A practical application of self-tuning generalized pole placement (GPP) is discussed. The application, which involves the control of a five-axis, electrically actuated robot manipulator, is presented for two reasons. First, it illustrates the performance of a novel neo-classical multi-step predictive self-tuner in an important area of applied research—namely, robot control. Second, since the manipulator in question is electrically driven through a harmonic gearbox, the investigation has a general relevance to the area of self-tuning electromechanical servomechanisms. Two forms of GPP algorithms are compared, one based upon a controlled autoregressive integrated moving average model and the other upon a controlled autoregressive moving average model. The relative merits are discussed in the context of (i) single-input single-output and multiloop robot joint control, (ii) programmed setpoint control, and (iii) the use of the performance tuning aids with which the GPP algorithm is equipped.