Provenance research in digitized memory institution collections is mainly devoted to documenting and mapping the trajectories of the physical source documents across time, place and contexts, primarily by developing metadata standards and data models. The provenance of the digital reproduction and its relation to one or several physical source documents is however not being subjected to much inquiry. A possible explanation for this is the face-value approach with which we tend to regard digital reproductions. Looking more closely at such reproductions and their complex digitization process suggests a far from straightforward and linear provenance relation, and begs the question of what it is we actually see on the screen and what source document the digital reproduction purports to reproduce. To ascertain this provenance bond and hence the value, authenticity and usability of the digital reproduction, users need to be provided with keys such as a thorough account of the digitization process, access to various states of the digital images, exhaustive metadata for the reproduction, and not least paradata documenting the digitization process, including conversion, editing, curation, and decisions made during the process.