By modifying some previously developed results for nonlinear identification and control using recurrent neural networks, the present authors propose a new neural network identifier in block form, and, based on this model, a control law is developed by. combining sliding mode and block controls. This neural identifier and control law allow satisfactory trajectory tracking for general nonlinear systems. Applicability of the new design is illustrated, via simulations, for robust tracking control of stepping motors. (C) 2003 Elsevier Science Ltd. All rights reserved.