This paper examines some of the deficiencies of the pole-placement self-tuning controller, particularly for the control of nonlinear processes. A comparison is made with switched multiple-model control and a local controller network, based on the local model network. The difference between the realisations of the local networks is also examined. All these approaches are applied for the control of the continuous stirred tank reactor. (C) 2010 Elsevier Ltd. All rights reserved.