Predicate transition networks are presented as graphical models of production rules. Nodes called places represent predicates, and nodes called transitions allow interconnection among places so that the production rules can be programmed. Although it is a model for distributed computation, there is no provision for priority or control. A matrix representation is used and some reachability properties are explored algebraically.

In the examples, the transition nodes can be programmed to do ands and ors, and change or reject certain variables, though their basic function is to move tokens from place(s) to place. The places are also considerably different for different examples. They can hold tokens as in Petri nets or they can hold n-tuples which need to be mixed and reproduced by the transitions.

The complicated representation of context-free and context-sensitive grammars give still another role for places and transitions. Also, the authors are using this formalism to construct a speech understanding system, but they don’t give any example of that in this paper.