This article compares the performances of two fuzzy modeling approaches such as Mamdani (linguistic) model and Takagi-Sugeno (clustering based) model in spatial interpolation of mechanical properties of rocks. For this purpose, both simulated and measured data sets are used and prediction error and data variability are considered as performance criteria. In addition, a new clustering validity approach is used for clustering-based modeling. The results indicate that prediction performance of the clustering-based fuzzy modeling approach is higher than that of the Mamdani model. (c) 2006 Elsevier Ltd. All rights reserved.