The Categories for the Description of Works of Art (CDWA)
comprise a set of guidelines for best practice in cataloging and describing works
of art, architecture, other material culture, groups and collections
of works, and related images. CDWA is not a data model, but it is arranged in a conceptual
framework that may be used for designing data models and databases, and for accessing information. CDWA includes around 540 categories
and subcategories of information. A small subset of categories are considered
core in that they represent the minimum information
necessary to identify and describe a work. CDWA includes discussions,
basic guidelines for cataloging, and examples. You may print
an overview of the CDWA categories and definitions as a PDF
(see left navigation).
The CDWA is maintained by the Getty Vocabulary Program. For comments and questions, please write to email@example.com.
CDWA and other metadata element sets
CDWA is mapped to other standards and metadata element sets
in the Metadata Standards Crosswalk.
What is CDWA Lite?
CDWA Lite was an XML schema to describe core records for works
of art and material culture based on CDWA and CCO. CDWA Lite
records were intended for contribution to union catalogs and
other repositories using the Open Archives Initiative (OAI)
harvesting protocol. In 2010, the CDWA Lite schema was enlarged and integrated into the Lightweight Information Describing Objects (LIDO) schema, available on the CIDOC site.
What is CCO?
Cataloging Cultural Objects: A Guide to Describing Cultural
Works and Their Images (CCO) includes rules and examples
for a core subset derived from the CDWA categories and the VRA Core
Categories. Its target audience is the image cataloging community. CCO is available on the CCO web site and in hardcopy from ALA and on Amazon.com.
History of CDWA
CDWA is a product of the Art Information Task Force (AITF),
which encouraged dialog between art historians, art repositories,
and information providers so that together they could develop
guidelines for describing works of art, architecture, groups
of objects, and visual and textual surrogates.
Formed in the early 1990s, the task force was made up of representatives
from the communities that provide and use art information:
art historians, museum curators and registrars, visual resource
professionals, art librarians, information managers, and technical
specialists. The work of the AITF was funded by the J. Paul
Getty Trust, with a two-year matching grant from the National
Endowment for the Humanities (NEH) to the College Art Association
CDWA has been regularly updated in order to remain current, to include guidelines for new media works, and to become ever more inclusive, with examples for cataloging works from many cultures.
Purpose of CDWA
CDWA provides a set of guidelines outlining best practice for documenting works of art, architecture, and other cultural works. CDWA also provides a framework to which existing art information structures may be mapped and upon which new data modeling may be referenced, in order to inform about the content of data elements. In addition, the discussions in CDWA identify vocabulary resources and descriptive practices that will make information residing in diverse systems and in the cloud both more compatible and more accessible.
CDWA has been mapped to or used as the basis for various art cataloging and information systems. CDWA is mapped to the Cultural Objects Name Authority (CONA) and it may be mapped to other data standards. (CONA compiles titles, attributions, depicted subjects, and other metadata about works of art, architecture, and cultural heritage, both extant and historical. CONA is linked to the AAT, TGN, ULAN, and the Getty Iconography Authority (IA). Through rich metadata and links, CONA may provide a powerful conduit for research and discovery for digital art history. The first large projects for CONA are under consideration by the Getty Research Institute, and may include using CONA as a resource to capture information about works depicted in visual surrogates.)
The use of the CDWA guidelines will contribute to the integrity
and longevity of data and will facilitate the inevitable migration
of data to new systems as information technology continues
to evolve. Above all, using standard content as described in CDWA will help to give end-users consistent,
reliable access to information, regardless of the system or data model in
which it resides.
It is our hope that these guidelines will provide a common
ground for reaching agreement on what information should be
included in art information systems and data models, and what information
will be shared or exchanged.
Authority files and data structure
As data moves into ever more linked and open environments, various issues regarding data structure are being addressed by the user community. CDWA originally recommended a relational data structure, where records
for objects/works are linked to each other in hierarchical
relationships. However, the CDWA guidelines may also be applied to graph data modeling or other systems.
CDWA recommends maintaining
separate local files or authorities for related visual works, related
textual materials, persons/corporate bodies, locations/places,
generic concepts, and subjects. An alternative approach is to use authorities already available in vendor systems. Authority information should be linked to or directly derived from the Getty Vocabularies or other standard authorities. Authority information about
persons, places, concepts, and subjects may be important for
retrieval of the work, but this information is more efficiently
recorded in separate authority files than in records about
the work itself. The advantage of storing ancillary information
in an authority file is that this information needs be recorded
only once, and it may then be linked to all appropriate work
records. Authorities described in CDWA are hierarchical;
given that authority entities often require multiple broader
contexts, a polyhierarchical structure is recommended.
Note that local authorities should be linked to standard resources, such as the Getty Vocabularies. It is recommended to contribute to these standard resources, where possible.
Linked Open Data (LOD)
A current trend in managing art information is to increasingly make data about art, architecture, and cultural heritage objects available as Linked Open Data (LOD). CDWA advocates the use of LOD. When data is linked and open, it means that data is structured and published according to the principles of Linked Data, so that it can be both interlinked and made openly accessible and shareable on the Semantic Web. The goal of linked open data is to allow data from different resources to be interconnected and queried, thus making it more useful. Relevant standards for LOD include the CIDOC Conceptual Reference Model (CRM), which provides an extensible ontology for concepts and information in cultural heritage and museum documentation. It is the international standard (ISO 21127:2014) for the controlled exchange of cultural heritage information. Also watch for developments from the LinkedArt community.
CDWA was formulated for the needs of those who record, maintain,
and retrieve information about art information, including
the academic researcher and scholar. The categories and subcategories
that are indicated as core are those that the
task force agreed represent the minimum information necessary
to uniquely and unambiguously identify and describe a particular
work of art or architecture. In the meantime, these assumptions have been periodically tested by analyzing common practice and dialog with the expert user community.
Note that which categories are considered core can and indeed
should vary depending upon the end-users whom the particular
art information system are intended to serve, the mission
of the specific institution, and a number of other factors.
Display vs. indexing
CDWA often deals with differences between information intended
for display and information intended for retrieval. Information
for display is assumed to be in a format and with syntax that
is easily read and understood by users. Such free-texts or
concatenated displays may contain all the nuances of language
necessary to relay the uncertainty and ambiguity that are
common in art information. In addition, CDWA assumes that
certain key elements of information must be formatted to allow
for retrieval, often referred to as indexing in CDWA. CDWA
advises that such indexing should be a conscious activity
performed by knowledgeable catalogers who consider the retrieval
implications of their indexing terms, and not by an automated
method that simply parses every word in a text intended for
display into indexes.
In CDWA, display fields are often described as free-text fields
(which may be alternatively be concatenated from controlled
fields, if necessary); indexing fields are intended to be
controlled fields. CDWA advises the use of controlled vocabularies;
CDWA describes when categories should be controlled by a simple
controlled list (e.g., Classification), an authority (e.g.,
Creator), or by consistent formatting of certain information
(e.g., Earliest and Latest Dates) to ensure efficient end-user
Specificity and exhaustivity
Specificity refers to the degree of precision or granularity used (e.g., campanile rather tower). However, catalogers should use terms only as specific as warranted by authoritative sources. As with all indexing, it is better to be accurate and broad rather than incorrect and specific.
Exhaustivity refers to the degree of depth and breadth that the cataloger uses, often expressed by using a larger number of indexing terms or a more detailed description. However, is it useful to index every possible applicable term? If not, where do you draw the limit? Index the most important or most prevalent characteristics.
Uncertainty and ambiguity
Explain any controversies or ambiguous issues. If an issue is in dispute, it is critical to the intellectual integrity of the record to not express it as a certain fact.
In order to correctly represent the information and allow scholarly research, indicate uncertainty and ambiguity as necessary. The cataloger should never assume, never choose one choice over another, and never state as a fact something that is debated among experts. Sources may reflect disputes about any number of characteristics of the work, including the attribution or dates for a particular work. When multiple suggestions have been made, include the most important, in the method allowed by individual elements.
Where a choice must be made for preferred information, prefer the information as accepted by the repository of the work. Other information, including conflicting opinions, should also be included provided the source is expert and authoritative. Always cite the source of the information.
Unknown and undetermined
What should the cataloger do if core information is limited or not available? When an element is indicated as required, this means that the element must be included. However, it is recognized that occasionally data for any element may be missing during the cataloging process.
Knowing that information is unknown or undetermined is important to users, particularly for the required core fields. Values for required fields must be supplied, even when the information is unknown or uncertain. When the information is unknown at the time of cataloging, include an appropriate designation indicating the state of knowledge or availability of information. This issue is discussed at various points in the CDWA subcategories, as appropriate. For fields that are not core, the cataloger may leave the field blank or null, or include unknown etc. if so desired by the cataloging institution. Note that null values will not be displayed to end-users and will likely be omitted in transfer of data; values such as unknown should be include in data exchanges and publications of the data.
Knowable vs. unknowable information
When information is unavailable at the time of cataloging, the cataloger may use values such as unknown, unavailable, undetermined, or not applicable, provided documentation or context explains to the user the meaning of these words for the given field. For required fields and in other contexts, including such values is better than omitting the information entirely, particularly when the possibility exists for the record to be enhanced in later passes at cataloging, or to provide clarity in retrieval and research. Has the cataloging institution simply forgotten to include the information? Or has the field been considered, but at this time the information is not available? Including such values for unavailable information clarifies the situation for users, while a blank field does not.
However, the cataloger must be careful not to imply that a fact is unknowable simply because the cataloger happens not to know it (generally because time and editorial priorities do not allow him or her to do the research required to resolve the issue). If a fact is knowable (but just not known by the cataloger), it is in some cases better to omit the fact entirely rather than to state it with qualifying phrases such as or or probably, because this implies more knowledge of the issue than the cataloger has.
In various subcategories in CDWA, suggestions are made regarding how to deal with unknown information, uncertainty, and ambiguity for the given field. One of the most common ways of dealing with such information is to state the vagary in a display field, and then to index with all authoritative, probable terms for that display. Another common method, for fields without accompanying display fields, is to index using a broader term that is known to be correct, rather than a narrow term that could be incorrect.
Example one: If two authoritative sources disagree on the date of creation of a work (one prefers ca. 1510, but another of equal authority prefers ca. 1525), this dispute may be referenced in a display date: created ca. 1510 or ca. 1525, and explained more fully in the Descriptive Note. Then the dates for retrieval on creation date should encompass the full range of possible dates, also estimating a range of a few extra years to include the uncertainty of ca.: Earliest: 1505; Latest: 1530. NB: If the work is in the possession of a repository, the opinion of the repository should take precedence over varying opinions by outside scholars; however, the full scholarly debate should be represented as possible, with methods of doing so varying dependent upon the field.
Example two: If the available authoritative information indicates a work is made of metal, but the cataloger consulting an illustration feels it looks like silver, the cataloger should never rely upon their own judgment with such lack of substantiating evidence. The cataloger should index the material as the general metal rather than risking the introduction of erroneous more specific information.
Example three: If one source calls the work by an anonymous artist Frenchand a second source calls it Flemish, for display, the cataloger should not necessarily state that the work is French or Flemish in a note field, because this implies that scholarship agrees it could be either. Instead, the cataloger should state the cultural origin of the work based on the most reliable, recent sources. Perhaps with further investigation, the cataloger will discover that although it was in the 19th century considered French, modern scholars agree it is Flemish. If indeed modern scholars differ on their opinions and are equally divided, then the cataloger may indeed state French or Flemish. If the work was formerly known as French, the cataloger should index French and Flemish for retrieval.
Disagreement among sources
Know your sources. When two sources disagree, prefer the information obtained from the most scholarly, authoritative, recent source.
Indexing important information
Descriptive notes and other text fields are not an access points for retrieval. Therefore, if a cataloger mentions important information in such a note, in order to facilitate retrieval, it must be indexed it in the appropriate controlled fields elsewhere in the record, using controlled terminology (such as AAT, TGN, ULAN, the CONA IA, CONA itself, or another controlled vocabulary such as Iconclass or Library of Congress Authorities). The recurring issue of correct indexing is discussed as appropriate in various CDWA subcategories.
It is critical for the cataloger to cite sources of information. In order for the information to be considered reliable, it must be derived from authoritative sources. Online sites to which any member of the public may contribute are not considered reliable. In general, authoritative sources are compiled or researched by verified, known scholars and experts, and published (online or in hardcopy) by reliable authoritative publishers. Scholarly catalogs, text books, monographs, encyclopedia, dictionaries, and journal articles authored by an expert are reliable sources. A scholar’s spoken opinion or email may be a source, if the person is a known expert on the topic (such sources must also be cited). Information may be derived from unpublished documents such as inventories, letters, bills of sale, photo mounts, and inscriptions on the work itself, if proven to be authentic by experts. Repository records are considered the preferred reliable source of information about a given object; if such records are reflected on the museum Web site, the site may be considered authoritative. Specific reliable sources are listed elsewhere in CDWA, in context for various subcategories.
Send questions and comments to us at firstname.lastname@example.org.
Revised 28 February 2019