A class of finite - state, finite - memory stochastic control problems is considered. A minimum principle is derived. Signaling strategies are defined and related to the necessary conditions of the minimun principle. Min - H algorithms for the problem are described.