Abstract: In order to study the hydraulic conductivity of unsaturated Q₃ loess, an equipment for sampling undisturbed loess is designed, four vertical and horizontal loess columns are acquired successfully and five remolded soil columns with different dry densities are prepared. The tests on horizontal soil columns of nine samples are carried out. TDR moisture meters and thermal conduction suction probes are embedded to monitor volumetric water content and matric suction in different sample sections. The results indicate that the infiltration velocity is influenced by crack direction and dry density of the soil columns. For the undisturbed loess samples, the infiltration velocity of vertical samples is greater than that of horizontal samples at the beginning stage of soaking tests. When the wetting front exceeds 50 cm, the infiltration velocities of vertical and horizontal samples get close to each other. The differences of diffusion coefficient between undisturbed vertical and horizontal samples exist distinctly below the saturation degree 0.6. When the saturation degree surpasses 0.6, the diffusion coefficients are nearly identical. As to the remolded loess samples, the effect of dry density on the diffusion coefficient mainly exists in the lower saturation degree area. Under the same dry density and moisture content, the unsaturated hydraulic conductivity of the remolded samples is greater than that of the undisturbed samples in the lower saturation degree area, however, in the higher saturation degree area, the unsaturated hydraulic conductivity of the undisturbed samples is greater than that of the remolded samples. The sampling method for the undisturbed loess and the research findings of those tests give some referential values to the same studies.