An adaptive extremum seeking controller is presented for the optimization of the production rate of a continuous stirred tank bioreactor with Monod's kinetics. This controller is saturated outside a domain of interest and a reduced-order observer is designed to estimate the substrate concentration in the bioreactor. It is shown that once a persistence of excitation condition is satisfied, the convergence of the parameter estimates to their true values is guaranteed. Semi-global asymptotic stability for the output feedback closed-loop system is proved based on Lyapunov's theorem. Simulation results are shown to illustrate the performance of the proposed approach. (C) 2004 Elsevier Ltd. All rights reserved.