Heterogeneous systems are systems that obey different functioning laws. For instance, during the design of embedded systems, it is generally necessary to study both the controller and the environment that it controls, these two subsystems being clearly different in nature. Moreover, data processing applications are also increasingly heterogeneous, mixing different technical domains such as telecommunications, man-machine interface, analog and digital electronic, signal processing algorithms. To combine these different technology domains, modeling languages and platforms generally use a hierarchical approach. This paper highlights how the hierarchy of the model and the changes of model of computation are coupled and why this coupling forbids the use of components that have inputs or outputs that obey different models of computation. In addition, this paper shows that what happens when data crosses the boundary between two domains depends on the modeling environment and it gives some means of managing component in the same level of the hierarchy.