Rachel Wilson and I would like to take up a discussion around what happens to our ideas about off-line collections when they take life online. This is something that we are nutting out as part of our work on The Screen Media Research Archive (an ANDS-funded project that aims to establish a national management standard for the digital archiving of higher education screen media research produced in a variety of formats). We would like to look at some of the underlying conceptual challenges for defining ‘collections’ in this context, including: What protocols already exist for defining collections by provenance, materiality or subject and how do these apply (or not) to a research archive? Are donated collections static by definition? What level of granularity applies to collections within collections? Are there a minimum/maximum number of items required to form a collection? Should collections describe items that are not currently included (that have been de-selected for example)? What value is there in defining collections in terms of their users rather than the items per se?