An essential motivation behind concurrent object-oriented systems is to exploit the software modularity and reuse potential of object-oriented features in the development of concurrent systems. However, the marriage between the two disciplines has not been a happy one. In this work we propose a model that reduces the problems associated with introducing concurrency into object-oriented languages. We explore techniques for reasoning about synchronisation requirements in composite object systems operating in a concurrent environment. Furthermore we investigate safe locking strategies, where the locks may be distributed over the components of the system and identify conditions for such locks to be non-redundant. Through the recognition of a Galois connection, we achieve an elegant mathematical characterisation of the lock-distribution problem, which allows us to identify efficient strategies and tools for assisting developers with lock selection.