In this paper, the decentralized minimal controller synthesis (DMCS) algorithm is shown to be extremely effective in controlling nonlinear multivariable plant with unknown parameters. Both the stability and robustness of this new adaptive algorithm are investigated. It is shown that if the plant coupling terms and nonlinearities are slowly varying, compared with each adaptive loop bandwidth, then the DMCS algorithm yields an asymptotically stable closed-loop system. However, if the coupling terms and/or nonlinearities are rapidly varying, tracking errors are shown to be bounded. An estimate of the upper bound of error convergence is derived. Implementation studies of the algorithm applied to a two-degree-of-freedom manipulator are included. The results of these experiments are closely aligned with the theoretical predictions.