A methodology for the synthesis of engineering Anticipatory Systems is presented. It accounts for both random and fuzzy aspects of uncertainty introduced in process control, and develops mathematical measures of performance. The random, probabilistic component quantifies the uncertainty of occurrence of an event. The fuzzy, possibilistic component quantifies the imprecision in the meaning of an event. Facts about the system are represented as fuzzy information granules of the form: g = (number of carts carrying parts) is (small) is (likely). The performance measures are used as an input to the diagnostic and control functions which are assumed by a Production-Rule System. Using a fuzzified Bayes formula as the link between present and future states, a model-based system estimates at any time t, current performance as well as anticipated performance at time t + At. A control action can be taken at time t, based on both current and anticipated performance. Accordingly, the system can change its current state on the basis of both the current and the anticipated future state. The agency for the prediction is a model of the system and/or its environment which is internal to the system. The synthesis of such system is demonstrated using a model of a nuclear reactor. The model of the reactor is constructed using procedural programming methods and is coupled to a symbolic program which assumes the overall control function.