We compare the expressive power of Hoare (i.e., CSP style) and Milner (i.e., CCS style) synchronizations for defining graph transformations in a framework where edges can perform actions on adjacent nodes to synchronize their evolutions. Furthermore, nodes can be communicated and merged. We show that the expressive powers of the two synchronization models are different, but no one is greater than the other. Finally, we show that in many interesting cases the behaviour of a synchronization model can be mimicked by the other one using suitable translations for the rewritten graphs. Key words: graph transformations, Hoare synchronization, Milner synchronization, Synchronized Hyperedge Replacement, mobility, expressiveness.