Soil aggregate stability influences several aspects of soil physical behavior, such as water infiltration and soil erosion (Amézketa, 1999). We investigated the soil aggregate stability characteristics in the framework of the ‘Grain for Green’ vegetation rehabilitation project at the Loess Plateau (China) by using the Le Bissonnais (1996) and the modified Yoder (1936) methods. Both non-grazed grassland and afforestation revegetations were considered. The size distribution mode was always >2 mm for the fast wetting test (FW) in the non-grazed grassland communities. This fraction accounted for approximately 40% to 80% of the total soil weight. The wet stirring (WS) test showed a distribution similar to that determined by the FW test. For the slow wetting (SW) test, 80% or more of the non-grazed grassland soil fragments was >2 mm. The mean weighted diameter (MWD) that was determined by the Le Bissonnais' method was different among the tests and land uses. For the FW test, all the plant communities were significantly different from that of the recently abandoned grazing on grassland at the 0–20 and 20–40 cm depths. In the Artemisia sacrorum community, the maximum MWD was approximately 3 mm for the 0–20 cm depth. There were no significant differences among the plant communities after 3 years of afforestation and 7 years of non-grazing of the grassland. The MWDs were lower in the afforestation area than in the non-grazed grassland area. The maximum MWD value from the FW test was approximately 1.8 mm and was significantly lower (b0.5 mm) for cropland. The MWD of the modified Yoder method was positively related to the slow wetting and wet stirring (WS) tests of the Le Bissonnais' method (n = 20 and r = 0.83 and 0.87, respectively). In the Loess Plateau, revegetations by non-grazed grassland and afforestation are efficient means of increasing aggregate stability and decreasing soil erodibility. The aggregate stability under non-grazed grasslands is higher than that under afforestation. The effect of revegetation is persistent, which makes it a suitable long-term management practice. Compared with the modified Yoder's method, the FW test of the Le Bissonnais' method is better at determining aggregate stability differences among land uses and is recommended for future studies.