In hierarchical system design, refinement allows to incrementally increase the level of detail in the system description. Several notions of refinement have been investigated in process theory and in object-oriented frameworks. From a process algebraic viewpoint, an object as the basic unit of structure and control can be understood as a process. Then action reification - the replacement of actions by transactions in object-oriented specification - corresponds to action refinement in process algebra where actions are replaced by process terms. Questions of distributed control, synchronisation, and serialisability translate naturally from the object-oriented framework into process theory and vice versa. In both frameworks, correctness critera exist which prevent the use of refinements if they would cause serious problems in synchronisation and distributed control. We compare an object-oriented and a process algebraic approach for action refinement, both based on an event structure semantics. We show how restrictions ensuring correct refinement known from the process algebraic framework can be applied in the object-oriented approach. Moreover, results from process theory on the preservation of system properties under refinement now become accessible for verification in the object-oriented setting. All concepts are illustrated by a small case study.