Quality aspects become increasingly important when business process modeling is used in a large-scale enterprise setting. In order to facilitate a storage without redundancy and an efficient retrieval of relevant process models in model databases it is required to develop a theoretical understanding of how a degree of behavioral similarity can be defined. In this paper we address this challenge in a novel way. We use causal footprints as an abstract representation of the behavior captured by a process model, since they allow us to compare models defined in both formal modeling languages like Petri nets and informal ones like EPCs. Based on the causal footprint derived from two models we calculate their similarity based on the established vector space model from information retrieval. We validate this concept with an experiment using the SAP Reference Model and an implementation in the ProM framework.