Within the framework of Functional Requirements for Bibliographic Records (FRBR), this paper explores how the entities of work and expression are manually identified for music MARC (Machine Readable Cataloging) data. The goal is to explore possibilities for identifying musical works and expressions automatically, based on MARC data, in order to "FRBRize" Online Public Access Catalogs (OPACs). As for music MARC data, Toccata MARC, a MARC file invented by a Japanese corporation, was selected for analysis. Consequently, the identification results contained features of music materials (mostly scores and music recordings) such as a "one medium multi pieces" (one material usually contains two or more pieces). The author found differences between the ways that records were described in classical and jazz/popular music. Because there were some problems in the ways that jazz/popular music was described (such as lacking the name of the composer), it was challenging to identify works and expressions in this genre. Moreover, it has become clear that collocating works and expressions in jazz/popular music is almost not necessary.