The aim of this paper is twofold. One is to propose a general framework for examining the effect of fuzzy partitions on the performance of fuzzy rule-based systems. The other is to compare fuzzy partitions with interval partitions using the proposed framework through computer simulations on realworld pattern classification problems.