The paper deals with an heuristic generalization of the traditional classication rules by incorporating within sample dependencies. The main motivation behind this generalization is to develop a new classication rule when training samples are not random, but, jointly equicorrelated.