted the terms and placed them in discrete fields or have built subject phrases without faceting. There are a couple reasons for this: though there is a MARC field for faceted topical headings, it has not been implemented in a sophisticated way by library systems; secondly, if a term includes a qualifier, it makes for an awkward or incomplete phrase. For example, if one wanted to use “architectural” as a modifier for “architectural schools,” the qualifier for architecture as discipline or object would be awkward if placed in the phrase. Because “architectural” is not listed as an alternate for the object genre, it should not officially be used in a subject phrase.

We have already talked about the Union list of artist names. Another Getty vocabulary, it is richly focused on art and architecture names. It however doesn’t include all of the names that you are likely to need in your cataloging of cultural objects. If your database includes historical figures as subjects of works of art or as patrons, they may not be in ULAN. Names are likely to travel pretty well from one authority file to another.

[slide: mapping] For subject and genre terms, using multiple vocabularies does present a range of problems but using more than one vocabulary can be valuable in that the access can be richer and more focused, as well as more representative of a discipline’s terminology. Because some vocabularies will contradict each other, results can be confusing to the user of the catalog when you have hits under a term in one vocabulary and a reference in another vocabulary. For example, LCSH uses “Motion picture plays” and AAT uses “screenplays,” with references in each vocabulary to the other form. If your user can search across subject vocabularies and you include references in your catalog, the references would appear to be contradictory. Searches limited to one vocabulary would, of course, yield more visible clarity.

If you try to map between subject languages, you may lose the benefit of the hierarchy. For example, many AAT terms will have LCSH equivalents but if you map the AAT to LCSH, it will not be possible to keep the hierarchies.

As much as I have loved working with controlled vocabularies, I also think there is much promise in social tagging. A number of futurist thinkers on this topic, such as Clay Shirky, have looked at how the community grows toward a consensus on the right terms to apply to an object. This happens in some formal settings as the STEVE project at the Metropolitan Museum of Art and also in the family photos on Flickr and similar shared picture sites. Collaborative content building, whether data or metadata, is much in the news. Wikipedia allows quite free editing of entries though especially controversial ones such as Hitler or George W. Bush are moderated. A performer and video maker in Brooklyn, Ze Frank, has opened his plots to the web, according to an article in last week’s New York times (June 18, 2006, p. ST 1-2). Social tagging collaboration is not all that different from literary warrant, our long-standing guide in controlled vocabularies. And we are already seeing interfaces that effectively mix metadata using controlled vocabularies and social tagging, with book jackets or other related material thrown in for good measure. Examples include Amazon’s A9 or the experimental catalog at Plymouth State in New Hampshire, using Word Press blog software.

[slide: beyond books] If an AACR/MARC environment works best for printed materials, there is an additional consideration in that our library opacs are mostly constructed of records for books and serials. If we have records for archival collections, they may have links to finding aids. What we don’t usually mix into our opac are entries from periodical databases, item-level descriptions of archival collections, indexing for the chapters of books or the maps within an atlas, and other contents. The users don’t necessarily understand this without instruction, but then they didn’t understand the difference when we were dealing with card catalogs and printed indexes. The web, yet again, provides considerable opportunity as well as challenges. We are all getting used to a wide variety of granularity when we look at responses to a web browser search. The ease of clicking into a resource and back out again makes this variety mostly tolerable.

[slide: other cat doc] So we have a variety of cataloging guidelines that we use in our collections. Adding CCO to the mix will allow further blending. Because the editors of CCO were informed by the various guidelines developed over the years by library colleagues and others, it should allow our savvy users and cross-format search engines to find reasonable responses.

[slide: amc] Archival and manuscript cataloging is a step away from cataloging of printed materials. Though MARC was developed partly to build databases for shared cataloging, the value of searching across collections was apparent in the area of unique materials too. The MARC AMC or archival and manuscript control format was developed in conjunction with APPM or the Hensen manual in order to get archival records into RLIN and OCLC, for resource discovery more than for shared cataloging. In theory, shared cataloging is impossible for archives and manuscripts though they are sometimes reproduced for wider distribution.

Early on, I talked about self-describing. Though archival collections don’t have a title page (usually), they are often verbal and the descriptive conventions provide a standard way to build records. [slide: ead] Item-level description is often provided through a finding aid, and the Encoded Archival Description or EAD has evolved into a strong tool for providing deep access to collections of materials. It is however really aimed at collections with strong support for hierarchical access. This can be very helpful in that access can be inherited from other parts of the hierarchy. Similarly, CCO is very aware of hierarchy with the relationship between groups and works, between works and images, between built complexes and individual buildings. While this difference in granularity may bear some similarity to books and chapters, or to serials and indexes, the expectation in a collection of cultural objects is that retrieval will include both results. I look forward to the day when our catalogs can build sophisticated and knowing access to materials at a variety of levels for our users.

We’ve reached the point where I would like to talk about the Functional requirements for bibliographic records. [slide: frbr] FRBR has now been around for almost a decade. It builds on earlier discussions of cataloging and informs the development of RDA. The principal discussion of FRBR among catalogers has revolved around the Group 1 entities: work, expression, manifestation, and item. This model seems to make the most sense when a resource appears in editions and versions. That is, a work such as Hamlet may be in Renaissance English or modern English, it may be translated, it may be published with pictures, it will come out in different bindings. The user will be well served if our catalog can keep these editions and versions of Hamlet together. Cataloging rules can go a long way toward putting them together.

The text of FRBR is available on the web. [slide: frbr url] For more information, the website of the IFLA FRBR Review Group will provide a good start.

[slide: iterations] The value of FRBR in pulling together the editions and versions is significant. Standard MARC records, using AACR2, would not lend themselves to FRBRization on the fly. At least one library system vendor, VTLS, has developed a FRBR ability which involves building an author/title work record on top of the manifestation record. Some list discussions have expended a good deal of energy on declaring something a work or expression or manifestation. It seems to me that the real advance will come when a retrieval system can be smart enough to look at bibliographic records and construct work-clusters with related clusters of expressions and manifestations. Various experiments have been undertaken in this direction. Examples include RedLightGreen from RLG and OCLC’s FictionFinder. There is additional information on OCLC’s experiments with FRBR on their website.

Perhaps what we need is a discussion of the functional requirements of a visual resources record. [slide: frvrr] For convenience sake, I’ll call it FRVRR or fervor. Users seeking cultural objects have the same need to cluster results meaningfully. The important relationship in FRBR is whole to whole. That is, a manifestation is one iteration of an expression which, in turn, is an iteration of the work. This model may work for the relatively simple case of a digital image of a painting but it falls apart pretty soon. In a recent email conversation with Jan Eklund, the chair of the VRA Data Standards Committee, we tried to envision a work/expression/manifestation/item model for an art work and didn’t get very far.

[slide: Lewitt] For example, here is a wall drawing by Sol Lewitt. The actual work of art is the conceptual plan for the drawing, that is what you would own if you owned this work. The conceptual plan is manifest in a particular space. Some Lewitts have been reinstalled in different spaces. In your image collection, you would probably want a work record for the concept which might be expressed as a text or a sequence of drawings with indications of size and color, or both. These, in turn, would be manifest as shapes or lines on a wall. In an installation at Dia:Beacon, the Lewitt drawings are on a gallery wall between the two rooms where the actual work has been executed. Clearly, one can use the FRBR Group 1 entity words (work, expression, manifestation) and the user seeking access to this work or images thereof will want the plan clustered with views of installations.

[slide: Rembrandt print, 2nd state] Another example is this Rembrandt print of “The three trees.” This image is of the second of four states, indicating separable stages in the development and printing of the image. Prints drawn from the same state or plate will also vary. This is also somewhat similar to editions of books and the user should be able to get clustered responses. [slide: Rembrandt, 4th state] Here is the fourth state of the same Rembrandt print.

While both of these examples do, I think, work rather like FRBR, the vast majority of the clustering you want to do in a visual resources collection is that of works and parts of the work, and works and views of the work.

[slide: diptych Memling] Here are two panels by Hans Memling, portraits of a Flemish couple with a continuous background. While many of us might be perfectly happy to live in either Berlin or Paris, it is important for the study of these portraits to know that though the man’s portrait is in Berlin and the woman’s is in Paris, the panels originally started as a diptych. Because the panels are now in different collections, I imagine almost every image collection would treat them as two works with metadata to relate the panels.

[slide: Villa Rotunda and drawings] Another category of works that need to be related to their specific context are preparatory or other studies for works. Here, we have a view of the Villa Rotunda in Vicenza, designed by Andrea Palladio, with the sheet from his writings that include a plan and section of the Villa Rotunda. Our records must be able to lead our user from work to related work. The Villa Rotunda has been an important influence on other architects. [slide: Jefferson] For example, Thomas Jefferson was inspired by the Villa Rotunda when he designed his competition design for the President’s House in Washington, D.C. And of course the user also needs to move from this Jefferson design to the finished White House.

[slide: relationship builder] So as with FRBR, the fundamental requirement of any visual resources record is to supply a record that will allow collocation between related works or surrogates. In library cataloging, we often use uniform titles to help us with this function. Cultural objects, particularly those without known creators or unique titles, have the same need for collocation but don’t necessarily have the same route to a concise uniform title that can expedite clustering. Standard identifying numbers have been suggested as a solution. The international music community has developed a scheme for identifying codes for musical works and this may provide a model. Meanwhile, in an individual database, you might use your database numbers as a certain link between related works.

[slide: Rembrandt “Three trees”] Before I move on, I’d like to show one more example of a relationship that is a little less direct. Sometimes, one art work is used in another, either as inspiration or as subject. One has seen parodies of the “Mona Lisa,” for example. Here is another print of Rembrandt entitled “Three trees.” [slide: Williams] It appears in this portrait of Katherine studying the print. In the background appears a print of a Rembrandt self-portrait and yet another work. Any of these works within works might be of significance to the user of our database.

[slide: frvrr] I trust that this has convinced you that the fundamental requirements of a visual resources record are somewhat like those of FRBR but not the same. The context is much more important, not only of the work to its cultural context, but also the work to related works, the parts of a work to the whole, the view to the work. While CCO is aimed at catalogers of both original objects and their surrogates, I anticipate that the first generation of users will be mainly those in those in image collections, whether slides, photographs, or digital images.

[slide: fror] If we need a FRVRR, we probably also need a similar model for objects, a functional requirements for object records. In a museum context, the work is usually determined by the repository and, as stated earlier, is usually based on the accession needs of the museum or other repository. I’ll leave it to others to determine the specifics of FROR which was actually coined by Murtha Baca who is also participating in this preconference.

Before we move on from this portion of the preconference, I’d like to review some of the special needs of the record for a cultural object, whether it’s the object itself or a surrogate. [slide: cco] Context is more explicitly important for cultural objects. It’s in the name itself; objects belong to their culture. Tomorrow, Sara Shatford Layne will talk to us about subject access to visual materials. This is another significant area of difference between cataloging of books and of visual materials. We do not usually explicitly talk about the paper and ink in a book. Until recently, we didn’t usually add headings for the genre or form of a work. In some areas, we specifically avoid genre headings and depend on broad MARC coding for delimiting by format such as electronic. Another subject difference is iconography. If you think about the portrait diptych by Memling, one cannot say that this is really ABOUT the couple depicted; it is pictures of them. I’ll leave it to Sara to explicate this difference but I want you to remember the importance of this distinction. In book cataloging, however, we have often glossed over this difference. For example, in art books, we usually give an author entry to an artist when the book includes illustrations, even if the author is long dead and the books includes none of his or her words.

One other function that your VR records might have to fulfill is related to use of images in a course or for other educational purposes. In one of the image databases at NYU, the course number is added as a sort of subject heading in order to collocate the images for that course.

[slide: nytw] As I was working on this paper, a message was sent out to subscribers of the New York Theatre Workshop which questioned whether a work of art can even stand on its own. This was the setting for a panel discussion on how the historical, cultural, political and social background of the artwork informs our understanding and appreciation of that artwork.

Here is one area where FRBR does resonate for cultural objects. [slide: frbr tasks] The FRBR user tasks are: find, identify, select, obtain. These are universal and are needed widely for our information gathering. And this returns us to our fundamental concerns. How can we provide the user with the critical information that is needed for effective retrieval. How can they find and identify the needed object? For printed materials described by AACR, it’s the item in hand. For cultural objects, it’s the artwork or material object in its context. We need to provide for the user the “who, what, where, and when” that will do this.

RDA is more aware of these FRBR tasks and will move our records in the direction of creating clusters of work, expression, manifestation, and item, clustering the results of a search. [slide: frbr & aacr] I am concerned about the movement within RDA for no cataloger intervention in the basic description. While this serves the purpose of matching the item to its description, library systems may need more uniform titles or other additions to build clusters for responses. And of course it doesn’t work so well if the item is not self-describing.

CCO is a giant step forward in common standards for visual resources cataloging. [slide: sharing dreams] As it is implemented by VR collections, I anticipate that idisosyncratic practices will lessen and image catalogers will move toward common behaviors, much like book catalogers did after OCLC and RLIN became ubiquitous. Shared practices can lead to shared cataloging. This will be further expedited by the sharing of standards, particularly vocabularies and other headings. Access will be enriched by such experiments as aka, a Getty project, now terminated, which attempted to build AAT hierarchies into a search interface. And I also think our access will be enriched by community cataloging or tagging.

As our interfaces grow more flexible and universal, and our mix of controlled vocabularies and social tagging grows more useful, the future seems to hold more promise than regret. And CCO will play a big part in image cataloging that works well with other sorts of information.

At the end of the slides, I’ve given a few citations for articles and sites that you might find interesting.

[slide: bibliography]


... go to other ALA reports ...