Based on an extended model reference control (MRC) theory developed here, a model reference adaptive control (MRAC) algorithm is proposed for not necessarily minimum phase digital or sampled plants. It is shown that the algorithm is capable of globally stabilizing the plants as well as achieving asymptotical model matching in a suitable sense without making a richness assumption on the reference signal. The only assumptions made on the plant are minimality and known order, and hence the delay in the plant is unknown, and moreever may be larger than the delay in the reference model.