This paper presents a formal grammar for discussing data flows and dependencies in audio processing graphs. A graph is a highly general representation of an algorithm, applicable to most DSP processes. To demonstrate and exercise the grammar, three central problems in audio graph processing are examined. The grammar is used to exhaustively analyze the problem of scheduling processing nodes of the graph, examine automatic parallelization as well as signal rate inferral. The grammar is presented in terms of mathematical set theory, independent of and thus applicable to any conceivable software platform.