(Partly because this has bothered me for ages and partly to try to resuscitate discussion of the DCMI Abstract Model draft....) The DCQ-RDF guidelines [1] and more recently the Abstract Model draft [2] highlight that a "Vocabulary Encoding Scheme" is the type of the resource which is the object/value of a statement. If DocumentX has a subject from LCSH (i.e. the encoding scheme used is dcterms:LCSH) then I say my:documentX a dcmitype:Document . my:documentX dc:subject my:subjectS . my:subjectS a dcterms:LCSH . my:subjectS rdf:value "some string" . i.e. the type of the object/value of the dc:subject triple is dcterms:LCSH Now suppose I want to express a relation between DocumentX and CollectionA of which it is part then I say my:documentX a dcmitype:Text . my:documentX dcterms:isPartOf my:collectionA . my:collectionA a dcmitype:Collection . i.e. the type of the object/value of the dcterms:isPartOf triple is dcmitype:Collection And the inverse my:collectionA a dcmitype:Collection . my:collectionA dcterms:hasPart my:documentX . my:documentX a dcmitype:Text . i.e. the type of the object/value of the dcterms:hasPart triple is dcmitype:Text So does that mean dcmitype:Collection and dcmitype:Text (the type of the object/value) are encoding schemes in these statements? Are the "terms" (classes) in the DCMI Type Vocabulary all also encoding schemes? (Of course when those classes are used as the object/value of a dc:type property, the encoding scheme in that statement is dcterms:DCMIType, not dcmitype:Text or dcmitype:Collection) Pete