I'd like to give my version of explanation of the relation between FRBR and Bibframe. The FAQ http://www.loc.gov/bibframe/faqs/ states quite briefly "BIBFRAME Work identifies the conceptual essence of something; a BIBFRAME Instance reflects the material embodiment of a Work; a BIBFRAME Authority identifies a thing or concept associated with a BIBFRAME Work or Instance; and a BIBFRAME Annotation provides a new way to expand the description of a BIBFRAME Work, Instance, or Authority. " Although close to the conceptual model of FRBR, Bibframe classes can be modeled by a sort of a "reduction" or "information extraction" procedure like this: - extract all conceptual core descriptions from the FRBR classes - that is, what is author-controlled and not owned or handled by publishers, vendors, sellers, companies, intermediates, readers, libraries, patrons. It's the author's name, the title of the author's work, the mode of the work (text, sound, image, ...) Repeat that for an artist or a translator who realizes other work versions of the original work. Put that into bf:Work classes. - extract all publisher-controlled information out of the FRBR classes - the name of the publisher, the decision about the publication structure (issuance, edition, version), the date and place of publication, the carrier, the physical extent, the tags like ISBN/ISSN, the prices for the product, all the addresses and names the publisher gave to the author's work etc. Put that into bf:Instance classes. - extract all library-controlled information out of the FRBR classes - the formal description, the classification, the subject cataloging, the call number, the shelf location, authority control information, (maybe also descriptions of the library service for access to printed and electronic resources, it's not clear right now) etc. Put that also into bf:Instance. - optionally, capture additional information that is created by companies/vendors/sellers/intermediates/readers and is somehow related to the author's work into bf:Annotation classes. - optionally, lift some of the information recorded in bf:Work or bf:Instance to a "higher level of trust" by adding bf:Authority classes. What helps me to understand the Bibframe model is the changing scope of information control over time, during the lifecycle of a work. It does matter who is allowed to add or modify information. This is reflected well in the Bibframe model. I hope this makes sense, and demonstrates that Bibframe modeling by using a reduction procedure can also be applied to all kind of information found in bibliographic resources, be it MARC or another format, of course also RDA information sets. Although, I would like a clearer separation of publisher-controlled and library-controlled information. It may help later to attach ONIX or something similar to Bibframe seamlessly, or an ontology for enterprise resource management. And most of the library's added information is metadata which is not changing the physical nature of the publisher's manifestation in bf:Instance. So bf:Instance sounds not like a good class name choice for that, maybe a new bf:Service class (as a subclass of bf:Instance) would be an option? It reveals also there is a "metadata production chain" from the author to the reader/user, completely independent of library catalog rules. It can be predetermined which Bibframe data sets must be touched for the purpose of efficiently creating or updating catalogs. The less data sets must be touched for updates, the better (e.g. adding holdings to a union catalog). And Bibframe can explain to the public what is happening with works in the library space. It helps tracking the path of a work from the author's idea of creating a work to the reader in the library who is searching for it. Still, Bibframe isn't complete yet, and needs a lot of contributions. Jörg Am 23.05.13 17:19, schrieb Laura Krier: > I apologize if this is an issue that has already been covered, but I > just want to get some clarity. > > In BIBFRAME, a work might be approximately the equivalent of a FRBR > expression? An instance might be approximately equivalent to a FRBR > manifestation? And items would be expressed in annotations?