The correctness of business process models is of paramount importance for the application on an enterprise level. A severe problem is that several languages for business process modelling do not have formal execution semantics which is a prerequisite to check correctness criteria. In this context, soundness defines a minimum correctness criterion that a process model should fulfill. In this paper we present a novel approach to reason about soundness based on so-called causal footprints. A causal footprint represents a set of conditions on the order of activities that holds for every case of a process model. We identify three kinds of error patterns that affect the soundness of a process model, namely the deadlock pattern, the multiple termination pattern, and the trap pattern. We use Eventdriven Process Chains (EPCs) and Petri nets to demonstrate the applicability of our approach for both conceptual as for formal process modelling languages. Furthermore, it can easily be applied to other languages, such as UML activity diagrams or BPEL. Based on the trap pattern, we prove that the .vicious circle., that is heavily discussed in EPC literature, is unsound.