This paper proposes a new virtual sensing approach for on-line monitoring and regulating process variables of the injection molding process. Based on the nonlinear observer theory, virtual sensors estimate process behavior using easily obtained measurements of machine variables by exploiting their dynamic interaction. A nozzle pressure virtual sensor during the so-called "nozzle resistance test" was developed. Results of experimental evaluation on a commercial injection molding machine establish the feasibility of the proposed virtual sensing approach. Furthermore, a sensitivity analysis indicates that the proposed virtual sensor delivered consistently better and more robust performance against parametric uncertainties than simple open-loop model prediction. (C) 2004 Society of Plastics Engineers.