DATE: May 30, 2002

NAME: Dealing with FRBR Expressions in MARC 21

SOURCE: Joint Steering Committee for the Revision of AACR, Format Variation Working Group

SUMMARY: This paper describes the work of the Joint Steering Committee's Format Variation Working Group to facilitate expression-level collocation in online systems, and discusses possible approaches to achieving this collocation using the MARC 21 Holdings, Bibliographic and Authority Formats.

KEYWORDS: Expression-level collocation; FRBR

RELATED: 2002-DP04 (January 2002), 91-13 (July 1991)


05/30/02 -- Made available to the MARC 21 community for discussion.

06/17/02 - Results of the MARC Advisory Committee discussion - Participants agreed that FRBR is a conceptual model of the relationships between bibliographic items in library catalogs that could be useful to both librarians and library users. There was, however, some confusion expressed over the difference between a work and an expression that participants felt should be resolved. Participants also felt that authority and holdings data should be investigated to ascertain where expression and manifestation data could be found. Although many participants expressed enthusiasm over the perceived improvements that FRBR displays may have on library catalogs, it was also acknowledged that most MARC 21 records are for single works with single manifestations. It is important that there be no adverse impact on them. Whether any refinements to any of the formats will be necessary to accommodate FRBR displays being discuss is not yet clear.

Discussion Paper 2002-DP08: Dealing with FRBR Expressions in MARC 21

I. Introduction

During discussions of Discussion Paper 2002-DP04 prior to and during the 2002 Midwinter MARBI meeting in New Orleans, the Joint Steering Committee's Format Variation Working Group expressed concern to MARBI that the approaches described in 2002-DP04 are not scalable to all situations involving multiple versions ("format variation"). At that time MARBI decided to delay further discussion on DP04, and requested that the Chair of the Format Variation Working Group report back to MARBI at ALA Annual on the group's progress. We have submitted this discussion paper in response to that request.

This paper contains a brief summary of the Working Group's current activities, followed by a comparison of several possible approaches to handling expression-level data in MARC 21. Several of these ideas have been presented elsewhere and by others (such as in DP04), and we have gathered them here to facilitate comparison.

Our work is still in progress and it is too soon for us to offer any specific recommendations for changes to MARC 21. However, we are far enough along to suggest to MARBI that a combination of different implementation solutions may be necessary in order to address adequately all situations related for format variation. We have posed several specific questions at the end of this paper related to this in an attempt to facilitate discuss in this area.

The Format Variation Working Group's work is focused upon the bibliographic entity expression, which is defined in the IFLA Functional Requirements for Bibliographic Records (FRBR) as "The intellectual or artistic realization of a work" The entity expression is not currently articulated as such in AACR, nor does it currently have a clear "home" within the MARC 21 formats. Tom Delsey's recent Functional Analysis of the MARC 21 Bibliographic and Holdings Formats illustrates the complexities that currently exist in the formats regarding the entity expression as well as the other FRBR entities of work, manifestation, and item (See Note #1). In Appendix A of his analysis, Delsey shows that while most data associated with the entity expression now appears in the bibliographic format, certain expression-level attributes (namely data frequency and regularity) currently appear in the holdings format. While the Delsey analysis does not discuss the authority format, it is clear by extrapolating the data in that study that some expression-level data (e.g. language of uniform title) now also appears in the MARC 21 Authority Format.

While on the one hand this current situation, where expression-level data is spread between the authority, bibliographic, and holdings formats, complicates discussions of how the MARC 21 formats can best facilitate use of expression-level data, it also provides us with an opportunity to look at several possible approaches. In this paper we present general descriptions of how the various MARC 21 formats may offer potential solutions to the problem of format variations. We hope that MARBI members will take this opportunity to comment on these ideas in a manner that will help to inform our future work in this area.

II. Background of Our Work

The Joint Steering Committee's Format Variation Working Group was originally charged in early 2001 to identify the practical issues associated with creating bibliographic records based upon expressions. In our original charge we were asked to create a test database that would demonstrate the feasibility of cataloging at the expression level, using as the basis of our work the report of the CC:DA Task Force on Rule 0.24. The 0.24 Task Force report outlines several possible data models for handling expression-level records (attached, as Appendix A).

The CC:DA 0.24 Task Force report presented several options for rewording AACR Rule 0.24 to deal with format variations. Of the options in the report, CC:DA recommended the Task Force's "Option C" to the JSC as one starting point for further discussions of the issues concerning format variation or multiple versions with respect to further revision of rule 0.24 (See Note #2). Specifically, Option C of the 0.24 Report was to:

  "Ignore any mere physical variation or any mere variation in distribution (i.e. any manifestation variation) in determining when to make a new record, " (See Note #3)

The 0.24 Task Force report then described five data models that could be used for handling expression-level records, and which the Format Variation Working Group then used as the starting point of our discussions. (The original models from the CC:DA 0.24 report are reproduced as Appendix A of this paper).

The Working Group began by using cataloging exercises to test the 0.24 data models. Several Working Group members had immediate difficulty determining how to assign data elements to the expression level. For example, members were concerned about how to determine (or construct) a title for the expression based only upon information from the manifestation in hand. While it is tempting to attribute such difficulty to the unfamiliarity of the task at hand, much of the real difficulty in this case was the lack of information about the expression available to the cataloger from simply examining the manifestation being cataloged.

From this exercise, Working Group members quickly concluded that it is contrary to the way most catalogers work to actually start the cataloging process by cataloging (that is, creating a bibliographic description for) an expression (See Note #4). It is also contrary to the way most library technical services departments operate to begin the cataloging process at the expression level. Rather, it is usually the acquisition of a particular manifestation (a book, an electronic resource, etc.) by a library that triggers the need to create a catalog record for that particular manifestation.

An impressive network of shared cataloging has been created over time by combining the contributions of libraries of all sizes and types, each contributing records for material in hand (i.e. manifestations) following common rules (See Note #5). These records have proven to be compatible precisely because the cataloging rules focus first on describing the material in hand on its own terms, recording titles, statements of responsibility, publication statements, series statements, all attributes of the manifestation, as found. This allows the cataloger, with a minimum of research, to produce a record that library users and library staff in other libraries can with confidence match with citations, for a variety of purposes such as acquisitions, interlibrary loans, or to identify as a basis for copy cataloging. Only once the description is created does the cataloger consider the content in an abstract way, to assign access points, including subject access.

Although the FRBR entities manifestation and item are concrete, the entities work and expression are often only discovered by a process of extrapolation based upon comparing similar manifestations. The boundary of the expression can be open to interpretation, and as more manifestations of a work are created over time, the understanding of their relationships can change, possibly requiring a reassessment of which manifestations belong to the same expression. In a situation where the catalog record is based on identifying the expression, recording the data pertinent to (i.e. "cataloging) a new manifestation would require revisiting and revising the existing catalog record for the expression. This is a more complex--and potentially more expensive--process than our current practice of creating a new record for the manifestation.

Delsey's functional analysis helps to elucidate another problem that the Working Group members found last year in attempting to "catalog an expression." In Appendix B of his report, Delsey maps the attributes of an expression to the MARC 21 Bibliographic and Holdings fields that are defined to contain that data. If one compares that table to Table 1.2 in Appendix D in his report (which lists these attributes) it is clear that of the 30 expression-level attributes defined in FRBR, exactly half of them do not have a specific MARC 21 tag defined to contain them. Some additional expression-level attributes may be included in a catalog record in non-specific textual fields (such as general notes). However, the lack of specific tags for so many of these data elements in the MARC 21 formats confirms our experience that the traditional cataloging process has generally not been focused upon recording expression-level attributes.

The Working Group concluded that while thinking about cataloging at the expression-level may be intuitive for catalogers in some specific settings (such as archives), it is neither logical nor practical as a starting point for most library cataloging. It may be a more viable option in archival collections, particularly when providing access to collections of unique materials. Based upon this, we then rejected those options in the CC:DA 0.24 report (Options 1-4) that describe the creation of expression-based, instead of manifestation-based, catalog records.

While rejecting the general notion of cataloging an expression, the Working Group nevertheless affirmed the value of collocation at the expression level in catalog displays in order to provide additional guidance to the users of our catalogs. In our first Interim Report to the JSC, (www.nlc-bnc.ca/jsc/forvarwg3.pdf) we recommended that our Charge be altered to focus upon the latter: facilitating expression-level collocation of separate manifestation-level records. This approach is represented by Option 5 of the original data models in the 0.24 Working Group, and also is mentioned by the Library of Congress in its action plan for bibliographic control of web resources:

2.4 Define functional requirements for systems that can manage separate records for related manifestations at the global level and consolidate them for display at the local level. Communicate the requirements to the vendor community and encourage their adoption.

In our initial interim report, we also described a possible table structure that could be used to link manifestation records at the expression level. We have now further refined the table structure model and, in the options discussed below, have tried to relate it to the existing MARC 21 formats.

During the past six months, under a new charge from the JSC (http://www.nlc-bnc.ca/jsc/forvarwg1rev.html), we have been working on specific changes to AACR to accommodate expression-level collocation. As a part of this, we have been considering how best this collocation might be achieved in the MARC 21 world. This has necessarily taken us away from thinking about how a single record might be displayed and toward thinking about how relationships between records can be displayed in an intelligible way in index displays.

The following is a summary of how various MARC 21-based solutions might facilitate the collocation of expression-level data. We have deliberately not included actual examples for any of these situations for three reasons:

III. The Single Record Approach: Double-Duty Bibliographic Records

The underlying goal in this approach is to reduce the creation of separate bibliographic records for multiple manifestations of the same expression by having one bibliographic record serve for more than one manifestation. Often this approach uses some kind of subrecord to hold data that varies between manifestations. Several different variations on this approach have been proposed over the past ten years using either the Holdings Format or another layer of subrecords (See Note #6), and are described below:

Variations on the Single-Record Approach

III.1 The bibliographic record contains only data common to all manifestations (0.24 Options 1 and 3) with data that varies recorded in a Holdings or other type of subrecord.


The bibliographic record contains data for all manifestations in repeatable fields, if necessary. (0.24 Option 2)


The bibliographic record for one manifestation is allowed to "stand in" for the bibliographic data of the other manifestations, with data that varies for other manifestations appearing in Holdings or other subrecords. (0.24 Option 4)

The following are some descriptions of these approaches:

III.4 (CONSER Single-Record Technique). The bibliographic record for one manifestation is allowed to "stand in" for the bibliographic data of the other manifestation. The bibliographic record for the print serial is edited to note the availability of the electronic version, and to include an 856 field to access the electronic version. While the CONSER technique does not specifically describe the use of the Holdings Format, libraries who use this technique may attach holdings records for both the print and electronic manifestations to the same bibliographic record so that this approach ends up looking like III.3 (above). Like the other variants on the single record approach, the CONSER technique is applied only in cases where the bibliographic description for one manifestation can be used for the other manifestation as well.

Advantages of a Single Bibliographic Record Approach:

  1. Techniques in this category have the advantage of avoiding the creation of full catalog records for each manifestation. This is potentially a great time- and money-saver for libraries, especially when the library's clientele is not primarily concerned with the bibliographic details of each manifestation in a library's collection.

  2. These techniques can provide very clear displays for users who see all manifestations on the same record display. Conceptually, it is a way to automate the "dashed-on" entry technique used in card catalogs. They provide an easy way to group (collocate) information about closely related items in a collection, such as known reproductions. It is not surprising that this approach is much favored by archives, for handling micro and macro-reproductions, and in situations where communicating the records between systems is not a concern.

Disadvantages of a Single Bibliographic Record Approach:

  1. The Holdings Format does not currently include data elements that would be needed to describe many of the variations that occur among manifestations of the same expression. As brought out in the discussion of DP04, it begins to become desirable to define a growing list of fields from the bibliographic format for the holdings format in order to accommodate more and more materials that might vary in extent, imprint, or even title.

  2. Issues relating to sharing the bibliographic information for the reproductions handled in holdings records for copy cataloging purposes have never been resolved.

  3. Libraries have had difficulty attempting to use the single-record approach when loading and manipulating large sets of records, such as those for serial titles in aggregator databases, in online systems.

  4. With the CONSER single-record technique, if differences between the print and electronic manifestations begin to occur (such as the print ceasing publication or the electronic changing title), it becomes difficult to describe such changes on the bibliographic record for the print. It may then become necessary to recatalog the electronic manifestation.
  5. It is difficult to use the holdings format to convey complex bibliographic relationships between manifestations.

  6. As described in DP04, if a single bibliographic record contains data for more than one manifestation, it is difficult to cluster all data that pertains to a particular manifestation for display.

IV. A Bibliographic Format Approach: Linking Entry Fields?

Linking entry fields already exist in the bibliographic format that can be used to establish relationships between manifestations when each has its own bibliographic record. Of these, only the 775 (Other edition entry) and 776 (Additional physical form entry) fields as are currently specifically defined to include linking manifestations at the level of an equivalence relationship and thus are used to link manifestations of the same expression (See Note #9). Bernhard Eversberg described how these fields (in particular field 775) might be used, with a linking field containing an ID number for the expression used in the bibliographic record for each corresponding manifestation. He suggests that one manifestation-level bibliographic record could function as the "expression record," with the ID number for that record used in all other records (See Note #10).

Advantages of the Bibliographic Format Approach:

  1. This approach is simple, would use fields that are already defined (although new subfields might need to be defined), and would not require the use of any other record or table structure to achieve linking. Thus, it could be used in libraries that do not wish to create authority records at the expression level (as will be described below).

  2. This approach would allow a cataloger the opportunity to explicitly define a link between two records even when the bibliographic data in the records does not easily allow implicit linking, in other words, to "force" a link between two records. In our first Interim Report, the Working Group discussed the necessity for catalogers to have this ability in situations where no specific linking data is available in the record but when the cataloger nevertheless knows that two manifestations represent the same expression and should be collocated for display. It would also be easy to remove a link that turned out not to be accurate after all.

  3. It is possible to see this approach as being compatible with the Working Group's Linking Table model in our first Interim Report, in that a system could store the linking numbers in a separate table or record structure.

Disadvantages of the Bibliographic Format Approach:

  1. If the ID number used to provide the links between manifestations is based on the bibliographic record control number from the local system, then the intellectual effort used to identify which links are needed would not be shareable between institutions.

  2. When a bibliographic record is created to represent a manifestation that contains several components or expressions, linking would be problematic. For example, a collection of texts published in one volume without a collective title or a sound recording containing performances of multiple works may require multiple linking fields within one bibliographic record if the individual components (analytics) themselves are represented in the catalog by records for other manifestations. There would be a need to link specific fields within the bibliographic record to the appropriate expression represented in another bibliographic record. In such cases, the bibliographic record number by itself might not be a suitable choice for the linking number.

  3. If the same bibliographic record (or at least its ID number) stands for both a manifestation and an expression, this approach would do nothing to clarify the distinctions between work, expression and manifestation in the MARC 21 format.

V. The Authority Format: Expression-Level Authority Records

The Format Variation Working Group's efforts to revise AACR to incorporate expression-level collocation have led us to focus upon Chapter 25 of AACR, on Uniform Titles. While uniform titles have traditionally been used in the past to identify and collocate works, there is already some precedent in AACR for using uniform titles to also identify particular expressions. For example, a uniform title constructed using the current rules in AACR for a particular version (expression) of the Bible (work) might look like this:

130 0# $a Bible. $p N.T. $l English. $s Authorized

The Working Group is currently working on incorporating additional rules into Chapter 25 that will allow for the creation of a heading for other types of expressions as well, should a library decide that it is desirable. If changes to AACR such as these are approved, it is logical to begin looking for ways that the authority format might accommodate these headings. Just as work-level authority records are currently used to collocate headings at the work level, expression-level authority records could be used to collocate headings at the expression level.

At this stage in our work, several things are undeveloped. For example, we haven't determined what such a heading would be called: Expression-level uniform title? Expression-level heading? Expression-level citation? The Working Group has also not yet determined exactly what such a heading/citation for an expression would look like or how it would be formed. While it is tempting to think in terms of stringing additional wording onto the end of a work uniform title to differentiate an expression, there will be times when this model might become cumbersome and confusing for catalogers.

Another possible model would be to think about an existing work heading used in conjunction with one or more other data elements (such as a name heading for a translator or an edition statement) to differentiate one expression from another. These other data elements could be added to the authority record for the expression, rather than stringing the data onto the end of the heading. In this case, a bibliographic record that matches on ALL of the relevant expression-level fields in the expression-level authority record would be identified by the system as a manifestation of that expression.

Some additions to the authority format would be needed to accommodate records for expression-level headings. An additional code might be needed in either the Leader or the 008 to distinguish these records from records that represent other entities (works, persons, etc.). Additional fields would need to be added for other expression-level data, and a mechanism for doing this would need to be determined. However, there is room in the authority format for such additions as, for example, 8XX fields.

Advantages of an Authority-Based Approach:

  1. This approach expands upon a use of the authority format that is already in place. Creating authority records for expressions is a logical extension of the work that catalogers currently do to create authority records for works.

  2. Authority records for expressions would only be created in certain cases:

    Because the creation and use of these records would be optional, the cost for implementing this approach, for a library that chooses to do so, would be kept to a minimum.

  3. This approach could be used for any situation involving format variations, even when the bibliographic information of one manifestation varies from another.

  4. By using textual fields, authority records for expressions could be used to explain complex bibliographic relationships.

  5. Expression-level authority records could be easily shared between systems. There are already mechanisms in place for updating and adding to authority records in a distributed manner.

  6. Catalogers would continue to create records for the manifestations in their collections.

Disadvantages of an Authority-Based Approach:

  1. When used on its own, this approach does not provide for the creation of explicit linkages when there is no obvious element of bibliographic data that can be used to identify the expression.
  2. In order to provide adequate collocation, online system developers would need to provide the ability to index these records and use them to collocate search results. Since some library systems still do not adequately make use of current work-level authority records for collocation of search results, it is likely that we will have difficulty persuading them to implement this approach too.

VI. Discussion Questions: Combining Approaches?

Format variation is a complex problem. When it was discussed ten years ago as "multiple versions," it was clear that different interest groups had different concerns. They still do. A solution that works well within a single system may not scale well in a shared database or utility, especially when records need to be communicated. A small public library and a large research library may have very different needs for distinguishing between manifestations. Luckily, we have several tools at our disposal now that we didn't have ten years ago: the FRBR conceptual model and more system options.

If a library or archive needs an inexpensive solution to managing reproductions and other bibliographically similar manifestations, a single-record solution may be very desirable. However, we believe that this approach is not scalable to handle ALL instances of format variation on its own. Therefore, it may be helpful to think in terms of several levels of solutions to this problem, as follows.

Question 1: Could a single-record approach be combined, in the same online system, with expression-level authority records that would provide additional collocation of manifestations of the same expression that are cataloged on separate bibliographic records? This might be of interest to libraries that have already cataloged their materials in a variety of ways. (e.g. some CONSER single records, some separate records).

Question 2: Could an approach based upon bibliographic linking fields to link expressions coexist in the same system with a single-record approach? This might be an attractive option for a library (perhaps with a smaller collection) that is not interested in creating authority records for expressions but would still like to link expressions in cases beyond what is possible using only a single-record approach.

Question 3: Could a linking subfield be used somewhere in a bibliographic record (such as in a field for an expression-level heading) that would explicitly link it to an expression-level authority record (perhaps linking on the authority record number)? This would combine some elements of the bibliographic linking approach with that of the authority-based approach. It could potentially allow linking of expressions in situations when a cataloger knows that two manifestations represent the same expression but when it is either difficult to construct an intelligible expression-level heading/citation or when a cataloger chooses not to do so.

Question 4: Assuming that explicit linkages to expression-level authority records are possible (as described in Question 3), could a mechanism be created in the authority format that would allow for the coexistence of multiple expression-level authority records with the same heading string, but that do not represent the same expression (as opposed to our current method of creating authority records for undifferentiated headings that may represent more than one person, etc. in a single authority record)? A cataloger could then add textual data to the authority record (e.g. Field 667) to explain the relationship between the two expressions. In this situation the authority record, and the explicit linkages to it in bibliographic records, would function primarily as a linking device, like the Table of Reference that the Working Group described in our first Interim Report.

Question 5: Could such a system of explicit linkages be used for linking serials at the expression level, possibly making use of the ISSN? In this case, would it be better to add the authority record number (a system-supplied sequential number) as an explicit link inside the bibliographic record for each manifestation, or would it be more meaningful to put the ISSNs of each serial manifestation into the authority record itself?

Appendix A

0.24 Task Force Models for Expression-Based Cataloging

"Cataloguing Examples Demonstrating Expression-based Records
Following Option C" (See Note #11)

There are several possible record structures that might underlie the 'single-record approach' to format variations that is recommended in Option C. These include

  1. The single record could be a multilevel description (rule 13.6 in AACR2) in which what is common to all manifestations represented by the record is described at the top level of the record, and the variations between manifestations are described in a subordinate level.

  2. The single record could summarize the variations in manifestations represented by the record in the notes or in repeatable physical description elements or both; ideally, multiple elements applying to one manifestation could be linked in some fashion.

  3. The single record could describe only what is common to all manifestations represented by the record, with variations between manifestations described in subrecords, either bibliographic or holdings records.

  4. The single record could describe what is common to all manifestations represented by that record, and not include any reference to the variations.

  5. The underlying bibliographic information could be stored in separate records for each manifestation. The virtual single record called for in the rules would be assembled by the system for display. The rules in this case would constitute a conceptual approach for the cataloger and a specification for display of information, but not a description of the data structures in which the data are stored and communicated.


  1. The FRBR entity item is not to be equated with the sort of item that is the object of the Item Information fields, 876-878, of the Holdings Format. An FRBR item corresponds to the term Holdings Item as found in the general information on Item Information fields.

  2. The other options related to format variations in the 0.24 Task Force suggest more minor changes to the wording of Rule 0.24. Option A suggested simply removing the word "physical" from the wording to facilitate cataloging reproductions based on the original; Option B suggested revisions that would have allowed catalogers to not create new catalog records for manifestations that had "minor variations", as defined in the report.

  3. ALCTS CCS Committee on Cataloging: Description and Access. "Overview and Recommendations Concerning Revision of Rule 0.24." (4JSC/ALA/30) August 16, 1999: 40.

  4. On the other hand, an archivist on the Working Group had no trouble creating catalog records for an expression based upon his experience with managing groups of related manifestations in his collection.

  5. While the 2002 revision of AACR Rule 0.24 removed the actual wording "item in hand", the use of the new wording "item being described" in this rule clearly keeps the focus of the bibliographic description at the manifestation level.

  6. BarbaraTillett, "Hierarchical Approach to Materials in Multiple Formats: General Description." Prepared for the LC/CLR Airlie House Multiple Versions Forum, 1989.

  7. Bruce Chr. Johnson, principal ed., Guidelines for Bibliographic Description of Reproductions (Chicago and London: American Library Association, 1995).

  8. Option 1, as described in DP04, discusses embedding information for more than one manifestation within a single bibliographic record, and corresponds to Option 2 of the CC:DA 0.24 Report. Because we had previously dismissed this option, and also because DP04 finds considerable fault with this approach, we have deliberately excluded it from this discussion.

  9. The 775 is applied to three types of horizontal relationships: language editions, regular-print reproductions, other editions; only the regular-print reproductions represent an equivalence relationship, the other two categories link manifestations of related expressions, as the content will differ.

  10. Bernhard Eversberg, Message to the MARCList, January 22, 2002 and subsequent personal electronic communication to Jennifer Bowen, February 12, 2002.

  11. ALCTS CCS Committee on Cataloging: Description and Access. "Overview and Recommendations Concerning Revision of Rule 0.24." (4JSC/ALA/30) August 16, 1999: 44.

Go to:

Library of
Congress Library of Congress
Library of Congress Help Desk (06/28/2002)