Scoped memory in Real-time Specification for Java achieves a level of predictability not found in applications that use the garbage collection. However, it creates new challenges for developers. First, the reference rules between scopes constrain the design of the application's memory model. Second, deciding on the appropriate number of scopes and which objects should be allocated into those scopes are not straightforward tasks. This paper presents an empirical study of this memory model using a simulation of a railway control system executed in Sun's RTS 2.2. Five different scoped memory design models were implemented in the case study and memory consumption for each was measured. Results showed that the number of scopes did not always indicate a good memory footprint, choosing the right objects/threads to be allocated in the right scopes is an important factor to be considered.