[Objective] The knowledge about soil hydraulic characteristics plays an important role in optimal soil water management and conservation practices, and hence is essential to evaluation of soil water storage capacity and prediction of water and solute fluxes in soil, which may affect crop yield, fate of agrochemicals, and eventually groundwater quality. Rock fragments are the main factor in rocky purple soils affecting hydraulic properties, such as saturated hydraulic conductivity and water characteristic curve.[Method] Two soils derived from different parent materials, i.e. purple shale and purple mudstone, separately, were studied in this paper. The soils were divided into groups, separately, different in rock fragment size (0.25~2, 2~5 and 5~10 mm) and in rock fragment content (0, 30%, 50%, 70% and 100%) for test. Soil particle size composition was measured with the pipette method. Microscopic morphologies of the purple shale, the purple mudstone, and the groups of soils were observed with a ZEISS EVO 18 scanning electron microscope. Water characteristic curves and saturated hydraulic conductivities of the soils were measured with the pressure film method and constant head method, respectively. Two transfer functions (PTF1 and PTF2) were established to predict saturated hydraulic conductivity with the aid of the BP neural network. PTF1 consisted of four input variables (rock fragment content, rock fragment size, initial soil bulk density, and texture), while PTF2 did of two more input variables (air entry value and S index (absolute value of the slope at the inflection point of water characteristic curve)) in addition to the four in PTF1.[Result] Results show that saturated hydraulic conductivity was significantly related to rock fragments content, S index, and air entry value. Saturated hydraulic conductivity and S index of the soil improved with increasing rock fragments content, while air entry value did reversely. Moreover, with increasing rock fragment size from 0.25~2 to 5~10 mm, saturated hydraulic conductivity increased by 2.3 times on average. Rock fragment size had little effects on air entry value and S index. The maximum error between predicted and measured values was 0.2225 and 0.1058, and the average error, 0.0756 and 0.0422, respectively for PTF1 and PTF2. The regression coefficients of PTF1 and PTF2 was 0.9416 and 0.9873, respectively. The geometric mean, geometric standard deviation, root mean square error, and AIC index of PTF2 and PTF1 was 1.27 and 1.17, 5.57 and 1.70, 0.16 and 0.060, 2.94, and 53.2808, respectively, and the correlation value of PTF2 was lower than that of PTF1. indicating that PTF2 performed better than of PTF1 in predicting saturated hydraulic conductivity.[Conclusion] In summary, the presence of rock fragments significantly affects the hydraulic properties of the purple soil, thus improving water conductivity and decreasing water holding capacity of the saturated purple soil. The transfer function PTF2 based on the neural network performs well in predicting saturated hydraulic conductivity of rock-containing soils.