Abstract: When specifying distributed real-time systems, decomposition and composition are the primary methods for coping with complexity. Typically, decomposition can be done in two orthogonal dimensions. First, the system is decomposed in the vertical dimension by partitioning the distributed system into loosely-coupled subsystems with well-defined interfaces. Each subsystem is then decomposed further in the horizontal dimension usually in a top-down fashion. However, with this approach it is easy to introduce unnecessary complexity due to two facts. Firstly, the interfaces between subsystems are defined before capturing their collective behaviour. Secondly, application-specific parts, that cross-cut the two dimensions, are scattered around in different components of the subsystems. Furthermore, they are tangled with other parts. In recent years the recognition of problems related with scattering and tangling in object-oriented software systems has led to an introduction of an aspect-oriented design methodology enabling different, possibly overlapping, design concerns to be decomposed separately before composing them together. In this paper specification architectures for distributed real-time systems are addressed which support capturing of collective behaviour in diverse dimensions of concerns at a high level of abstraction.