A new neural linearizing control scheme (NLCS) is proposed for controlling nonlinear chemical processes. In the proposed NLCS a radial basis function (RBF) network is used to linearize the relation between the output of the linear controller and the process output. The learning of the RBF network proceeds adaptively to minimize the difference between the output of the pre-defined linear reference model and the process output, After the neural network is fully trained, the apparent dynamics of the process becomes linear. The determination of the linear reference model and the heuristics about avoiding the effects of the unmeasured disturbances are considered. Simulation studies on a continuous stirred tank reactor and a pH process are carried out to evaluate the performance of the proposed control scheme. Copyright (C) 1996 Elsevier Science Ltd