Abstract: An experiment with milk vetch input to replace chemical nitrogen fertilizer was established in 2008. This study investigated the soil aggregate composition and soil organic carbon distribution with the wet screening method to classify the aggregates and determine the soil organic carbon content of aggregates with different particle sizes. Four treatments were selected, i.e. rice-rice-winter fallow conventional fertilization, NPK (fallow), rice-rice-milk vetch conventional fertilization, NPK (milk vetch), rice-rice-milk vetch 20% chemical nitrogen reduction for both early and late rice, 0.8N+PK (milk vetch) and rice-rice-milk vetch 40% chemical nitrogen reduction for both early and late rice, 0.6N+PK (milk vetch). The results showed that the soil aggregate composition was significantly changed by adjusting nitrogen application rate. Mean weight diameter (MWD), geometric mean diameter (GMD) and >0.25 mm soil aggregate content were the highest in the treatment of NPK (milk vetch). Compared with winter fallow treatment, the contents of >2 mm and 0.25 ~ 2 mm aggregate in the treatment of NPK with milk vetch returned to the field increased by 15.3% and 3.81%, respectively, but there was no significant difference. Compared with the normal nitrogen application, the content of >2 mm aggregate was significantly reduced by 41.8% ~ 57.6% in the treatment of 0.6N+PK (milk vetch) and 0.8N+PK (milk vetch) (P<0.05). Compared with winter fallow treatment, amendment of milk vetch to field together with the reduction of nitrogen application significantly reduced soil organic carbon content in aggregates of all particle sizes (P<0.05). However, there was no significant difference in soil organic carbon content in aggregates among treatments after milkvetch returned to the field. The enrichment coefficients of organic carbon in aggregates with different particle sizes treated with NPK (fallow) were all greater than 1. The enrichment coefficients of organic carbon in aggregates >2 mm and 0.25 ~ 2 mm treated with NPK (milk vetch), 0.8N +PK (milk vetch), and 0.6N +PK (milk vetch) were greater than 1. Soil MWD and GMD were positively correlated with the contents of >2 mm aggregate and 0.25 ~ 2 mm aggregate (P<0.05), but not significantly correlated with the content of soil organic carbon in aggregates. In conclusion, the return of milk vetch to red soil under two cropping rice per season may be beneficial to soil structure, but no matter whether nitrogen application is reduced or not, it is still necessary to adjust the carbon to nitrogen ratio of exogenous input to achieve the stability of soil structure and improvement of soil fertility.