ABSTRACT: Numerical methods combined with centrifuge tests are used to investigate the distribution and development of soil stresses and reinforcement tensile loads in geosynthetic-reinforced soil (GRS) structures. In this study, system stability indicated by the factor of safety (FS) of GRS slopes is calculated by limit equilibrium analysis. Stress information under various stress states is evaluated using finite element analysis. Advanced models and an integration algorithm are implemented in finite element code to enhance the simulation results. The proposed numerical models are validated by centrifuge tests of two GRS slopes with different backfill densities. Numerical results indicate that soil stress mobilisation can be described by the soil stress level S, which is defined as the ratio of the current stress status to peak failure criteria. For both slope models, as loading increases, backfill stresses develop and propagate along the potential failure surface. Mobilisation of soil stress was non-uniform along the failure surface. Immediately after the stress level reaches peak (S = 1), strength softening initiates at the top and toe of the slope at approximately FS = 1.2. The slope settlement rate and reinforcement tensile load increase significantly when soil softening begins. The softening occurs randomly and irregularly along the failure surface, and the formation of the soil-softening band completes at approximately FS = 1.1. The failure surface corresponds to the locus of intense soil strains and the maximum tensile loads at each reinforcement layer. Additionally, the numerical results show that the initiation of soil softening and the failure of the slope occurred earlier in the slope model with low backfill density. The numerical results support the view that peak shear strength, not residual shear strength, governs system stability. Last, the distribution of maximum reinforcement tensile loads with depth was highly uniform at low g-level and became trapezoidal at high g-level. The peak value was located at approximately mid-height of the reinforced slopes. This observation contradicted the triangular distribution with depth assumed in current design methodologies for geosynthetic structures.