Nabil Kashyap, Swarthmore College, and Lindsay Van Tine, University of Pennsylvania In this presentation, Early Novels Database project (END) collaborators Nabil Kashyap and Lindsay Van Tine will offer perspectives on the possibilities and perils of reframing the special collections catalog as a collaborative datastore for humanities research. Among other activities, the END project includes curating records from regional special collections, developing standards for enhancing catalog records with copy-specific descriptive bibliography, and publishing open access datasets plus documentation. Work on END therefore excavates basic questions around what thinking through library holdings as data might actually entail. What ultimately constitutes “the data”? What do they do? For whom? Starting from Leigh Star’s notion of the boundary object, this presentation explores the theory and praxis of MARC as a structure of knowledge that can allow “coordination without consensus.” The MARC records at the core of the END dataset, the result of meticulous work on the part of institutional catalogers, serve as “boundary objects”–that is, they serve as a flexible technology that both adapts to and coordinates a range of contexts. These contexts, in turn, can have very different needs and values, from veteran catalogers to undergraduate interns, special collections to open source repositories, and from projected to actual uptake and reuse of the data in classrooms and research. These shifting contexts call into question just what the “data” is. It will look different to a cataloger, an outside funding organization, a sophomore, a programmer, or an 18th c. scholar. What might appear straightforward–creating derivatives, for example–instead reveals a host of issues. Transforming nested into tabular data brings to light frictions between disparate assumptions as to the unit of study, whether a work or volume or a particular copy. Privileging certain fields either effaces the specificity of transcription or sacrifices discoverability. There is no transparent “data dump”; instead, every act of transformation reinscribes a set of disciplinary and institutional values. Viewing collections as data is as much about opening up data as about actively demonstrating and to an extent prescribing research possibilities.