【Objective】 Water erosion is one of the factors driving land degradation, water non-point source pollution and soil erosion, which involves a complex processes that include soil detachment, sediment transport and deposition owing to the interaction of raindrops and overland runoff. Soil detachment refers to the dislodging of soil particles from the soil surface triggered by water flow, and soil detachment rate (or soil erosion rate at erodible beds) is a key parameter in process-based erosion models. However, the study on the pattern of soil erosion rate among different texture soils is still unclear. 【Method】Therefore, an indoor experiment of running overland flow on erodible beds with remoulded soils of different textures was conducted. Thirty five combinations included 2 slopes (5° and 15°), 4 discharges (2, 4, 6 and 8 L•min^-1 ) and 5 types of remoulded soil (0, 30%, 50%, 70% and 100% in sand contents) were carried out on a steel flume (3 m long, 1 m wide and 0.35 m high) adjustable in slope gradient. The five types of remoulded soil were prepared by mixing red soil and normal engineering sand at different mass fraction and were labeled as S1, S2, S3, S4 and S5 treatment, respectively. Those texture were silty clay, clay, sandy clay, sandy loam and sandy soil from S1 to S5, respectively. The prepared soil was packed layer by layer (5 cm depth per layer) in the flume to ensure the soil uniform in bulk density (1.35 g•cm^-3). The packed flume was pre-wetted (30 mm•h^-1) with a rainfall simulator to make the soil moisture saturation. When the moisture content in the soil reduced to 30%, the test started, and ended 15 minutes after surface runoff was initiated. During the test, flow surface velocity was measured with the dye tracing method (using potassium permanganate), and flow width, mean rill depth and mean rill width were measured by using a steel ruler at 20 cm intervals along the tick marks along the edge of the flume (amount to 15 sections). Runoff volume and sediment yield were measured with plastic bottles (at 30 s intervals) and plastic buckets (at 1 min intervals) at the outlet of the flume. 【Result】Results show: (1) Slope gradient and flow discharge were the two basic factors affecting soil erosion rate, both show a close relationship, but the former as more significant. Besides, soil texture had a significant effect on the soil erosion rate, making the latter appear in “monomodal” distribution pattern in the same test condition, with the peak being the maximum in S3 treatment (sandy clay, 50% in sand content); (2) The mean rill depth and rill cross-section width-depth ratio represented obvious regularity with the variations of slope gradient, flow discharge and soil texture. Furthermore, soil erosion rate closely correlated with mean rill depth and the rill cross-section width-depth ratio (r=0.943 and -0.865, P<0.01, respectively), which could be used as indicators of slope erosion and sediment yield; (3) Slope gradient, flow shear stress, stream power and unit stream power were all remarkably related to soil erosion rate, especially the unit stream power was the best hydraulic parameter to describe soil erosion rate (r=-0.911, P<0.01), and their relationship can be well described quantitatively with a power function (D_r =49.96P_r^2.07, R²=0.795); and (4) Considering the effect of soil properties on slope erosion, soil cohesion was introduced into quantitating the relationship between unit stream power and soil erosion rate (D_r =165.22P_r^2.36 C ^-0.44，R²=0.816), so as to make the quantitation more reliable. The new equation based on soil cohesion make the tests more dependable in predicting soil erosion rate regardless of soil texture.【Conclusion】All the above described findings show that soil erosion on slopes varies regularly with soil texture, and unit stream power and soil cohesion are proper parameters in quantifying soil erosion patterns. Since acquisition of the parameters in the equation is relatively simple, the equation has a wider adaptation range and a higher reference value in practical applications.