This paper presents a stochastic study of slope stability in unsaturated soils for steady state flow conditions. In order to account for material non-linearity and the spatial variability of soil properties, an elasto-viscoplastic finite element program has been combined with random field theory within a Monte Carlo framework. The results of multiple realisations have been evaluated in terms of reliability. Focusing on the influence of spatial variability of matric suction, the results highlight the importance of spatial correlation length and degree of anisotropy in the assessment of unsaturated slope stability.