Abstract: The spatial variability of gravel particle size in residual-colluvial slopes is an objective reality that significantly impacts slope deformation and failure. To investigate the influence of particle size spatial variability on the deformation and failure of residual-colluvial gravel slopes, this study first conducted a spatial variability survey and then used random field theory to characterize the particle size spatial variability through stochastic realizations. By varying the coefficient of variation(COV),distribution trends, and correlation distances, different spatial variability characteristics of particle size distributions were generated. Numerical slope models with these varying characteristics were then developed using the Particle Flow Code(PFC), and numerical experiments were conducted. The results showed that in PFC simulations, areas with coarse particles had fewer particles per unit area, fewer connections, and lower equivalent strength, making them more prone to local instability compared to areas with finer particles. Slope failure typically initiated in regions dominated by coarse particles. Along the direction of stress adjustment, local factors of safety(LFS)were lower at the transition from fine to coarse particles, showing tensile characteristics, whereas LFS were higher at the transition from coarse to fine particles, bearing more downward forces and resisting deformation, showing compressive characteristics. The spatial variability characteristics of particle size in the residual and colluvial parts of the slope, such as directional trends, correlation distances, and COV,controlled the distribution of equivalent strength across the slope, forming different spatial variability structures. These structures influenced local stability, potential failure surface distribution, deformation patterns, and failure modes. This study provides a detailed analysis of the deformation and failure patterns of residual-colluvial slopes with different particle size spatial variability, contributing to more refined disaster investigation, prediction, evaluation, and prevention in residual-colluvial slopes.