Direct multivariable model reference adaptive control (DMMRAC) procedures have been successfully used in continuous-time systems. Previous discrete versions of the algorithm required apriori knowledge of a stabilizing feedback gain matrix for the controlled plant and also an estimate of the ideal model following control in order to guarantee a bounded output tracking error.In this paper a discrete version of the DMMRAC algorithm is shown to guarantee stability of the adaptive system as well as asymptotically perfect output model following provided that the system satisfies a positive realness condition. No a priori information is needed for implementation.