In this paper, an adaptive neural network controller for the control of nonlinear dynamical systems is proposed. The new approach is adaptive in structure, and unlike standard adaptive controllers, uses no explicit model of the process in the design. Traditional neural networks are not practical in adaptive environments because of the large number of weights normally associated with them. In the proposed structure, the controller network has very few connection weights and hence is well suited for real-time implementation. This new neural network controller is shown to exhibit very good performance as shown by the control simulation of a nonlinear continuous stirred tank reactor and a pH neutralization process.