Bracket languages play an important role in the syntax analysis of programming languages. We investigate the parallel recognition and analysis of these languages as a first step towards a parallel working compiler. The main result consists in the design of an appropriate algorithm, which can be executed on hypercubes as well as on related bounded degree networks. In the average case we can achieve an optimal speed-up, meaning that q processors together can analyse bracket words of length N in time O(N/q), if we restrict ourselves to employing no more than √ N processors.