||Stress-dependent soil-water characteristics, conjunctive surface and subsurface flows and slope stability are governed by various parameters. In this study, laboratory testing was carried out to study the influences of stress level, drying and wetting history, as well as the particle size distribution on the stress-dependent soil-water characteristic curve (SDSWCC) of a completely decomposed granite (CDG) from a fill slope located behind Hong Kong Sanatorium and Hospital in Hong Kong. Results are also compared with the SDSWCCs of a completely decomposed volcanic (CDV) and a completely decomposed tuff (CDT) in Hong Kong. For both of the intact and recompacted specimens, hysteresis loop size decreased when the specimen was subjected to a higher stress level and in the second drying and wetting cycles. The air-entry value of a specimen with smaller maximum particle size is higher and the influence of D60 and D50 values on SDSWCC is larger at suction lower than 5 kPa while the values of D30 and D10 have greater influence on SDSWCC at suction higher than 10 kPa. The SDSWCC of the CDG specimen had a lower air-entry value, a larger hysteresis loop and higher desorption and adsorption rates at low suction range when compared with those of CDV and CDT specimens. The influences of conjunctive flow, stress level and drying and wetting history of soil-water characteristics and rainfall pattern on numerical analyses were also investigated. Numerical transient seepage analysis was carried out by the Conjunctive Surface-subsurface Flow (CSSF) program and slope stability analysis was determined by a new numerical program Limit Equilibrium Conjunctive Slope (CSLOPE) program developed in this study. With the consideration of the conjunctive flow effect, the depth of the wetting front was deeper and the factor of safety was lower than the analysis without considering the conjunctive flow effect. A wetting soil-water characteristics also gave a deeper wetting front and a lower factor of safety than a drying soil-water characteristics in the analyses.