An art collection rarely arrives as a set of works waiting to be described. More often it arrives as a work, the photograph that reproduces it and the book or catalog where that photograph was published. A mural that never leaves its wall is known through its photographs, an easel painting comes back from an exhibition with a new photographic record, and a work on paper is almost always studied in reproduction. Describing the collection requires, first of all, keeping those layers separate.
A bibliographic catalog describes publications, while a work of art has other data: a technique, a support, dimensions and a creator whose role matters, because in a print there can be one person who drew, one who engraved and one who printed. That is what CDWA is for: the categories the Getty defined for works and cultural objects, and it is the format in which Janium Collect catalogs this type of collection. How CDWA models each data point is covered in Works of art in CDWA with the Getty vocabulary, and what follows deals with the decisions the collection imposes.
The three layers of a published work
When describing a work that appears in an exhibition catalog, three things come up that are easy to reduce to one:
- The work and who made it. An oil painting, a print or a fresco, with its artist and the role they played. In a print, the draftsman and the printer can be different people, each with their own dates.
- The photograph and who took it. The image that reproduces the work has its own author and its own rights, and the photographer who reproduces a mural is not the muralist who painted it.
- The book where it is published. The page of a catalog or a study is a reference and does not indicate where the work is held.
When these layers are mixed, the catalog ends up attributing to the photographer rights over the painting, or recording the book as if it were the location of the work. CDWA keeps each data point in its category, and Collect respects that separation when it extracts: the artist, the photographer and the reference to the book each go to their own field, which is what keeps the catalog from asserting something false about ownership or provenance.
The image and the catalog card
A digitized art collection usually comes with two sources per work: the image and some description, which may be a catalog card, an entry in an index of illustrations or a gallery handout. Collect reads both and takes from each what it can give. From the image comes the description of what is seen, such as the type of object, what is depicted and the composition. From the card come the documentary data that the image does not contain: the exact date, the measurements, the inventory number and the stated technique. The same division, applied to other materials, is explained in From the image and the catalog card to the record.
When both arrive together, the record is completed. When only the image arrives, the record comes out partial, with the documentary fields pending the card or the cataloger.
What the image does not allow us to assert
Three cases account for most of the errors when describing from a reproduction, and in all three Collect marks what was inferred instead of presenting it as a fact:
- Attribution without a signature or a document. A work may closely resemble the style of a known artist and be a copy, a workshop piece or the work of a follower. Without a signature or a documentary mention, the probable attribution is noted as a hypothesis and the creator field is not filled in by resemblance.
- The technique. Oil, tempera and encaustic cannot be told apart reliably in a reproduction, even less in a mural photographed under the room’s lighting. If the card does not state the technique, the field is left for whoever knows the work.
- The culture of origin. A work that depicts a pre-Hispanic scene does not therefore belong to a pre-Hispanic culture: what is depicted is the subject of the work and not its origin. It is an error that is hard to detect later, because the wrong value sounds reasonable.
A record made from the image alone is completed when the card appears or when a cataloger reviews it. Collect leaves those fields visible as pending, because an empty field gets filled later, while a plausible but false data point goes into the catalog without anyone noticing. How AI contributions are marked is covered in Collect marks AI-made metadata.
The Getty vocabulary
CDWA relies on the Getty vocabularies: ULAN for creators, AAT for types and materials and TGN for places. After extraction, Collect links each term to its entry only when the match is exact and leaves no room for doubt. That is why many terms remain unlinked, which is the intended outcome, because a wrong identifier leads whoever consults the catalog to another work or another artist. The authority control used to resolve names is described in Authority control: each name enters the catalog in one form.
The specialist’s work
Collect proposes records that a specialist reviews, with the layers already separated, the documentary data in the place where the card supplied it and the gaps marked where the image fell short. Less time goes into typing each field and more into resolving doubtful attributions and the decisions that only someone who knows the work can make.
To continue the conversation
If your institution has a collection of paintings, murals or works on paper and you are thinking about how to describe it without losing the distinction between the work, its reproduction and the book, write to us at info@janium.com. We are interested in knowing how you resolve those decisions in your catalog today.