The emphasis in distributed multimedia computing has been on developing and building systems, rather than on formally defining the precise behaviour of developed systems. We assess the suitability of a spectrum of formal description techniques for expression of distributed multimedia structures. Then we focus on three particular approaches; extended finite state machines, synchronous languages and process algebras. Representative techniques for each are assessed against multimedia requirements. In particular, we consider to what extent each technique satisfies the real-time requirements of distributed multimedia computing. The conclusions of the paper centre, firstly, on the relative benefits of these specific techniques and, secondly, more broadly, on the limitations of the standard single language based approach for formal description of distributed multimedia systems.