An improvement is presented for the tunnel algorithm introduced in [17] for determining the minimum number of rules required by a fuzzy system for modeling and identification in the SISO case. The original tunnel algorithm can deal with only Mamdani type fuzzy systems, while the Testing Point algorithm proposed in this paper, can be used to construct Takagi-Sugeno type fuzzy systems with consequent parts of any order and has the potential to be further extended to the MISO case. Numerical examples are given to illustrate the idea and improvement.