In this paper, a new local measure of nonlinearity is introduced to quantify the sensitivity of controller performance for linear controllers operating on a nonlinear plant. The measure, called the performance sensitivity measure, quantifies the departures from optimality of locally linear quadratic regulators. A root mean squared measure, the RMSPSM, is proposed to summarize average nonlinearity effects over a region of interest. The measure applies to nonlinear systems that admit a controllable and observable linearization. (C) 2004 Elsevier Ltd. All rights reserved.