Model checking of concurrent systems is usually done from the viewpoint of a sequential observer outside the system. We show that it may be useful to employ several communicating observers where the observers may be situated outside or inside the system. Since each observer can be checked separately, state-space explosion can be reduced to some extent. Communication among observers, however, may become complex and is not included in the checking. On the other hand, this separation of concerns opens the possibility to check communication infrastructure (like middleware) separately without checking it over and over again with the object systems built on it.