Hammerstein systems are nonlinear models that are used in many domains for their simplicity and physical meaning. This paper presents a simple iterative method for generating good starting values which can be used to initialize the numerical nonlinear optimization of the cost function. In the absence of model errors, the presented method is proven to converge locally to the true model values in the noise-less case. (C) 2004 Elsevier Ltd. All rights reserved.