Rill erosion is a serious environmental problem threatening the agricultural production safety and sustainable societal development in the Loess Plateau, China. It is, therefore, very important and urgent to quantify dynamic process of the rill erosion. To that end, an indoor simulation experiment was carried out the scouring and backfilling method. Erosion process was analyzed through observing variation of the shape of rills and measuring volume of erosion along the rill segment by segment. Cumulative amount of erosion along the rills was worked out by refilling the separated rill segments with water to their original elevation. The sediment concentration was calculated by dividing the cumulative amount of erosion at each segment by the total volume of the water flow during the erosion period. A typical silty-loam soil was sampled from the Loess Plateau of China (109° 19′ 23″ E, 36° 51′ 30″ N) and used as experimental soil in this study. The soil contains 16.31% clay (<0.005 mm), 61.35% silt (0.005 to 0.05 mm), and 22.34% sand (>0.05 mm). The soil was air-dried and then passed through an 8 mm sieve. In order to a complete continuous erosion process, an experimental trough, 3.0 wide and 12.0 m long, filled with loess, was used. In the trough six experimental rills, 0.1 m wide and 12 m long, were built to simulate well-develped rills. The trough was put at horizontal position for soil packing. At the bottom, the trough was densely packed with a 5cm thick layer of local clay soil, made compact with a hammer to simulate the plow pan layer, 15 g cm^-3 in bulk density. The upper 20 cm of the trough was packed with loess in layers, about 5 cm each, to make the soil even in bulk density, ranging approximately between 1 150 to 1 200 kg m^-3. The soil surface was raked to make it as rough as natural condition. Cumulative volume of erosion of a rill was the sum of erosion volumes of the eleven rill segments (0～0.5 m, 0.5～1 m, 1～2 m, 2～3 m, 3～4 m, 4～5 m, 5～6 m, 6～7 m, 7～8 m, 8～10 m, 10～12 m) of a rill. Total amount of erosion sediment was measured volumetrically by a collecting vat at the outlet of each rill during the experiment. The experiment was designed to have five slope gradients, i.e. 5˚, 10˚, 15˚, 20˚ and 25˚ for the trough, three flow rates, i.e. 2, 4 and 8 L min^-1 and three replicates. Results show that the process of rill erosion was not constant along a rill, with cumulative rill erosion and sediment concentration increasing exponentially and margin of the increase declining with rill length till the extreme in the end, which means that cumulative amount of erosion and sediment concentration increases rapidly at the initial segments of the rills, and the increase rate (the slope of the curve) attenuates gradually to approach zero along the rill. Flow rate and slope gradient were the two major factors affecting rill erosion, but the former seemed to have more influence than the latter. The comparison of the cumulative amount of erosion measured with the volumetric method with the result of direct measuring at the rill outlet validated the accuracy of the former, and the comparison of the sediment concentration measured with the water backfilling method with the result of direct measuring demonstrated that the former was quite accurate in and applicable to investigating distribution of erosion sediment along a rill. The findings of this study may serve as certain basis for consummating the research on soil rill erosion process, assessing sediment concentration from rill erosion and establishing a prediction model for farmland rill erosion.