Model-based experiment design techniques are becoming an essential tool for the rapid development and refining of process models. One of the areas where an effective design can be most useful is the identification of the kinetic parameters of a model. When complex kinetic networks (i.e., parallel, consecutive reactions) are involved, parameter correlations play a significant role because they often prevent the solution of experiment design calculations, make parameter identification more difficult, and decrease the statistical validity of the resulting models. It is therefore important to obtain estimates of the parameters as uncorrelated as possible and this article presents new optimal experiment design criteria that include explicit measures of correlation as objective function or as constraints and are able to target the experiments to the improvement of specific parameter(s). Through an illustrative application to an epoxidation example, the new approach proposed is demonstrated to be very successful, highly flexible, and more effective than the standard experiment design criteria in both reducing the uncertainty regions of the parameters and improving the reliability of the estimates. (c) 2008 American Institute of Chemical Engineers.