In this work, the effect of the estimation structure on the performance and robustness characteristics of a geometric nonlinear estimator is studied for free-radical continuous solution copolymerization reactors where conversion, composition, average molecular weight, solid mass fraction, and production rate are inferred from temperature, density, refractive index, volume, and flowrate measurements. The case study is based on the geometric characterization of the robust delectability property, yielding a set of admissible estimator structures that represent a degree of freedom in the estimator design. The choice of structure determines the estimator reconstruction rate and error propagation mechanism. Measures of estimability (i.e., robust detectability) are introduced to discriminate structures. The estimator functioning results confirm the ones drawn from the structural assessment stage, and yield that the best estimator behavior is attained with a structure different from the nominal detectability structure that underlies the nonlinear fixed-structure EKF and Luenberger techniques employed in previous polymer reactor estimation studies. (C) 2003 Elsevier Ltd. All rights reserved.