This paper presents a new fuzzy interpolative reasoning method in the sparse fuzzy rule bases based on so-called similarity relations of fuzzy sets. By this reasoning method, an inference consequence can be simply obtained, and is a normal and convex fuzzy set without any limitation, which shows the potential ability of the proposed method in real-world fuzzy applications.