Haptic rendering complex objects in virtual environments is computationally intensive. In this paper we start the investigation of a new category of approach to reducing the computation in haptic rendering. Our approach is based on the hypothesis that the accuracy of haptic perception might be limited. Results of the experiments described in this paper suggest that subjects might not be able to distinguish two haptic objects if they are beyond some refinement level. This limitation of haptic perception may be taken advantage of in haptic rendering by replacing a fine object with a coarser object to reduce scene complexity.