A parametric study on the reliability of a cohesive slope is carried out to investigate the influence of spatial variability of undrained shear strength (cu). The random finite element method (RFEM), which uses random field theory and elasto-plastic finite element analysis, is adopted in this study. This study concentrates on the effect of soil variability, which is commonly measured by the coefficient of variation (COV) and scale of fluctuation (θ), on the reliability of slopes with different geometries. Various slopes having combinations of slope angles (β) and depth factors (D) are considered. The numerical analyses are carried out using Monte Carlo simulations to enable the probabilities of failure (Pf) to be estimated. The deterministic factors of safety (FOS), based on the mean values of cu, are also computed using the finite element method. The results of comparisons between the Pf and the FOS values show that θ has a significant effect on Pf for marginally stable slopes (1 ≤ FOS ≤ 1.5), even those slopes having low to intermediate values of COV (e.g. 0.1 – 0.3). Slopes having higher values of COV (e.g. 0.5 – 1), which have high FOS values (e.g. 1.5 – 5), are also vulnerable to failures depending on the values of θ.