A filtering technique is proposed for on-line estimation of the temperature, monomer conversion, initiator conversion, and the entire molecular weight distribution in a batch methyl methacrylate polymerization reactor. The technique uses a detailed polymerization model combined with on-line measurements of conversion, temperature, and the molecular weight distribution, taken at different discrete time intervals. The polymerization model includes a chain-length-dependent termination rate constant which allows the prediction of the molecular weight distribution for common free-radical polymerization conditions. Comparisons between modeling and experimental results show that the polymerization model gives good predictions of the monomer conversion and the molecular weight distribution in the polymerization system. The performance of the estimation scheme is tested for cases of strong gel effect conditions leading to a bimodal molecular weight distribution, and poor initial conditions. Finally, off-line experimental data are used to test the algorithm under actual reactor operating conditions.