We investigate process mining for the declarative Dynamic Condition Response (DCR) graphs process modelling language. We contribute (a) a process mining algorithm for DCR graphs, (b) a proposal for a set of metrics quantifying output model quality, and (c) a preliminary example-based comparison with the Declare Maps Miner. The algorithm takes a contradiction-based approach, that is, we initially assume that all possible constraints hold, subsequently removing constraints as they are observed to be violated by traces in the input log.