Problems caused by concurrent access to data can be solved by locking. This, however, diminishes efficiency and may cause deadlocks among the accessing processes. In this paper the structuring of data sets is deduced from assertions about accesses which may be performed simultaneously. This structuring supports the development of operations which synchronize concurrent access to these data. Thus the provisions for locking are transferred from programs to data structures where the requests for locking actually arise. The synchronization operations will exclude all undesired interferences and will maximize the number of possible simultaneous accesses in accordance with the original assertions.