Abstract: It has been recognized that the spatial variability of soil properties exhibit obvious anisotropy. Generally, the mechanical parameters of soils vary more rapidly in vertical direction than in horizontal one. In order to reflect this point of characterization and to study the effect of anisotropic variability on slope stability, it is more rational to simulate soil properties in anisotropic random fields than in isotropic ones. Applying the local average theory can realize the discretization of 2D anisotropic random fields efficiently, and using the shear strength reduction method can obtain the safety factor of slopes without giving a predefined sliding surface. Combining these two methods with Monte-Carlo simulation can easily get slope reliability. So a Monte-Carlo stochastic finite difference method is formed based on the idea, and it is adopted to investigate the effects of anisotropic spatial variability of soil properties on the slope reliability. The numerical results show that the anisotropy of random field has a significant impact on the slope reliability, and ignoring the feature will result in varying deviation. In addition, the vertical variation of soil properties has a more significant influence on the slope stability.